Stumped By New Fake Chips Scam
ฝัง
- เผยแพร่เมื่อ 31 พ.ค. 2024
- Let's fix those last strange errors on the modern Commodore 64 replica board. They look innocent enough, but they hide something subtle and malicious. As we unravel the mystery, we'll deal with mains voltage frequencies, logic trigger levels, and fake Chinese chips again. But this time the scam seems a bit different.
Chapters:
00:00 Introduction
00:45 CIA
03:07 ToD
07:04 TLL
09:10 CMOS
10:29 TTL vs CMOS experiment
14:08 Fake chip tests
17:30 Zener fix
20:16 Final touches
21:34 Conclusion
Support Noel's Retro Lab on Patreon: / noelsretrolab
Music tracks:
Funky Stars by McKlain mcklain.bandcamp.com/track/fu...
Battro OST by McKlain mcklain.bandcamp.com/track/ba...
More awesome music by McKlain: www.mcklain.com
You can also support Noel's Retro Lab on TH-cam by joining this channel:
/ @noelsretrolab
🛠 Tools I use ➤ noelsretrolab.com/tools.html
Connect with Noel's Retro Lab:
Discord ➤ / discord
Facebook ➤ / noelsretrolab
Twitter ➤ / noelsretrolab
Instagram ➤ / noelsretrolab
Mailing list ➤ noelsretrolab.com - วิทยาศาสตร์และเทคโนโลยี
I think I’ve mentioned before that your probe graphics overlaid on the schematics is really helpful for those closely following along. Another fantastic episode. It’s obvious that you are doing your homework for these and I’m really enjoying them.
Thanks so much! One thing I'm trying this time is making the probe on the graphics smaller and put it on the inside since several people mentioned that they didn't like how the probe was covering up the pin label. Hopefully this works better. Cheers!
De productie waarde is inderdaad fantastisch!
Ik zie jou vaker in retro kanalen :D Groetjes uit Nieuwegein.
I agree totally, that shows care and respect. Coming prepared makes a world of difference, saves a ton of time and makes everything much easy to understand. Another great example is Ben Eater. His pre-shooting prep work is also outstanding.
Fake components started to become a problem in the early 2000s, when production was basically outsourced to Asia. A company I used to work for was seriously hit several times by the infamous capacitor plague. Even when sourcing supposed top notch products from very serious distributors, that ended reviewing their own supply chains, and, for what i understand, involved the police as well. It meant millions of pounds to these guys.
The real cash cow appeared after 2004 when RoHS laws were passed. Thousands of companies found themselves with a huge inventory of components with lead and almost all were bought by China at super low prices. These are now resold everywhere with new markings all dated after 2004 as RoHS compliant parts.
It was happening in the 90's. We had some Pentium chips that had been skimmed and re-lettered. Some would work for a while as the clock was out (A Pentium 75 clocked at 90 for example) but would crash after a while....
It would be interesting to see the current consumption of the suspect chip. A CMOS one will consume essentially zero (just a few μA) unless it is switching rapidly. A LS TTL one will consume mA or tens of mA in the same state.
That's a good idea. Maybe I need to do a followup video running some more tests on those chips and maybe cracking them open too. I've been meaning to do that for a while.
Also, TTL chips behave like they have a built-in pull-up resistor, and source some current when you input "zero" into them. CMOS chips are high-impedance in either low or high input state, you only need some current to (dis)charge internal capacitances when you switch the input from high to low or the other way around. Measuring input pin current can be used to tell TTL-like logic from CMOS-like logic.
I've seen that... back in the day, a friend helped me make a home-made RAM upgrade to a Tandy 100. Using a standard RAM chip meant adding a logic gate, and without any HCT handy used a LS. That one chip used as much power as the entire computer!
Excellent demonstration of the TTL/CMOS differences. That really made a difference in my understanding.
Thank you! Glad you enjoyed it.
@@NoelsRetroLab Thought, unfortunately, I am not wearing the same shirt today. Maybe next time. ;)
I just got pointed to this video by a friend and O M G, I'm having a similar issue and now ALL makes sense, I even replicated your test to identify the TTL parts. Thank you SO much for the detailed explanations, great video.
I'm really glad to hear that helped, but wow, I guess the re-labeling of HC chips goes further than I would have expected. Can't trust anything these days it seems!
Wow, that's shocking that they've have gone so low to start rebadging 74 series...
Seriously! Although I really don't think this is rebadged in the sense of produced one way and modified, but I really think it was manufactured this way. Crazy!
@@NoelsRetroLab There are several Chinese manufacturers that produce these chips legit (i.e. HGSEMI, Xinluda and more). Very likely someone paid one of those factories to produce chips with a TI label (because then they sell for a higher price) and they were labeled LS because that is what people are searching for. If I buy from China I rather have a Chinese brand on the chip, because then it is much less likely a fake.
@@danielmantione . I am think your assessment is correct. Buying from AliExpress I have gotten new 74 chip with the wrong family printed on the top. Usually for me they will mark as 74LS when it is actually 74HC variety. They were for sure new chips because the numbering on the tooling marks was consistent in the tubes. With the explosion of places like LCSC and pressure on China I think this kind of counterfeiting is getting less likely and you will see a China brand instead of countefeit TI brand (or whatever is most popular).
I have used HGSemi TDA2003 and XINLUDA LS93, 4514 and they where fine. HLF is another china brand from LCSC that has been OK (op amps). I am kind of happy some of these companies like XINLUDA is making some of the obsolete chips like dip24 4514. At least they are not faking the brand...
@@danielmantione I use CMOS a lot, both 74HC, 74HCT and 40 series. I think people should move to CMOS as they spend less electricity. Also, I have no static protection, and none have failed so far. So people should not be afraid of them.
It happens quite a bit. I found it when developing Pi1541. I use a 74LS06 to take a 3.3v signal from the Pi and invert it to drive the open collector IEC bus signals on the c64 low. Some cheap Ebay sellers buy their components in bulk from China and most of the time they don't work as well as the official TI ones from Mouser, Digikey or Element14. Then the end users come to me complaining about my software. After I tell them to replace the 74LS06 with one from a credible source their problems go away. Thanks Noel, now I can refer them to this video.
What a bummer that it's happening all over the place. I really need to find a reliable source of chips that has reasonable shipping for small orders. That's the difficult part!
TI sells direct to consumers now in their web store and I found some good prices. Arrow usually has good prices and free shipping, good rewards program too. Rochester Electronics will have active and obsolete stuff, sometimes high MOQ tho. Will get genuine parts from those places.
Great video. Just as a side note, though the 74HC chips may or may not work in those types of TTL circuits, 74HCT chips(the T standing for TTL levels) are specifically made to be backwards compatible with TTL LS and ALS chips, and the input threshold voltages would be the same for both.
I really enjoy these videos. It's all good: content, production quality, preparedness, but the outstanding feature is pacing, Noel does it perfectly. Never do the videos seem to drag on. Not a second is wasted.
Thank you so much! My goal is definitely to be as respectful of people's time as possible, so I do sweat over every 5-second sequence during editing. Glad it's working out for you! 😃
Time Error Correction is applied every day to ensure that the frequency of mains supply is correct over every 24 hour period. This ensures that mains-synchronised clocks are extremely accurate.
The analog behaviour of 7400-series logic can vary widely from device to device. I've noticed this myself with multivibrator circuits which were built using 7400 quad AND gates.
Well, they're not analog chips, after all. You can get a hysteresis filter using a Schmitt trigger part, such as 74x132 instead of 74x00.
In the 80' days a lot of clocks worked using ac signal as a clock freccuency because it is very accurate in long term.
Remember the "radio reloj despertador" all of them worked using the 50/60Hz from mains, in fact most of them have a switch for selecting 60 or 50 Hz.
Makes sense, but somehow it never dawned on me back then. I assumed they just had something internally that acted as a clock, but the 60/50Hz switch is a good giveaway 😃`
Power companies actually keep track of any clock skew on the system and slowly skew the other way to bring the total number of cycles back in line with correct time. That's why it's so accurate. Before widespread cheap quartz oscillators, mains-timed clocks were actually the most accurate you could have at home for a reasonable price. Sure they weren't accurate to the millisecond, but a well-built simple mains clock would *maintain* accuracy over months as long as there was no interruption of power. It would even beat cheap low-grade quartz oscillators.
I kind of wonder in an age where almost nothing uses mains power for timekeeping if power companies even bother now. Especially now that power outages/flickers are more common.
I used to have 3 C64 computers but only one reset when the kitchen fluoro was turned on or off. The fast spike went straight through the psu and any filter Cs. If I remember correctly, tantalum caps filter out higher frequencies than electrolytics so would be better for HF noise. Once reversed, I wouldn't risk using a tant again...I have seen some light up even when installed correctly.
Interesting videos , Noel, and nice looking C64 board.
Another giveaway that the chip is CMOS is that the OUTPUT voltage goes at or near 5V. A real LS one would barely reach 4V. Not that this affects the circuit any; I'm just bringing it up to add one more way of identifying the line of chip. Excellent video, although I wish that the scope had more prominent lines for each volt.
Good point! And I agree about the volt lines. I don't think I can tweak those, but I'll check just in case. I'd love to have them more visible as well.
I repaired a ZX Spectrum programmable joystick interface recently and accidentally fitted a HC ic by accident with the others, instead of a LS. Had me looking for a very intermittent crash for ages, until I realised my mistake. The tipping point for a port crash was just on the edge and was corrupting the data line.
To elaborate on what was mentioned near the end of the video, there are actually two kinds of 74 series cmos chips. The ones that are labeled 74C, 74HC, 74AC etc, are the kind you have in the video. They are made with the functions and pinouts of 74 series TTL chips, but with CMOS voltage levels for interfacing to CMOS logic chips like the 4000 series. There are also chips labeled with numbers starting with things like 74CT, 74ACT or 74HCT, etc. These are made with CMOS but they are made to use TTL voltage levels to interface with TTL logic, but with the low power consumption of CMOS.
There are also parts labeled with numbers starting with 54 , such as 54LS08, 54ACT08, or 54HC08. These are the same as their 74 series equivalents, but are mil spec components with an increased temperature range.
Something worth mentioning, re: mains frequency synched clocks..
For several decades, clocks have been synched to mains frequency. Before electronic clocks with digital logic, clocks like the one pictured used synchronous motors - a bit like the one that drives my belt-drive turntable. Here in Australia, the lights still flicker around 2am each morning as power utilities trim the rotating machinery driving the grid such that the number of complete cycles in the 24 hour period is (as far as I know) within a certain percentage of 24hrs x 3600 seconds x 50 cycles/sec = 5,184,000 cycles. That is mandated by the fact that traditionally, so much time-keeping relied on it. Nowadays, NTP (the network time protocol servers) means our computers and phones are almost always in sync.
I love the detail of the lights flickering at 2am because of that! I never thought of that, but it makes sense to resync the pulse count. I wonder if the same is true in other countries or they handle drift differently in a more progressive way (skip a cycle every X hours).
@@NoelsRetroLab It came up in a subject I was studying. It blew my mind that all those machines connected to the grid all run in phase, in sync with each other. Has to be the case of course, otherwise there would be massive failure currents. Good job on the video, btw. I had a heck of a time trying to source some HD63C09E chips that weren't fake. Two orders out of China were then a UK vendor sourced some, (also from China) tested them first then advertised on eBay. Got to upgrade my FLEX system but I haven't done much experimentation with the chip's native mode as yet.
Great video, many thanks, love the logical fault finding and 'proving' technique...now that's how to find faults, guys. Servicing TVs in the 90's we had a batch of high power 2N3055 TO3 transistors immediately blowing up when switched on ( put on Blast hats and dive under the bench!!) Cutting a 'new' one open, we found that the silicon die was a fraction of the size it should be, yes it would test OK and run in a low power circuit fine... But, give 'em some whoomf and Bang!! .. Yes, they were 'new' from China...... I hope you like the highly technical language and description... 🙂
with the latest version of xgpro (11.71) you can actually choose the test voltage (5.0, 3.3, 2.5, 1.8). one thing to note is the test menu is no longer listed as an ic option. it has its own menu and dialog window
Your video quality and graphical explanation are getting better on each video Noel. Good work.
Much appreciated! 👍
Thanks for showing that breadboard circuit to test the TTL chips. I will check my Chinese so called TTL chips with it. Best wishes.
Glad to help!
There are two types of 74xx CMOS logic gates, the 74HC range, like the ones you were experimenting with, but there is another range 74HCT specifically designed to have the same transition voltages as standard 74TTL chips. The “T” after the HC denotes standard TTL logic levels.
I think if you put a 74HCT08 it would work.
Great diagnostic video!
Yeah, conclusion is shocking. There is no logical sense to fake 74 logic chips, but, apparently, someone gets money from that.
One option is, that assembly factory purchased lots of unmarked chips and after job was done, they resell these leftover chips very cheap or thrown away and rebranding was done by some underfloor factory in China who picked these up..
Now I need to verify my LS stock. I think that I have at least one fake, because of strange behavior in one new circuit - replaced with old used IC from different manufacturer and problem disappear..
This reminded me situation with TDA1524, back in days. Purchased originals - both was faulty. For one chip worked only one channel, for other chip worked only other channel. From different local seller I purchased other pair of chips. Seller warned me, that these chips may be fakes. I get home, plugged first chip in socket and it worked as it should.. Both chips, claimed as fakes, worked, but originals didn't.. 😂
Exactly. I never even considered that 74xx chips would be fake. But now... ugh!
Maybe you can just order them with whatever label you want
CuriousMarc did some videos the first part of this year where he was repairing a bunch of timer modules for an old HP computer. The clock circuit used was similar to the way the squaring up circuit for the CIA timer was done by Commodore, except it was crystal driven.
He found that even if he replaced the chip with the same part number from a different manufacturer it would not work properly and he had to change a feedback resistor to a lower value. Evidently this type of operation is 'quasi-analog' (made up but descriptive term) and depends very much on the analog properties of how the exact chip was fabricated. So, in a typical digital signal operation it would work fine but not in this quasi-analog state. Interestingly Ken Shirriff found out from decapping and reverse engineering flip-flip chips that they have a quasi-analog design as well which explains why they tend to fail more often than other logic chips.
Potentially this might happen even with chips from the same manufacturer if they ever changed the design.
Something which was clear is that the LS chip produced a lot less "noise" than the HC chips when given a "midpoint input".
I can imagine a low-volume manufacturer using only chips that "passed" testing for their particular application, but HP isn't a low-volume producer. They must have had success with their default supplier's chips, and never knew how close to failure they were.
@@jpdemer5 I suspect HP knew full well. They designed the circuit knowing it would work with the ICs their suppliers made and that was that.
I taught digital electronics for almost 25 years. Our supplier of lab parts, the lowest bidder, supplied fake chips. It was a nightmare for me because we were analyzing the performance of various families and they were all the same (CMOS) yet labeled differently. Also, several of the kits had chips that were labeled upside-down so had many burnups. Also, some chips were NAND for NOR chips, and NOR for NAND chips. We had to collect all the parts and purchase replacements from a reputable supplier.
I have never seen a stand-alone TTL produce an output high greater than 4.4 volts, and a CMOS no less than nearly 5 volts. This is the easiest way to test. Also, the input current unloaded for CMOS is nearly unmeasurable whereas TTL will have a few mA.
I think that Commodore engineer's decision to use the chip in its low threshold mode is a bad design by itself.
Agreed! That was really puzzling when I studied the circuit. Especially relying on the very bottom of the acceptable TTL threshold.
Well, the TTL spec guarantees the output will switch as long as the input is 2V or greater. So it may also very well switch at a lower voltage.
But 2.7V is safely above 2V, so it's perfectly fine IMO.
Perhaps they had a bunch of them in stock. Or maybe someone got a real good deal on 2.7V zeners that month.
Back in the 80's we had a very similar problem when card production swapped the manufacturer of an LS chip in the Power reset circuitry for a banking terminal (Texas Instrument vs Nat SemiConductors I can't remember which way round it was). Needless to say when turning the machine off the replacement chip enabled the processor whilst the voltage was dropping and the processor corrupted the backup memory - it took some time to find that problem. Nothing fake - just manufacturing tolerances.
CSI:Retro Lab ... Where no one has gone before... ;)
Hi Noel, me the Chinese mythbuster again! From my China factory experience, they may have shortage on the LS type of chips but not the HC, so they need to fill the shipment, otherwise the business will be ruined. Or someone did placing the wrong order carelessly. Since Chinese character is totally different from English, and Chinese factory is not a place with people with high level education, the foreman may barely can use English and make a batch of HC but customer want LS, so what? Mark them as LS!! Remark chips can cheat TLS866II and work in some situation, just this kind of circuit which require tight voltage level,fail.
Right, that makes a lot of sense. It's such a shame those things happen!
@@NoelsRetroLab I bought chips from AliExpress too, I test every single one, this is the first time you discover this new way to cheat, learned a lesson, thank you for your video, both entertaining and informational.
@@johnsonlam Do you think they feel bad about cheating or do you think they feel good that they succeeded in getting away with it? =(
I'm of the same opinion as you. The chip was bought from a reputable Chinese seller (they don't deal with NOS, or refurbished chips, only new stock), so the problem was clearly at the factory. I, too, guess they had this order, run out of LS and said "whatever, rebadge some HC and call it a day, no one will notice", and sent the chips to the seller.
Great stuff, love your systematic approach. Your explainations are crystal clear. Like to build these repos myself, and this is a great source for learning commodore stuff, the best channel in my opinion!
Great to hear! 😃
Two things:
1. Using the mains frequency as a time base is a very good idea if one does not mind losing the time when power is off. The mains is more stable than most crystal oscillators over time periods of days or weeks.
2. I would be cautious of triggering a digital circuit using marginal logic levels. The trigger voltage can change with power supply voltage, temperature and time. What works today may not work tomorrow if one or more of those parameters change. If LS chips are not available you could try HCT chips as those are meant to have TTL compatible input ranges at the expense of lower output drive.
Love your videos and the effort you make to educate. Never stop
Thank you so much! Glad you're enjoying them.
@@NoelsRetroLab every single video :)
You can also create a zener diode with an higher voltage by adding a normal diode to it in the reverse direction so its drop adds to the zener voltage. To keep the normal diode conduction mode you have to add another normal diode ok parallel to the zener and the first diode.
Difficult to do that on a premaid circuitboard.
Another great video!. I really like your way of explaining this stuff. You make it understandable, even for people like me, with almost 0 knowledge of electronics. Thanks!
So glad to hear that! Thank you.
Again great cap from your serial! :P and my gosh! how great video quality and edition!
Hello Noel! If you also change the resistor R100, not only the zener diode, to a higher value, you should get a higher voltage level on the AND gate input. So then it should also work with the "original" 74(HC)08.
Greetings, Doc64!
Yes, good point. That was going to be the next thing I would have tried this this hadn't worked. I don't think it would have mattered too much to put a slightly higher value.
Great explanation on the zener diode. I had always wondered what an application for one would look like - little did I know I had one next to me!
Another application is to reduce a "TTL" serial port signal from "5V TTL" to "3.3V TTL" signal levels when connecting to the hidden serial ports in various gadgets using USB adapters designed for 5V arduinos.
Excellent detective work Noel! Very well researched and clearly explained with great diagrams as always. 👍
Glad you enjoyed it! Means a lot coming from you 👍
For time-of-day keeping, the mains' frequency, at least in the US, is quite accurate over time. Traditionally, the multiple grids in the US have been kept it tight synchronization, partly to ensure clock accuracy.
Yes, I never realized that before I learned about this. But I guess it makes sense, especially when they used that signal much more a few decades ago.
Again a super informative episode, expertly presented! Thank you!
Excelente job! You are clear and gave the full information. Thank you!
Glad it was helpful!
Amazing video! The investigation was awesome. Thanks!
You're welcome! Glad you enjoyed it!
Great catch, can't say I would have noticed that re-mark!
I wasn't so about all of being fake but good info as mentioned here about differences in those signal levels. Thanks!
Great episode! Learned something - again!
Thank you! I also learned new stuff preparing this episode 😃
I mean, you can still get brand new 74LS-series chips from mainstream component sources (Mouser, Digikey, etc). I think this is more of a story about buying from reliable sources.
That's true. But it's hard to find a reliable source with reasonable shipping (or no minimum order) for when you just need a small component.
@@NoelsRetroLab I mean, I guess? Most places have an economy shipping option. I know Mouser will ship as little as one resistor and you can choose their economy shipping for $3.50, so I pretty much use them exclusively. It really comes down to, "how much is my time worth?"
@@An_Onion $3.50 to Europe? Seriously? No. We have a very fragmented market here and lots of companies simply don't sell to hobbyists.
@@An_Onion Sadly when you order from Mouser in Europe, shipping starts (!) at 20 Euro ($23,61). Only when you spend more than 50 Euro ($59,02) it is free shipping. So it only makes sense when you spend more than 50 Euro each order. So wait and stack all the ICs up in your card :(
Thanks for the lesson on CMOS and TTL differences. As always a very well put together video.
Quite weird, as 74LS isn't the original TTL, and 74HC isn't the usual Motorola 40xx CMOS. 74HC is a modified CMOS internal circuit that tries to approximate the original 74 series voltage levels.
Noel, this explains why I was told that you could use a TTL chip for a CMOS circuit, but you cannot use a CMOS chip for a TTL... due to the low trigger voltage levels. Blows my mind that a manufacturer would do this!
Well, this is not exactly the reason, because this is a very specific analog signal. But even in the case of digital signals, it's possible for a TTL gate to output 2.7V, which would not be recognized as high by a CMOS gate. Although I suspect that in practice that's not very often a problem.
I love the research you have done on the and gate. Very didactic.
I am planning to build a C64 replica, so I will love to see the upcoming videos on the alternative chips.
Thanks for the video...
You're welcome. Good luck with your build! It's so much fun. 😃
This was great, Noel! Thank you very much!
Glad you liked it!
Interesting and a great vid as usual. Thanks Noel.
Might note that 74HCT08 uses TTL voltage levels. I use those.
Usually LS TTL outputs are
I fixed the exact same issue fixed it with a higher zener thanks to you. Exact same label as well. I actually also had a similar fake 193, causing wrong clock. Pretty irritating. It is easy to see via the power consumption, around 4ma is too low.
That was interesting. It seem to behave like a CMOS chip with the trigger level. Did you check the current consumption?
i always look forward to your new video's and this video certainly didnt disappoint. Keep up the video's they are great
Glad to hear it!
Great job Noel, every time better videos!
Thanks a ton!
For TTL, the output transistors draw current at any input level.
For CMOS, the input voltage outside of thresholds, i.e. 2.5 volts, neither P or N transistor conducts so output is indefinite (high resistance).
Nice repair video and discovery of a mislabeled or (more likely) FAKE TTL chip!
One thing you did not mention, Noel, is that CMOS power consumption goes up dramatically as the input voltages are farther from the rail (like 2.5v for high on a 5v rail), and likewise TTL power consumption increases quite a bit as the input levels get closer to the voltage rails (such as 0.1v low input instead of .7v low input). The good thing is that for this application it shouldn't really matter as far as the input from the zener shunt is concerned. However, due to the reliability issues of (seemingly) all Commodore/MOS chips I would personally be concerned about the CMOS 74HC08 driving an input on each of the 6526 CIA chips. It's one more reason not to mix CMOS and TTL logic chips. If you have to do that a series resistor between the CMOS output and the TTL input is a good idea to prevent blowing out either the CMOS output FET's or the TTL input transistor(s).
Commodore chips are NMOS, not TTL. They can handle 5V on inputs without problems and their output is usually slightly above 4V. There is no increased power consumption for high input voltage on NMOS as the input is connected to the gate of a MOSFET, which doesn't conduct.
Changing the zener diode with one having higier zener voltage is meaningles. With ~ 2 volts on the rail you were not nearly close to that 2.7 volts on the first one so puting one with 3.3 volts zener voltage doesnt make any sense. Placing the second 74LS08(HC actually) may have solved the problem byt that is because of the tolerance of that chip and it seams that it works at the edge of its specification, which is not good because in time as the chip degrades it will probably go out of that range and the clock failure will reaccure.... Better find a genuine 74LS08 as soon as possible. The only thing that I think may help here is changing resistor R5 (560 ohms) with one with lower value (240, 270, 300 or 330 ohms). In that way the voltage drop acros the resistor will be smaller and the rest will go accros the Zener diode wich will reach its zener voltage
I wouldn't say it's meaningless. The 2V was due to the feedback resistor. The higher the input level, the higher that voltage would go. You can even see it in the oscilloscope that voltages are higher with the new zener. And I did try the HC by itself before I measured things that it didn't come close to triggering. But yes, changing the resistor would have helped as well.
It is the feedback resistor that lowers the voltage. Therefore if you aim a bit higher with the zener diode (1 to 1.5 volt higher), it might work. However, you then get a smaller pulse width.
Very nice video as usual, very well researched.
The "fake" chips could be regular chips that fail some test (temperature or other) and are therefore rejected on the standard line.
The factory could be tempted to sell them anyway where it could be assumed that they "should" work.
As far as I remember the voltage specification for high speed CMOS is as follows
HC (Standard CMOS) VCC/2
HCT CMOS with TTL 0.8 / 2.0
The reason for the threshold not being absolute values like like 0v and 5v is that TTL uses bipolar transistors.
This in turn means that currents flow in and out of circuits which changes the voltage of the logical levels. So there is a safety margin in the TTL levels.
CMOS ICs are not susceptible to these level issues as they can drive to "rail to rail" and they use virtually no current on the inputs.
CMOS chips will use less power then the equivalent chip in TTL technology but this is only valid in a stable state (not switching).
They have a pretty linear consumption to frequency ratio (consumption goes up with frequency). This is much less the case for TTL.
TTL or TTL compatible chips (74HCT) are meant to work at 5v +/- 10% (from 4.5v to 5.5v)
CMOS chips are not limited to 5v or 3.3v in fact many of the 74HC chips are specified at 6v (from 3v or even 2v to 6v)
One last thing: using logic chips in an 'analogish' way like crystal clocks is always risky as some of the characteristics of the circuit will depend on the implementation technology.
The problem here is with the original design. The extra feedback resistors are an attempt to create a Schmitt-Trigger to overcome AC line noise, however relying on the linier characteristics of a digital part is very bad practice. My guess is that the designer found he (she) had a spare gate and so decided to save the expense of adding another logic package (with a real Schmitt-Trigger) such as an SN7414. (One extra DIP package took up a fair amount of space)
Unfortunately this was done a lot in the early days but the variations in manufacturing process from manufacture to manufacture or even run to run can be large.
That's why the data sheets list typical values with a fairly wide range from Min to Max.
The moral to the story is don't design a digital logic circuit that depends on the linier characteristics of your logic elements.
All of that said, I'm guessing that the zener diode voltage was chosen to keep the transition close to the zero crossing of the AC line where power supply rectifiers (and other AC loads) might generate noise into the wave form.
I agree! That clearly shows it was added last minute because they had a spare gate already on the board 😃 Typical for computers of the time. The amazing thing is that it was never changed in subsequent revisions of the board.
As for the value of the zener, remember that the original AC signal goes all the way to 9V, so the slope is really steep over a whole range, so I think they could have safely picked lots of different value zener (maybe even up to 4-5V) and had the same effect (just a guess, I haven't really tested it).
If you really want to troubleshoot a power line derived time issue, wait until you have something in a building with solar and battery backup. Both the solar and battery systems use a pll to lock to the grid frequency. If the grid drops, the solar and battery inverters are locked together but may drift a fraction of a second relative to the missing grid power. If you have many power outages, you may find that clock circuits that rely on line sync may be minutes off over time.
Taking the timing signal from the AC mains may give a time that is ahead of or behind the real time by up to a minute or so, but the deviation does not get bigger than this. The speed of the generators at the power stations is controlled so that on average over a long period the frequency is exactly the nominal frequency, but at any given moment it could be slightly lower or higher.
Another illuminating video! Thanks!
You're welcome! 😃
ttl logic chips that brings back memories... only it was 54 series chips that were used where I worked in the mid 80's...
Good video. I work in retro pinball circuit boards and I have dealt with the same thing, usually when trying to source out obsolete parts from China.
Wow! That wasn't unexpected, as I'm following you on Twitter, but it's another example of great troubleshooting!
Thank you! Yes, that came as a surprise for me too. Wasn't expecting that!
Great video, Noel. Could you tell me where you got the solder sucker that you use. It really works very well! The ones that I have used don't work that well. Thanks.
Wow, that was very informative!!! So good to know, thank you for sharing!! Michael
My pleasure!
You cold try to measure gate input current to see if it is CMOS or TTL. At least if they haven't gone through the trouble of putting in a resistor to the input. Great video btw. 👍
The Atari 8-bit series uses a division of the color clock for the RTC functions. Since this function is subject to interrupts due to bus-mastering by the graphics chipset, the clock on these machines is not very accurate over long periods, something on the order of a superb chronometer (mechanical). The 60hz signal that drove the clocks in my schools as a boy produced much better results over long terms. I was fascinated to find this in use on the otherwise uninspiring C:64. Now I have a good reason to admire it.
Hi, question: can you still get commadore sound hips?
Another fantastic video!
Plus I enjoy seeing you use a simple handheld solder-sucker, as it's all I can afford. 🇨🇦🐧
Glad you like them!
Great video. That's really annoying when a new component is out of spec. It might be safer to go with HCT as I can imagine the demand for LS is very small these days.
Thank you! Yes, using HCT would be perfect (I just didn't have any in hand and had to improvise 😃).
20:12 Well, you accidentally found a viable replacement path for at least one LS chip in the C64! Started diagnosing a fake chip, end up slightly improving on the C64 design...
Another possibility:
Maybe those are legitimate HC ones, which were just mislabeled as LS by accident and were supposed to be disposed of .. but someone "rescued" and sold them?
I think this happens. The far East vendors get these deliveries as ewaste and they just resell them.
Could you swap out the zener diode for a 3.5v diode and have both CMOS and TTL work?
Good way to distinguish TTL logic ICs from CMOS is to measure powier consumtion at idle. TTL will drain few to even 20mA, while CMOS ICs consumpition will be about 0.
Wow, I found this channel recently, it's so underrated, I hope videos like these receive a lot more views than now, success for you, have a great day :)
Thank you so much and welcome on board! 😃👍
Another possibility is that these chips were manufactured as LS chips, but failed testing and were dumped into a reject pile that someone bought and is now selling.
That's a possibility I hadn't considered. But seeing how they trigger just like an HC, I'd be surprised. Still, I could do some other tests to find out if they're really HC.
@@NoelsRetroLab You already did: 13:34. An LS would never reach an output of 5V, you would expect something between 3.0 and 3.4 volt if it was LS.
Thanks for great content !!!
Just a small clarification: LS does not directly identify the technology. LS stands for L-ow power, S-chottky diod. 74LS08 takes less power and is much faster than 7408. There are also variants 74L08 or 74S08. Some of the chips from S series can even work well above 100MHz.
Correct. It's one very specific type of implementation, but always using TTL. I didn't mean to imply that all TTL chips were LS, I should have made that clearer.
Bring electronics production back to our countries and make them great again
Where the un-ANDed signal voltage goes under 0V, I think that can be expected. At he "unused" half wave of the 9VAC the zener diode is working in the forward region with the resistor, leaving 0,6ish V under the zero threshold. I believe that is not a problem for most logic chips, but it could be avoided with a standard diode between the 9VAC and the ballast resistor.
i love these little mysteries. gj
I was wondering if they accidently mislabled them at the factory. Great vid, and something we will have to watch out for in the future.
That's a very generous way of putting it about the factory 🤣
@@NoelsRetroLab I figure, with them being so cheap and readily available, why mislabled them on purpose ?
Very interesting channel! I'm a subscriber now!
Welcome aboard!
Another reason the CMOS part would not work is that the output swings pretty much to 5V, whereas the LS part would only be maybe 4V or so when high. So the feedback resistor on the CMOS part will raise the threshold more than the TTL one would have.
I hadn't thought of the effect of the higher output level. That's interesting. I think it might still have worked because it would always lower it enough to go low though (and clearly in the oscilloscope it just never managed to trigger the high level), but now you have me wondering what it would have done if it went high. Hmm...
I'm wondering, when you swapped the 3.3 v zener, and the first cmos chip didn't trigger, could you lower the value R37 from 2.7k to a 2.4k or 2.2 k? it would slightly increase the current flow back though the zener/R100 and pull the voltage up ever so slightly....
Won't work, R37 doesn't pull up until the chip has detected the input as high.
Excellent job!
Thank you very much!
Tandy had a similar issue in Australia in the early 80's with their model 16 B running Xenix (a derivative of Unix from Microsoft that latter became SCO Unix) in that they had set the system clock run off mains frequency which is 60 Hz in the USA and 50 Hz in Australia. It was fixed with a software patch.
Don't forget that the CIAs are also responsible for reading the keyboard and joysticks!
As for why the CIA has a separate input for the TOD clock: one issue is that while every computer system using these will have a much faster system clock, that system clock will vary from system to system depending on what was convenient for the designer. (Remember, unlike the VIC-II, the CIA was not designed just for the C64; it was designed to replace the 6522 VIA and add additional capabilities, much as the 6522 VIA was designed to replace and add capabilities to the 6520 PIA, which in turn was a clone of the widely used Motorola 6821 and 6820 PIAs. It's perfectly plausible that they hoped to sell these into industrial and embedded systems where those earlier chips were already being used.) Rather than adding more circuitry to deal with dividing down fast clocks of random rates which still won't be exact in some situations, and yet still restrict designers who might want to be able to vary the system clock rate or even pause it for various reasons, it's much easier just to accept a separate 50 Hz or 60 Hz input, especially since a _very_ reliable source of that is widely available all over the world.
08:15 was where I started yelling at my screen, by the way. Did you not notice that your tester said 74 __HC__ ? Coming immediately after an explanation of TTL logic levels, that was the problem (or at least potential problem) staring at you right there. (I'm guessing that it your tester said "HC(LS)" because it can't tell the difference, but that's a big reminder that you need to check that.) But then again, maybe I'm just already primed for this sort of thing because I get CMOS parts relabeled as TTL from AliExpress all the time. (Probably about three quarters of the 6502 CPUs I've bought from them have actually been remarked CMOS parts.)
BTW, it's also worth putting an ammeter on the power input of the chip when comparing TTL and suspected CMOS. You'd probably see a massive difference there; my CMOS 6502s draw literally less than a tenth the power an NMOS 6502 draws.
Nicely done.
It'd be interesting to take this further...
1: Measure the unloaded and loaded high and low OUTPUT voltages to see what the levels and drive currents are
2: Measure the quiescent current consumption of the chip with all I/O pins not connected (CMOS should be near zero)
3: Measure the current consumption with one input grounded (If the input is a BJT as in 74LS08, then the current consumption will increase)
etc etc
I had a similar problem trying to control a HUB75 display from a Blue Pill board. The display used CMOS chips, and the datasheet said they wanted 70% VCC for a logic 1. That's 3.5 volts, which you aren't going to get from a chip that runs on 3.3 volts. It was close enough to work most of the time, but when actually using it, sometimes a few pixels on the display would glitch because it was right on the noise threshold. I made one failed attempt at a level shifter before I moved on to other things.
I've been experiencing this quite a bit myself. Recently I got a lot of 5 MC6847 VDG chips. 4 out of the 5 worked. The one that didn't work had a notch that was different than the 4 that worked. So I did the acetone wipe and sure enough it was a rebadged chip. But it was rebadged from a 100% comaptible! It just didn't work. Why did they feel like they needed to rebadge that? So I tried the test on the other 4 chips just to see and sure enough they were rebadged too....but they were still all 6847's! I could see the original labeling underneath the paint. They didn't bother sanding them down. The only reason I could think of them rebadging them was to make the date look like they were new old stock. We all know they haven't made these in a long time, so they're not fooling us. It's just a waste of time.
I got 5 63C09P's that didn't work, and did the acetone test and I got a nice black q-tip, meaning the badging was fake, but they sanded them down so much, I couldn't tell what they were originally. These printing jobs on the labels even indent into the plastic a little bit, so you can't completely get rid of the badging, but that copperish color sure faded away.
Yes, things are really bad with larger, more expensive chips (even though they aren't THAT expensive), but this came as a surprise to me. Such an unfortunate situation that we can't trust chips like this.
@@NoelsRetroLab You can't trust CHINA!
Only manufacturers who have Americans watching over the quality control from beginning to landing in America get what they pay for.
It's a problem in practically every industry in china.
Booooom! BINGO! Fakeness explained for the masses.
Except the winner is some boss in China.
*Noel, you are the MAN.*
There are also significant differences in current draw, input impedance, and output impedance between HC, LS, etc.... chips. Also note that Zener diodes below 5.6V are very sensitive to current... they need a higher minimum current to actually pull the voltage down to the rated value. This is handled by R5 on your schematic... looks like. Mmm. 9V - 2.7V = 6.3V. 6.3V across 560 ohms is roughly 11mA. Should be enough for the zener itself. It probably needs at least 5mA to work properly.
So, that leaves current draw from U27 and R37 to cause the voltage to dip below 2.7V. The current is drawing through a 1K resistor after the zener... it doesn't take much to cause the voltage to go well below 2.7V, so the components wind up mattering a lot.
Your attempt of replacing the zener with a higher-voltage zener is reasonable, but note that the current draw from the 9V AC through the 560 ohm resistor will be lower and that can cause major problems for voltage drops across that 1K resistor. The zener's generated voltage will be too 'soft'.
Other possible solutions... reduce that 1K resistor a bit too, so the current draw from the rest of the circuit doesn't draw the zener voltage down as much. But the feedback circuit will also be sensitive to that resistor so... you might have to do some math. The whole circuit is very finely tuned. It would have been a lot easier if they had just used a schmitt trigger AND gate there, then they wouldn't have needed to square up the input with the feedback resistor from the output of the gate back to its input.
--
Yes, tants tend to explode after a while if you apply reverse voltage to them, and/or cause a permanent short.
Cool stuff. I haven't played with the PET/C64/Amiga stuff for decades(!).
-Matt
Uh that's very interesting, i'm excited. I got a KC compact which is a kind of duplicate from the amstrad cpc. It is working fine with his own CPU - a U880 (the gdr type of the z80 processor). Interestingly, this CPU works also fine in a cpc but the Z80 CPU from this cpc doesn't work in the KC compact. This Z80 CPU btw. is a brand new one. I think i take the board out on the weekend and check a few voltage levels with my oszi again. Thanks Noel for this informations. Thumbs Up.
Oh wow, I didn't know about the KC Compact! Very interesting. I'm going to have to look into getting one given my love of Amstrad CPCs 😃 I'm not sure this is the problem you're having though since the original parts (I'm guessing) are all TTL. But it's worth checking. Good luck!
@@NoelsRetroLab KC Compact are quite hard to get your hands on as they came very late in the existence of the GDR - it's almost only a C65 like batch that found its way to collectors.
Interestingly enough sometimes in those East German Z80 computers they had to use original Z80 CPUs as the demand for their clones in the industry had been way too high. This way original Zilogs found their way into my possession years before the Berlin wall fell.
Being a young greenhorn back then, I did only wonder shortly when first looking at the documentation ("Where is that U 880 D they talk of here?") but in the past few years I came to a new appreciation of this little detail of history.
Wow, I am facing a similar problem with my Z80 system (AEA PK-232), when swapping the NMOS Z80 CPU with a CMOS it doesn't work anymore. This system has also an NMOS Z8536 (in your KC compact it is the U82536) inside like the KC compact. You say your CPU is a new one. Nowadays only CMOS Z80 CPUs are made, so I guess your problem is coming from CMOS which requires and produces slightly different logic voltage levels. But even if the CPU is not CMOS, coming from a different makers it might require or produce different logic level voltages which the Z8536 will not understand and as I understood the KC compact initializes the U82536 right at the start. Applying 4.7K or 10K pull-up resistors on the data bus might help sometimes, but not in my case.
ummmm, I just ordered 74LS08's from Jameco and 4 of them look exactly like that 'fake' one. Time to test them I guess. Good job catching something so subtle with these btw!
Interesting. Let me know what you find out!