Several people have asked if these fixes address the jail bars in the C128. They definitely don't. The jailbars come from the VIC-II itself and the problem is especially bad on the C128. I did look at my 128 last night and opened the modulator. It's already sharp so needs no mods there. The NTSC version has color smearing but you can't just bypass the L1 inductor otherwise it lowers the color level too much. (It does fix the smearing though.) So for now, just run a wire to the RCA jack to convert it to composite if you want to work on the C128. :-)
Hi Adrian, it's good to see my old PAL C64 out and about again. I actually bought myself a PAL C64C from eBay now to replace it. But i didn't get any video cable to plug it into anything. I ordered a c64 video to SCART cable a while back but it never got dispatched due to the COVID-19 lockdown (still hasn't). So i just watched this and now 20 minutes later i've converted the RF output to composite. Works like a charm! But i do notice jail bars on the composite output on this C64C. Not sure if this is a similar issue to the one described on the C128. For reference the sticker on the modulator says "P/N: 251916-03A" & "1829F" written underneath in bold.
@HalfSpeedMastering Yep - I'm also here to mention the possibility of adding a Schmitt Trigger to the output of the old VIC-II to clean up the signal :)
I cleaned up the pins of the vic chip (and the socket) and saw a great improvement, though it wasn't gone. I also have a hunch the oscillator is aging.
L1 and C3 are in series which forms a band pass circuit. I am assuming that C3 is not being used exclusively as a DC blocker for Q3. Most likely the resonant frequency is 3.58 MHz, which is the center frequency of the color DSB (double side band) sub-carrier. It was probably placed there to make the RF output NTSC compliant to meet FCC regulations. In a compliant NTSC signal, color bandwidth is less than half that of the luma. This was done to help reduce dot patterns when watching the RF output on a B/W TV and to eliminate adjacent channel interference. Combined, the Y (luma) and C (Chroma) need to fit inside the 6 MHz window to meet FCC requirements. By eliminating the choke, you are allowing the chroma signal to have as much "sharpness" as the luma. So it must be there for regulatory purposes. This will make the composite and the RF output look great, but I bet if you used an old B/W TV you would see dots in the colored areas. I used to use a B/W TV growing up in high school for my Atari 800, so there were people that used B/W TV's. Money was tight... LOL
Spot on reply - you nailed it. Cross channel interference in the analogue tv days would have been a big no-no, in the crowded (and strictly regulated) RF spectrum.
@@Petertronic I realized after I posted my comment that "most" commodore 64 RF installations would be closed - The TV would only be connected to the C64 with no other channels present via the RF TV/GAME switch. If you were to mix / combine the output of the modified C64 with other channels, like various city information channels used to do, there would be severe interference. I set one up a long time ago and had to purchase really expensive filters to make it work.
@@Craig1967 Adrian wasn't interested in the RF output, but I wonder if removing the BP filter might result in bleed into the audio subcarrier too. That might result in hissy audio from the TV used on the RF output.
You are most likely right, question is: Why is a bandpass- filter required for NTSC and not for PAL? Could it be the larger distance between channels of n n the UHF spectrum (8MHz in W.Europe)? Further, rather than kill the filter, would it be possible to improve it to have a sharp cut-off at exactly the right frequency? (A standard value of 22mH sounds like a rough approximation.)
L1 and C3 in the chroma circuit form an L/C “tank” circuit. You can tell this, because they’ll either both be in series with the signal path (as here), or they’ll both be in parallel across the signal path, with one end tied to ground or a supply rail. From basic electronics, we know that inductors pass DC, and capacitors pass AC. In series like this, we therefore know that DC (and low frequencies) will be blocked, and high frequencies will be blocked by the inductor, so our response curve (Bode plot) has to have minima at DC and high frequencies. Therefore, there must be some frequency that maximizes the signal flow, so this is a band-pass filter. This is followed by a similar low-pass filter composed of C4 and R4 to the one we saw with Q1/Q2 above. This likely attenuates the frequencies on the high side of the pass-band, resulting in an asymmetrical response curve, though it might attenuate on the low-side of the pass-band.
I just modded my PAL version with MD6-UE3603 - definitely a visible improvement! I also added the Comp. Out removing the RF out link but instead of soldering directly the wire to the right side connector pin and "smearing it" I soldered onto the 140 Ohm resistor leg directly above it (around 2cms) just in front of the 2 Caps - it's easy to locate the leg using a multimeter. Cheers!
Hey Adrian, one issue with cutting the LPF caps and shorting the inductor. The reason they are there is because transistors don't act well when they receive signal that has power above the transistor's bandwidth. So if your transistor goes up to 1MHz at unity, and you put 2 MHz into it, you can get ringing or even oscillation, and reduction of dynamic range - the parasitic oscillation will eat into the total (Vpp) dynamic range that it shares with the desired signal. So what you want is: 1. a transistor that can manage the bandwidth you need for your signal and 2. an lpf filter that will filter out everything above, and leave everything below. I bet the commodore factory subbed in cheaper transistors that didn't quite manage full video bw, and that's why they added the severe lpf. You want to check the GBW of the transistors in there, and then replace with something better if necessary. However! bear in mind in general in electronics mfr companies those parts would get replaced for the cheapest part weekly, so what you test will be different to what others have. So it's best to just find a good substitution and use that. Then, build an lpf around it that passes the right bandwidth. you want to check by using an RF signal generator, to see what freqs get passed. Just a guess, but i think you want -3 dB at the top freq your signal is supposed to carry. I think you can just go by LC and RC low pass filter equations to figure out the capacitor values - or just use a few values and experiment, probably simpler. You might take the inductors and just unwind them a bit. But if you completely remove the lpf, you're asking for trouble :-) Hope this helps. Enjoy!
I think that's not an issue here, we are dealing with 4-6MHz max, I bet those transistors can handle up to 100MHz or multiples of that. But shorting the inductor is not fine, because it is there to filter out luma signal from the input of the chroma buffer. I'm surprised that shorting it seemingly has no negative effects. I agree, probably reducing the inductance would be more elegant. Maybe it's enough to put a ferrite bead on the place of it. The original 22uH probably delays the chroma signal too much, causing color smearing.
Fantastic video. One of your most interesting ones yet.This was a much needed modification and I really like the way you explain the circuitry theory. Learning a lot from you!
I'm sure that those capacitors and one inductor you removed to get better sharpness were mainly there for FCC type acceptance. The filter was probably for limiting the bandwidth for NTSC to insure the signal fits inside a standard TV channel.
The caps almost certainly had to do with FCC acceptance, but it may be worth noting that computer RF modules are absolutely terrible at keeping their output within a single channel, and are only suitable for uses where they're the only thing feeding an RF cable. Even if one wants to hook a computer or video game system up to an RF TV, one will get a better picture feeding S-video into a a 2000s-era video switchbox with a built-in RF modulator than one would get by using the built-in RF connection.
I took the plunge and did this today. I've had this Rev 250407 for 37 years. It made quite a difference. Great research and excellent explanation of this fix! Bravo.
I've now modded my PAL C64 (but I went the extra hassle and removed the modulator to be able to remove the cap from behind!) and indeed picture quality has improved! Thanks a lot!
Regarding Hanover bars - PAL works by inverting the phase of a portion of the colour carrier every other line. The idea is to eliminate hue errors that occur in NTSC. PAL receivers don't have a hue control for this reason. Unfortunately, to invert the colour carrier accurately, with no phase errors, is quite a difficult to achieve. These phase errors would normally be canceled out in the receiver with the delay line - at the expense of vertical colour resolution - but without that, you'll see the slight hue errors.
It's probably totally a non issue for normal video especially as you sit across the room, but sitting close to a computer display with solid colors, it really looks disturbing. (to someone who isn't used to it)
@@adriansdigitalbasement That's exactly it. The first PAL receivers in the 1960's used the PAL-S standard, S for 'simple', which didn't use a delay line for cost reasons and relied on the human eye to average out the colours. On a related topic; I wonder if the ZX Spectrum has any problems with Hanover bars. The ULA in the system outputs unmodulated video which is converted to PAL externally. However, the colour carrier isn't inverted, the baseband video is inverted on the ULA's output. This makes the colour modulator simpler. It shouldn't have the phase issues I mentioned because the colour carrier should be the same phase every line.
@@256byteram the main problem with the spectrum is the dot crawl caused by using two separate oscillators. They fixed this in the later 128 design by 1) having the ULA generate RGB output and 2) clocking everything from a single 17.7345MHz oscillator.
Did this to my 250407 with the MD6-VA3403. Looks great on my 1702 monitor. Simple composite output is a bonus also. Thanks Adrian! Just a warning to anyone trying this: If you prefer the softer output do not do this! Perhaps just bypass the inductor and don't cut the caps.
Adrian, the mods you're doing here are common in the Atari 8-bit world for the XL machines. They have much worse video output than C64's do out of the box, and they even recommend installing a switch to disable composite video to help the luma-to-chroma leakage in s-video. I'm so glad you've done the legwork to finally bring similar mods to the C64. I did your capacitor clip and inductor bypass to my 250407 tonight and the improvement is outstanding. This is an excellent video!
Although this video was released a while ago, I didn't want to do the mod until I educated myself enough to understand what's actually going on here. Over the last couple years I spend hundreds of hours working on these computers and finally understand now what you're trying to do. I did the full 4-change mod on the 3403 and the picture is much clearer and sharper than it was before. For years, I blamed it on the digital conversions that occur when displayed on a modern monitor. The RF to Composite mod is pretty handy too, since most people don't have an RF compatible TV around anymore... Regarding the 3405 RF Modulator, bypassing/removing the inductor does also make the image clearer. Excellent work Adrian and Thank You!
D1 and D2 are for biasing the transistors, otherwise they would be completely off at the bottom of the waveform during sync pulses. The capacitors and inductor you've removed are most likely for EMI suppression (FCC compliance).
I like these circuits with analog tricks, today we tend to use chips for everything, it is useful to see how you achieve stuff with simple transistors.
After watching several of your videos, I decided to try out some repairs/mods myself. I have bought a non-working C64 to repair and restore, now I just have to wait for parts. Thanks!
PAL viewer here! After seeing the unknown modulators for PAL machines, I dissassembled my C64s to check them. I don't have the proper cables to see the video output, so I'll just confirm the model numbers. Here's what I got: OG C64 250425 board, Mitsumi MD6-UE3605 modulator; C64C 250466 board, Commodore 51A1-10 Rev. 2 modulator. On my machines the model number wasn't on the steel can, but printed directly on the PCB.
My PAL 250425 has a RF-modulator called "251025-05", and it also has another label reading "5095". It's not like any of the ones shown in the video, and features a big metal can (inductor?) in the middle which has its screen soldered to the lid. Having a closer look inside is therefore impossible without a soldering iron at hand. Another detail, instead of a channel switch, it has a trimming-screw in its place.
This is amazing! I installed Perifractic's VIC-II² PAL/NTSC Switcher Kit in my main C64 and I was getting really bad color smear on the PAL side (it was originally an NTSC machine). With the inductor bridged in the MD6-VA3405 modulator the PAL output is now fantastic. Unbelievable difference considering how much smear I was getting!
I'm about to install a VIC-II² PAL/NTSC Switcher Kit in my main C64, an original NTSC one like yours, so that's great information to know. Did you install the Lumafix components? I've left mine off for now and thought I'd see how it looks on my setup.
@@Ikrananka I did install the Lumafix components. It helped with the jailbars as you'd expect once adjusted, but obviously did nothing for the color smearing issue.
Just modded my (entirely stock) 250407 with MD6-VA3404. Looks fantastic! RF can now proudly claims to have Adrian's Digital Basement 2020 mods applied.
I used to service TV's in the UK. Hanover bars are almost never seen. Only ancient early 70's tv's or very cheap tv's would exhibit them. 99.9% of TV/monitors from the 80's onwards used PAL decoder IC's (with delay line) that accurately deal with and correct the alternate line phase errors and they cannot be seen. If they are showing up on a signal that's going into a converter box, that converter is going back to the 70's by taking shortcuts : i.e. not bothering to do the phase detection and correction properly (which requires some considerable DSP processing by the way).
Great timing. I just finished soldering the same exact rf modulator bypass board that you showed in the beginning of the video. Oh well, I guess I'll try this mod first and then experiment further with the replacement board to see if it makes any difference. Thumbs up for going through the PAL version too for us here on the other side of the pond.
I just came across this video, as well as part 1. This is a fantastic mod. Simple but effective. Thanks for all the effort you make creating this for everyone. 🙏
This is great, thanks Adrian. I just wheeled out my C64 two use in a nerd project, I used it as a twitter client 7 years ago but thought I create a little Weather BASIC program pulling from an online API. The video on mine looks quite blurry to my eye on a TV across composite, so I am looking forward to trying out some of your mods.
There are a few pcb designs out there for replacing the rf modulator including the one he tried. It worked, he just didn't see the need to make that vs just doing a few minor mods on the existing one. He could make his own pcb variation but I'm not sure there's a need out there for another option. This was something that wasn't out there and that anyone can do.
Mods for the 5 pin DIN 326298 board The 326298 generates the composite video on the main board instead of in the modulator. No mods need to be done on the modulator. Remove C73 to increase sharpness. It makes a big difference. Check the value of R10. If it's 300 ohms, replace it with a 120 ohm resistor or parallel it with a 220 ohm resistor. That will increase the brightness. C73 and R10 are inside the RF shield near the VIC chip. While you have the case open, check the value of C10 and C11. They are the filter capacitors for the SID. If they are 2.2nf, replace them with 470pf. That will improve the audio.
I suggest simply jumping the inductors. Don’t snip them. By snipping, the remaining leads stumps could move around and slip down into the bottom of the unit when you heat them up and solder the jumper to them. Jumping across the component takes it out of the circuit anyway, and jumping it would make reversing the mod very simple in the future as well.
Adrian, love the channel, but stop looking at yourself in the monitor! lol. Seriously though, viewers want to feel as though you're talking to them directly. When you look straight into the camera lens, you accomplish this. When we see you looking at yourself instead of us we're not as engaged as we could be. Not engaged means not invested. I understand that you just want to make sure you remain in frame the whole shot through. I also understand that it can be difficult for some to get into the habit of looking into the lens and not the monitor. Trust me, I've been in video production a long time. It may seem like a tiny thing but it matters to viewers more than you think. You're a great presenter. Best of luck in the future! Glad to see the channel growing and getting some much deserved love. :)
Great video Adrian, I have the MD6-UE3603 PAL modulator on my 250407 board too and I found chopping out that cap improved the sharpness of my video out very nicely!
Best simple C64 mod ever.... why didn't somebody figure this out 35 years ago? I did not add the 100nf cap, as it was a bit tight for my soldering iron and skills. The image is much sharper, and the reduction in color bleed is very apparent. Thanks again for these mods.
I did the 100nf cap, just because i was there anyway, but I'm not sure there's a need. As for why they didn't do it 35 years ago, good question.. There have been composite replacements, maybe anyone who has the skill, opted to make and sell those instead of simple DIY options.
I run svideo (YC) only. 250407 ntsc breadbin. VICII2. I did this mod and it’s good all around (NTSC & PAL) on both my HP 4:3 elite display port medical document viewing monitor and Sony PVM. Side note: yes, the VICII2 does fit with SIDFX on this board. Even had room for mechboard64 and FU32 dual tetrode vacuum valve headphone preamp and heat sinks all around. Considering cramming an OLED audio spectrum analyzer in as well.
The Hanover bars are created due to the nature of PAL fixing the NTSC tint issue by alternating the phase of the chroma signal per line. That's what PAL stands for: Phase Alternating Line. This has the effect that phase changes are cancelled out so the tint of the colors is stable. But at the cost at those artifacts. The proper way to eliminate these is not to average the colors between the two lines, but to use a comb filter. SECAM by the way has half the vertical chroma resolution because it actually uses a delay line for the chroma signal as it is transmitted every even (or odd, not sure) line. That's why it's called "Sequential Color And Memory".
The quality loss with half resolution chroma is the same as in digital video 4:2:2 (with chroma information each line) and 4:2:0 (with chroma information every odd line).
I did the mod on a PAL MD6-UE3603 modulator, this modulator has been taken from a 250407 board and used on a new SixtyClone board which I built myself. I am feeding the signal through a RetroTink X2 and the output was a bit blurry. Not too bad, though. After this mod, I can definitely see a difference in the sharpness. It is a lot less blurry. Thanks for this tip!
Thanks a million Adrian ! I was looking for a way to improve my 250407 pal board picture quality on my crt monitor and it improved so much ! I am very grateful THANKS !!
Thank you Adrian for a simple fix too clean up the svideo signal. I just performed this on my PAL model ASSY NO 250407 - with VIC 6569R1 1183 video looks much better as you showed. Keep up the videos, I love all your content, how you transfer your knowledge to us all and presenting style.
It's been forever since I've done analog electronics but L1/C3 looks like a classic bandpass filter. The combination is filtering out frequencies below and above a certain range. Probably intended to clean up noise before amplifying the signal.
Future video suggestion: Improving VIC-20 video quality. I know that are at least a couple of S-Video mods that can be done but I wonder if anything similar to what you've done for the C64 is also possible.
That's funny, I actually noticed I got a color image from my pal c64 on the Luma pin yesterday. And was quite puzzled about it. But your comments around 27.45 explains it! Thanks
It's not hard to automatically eliminate hanover bars WITHOUT loosing resolution these days, IIRC the method was regularly used from the early 90s on better PAL TVs but is mostly forgotten because pretty much no one discuss "high end analog video" any longer. I suspect a manual color (phase) dial exactly like those often seen on old NTSC TVs would allow manual elimination too. The alternating phase shift was added to combat the phase-related color shift that plagued NTSC (NeverTheSameColor) and meant that without compensation any phase errors would results in the hanover bars which on the TVs of the day was mostly not visible anyway. The cheapest way to remove them is to average the two lines as you mention, but since the phase error is either static or change very slowly (minutes if not hours) it's possible to instead compute the phase error over time and then use that to get correct colors and full resolution without hanover bars. The obvious way today (and I suspect in the 90s) would be to compute a phase error estimate per-frame (compare each line-pair and average, possibly throw out outliers) and then do a slow/long decaying average on that (to reduce the chance of it introducing glitches) to get the compensation value for display purpose. This results in very good resistance to slow phase error changes without loosing resolution or even showing the wrong colors after the first few frames (say less than 0.2 seconds).
I promptly went down stairs and did this mod to my 250407 board. The quality has improved, but it seems like when using the s-video the checkerboard pattern and jail bars seem more pronounced, it may just be all in my head though. I didn't take a before and after pictures so I cant be sure. The old LCD TV I'm using doesn't have the best picture quality, but its the only thing I have with s-video. This is definitely a worth while mod that doesn't cost anything but a little bit of time, I wish this knowledge was around back in the day.
Very well done video, Adrian. And thanks for pointing out the quick jump links in your comments, as I may very well come back to this when/if I mod my machines.
The most interesting video about C64 color problems! But I would like to ask you where can you find the 'yellow PCB' placed on top of the TEBL PCB? Thanks so much for an answer ...
Now that I've seen the schematics of the modulators, I have an idea: I had noticed once, that when I connected the chroma output to my TV's composite input I get a picture. Not worse than using the real composite output. My theory is, that the color smearing/fringing on these larger contrast areas might be due to the luma signal passing back through the capacitors C6/C7 on the NTSC modulator into the chroma output. (C11/12 on the PAL version) I haven't tested it, but if you intend to use S-Video only, you might try cutting those two capacitors out. Or, if you want to keep the color composite output, you could add a fast Schottky diode in series with each of these two capacitors.
Wow that is some serious backfeeding of luma and sync into the Chroma! I've never seen that work on any of my machines before. Yeah S-Video only it likely would improve quality if you break the connection like it did on my little bypass board. Since I don't really have any PAL machines to play with I can say for sure.
The coil, capacitors and resistor in the chroma input form a tuned circuit that was probably designed to resonate at the color carrier frequency. So if you bridge the inductor you also inadvertently reduce the color signal.
Thanks for the amazing video. Great to see someone spending a lot of time to get a perfect result. Amazing quality at 25:30 and (monchrome) at 27:05. Keep it up!
L1 - C3 is a band-pass filter, likely for the 2 MHz chroma band, which has a carrier at 3.579 MHz above the luma carrier. C4 would be a low-pass filter.. the only thing above the chroma is the FM audo band though... it may be needed to prevent the transistor from oscillating at a higher frequency, usually a good practice in designing HF circuits to damp the input that way. By cutting the inductor out, the circuit is changed from a band-pass filter to a low-pass filter with a -3 dB point relative the the base resistor R4. You may get some additional noise coming in from the main PCB (digital logic noise) but it does not seem to affect the signal so it seems ok... I think some considerations were put in the design of the circuit to meet FCC emissions guidelines for NTSC signalling... not really relative anymore... Cheers,
Great video. I did this mod to one of my C64s with great results, similar to yours. I also tried to do something similar to one of the early 1982 motherboards (326298 Rev A) that don't have the the chroma output. On those motherboards, the video out does not go through the RF modulator. I lifted some of the filter caps in in the video circuit, which improved the sharpness somewhat, but the overall video output still looks pretty bad. It 's probably due to the older ceramic VIC II chip. I'll try swapping in a newer chip later this week when I have some more time.
The 326298 generates the composite video on the main board instead of in the modulator. No mods need to be done on the modulator. Remove C73 to increase sharpness. It makes a big difference. Check the value of R10. If it's 300 ohms, replace it with a 120 ohm resistor or parallel it with a 220 ohm resistor. That will increase the brightness. C73 and R10 are inside the RF shield near the VIC chip. While you have the case open, check the value of C10 and C11. They are the filter capacitors for the SID. If they are 2.2nf, replace them with 470pf. That will improve the audio.
Yeah Mr Lurch did some testing for me and he had one PAL VIC that was also really soft and another that is better ... On the soft one, the improvement was minor as it was for me.
There is indeed a somewhat different type of NTSC in Japan called NTSC-J. You'll notice it on a US TV when, trying to tune in to Channel 3/4 and connecting the C64 via RF, you get absolutely nothing of a signal. You'd have to tune in to Channel 95 or something on a US NTSC TV. That is somewhat modable though.
As I understand it, that is due to broadcast frequency differences and not NTSC format differences. What was call Channel 3 in the USA isn't the same frequency they use in Japan.
US CATV channel 95 is around 91.25 mhz which is just above broadcast channel 6. (US/Canada/Mexico Channel 6 is 83.25mhz and channel 7 is 175.25mhz.) In Japan, what was called Channel 1 there was 91.25mhz right smack in the US FM radio band. This is more to do with channel allocations than anything else.. not a difference in NTSC vs NTSC-J.
@@emmettturner9452 Indeed, the frequency behind the channel. Here in Costa Rica, we used US NTSC without changes, so we used Channel 3 as well when using an old VCR via RF. ;) And with moddable, I meant that e.g. you can open a Famicom (the Japanese version of the NES) and change its RF broadcast channel to Channel 6 with just a screwdriver. I saw that in another vid.
Awesome work! I'm going to have a closer look at the modulators in my different PAL machines to try to figure out what's going on. But yeah as has been noted by others, the smoothing of hanover bars gives us using PAL machines a vastly increased palette. So I think it's more of a feature tbh :)
I just watch the whole thing even if I am not able to do the kind of stuff you do because you have a great voice and I find all of this interesting and I need background noise on graveyard shifts. I do have some electronics projects I've been meaning to get to but I've been more of a software guy than hardware guy so I don't have much confidance in my soldiering ability etc. I'm glad I found you (I come via 8 bit show and tell some weeks back) but I also do follow 8bit guy as well.
Hey Adrian I did these mods to my c64 I recently got from my uncle it was his's when he was a kid and passed it to me (still in shock wasn't expecting that). anyway holy cow these mod made a night and day difference. was super easy to do which is always a plus. Oh my board is a 250407 "Rev. A-CR"
17:51 About the Zener diode and the capacitor, look it´s a yes and no really... Zeners are modeled as voltage sources in series with a resistance (r-zener) in their simplest form. Thing is, this Rzener when the zener is forward biased (or reverse biased based on the convention you want to follow) is quite small, between 10 - 20 Ohm small. All though that might seem significant, it is not really. Adding a capacitor there, sure, hands down helps. I would go with Resistor - Zener - Resistor - Capacitor, but that would need a bit of calculation.
It's more often you see resistor - capacitor - resistor - zener feeding the base of the pass transistor. This works extremely well. Although today, you would just put an active regulator such as a 7805 there. At the time, the resistor/zener/transistor combo was the cheapest option
@@jaycee1980 It´s more about the frequency component of the PSRR rather than the stability. The transistor - Zener solution is not regulated (there is no feedback) but the transistor, acting as a current source, is the closest that they could have and still have today for open circuit in AC. A.k.a. no noise in, no noise out.
And actually, the transistor - zener solution could potentially also become unstable, if you don´t bias the zener correctly. The is the potential for positive feedback. So you got to know what you are doing.
So interesting they made the circuit like they did, Commodore was all about saving costs so not having more components (which actually make the signal worse), seems the right choice from the start. Which makes me wonder if it's actually, in certain situations, "dangerous" or a disadvantage?
Another excellent and detailed video Adrian. I'm not a C64 owner, but I still enjoy each one of your videos. With my ex-radio engineers hat on, I wonder if the issues with the RF modulators is because they were just bought in modules that might be generally used in say analogue video recorders where their deficiencies are not so noticeable with moving pictures compared to still images like the ones you were showing. Just a guess. May search on the part numbers to see what else they are used in.
Great research and explanation. I was prepared to make this RF modulator fix but noticed the R5 resistor on your schematic is a 150 Ohm resistor. On another schematic for the same RF module, it is shown as a 220 Ohm resistor for R5. I'm wondering which value and wattage is best for video quality?
Thanks for sharing! I hoped there was a way to make the output better without drastic changes (like removing the whole modulator). Now I know what I can try to improve video output. Great stuff! :-)
Great vid, I always knew PAL decoders had a delay line in there, but I never found out what it did, till now. I assumed it was for the decoder to be able to somehow mix the alternating phase's to find an average?, but not to purposly reduce the vertical chroma resolution.
So on one C64 that i own, i have the original board, 326298 Rev A, that you mentioned you were not referring to in this video. it does have a 5 pin din. It doesn't have a sticker on the RF modulator model "MD6..." on it, but it does say a part number. The part number listed on the RF modulator is "PN 326130-01". I am wondering if cutting the caps on it will improve it. I haven't tried it yet though...
In your NTSC modulator schematic, are diodes D1 and D2 the right way around? I don't see how any current is going to flow into the bases of transistors Q1 and Q2... (edit) And similarly D3 and D4 on the PAL modulator.
Yes they are, there to both provide a 0.6v step to compensate for the base emitter voltage in the transistor, and there is a base pull up resistor there to provide bias.
This relates to the the schematic at 15:17. The reason for the diode is to match the voltage coming off the the Q1 or Q2 follower to the voltage output by the VIC-II on LUMA_IN. R3/D2 ensures the the voltage at Q2.BASE is a diode drop (0.6V) above LUMA_IN. The transistor's integral diode from Q2.BASE to Q2.EMITTER drops the voltage back down to the same level as LUMA_IN. This transistor is connected as an "emitter follower" (google if you want to learn more). Basically, the VIC-II Luma output driver only needs to drive against R3, instead of powering all of the subsequent stages. I can speculate that there is some impedance in the system unaccounted for by the RF modulator designer (likely the VIC-II output impedance) which is causing C2 to become part of a low pass filter. As for L1 in this schematic that is being bypassed, L1 and C3 form a bandpass filter. This is almost certainly constructed to allow only the colorburst frequency (3.58Mhz) to pass through, so that other frequencies picked up can't find their way into the composite/RF signal. Bypassing L1 effectively makes this a highpass filter. The impact on image quality suggests that passband might be a little too tight (or maybe they didn't consider the series resistance of L1 in their design, or the impedance of the VIC-II driver is significant), and it is affecting the phase of the signal propagated to Q3.
I noticed the same thing. The arrows on the diodes points toward the RF input. How would RF signal pass through? Maybe RF can pass backwards through a diode but not current?
@@brancarr1 The resistor pulls the base up, and the diode allows the output of the chip to pull it low, setting the voltage. There is no need for any pull up on the chip side, just an open collector output, or at least an opamp that can sink a small amount of current instead of sourcing it. The IC was likely designed that way, knowing that it only has to sink to a voltage level eases the design, as they do not need to have active pull up devices in it, which (as Commodore was so want to do in their own chip designs) saves a lot of space on the chip itself. A diode will pass current when forward biased, and will act like a resistor with a fixed voltage offset, which is why they also can be used as switches, irrespective of frequency, so long as the applied voltages and currents keep them on.
Hi Adrian, not sure if you will be reading this but I'd like to tell you that I have a 250425 board with a (replacement) 6569R5 and a MD6-UE3603 modulator.... the composite picture is very reasonable. It's possible that board carried several versions of the modulator. I also have a 250407 board with a 3603, but with a 6569R1. Really crappy picture on composite. The funny thing is, I have all kinds of inputs on the LCD that I use, EXCEPT S-video! I am going to try your mods to see if the picture will clean up. I'd like to add that although both modulators are a 3603, I see subtle differences in the placement of the components. Did not check for differences in schematic.
When you decide to skip the early rev A boards (326298), is that because the Vic ii's on those are already so bad that it's not worth it? or is this modification not possible on those earlier boards for other reasons? I'm looking to improve this one, but I don't know if I can just drop a newer vic ii in here and have it work / make a difference, or if it's more trouble than it's worth and I should just get a newer machine.
The 326298 generates the composite video on the main board instead of in the modulator. No mods need to be done on the modulator. Remove C73 to increase sharpness. It makes a big difference. Check the value of R10. If it's 300 ohms, replace it with a 120 ohm resistor or parallel it with a 220 ohm resistor. That will increase the brightness. C73 and R10 are inside the RF shield near the VIC chip. While you have the case open, check the value of C10 and C11. They are the filter capacitors for the SID. If they are 2.2nf, replace them with 470pf. That will improve the audio.
It occurs to me that if you wanted to get a cherry image of a CRT you could break out a film-emulsion camera. I recall one of the peripherals I lusted after in the '80s was a gizmo that consisted of a dark shroud over a CRT with a fixed-focus camera taking snaps of the image.
From what I see, the C128 modulator (NTSC at least) is already as sharp as it can be. The main issues are the jail bars and those are from the VIC-II itself, so not easy fix there sadly.
Looks like C1 and C2 are used to filter out any chroma cross-talk coming from anywhere else in the circuit, as well as preventing the luma from interfering with the chroma carrier, and the cutoff frequency is fairly low which causes the blurriness. I suspect the PAL version does not have as much trouble with colour smearing or rainbowing because of the way chroma is encoded in PAL.
Although this kind of work is not my thing at all (and to be honnest i never tried/bothered to do stuff like this) I do find it very interresting! a Thumbs Up for you Adrian! :-)
I remember years ago similar efforts were put forth to modify the Atari 8-bit series video output to dramatically improve the quality. I've never seen anyone do it.
in theory removing the resistors for either collector +V or emitter to ground on any transistor in the RF modulation section might slightly reduce the noise that their current consumption puts on the nearby power rails (since they now aren't operating and therefore aren't using any current), making the composite output a very tiny bit cleaner. i'm not confident that this would actually have any real effect given that 1. everything's real close together and has sufficient bypass capacitance on the rails 2. the noise floor of composite is pretty high anyway and 3. they'd all be operating with similar waveforms except for i think Q4, so their collector currents should also be dipping and spiking in sync, meaning no interference.
in any case it might be wise to add some extra bypass capacitors to the power rail near Q2,Q1. i would assume that the reason why c1,c2 were there is to help stabilize the base bias voltage, so that long areas of bright white don't start to dip down in brightness. alternatively you could put some resistance between the diode and the base of each transistor to increase the effective cutoff frequency of the LPF created by the capacitor and the ESR of the diodes. since base currents are pretty low, this shouldn't cause too much voltage drop.
admittedly since these modifications would involve more than just shorting or opening components, it gets into the realm where it would be better to design a new RF modulator replacement board from scratch using new components, which has already been done by other people.
Maybe those caps and coil are for limiting the bandwidrh usage , so, when mixing with sound band and outputing the RF, the bandwidth will be within specs, and not interfering with the sound RF.
Thanks for the bookmark links (even though I always watch 100% of your excellent videos anyway) and other links in your show notes. You should also put an obvious link to the previous part(s) of a continuation video at the top of the show notes.
I know this is an old video, but at 8:58 you list the motherboards that you tested, and you mention 240441. What is that exactly? I have never seen such a board, and no pictures seem to exist. Yet, it is mentioned in various places
Several people have asked if these fixes address the jail bars in the C128. They definitely don't. The jailbars come from the VIC-II itself and the problem is especially bad on the C128. I did look at my 128 last night and opened the modulator. It's already sharp so needs no mods there. The NTSC version has color smearing but you can't just bypass the L1 inductor otherwise it lowers the color level too much. (It does fix the smearing though.) So for now, just run a wire to the RCA jack to convert it to composite if you want to work on the C128. :-)
Seems like, to fix that, you'd have to tweak the signal at the VIC-II itself. Don't know if that's even possible.
I would love to see a fix for the checkered pattern in my C64's S-Video signal.
Hi Adrian, it's good to see my old PAL C64 out and about again. I actually bought myself a PAL C64C from eBay now to replace it. But i didn't get any video cable to plug it into anything. I ordered a c64 video to SCART cable a while back but it never got dispatched due to the COVID-19 lockdown (still hasn't). So i just watched this and now 20 minutes later i've converted the RF output to composite. Works like a charm! But i do notice jail bars on the composite output on this C64C. Not sure if this is a similar issue to the one described on the C128. For reference the sticker on the modulator says "P/N: 251916-03A" & "1829F" written underneath in bold.
@HalfSpeedMastering Yep - I'm also here to mention the possibility of adding a Schmitt Trigger to the output of the old VIC-II to clean up the signal :)
I cleaned up the pins of the vic chip (and the socket) and saw a great improvement, though it wasn't gone. I also have a hunch the oscillator is aging.
L1 and C3 are in series which forms a band pass circuit. I am assuming that C3 is not being used exclusively as a DC blocker for Q3. Most likely the resonant frequency is 3.58 MHz, which is the center frequency of the color DSB (double side band) sub-carrier. It was probably placed there to make the RF output NTSC compliant to meet FCC regulations. In a compliant NTSC signal, color bandwidth is less than half that of the luma. This was done to help reduce dot patterns when watching the RF output on a B/W TV and to eliminate adjacent channel interference. Combined, the Y (luma) and C (Chroma) need to fit inside the 6 MHz window to meet FCC requirements. By eliminating the choke, you are allowing the chroma signal to have as much "sharpness" as the luma. So it must be there for regulatory purposes. This will make the composite and the RF output look great, but I bet if you used an old B/W TV you would see dots in the colored areas. I used to use a B/W TV growing up in high school for my Atari 800, so there were people that used B/W TV's. Money was tight... LOL
Spot on reply - you nailed it. Cross channel interference in the analogue tv days would have been a big no-no, in the crowded (and strictly regulated) RF spectrum.
@@Petertronic I realized after I posted my comment that "most" commodore 64 RF installations would be closed - The TV would only be connected to the C64 with no other channels present via the RF TV/GAME switch. If you were to mix / combine the output of the modified C64 with other channels, like various city information channels used to do, there would be severe interference. I set one up a long time ago and had to purchase really expensive filters to make it work.
@@Craig1967 Adrian wasn't interested in the RF output, but I wonder if removing the BP filter might result in bleed into the audio subcarrier too. That might result in hissy audio from the TV used on the RF output.
Came here to say the same thing: FCC Compliance.
You are most likely right, question is: Why is a bandpass- filter required for NTSC and not for PAL? Could it be the larger distance between channels of n n the UHF spectrum (8MHz in W.Europe)?
Further, rather than kill the filter, would it be possible to improve it to have a sharp cut-off at exactly the right frequency? (A standard value of 22mH sounds like a rough approximation.)
I tried it on a PAL longboard with MD6-UE3603 modulator and it definitely made a difference. Cool!
L1 and C3 in the chroma circuit form an L/C “tank” circuit. You can tell this, because they’ll either both be in series with the signal path (as here), or they’ll both be in parallel across the signal path, with one end tied to ground or a supply rail. From basic electronics, we know that inductors pass DC, and capacitors pass AC. In series like this, we therefore know that DC (and low frequencies) will be blocked, and high frequencies will be blocked by the inductor, so our response curve (Bode plot) has to have minima at DC and high frequencies. Therefore, there must be some frequency that maximizes the signal flow, so this is a band-pass filter. This is followed by a similar low-pass filter composed of C4 and R4 to the one we saw with Q1/Q2 above. This likely attenuates the frequencies on the high side of the pass-band, resulting in an asymmetrical response curve, though it might attenuate on the low-side of the pass-band.
I think this might be my favorite video of yours, Adrian. Well done.
Adrian deserves WAY more subscribers! Such quality content.
I agree
I just modded my PAL version with MD6-UE3603 - definitely a visible improvement! I also added the Comp. Out removing the RF out link but instead of soldering directly the wire to the right side connector pin and "smearing it" I soldered onto the 140 Ohm resistor leg directly above it (around 2cms) just in front of the 2 Caps - it's easy to locate the leg using a multimeter. Cheers!
Hey Adrian, one issue with cutting the LPF caps and shorting the inductor. The reason they are there is because transistors don't act well when they receive signal that has power above the transistor's bandwidth. So if your transistor goes up to 1MHz at unity, and you put 2 MHz into it, you can get ringing or even oscillation, and reduction of dynamic range - the parasitic oscillation will eat into the total (Vpp) dynamic range that it shares with the desired signal. So what you want is: 1. a transistor that can manage the bandwidth you need for your signal and 2. an lpf filter that will filter out everything above, and leave everything below. I bet the commodore factory subbed in cheaper transistors that didn't quite manage full video bw, and that's why they added the severe lpf. You want to check the GBW of the transistors in there, and then replace with something better if necessary. However! bear in mind in general in electronics mfr companies those parts would get replaced for the cheapest part weekly, so what you test will be different to what others have. So it's best to just find a good substitution and use that. Then, build an lpf around it that passes the right bandwidth. you want to check by using an RF signal generator, to see what freqs get passed. Just a guess, but i think you want -3 dB at the top freq your signal is supposed to carry. I think you can just go by LC and RC low pass filter equations to figure out the capacitor values - or just use a few values and experiment, probably simpler. You might take the inductors and just unwind them a bit. But if you completely remove the lpf, you're asking for trouble :-) Hope this helps. Enjoy!
I think that's not an issue here, we are dealing with 4-6MHz max, I bet those transistors can handle up to 100MHz or multiples of that. But shorting the inductor is not fine, because it is there to filter out luma signal from the input of the chroma buffer. I'm surprised that shorting it seemingly has no negative effects. I agree, probably reducing the inductance would be more elegant. Maybe it's enough to put a ferrite bead on the place of it. The original 22uH probably delays the chroma signal too much, causing color smearing.
Just tried this on my PAL C64 with UE3603 RF Modulator, the difference is amazing, text is so much clearer!
Fantastic video. One of your most interesting ones yet.This was a much needed modification and I really like the way you explain the circuitry theory. Learning a lot from you!
I'm sure that those capacitors and one inductor you removed to get better sharpness were mainly there for FCC type acceptance. The filter was probably for limiting the bandwidth for NTSC to insure the signal fits inside a standard TV channel.
You are probably totally right -- likely had more to do with the RF signal than anything else.
The caps almost certainly had to do with FCC acceptance, but it may be worth noting that computer RF modules are absolutely terrible at keeping their output within a single channel, and are only suitable for uses where they're the only thing feeding an RF cable. Even if one wants to hook a computer or video game system up to an RF TV, one will get a better picture feeding S-video into a a 2000s-era video switchbox with a built-in RF modulator than one would get by using the built-in RF connection.
I took the plunge and did this today. I've had this Rev 250407 for 37 years. It made quite a difference. Great research and excellent explanation of this fix! Bravo.
I've now modded my PAL C64 (but I went the extra hassle and removed the modulator to be able to remove the cap from behind!) and indeed picture quality has improved! Thanks a lot!
Regarding Hanover bars - PAL works by inverting the phase of a portion of the colour carrier every other line. The idea is to eliminate hue errors that occur in NTSC. PAL receivers don't have a hue control for this reason. Unfortunately, to invert the colour carrier accurately, with no phase errors, is quite a difficult to achieve. These phase errors would normally be canceled out in the receiver with the delay line - at the expense of vertical colour resolution - but without that, you'll see the slight hue errors.
It's probably totally a non issue for normal video especially as you sit across the room, but sitting close to a computer display with solid colors, it really looks disturbing. (to someone who isn't used to it)
@@adriansdigitalbasement That's exactly it. The first PAL receivers in the 1960's used the PAL-S standard, S for 'simple', which didn't use a delay line for cost reasons and relied on the human eye to average out the colours.
On a related topic; I wonder if the ZX Spectrum has any problems with Hanover bars. The ULA in the system outputs unmodulated video which is converted to PAL externally. However, the colour carrier isn't inverted, the baseband video is inverted on the ULA's output. This makes the colour modulator simpler. It shouldn't have the phase issues I mentioned because the colour carrier should be the same phase every line.
@@256byteram the main problem with the spectrum is the dot crawl caused by using two separate oscillators. They fixed this in the later 128 design by 1) having the ULA generate RGB output and 2) clocking everything from a single 17.7345MHz oscillator.
Did this to my 250407 with the MD6-VA3403. Looks great on my 1702 monitor. Simple composite output is a bonus also. Thanks Adrian!
Just a warning to anyone trying this: If you prefer the softer output do not do this! Perhaps just bypass the inductor and don't cut the caps.
Adrian, the mods you're doing here are common in the Atari 8-bit world for the XL machines. They have much worse video output than C64's do out of the box, and they even recommend installing a switch to disable composite video to help the luma-to-chroma leakage in s-video. I'm so glad you've done the legwork to finally bring similar mods to the C64. I did your capacitor clip and inductor bypass to my 250407 tonight and the improvement is outstanding. This is an excellent video!
Although this video was released a while ago, I didn't want to do the mod until I educated myself enough to understand what's actually going on here. Over the last couple years I spend hundreds of hours working on these computers and finally understand now what you're trying to do. I did the full 4-change mod on the 3403 and the picture is much clearer and sharper than it was before. For years, I blamed it on the digital conversions that occur when displayed on a modern monitor. The RF to Composite mod is pretty handy too, since most people don't have an RF compatible TV around anymore... Regarding the 3405 RF Modulator, bypassing/removing the inductor does also make the image clearer. Excellent work Adrian and Thank You!
D1 and D2 are for biasing the transistors, otherwise they would be completely off at the bottom of the waveform during sync pulses. The capacitors and inductor you've removed are most likely for EMI suppression (FCC compliance).
I like these circuits with analog tricks, today we tend to use chips for everything, it is useful to see how you achieve stuff with simple transistors.
They also partially compensate the Vbe drop, so the circuit provide a better buffering.
Removing EMI related stuff? That's Adrian's middle name!
After watching several of your videos, I decided to try out some repairs/mods myself. I have bought a non-working C64 to repair and restore, now I just have to wait for parts. Thanks!
That's the fastest mod I've ever seen. Holy smokes, this is game-changing wizardry!
What an incredible improvement. Now I wish I had known to do this back when I bought my original C64.
Brilliant! It's good to have this issue covered in a single video. The other guides, scattered around are either vague and/or incomplete.
Just cut the cap on my PAL 250407. Even on the main basic screen I was blown away by the difference. Thanks a lot Adrian.
PAL viewer here! After seeing the unknown modulators for PAL machines, I dissassembled my C64s to check them. I don't have the proper cables to see the video output, so I'll just confirm the model numbers.
Here's what I got: OG C64 250425 board, Mitsumi MD6-UE3605 modulator; C64C 250466 board, Commodore 51A1-10 Rev. 2 modulator. On my machines the model number wasn't on the steel can, but printed directly on the PCB.
My PAL 250425 has a RF-modulator called "251025-05", and it also has another label reading "5095". It's not like any of the ones shown in the video, and features a big metal can (inductor?) in the middle which has its screen soldered to the lid. Having a closer look inside is therefore impossible without a soldering iron at hand. Another detail, instead of a channel switch, it has a trimming-screw in its place.
This is amazing! I installed
Perifractic's VIC-II² PAL/NTSC Switcher Kit in my main C64 and I was getting really bad color smear on the PAL side (it was originally an NTSC machine). With the inductor bridged in the MD6-VA3405 modulator the PAL output is now fantastic. Unbelievable difference considering how much smear I was getting!
I'm about to install a VIC-II² PAL/NTSC Switcher Kit in my main C64, an original NTSC one like yours, so that's great information to know. Did you install the Lumafix components? I've left mine off for now and thought I'd see how it looks on my setup.
@@Ikrananka I did install the Lumafix components. It helped with the jailbars as you'd expect once adjusted, but obviously did nothing for the color smearing issue.
Just modded my (entirely stock) 250407 with MD6-VA3404. Looks fantastic! RF can now proudly claims to have Adrian's Digital Basement 2020 mods applied.
Tried this on my MD6-VA3405, worked perfectly. Thanks Adrian!
I used to service TV's in the UK. Hanover bars are almost never seen. Only ancient early 70's tv's or very cheap tv's would exhibit them. 99.9% of TV/monitors from the 80's onwards used PAL decoder IC's (with delay line) that accurately deal with and correct the alternate line phase errors and they cannot be seen. If they are showing up on a signal that's going into a converter box, that converter is going back to the 70's by taking shortcuts : i.e. not bothering to do the phase detection and correction properly (which requires some considerable DSP processing by the way).
Great timing. I just finished soldering the same exact rf modulator bypass board that you showed in the beginning of the video. Oh well, I guess I'll try this mod first and then experiment further with the replacement board to see if it makes any difference. Thumbs up for going through the PAL version too for us here on the other side of the pond.
This is one of the single best video videos I've ever seen. Going to mod my C64 and SX64
I just came across this video, as well as part 1. This is a fantastic mod. Simple but effective. Thanks for all the effort you make creating this for everyone. 🙏
This is great, thanks Adrian. I just wheeled out my C64 two use in a nerd project, I used it as a twitter client 7 years ago but thought I create a little Weather BASIC program pulling from an online API. The video on mine looks quite blurry to my eye on a TV across composite, so I am looking forward to trying out some of your mods.
And did it work, did you see improvement ?
Just amazing work.
I wrote a couple questions in the comments but had to delete both, all answers were in the video.
Amazing, thanks for sharing.
I was thinking he should make a better RF module replacement board, but the fixes he then shows are really easy to do. Good stuff.
There are a few pcb designs out there for replacing the rf modulator including the one he tried. It worked, he just didn't see the need to make that vs just doing a few minor mods on the existing one. He could make his own pcb variation but I'm not sure there's a need out there for another option. This was something that wasn't out there and that anyone can do.
Mods for the 5 pin DIN 326298 board
The 326298 generates the composite video on the main board instead of in the modulator. No mods need to be done on the modulator. Remove C73 to increase sharpness. It makes a big difference. Check the value of R10. If it's 300 ohms, replace it with a 120 ohm resistor or parallel it with a 220 ohm resistor. That will increase the brightness. C73 and R10 are inside the RF shield near the VIC chip. While you have the case open, check the value of C10 and C11. They are the filter capacitors for the SID. If they are 2.2nf, replace them with 470pf. That will improve the audio.
I suggest simply jumping the inductors. Don’t snip them. By snipping, the remaining leads stumps could move around and slip down into the bottom of the unit when you heat them up and solder the jumper to them. Jumping across the component takes it out of the circuit anyway, and jumping it would make reversing the mod very simple in the future as well.
Adrian, love the channel, but stop looking at yourself in the monitor! lol. Seriously though, viewers want to feel as though you're talking to them directly. When you look straight into the camera lens, you accomplish this. When we see you looking at yourself instead of us we're not as engaged as we could be. Not engaged means not invested. I understand that you just want to make sure you remain in frame the whole shot through. I also understand that it can be difficult for some to get into the habit of looking into the lens and not the monitor. Trust me, I've been in video production a long time. It may seem like a tiny thing but it matters to viewers more than you think.
You're a great presenter. Best of luck in the future! Glad to see the channel growing and getting some much deserved love. :)
I can confirm that my PAL 250466s have a MD6-UE3605. PAL 466 boards generally have the best video quality.
Great video Adrian, I have the MD6-UE3603 PAL modulator on my 250407 board too and I found chopping out that cap improved the sharpness of my video out very nicely!
I always enjoy that intro. I’ll open my PAL C64 and check which modulator it has. Fantastic video; thanks for all your hard work.
Best simple C64 mod ever.... why didn't somebody figure this out 35 years ago? I did not add the 100nf cap, as it was a bit tight for my soldering iron and skills. The image is much sharper, and the reduction in color bleed is very apparent. Thanks again for these mods.
I did the 100nf cap, just because i was there anyway, but I'm not sure there's a need. As for why they didn't do it 35 years ago, good question.. There have been composite replacements, maybe anyone who has the skill, opted to make and sell those instead of simple DIY options.
Yeah I have to wonder why no one figured this out at some point between then and now too! But I'm glad people are benefitting!
I run svideo (YC) only. 250407 ntsc breadbin. VICII2. I did this mod and it’s good all around (NTSC & PAL) on both my HP 4:3 elite display port medical document viewing monitor and Sony PVM. Side note: yes, the VICII2 does fit with SIDFX on this board. Even had room for mechboard64 and FU32 dual tetrode vacuum valve headphone preamp and heat sinks all around. Considering cramming an OLED audio spectrum analyzer in as well.
The Hanover bars are created due to the nature of PAL fixing the NTSC tint issue by alternating the phase of the chroma signal per line. That's what PAL stands for: Phase Alternating Line. This has the effect that phase changes are cancelled out so the tint of the colors is stable. But at the cost at those artifacts. The proper way to eliminate these is not to average the colors between the two lines, but to use a comb filter. SECAM by the way has half the vertical chroma resolution because it actually uses a delay line for the chroma signal as it is transmitted every even (or odd, not sure) line. That's why it's called "Sequential Color And Memory".
The quality loss with half resolution chroma is the same as in digital video 4:2:2 (with chroma information each line) and 4:2:0 (with chroma information every odd line).
I love seeing u agin and again, specially in this time
@I Blow Bubbles hahahaha
@I Blow Bubbles Love might be too strong a word :)
love this series...has encouraged me to repair my pile of acorn electrons...about 2 dozen working...3 still to repair...2x pres plus 1 outstanding
You absolutely can -- and I think it lets you launch both copies at the same time.
In Hannover they just call them bars 🙂
LOL
I did the mod on a PAL MD6-UE3603 modulator, this modulator has been taken from a 250407 board and used on a new SixtyClone board which I built myself. I am feeding the signal through a RetroTink X2 and the output was a bit blurry. Not too bad, though. After this mod, I can definitely see a difference in the sharpness. It is a lot less blurry. Thanks for this tip!
Thanks a million Adrian !
I was looking for a way to improve my 250407 pal board picture quality on my crt monitor and it improved so much ! I am very grateful THANKS !!
Thank you Adrian for a simple fix too clean up the svideo signal.
I just performed this on my PAL model ASSY NO 250407 - with VIC 6569R1 1183
video looks much better as you showed.
Keep up the videos, I love all your content, how you transfer your knowledge to us all and presenting style.
It's been forever since I've done analog electronics but L1/C3 looks like a classic bandpass filter. The combination is filtering out frequencies below and above a certain range. Probably intended to clean up noise before amplifying the signal.
Future video suggestion: Improving VIC-20 video quality. I know that are at least a couple of S-Video mods that can be done but I wonder if anything similar to what you've done for the C64 is also possible.
That's funny, I actually noticed I got a color image from my pal c64 on the Luma pin yesterday. And was quite puzzled about it. But your comments around 27.45 explains it! Thanks
It's not hard to automatically eliminate hanover bars WITHOUT loosing resolution these days, IIRC the method was regularly used from the early 90s on better PAL TVs but is mostly forgotten because pretty much no one discuss "high end analog video" any longer. I suspect a manual color (phase) dial exactly like those often seen on old NTSC TVs would allow manual elimination too.
The alternating phase shift was added to combat the phase-related color shift that plagued NTSC (NeverTheSameColor) and meant that without compensation any phase errors would results in the hanover bars which on the TVs of the day was mostly not visible anyway.
The cheapest way to remove them is to average the two lines as you mention, but since the phase error is either static or change very slowly (minutes if not hours) it's possible to instead compute the phase error over time and then use that to get correct colors and full resolution without hanover bars. The obvious way today (and I suspect in the 90s) would be to compute a phase error estimate per-frame (compare each line-pair and average, possibly throw out outliers) and then do a slow/long decaying average on that (to reduce the chance of it introducing glitches) to get the compensation value for display purpose. This results in very good resistance to slow phase error changes without loosing resolution or even showing the wrong colors after the first few frames (say less than 0.2 seconds).
I'm never going to any of this stuff but find it incredibly fascinating to watch
I promptly went down stairs and did this mod to my 250407 board. The quality has improved, but it seems like when using the s-video the checkerboard pattern and jail bars seem more pronounced, it may just be all in my head though. I didn't take a before and after pictures so I cant be sure. The old LCD TV I'm using doesn't have the best picture quality, but its the only thing I have with s-video.
This is definitely a worth while mod that doesn't cost anything but a little bit of time, I wish this knowledge was around back in the day.
Thank you for existing in this video space.
No idea why I've watched this, but it was fascinating!
Very neat explanations, nice filmmaking too, thanks you.
Very well done video, Adrian. And thanks for pointing out the quick jump links in your comments, as I may very well come back to this when/if I mod my machines.
Very nice! I only clipped the capacitors and it did wonders for my C64's picture. Thanks for sharing this!
The most interesting video about C64 color problems! But I would like to ask you where can you find the 'yellow PCB' placed on top of the TEBL PCB? Thanks so much for an answer ...
That was excellent! Really appreciate the all the well explained information. Will be doing this soon. Thank You, Adrian!
Now that I've seen the schematics of the modulators, I have an idea:
I had noticed once, that when I connected the chroma output to my TV's composite input I get a picture. Not worse than using the real composite output.
My theory is, that the color smearing/fringing on these larger contrast areas might be due to the luma signal passing back through the capacitors C6/C7 on the NTSC modulator into the chroma output. (C11/12 on the PAL version)
I haven't tested it, but if you intend to use S-Video only, you might try cutting those two capacitors out. Or, if you want to keep the color composite output, you could add a fast Schottky diode in series with each of these two capacitors.
Wow that is some serious backfeeding of luma and sync into the Chroma! I've never seen that work on any of my machines before. Yeah S-Video only it likely would improve quality if you break the connection like it did on my little bypass board. Since I don't really have any PAL machines to play with I can say for sure.
The coil, capacitors and resistor in the chroma input form a tuned circuit that was probably designed to resonate at the color carrier frequency. So if you bridge the inductor you also inadvertently reduce the color signal.
Thanks for the amazing video. Great to see someone spending a lot of time to get a perfect result. Amazing quality at 25:30 and (monchrome) at 27:05. Keep it up!
Watching this while writing modifications to Mister C64 looking for inspiration for how to simulatewhat you are trying to fix :)
That was a great video. Those are simple fixes that have some dramatic results thanks Adrian.
L1 - C3 is a band-pass filter, likely for the 2 MHz chroma band, which has a carrier at 3.579 MHz above the luma carrier. C4 would be a low-pass filter.. the only thing above the chroma is the FM audo band though... it may be needed to prevent the transistor from oscillating at a higher frequency, usually a good practice in designing HF circuits to damp the input that way.
By cutting the inductor out, the circuit is changed from a band-pass filter to a low-pass filter with a -3 dB point relative the the base resistor R4. You may get some additional noise coming in from the main PCB (digital logic noise) but it does not seem to affect the signal so it seems ok...
I think some considerations were put in the design of the circuit to meet FCC emissions guidelines for NTSC signalling... not really relative anymore...
Cheers,
Great video. I did this mod to one of my C64s with great results, similar to yours. I also tried to do something similar to one of the early 1982 motherboards (326298 Rev A) that don't have the the chroma output. On those motherboards, the video out does not go through the RF modulator. I lifted some of the filter caps in in the video circuit, which improved the sharpness somewhat, but the overall video output still looks pretty bad. It 's probably due to the older ceramic VIC II chip. I'll try swapping in a newer chip later this week when I have some more time.
The 326298 generates the composite video on the main board instead of in the modulator. No mods need to be done on the modulator. Remove C73 to increase sharpness. It makes a big difference. Check the value of R10. If it's 300 ohms, replace it with a 120 ohm resistor or parallel it with a 220 ohm resistor. That will increase the brightness. C73 and R10 are inside the RF shield near the VIC chip. While you have the case open, check the value of C10 and C11. They are the filter capacitors for the SID. If they are 2.2nf, replace them with 470pf. That will improve the audio.
Thanks for the wizardry! I like this kind of well explained, in-depth look of schematics and components. Congrats!
Just did my 407 with the UE3603, deffo better, shame its one that suffers from the rise and fall issue. So a huge thank you for doing this tutorial
Yeah Mr Lurch did some testing for me and he had one PAL VIC that was also really soft and another that is better ... On the soft one, the improvement was minor as it was for me.
@@adriansdigitalbasement Its better so I'm more than happy, thanks Adrian..
There is indeed a somewhat different type of NTSC in Japan called NTSC-J. You'll notice it on a US TV when, trying to tune in to Channel 3/4 and connecting the C64 via RF, you get absolutely nothing of a signal. You'd have to tune in to Channel 95 or something on a US NTSC TV. That is somewhat modable though.
As I understand it, that is due to broadcast frequency differences and not NTSC format differences. What was call Channel 3 in the USA isn't the same frequency they use in Japan.
US CATV channel 95 is around 91.25 mhz which is just above broadcast channel 6. (US/Canada/Mexico Channel 6 is 83.25mhz and channel 7 is 175.25mhz.) In Japan, what was called Channel 1 there was 91.25mhz right smack in the US FM radio band. This is more to do with channel allocations than anything else.. not a difference in NTSC vs NTSC-J.
@@emmettturner9452 Indeed, the frequency behind the channel. Here in Costa Rica, we used US NTSC without changes, so we used Channel 3 as well when using an old VCR via RF. ;) And with moddable, I meant that e.g. you can open a Famicom (the Japanese version of the NES) and change its RF broadcast channel to Channel 6 with just a screwdriver. I saw that in another vid.
@@bundesautobahn7 Yes. That was the Displaced Gamers video. I did similar with my Japanese PC Engine before I removed the RF modulator entirely. :)
Awesome work! I'm going to have a closer look at the modulators in my different PAL machines to try to figure out what's going on.
But yeah as has been noted by others, the smoothing of hanover bars gives us using PAL machines a vastly increased palette. So I think it's more of a feature tbh :)
I just watch the whole thing even if I am not able to do the kind of stuff you do because you have a great voice and I find all of this interesting and I need background noise on graveyard shifts.
I do have some electronics projects I've been meaning to get to but I've been more of a software guy than hardware guy so I don't have much confidance in my soldiering ability etc.
I'm glad I found you (I come via 8 bit show and tell some weeks back) but I also do follow 8bit guy as well.
Hey Adrian I did these mods to my c64 I recently got from my uncle it was his's when he was a kid and passed it to me (still in shock wasn't expecting that). anyway holy cow these mod made a night and day difference. was super easy to do which is always a plus. Oh my board is a 250407 "Rev. A-CR"
17:51 About the Zener diode and the capacitor, look it´s a yes and no really... Zeners are modeled as voltage sources in series with a resistance (r-zener) in their simplest form.
Thing is, this Rzener when the zener is forward biased (or reverse biased based on the convention you want to follow) is quite small, between 10 - 20 Ohm small.
All though that might seem significant, it is not really. Adding a capacitor there, sure, hands down helps. I would go with Resistor - Zener - Resistor - Capacitor, but that would need
a bit of calculation.
It's more often you see resistor - capacitor - resistor - zener feeding the base of the pass transistor. This works extremely well.
Although today, you would just put an active regulator such as a 7805 there. At the time, the resistor/zener/transistor combo was the cheapest option
@@jaycee1980 There is an 7805 on the C64 mainboard, so they were available to Commodore.
@@danielmantione Yes, they were available, but more expensive... also the modulators were subcontracted to Mitsumi
@@jaycee1980 It´s more about the frequency component of the PSRR rather than the stability. The transistor - Zener solution is not regulated (there is no feedback) but the transistor, acting as a current source, is the closest that they could have and still have today for open circuit in AC. A.k.a. no noise in, no noise out.
And actually, the transistor - zener solution could potentially also become unstable, if you don´t bias the zener correctly. The is the potential for positive feedback. So you got to know what you are doing.
So interesting they made the circuit like they did, Commodore was all about saving costs so not having more components (which actually make the signal worse), seems the right choice from the start. Which makes me wonder if it's actually, in certain situations, "dangerous" or a disadvantage?
Another excellent and detailed video Adrian. I'm not a C64 owner, but I still enjoy each one of your videos.
With my ex-radio engineers hat on, I wonder if the issues with the RF modulators is because they were just bought in modules that might be generally used in say analogue video recorders where their deficiencies are not so noticeable with moving pictures compared to still images like the ones you were showing. Just a guess. May search on the part numbers to see what else they are used in.
Great research and explanation. I was prepared to make this RF modulator fix but noticed the R5 resistor on your schematic is a 150 Ohm resistor. On another schematic for the same RF module, it is shown as a 220 Ohm resistor for R5. I'm wondering which value and wattage is best for video quality?
Thanks for sharing! I hoped there was a way to make the output better without drastic changes (like removing the whole modulator). Now I know what I can try to improve video output. Great stuff! :-)
Great vid, I always knew PAL decoders had a delay line in there, but I never found out what it did, till now.
I assumed it was for the decoder to be able to somehow mix the alternating phase's to find an average?, but not to purposly reduce the vertical chroma resolution.
So on one C64 that i own, i have the original board, 326298 Rev A, that you mentioned you were not referring to in this video. it does have a 5 pin din. It doesn't have a sticker on the RF modulator model "MD6..." on it, but it does say a part number. The part number listed on the RF modulator is "PN 326130-01". I am wondering if cutting the caps on it will improve it. I haven't tried it yet though...
In your NTSC modulator schematic, are diodes D1 and D2 the right way around? I don't see how any current is going to flow into the bases of transistors Q1 and Q2... (edit) And similarly D3 and D4 on the PAL modulator.
It's acting as a half wave rectifier, IIRC.
Yes they are, there to both provide a 0.6v step to compensate for the base emitter voltage in the transistor, and there is a base pull up resistor there to provide bias.
This relates to the the schematic at 15:17.
The reason for the diode is to match the voltage coming off the the Q1 or Q2 follower to the voltage output by the VIC-II on LUMA_IN. R3/D2 ensures the the voltage at Q2.BASE is a diode drop (0.6V) above LUMA_IN. The transistor's integral diode from Q2.BASE to Q2.EMITTER drops the voltage back down to the same level as LUMA_IN. This transistor is connected as an "emitter follower" (google if you want to learn more). Basically, the VIC-II Luma output driver only needs to drive against R3, instead of powering all of the subsequent stages.
I can speculate that there is some impedance in the system unaccounted for by the RF modulator designer (likely the VIC-II output impedance) which is causing C2 to become part of a low pass filter.
As for L1 in this schematic that is being bypassed, L1 and C3 form a bandpass filter. This is almost certainly constructed to allow only the colorburst frequency (3.58Mhz) to pass through, so that other frequencies picked up can't find their way into the composite/RF signal. Bypassing L1 effectively makes this a highpass filter. The impact on image quality suggests that passband might be a little too tight (or maybe they didn't consider the series resistance of L1 in their design, or the impedance of the VIC-II driver is significant), and it is affecting the phase of the signal propagated to Q3.
I noticed the same thing. The arrows on the diodes points toward the RF input. How would RF signal pass through? Maybe RF can pass backwards through a diode but not current?
@@brancarr1 The resistor pulls the base up, and the diode allows the output of the chip to pull it low, setting the voltage. There is no need for any pull up on the chip side, just an open collector output, or at least an opamp that can sink a small amount of current instead of sourcing it. The IC was likely designed that way, knowing that it only has to sink to a voltage level eases the design, as they do not need to have active pull up devices in it, which (as Commodore was so want to do in their own chip designs) saves a lot of space on the chip itself.
A diode will pass current when forward biased, and will act like a resistor with a fixed voltage offset, which is why they also can be used as switches, irrespective of frequency, so long as the applied voltages and currents keep them on.
Tons of useful information even for non-Commodore tinkerers!
My longboard 250425 has a P/N:251697-02 RF modulator i dont know if that helps you or not, but thats the only long number on it
Hi Adrian, not sure if you will be reading this but I'd like to tell you that I have a 250425 board with a (replacement) 6569R5 and a MD6-UE3603 modulator.... the composite picture is very reasonable. It's possible that board carried several versions of the modulator. I also have a 250407 board with a 3603, but with a 6569R1. Really crappy picture on composite. The funny thing is, I have all kinds of inputs on the LCD that I use, EXCEPT S-video! I am going to try your mods to see if the picture will clean up. I'd like to add that although both modulators are a 3603, I see subtle differences in the placement of the components. Did not check for differences in schematic.
Hanover bars are used to create extra colours, especially in demos.
Yeah really the delay line that reduces the color information does that as seen near the end of my video.
When you decide to skip the early rev A boards (326298), is that because the Vic ii's on those are already so bad that it's not worth it? or is this modification not possible on those earlier boards for other reasons? I'm looking to improve this one, but I don't know if I can just drop a newer vic ii in here and have it work / make a difference, or if it's more trouble than it's worth and I should just get a newer machine.
That machine has no S-video support -- and composite video isn't going to be sharp no matter what you do.
The 326298 generates the composite video on the main board instead of in the modulator. No mods need to be done on the modulator. Remove C73 to increase sharpness. It makes a big difference. Check the value of R10. If it's 300 ohms, replace it with a 120 ohm resistor or parallel it with a 220 ohm resistor. That will increase the brightness. C73 and R10 are inside the RF shield near the VIC chip. While you have the case open, check the value of C10 and C11. They are the filter capacitors for the SID. If they are 2.2nf, replace them with 470pf. That will improve the audio.
It occurs to me that if you wanted to get a cherry image of a CRT you could break out a film-emulsion camera. I recall one of the peripherals I lusted after in the '80s was a gizmo that consisted of a dark shroud over a CRT with a fixed-focus camera taking snaps of the image.
This is fantastic improvement! Any chance you can look at the C128 modulator circuit?I suspect there's a bit of room for improvement there, as well.
From what I see, the C128 modulator (NTSC at least) is already as sharp as it can be. The main issues are the jail bars and those are from the VIC-II itself, so not easy fix there sadly.
Looks like C1 and C2 are used to filter out any chroma cross-talk coming from anywhere else in the circuit, as well as preventing the luma from interfering with the chroma carrier, and the cutoff frequency is fairly low which causes the blurriness.
I suspect the PAL version does not have as much trouble with colour smearing or rainbowing because of the way chroma is encoded in PAL.
Although this kind of work is not my thing at all (and to be honnest i never tried/bothered to do stuff like this) I do find it very interresting! a Thumbs Up for you Adrian! :-)
I remember years ago similar efforts were put forth to modify the Atari 8-bit series video output to dramatically improve the quality. I've never seen anyone do it.
Awesome work Adrian! Thank you so much for releasing the source code too - perfect :)
Wow, that was a lot of good information Thanks Adrian!
A quick and simple fix... 40 minutes long - I love it)))
in theory removing the resistors for either collector +V or emitter to ground on any transistor in the RF modulation section might slightly reduce the noise that their current consumption puts on the nearby power rails (since they now aren't operating and therefore aren't using any current), making the composite output a very tiny bit cleaner. i'm not confident that this would actually have any real effect given that 1. everything's real close together and has sufficient bypass capacitance on the rails 2. the noise floor of composite is pretty high anyway and 3. they'd all be operating with similar waveforms except for i think Q4, so their collector currents should also be dipping and spiking in sync, meaning no interference.
in any case it might be wise to add some extra bypass capacitors to the power rail near Q2,Q1. i would assume that the reason why c1,c2 were there is to help stabilize the base bias voltage, so that long areas of bright white don't start to dip down in brightness. alternatively you could put some resistance between the diode and the base of each transistor to increase the effective cutoff frequency of the LPF created by the capacitor and the ESR of the diodes. since base currents are pretty low, this shouldn't cause too much voltage drop.
admittedly since these modifications would involve more than just shorting or opening components, it gets into the realm where it would be better to design a new RF modulator replacement board from scratch using new components, which has already been done by other people.
Maybe those caps and coil are for limiting the bandwidrh usage , so, when mixing with sound band and outputing the RF, the bandwidth will be within specs, and not interfering with the sound RF.
Thanks for the bookmark links (even though I always watch 100% of your excellent videos anyway) and other links in your show notes. You should also put an obvious link to the previous part(s) of a continuation video at the top of the show notes.
Thanks!
I know this is an old video, but at 8:58 you list the motherboards that you tested, and you mention 240441. What is that exactly? I have never seen such a board, and no pictures seem to exist. Yet, it is mentioned in various places