I have recently tried to test DRAM in a chip tester, and I found that you get a lot of false failures if you don't have a good 100nF-400nF decoupling capacitor right across the power lines on the chip being tested. These chips draw a lot of current very quickly when you pull the /RAS signal, and if the decoupling is too far away and the current is delayed by stray inductance in the power-leads, you're bound to loose bits in a perfectly good chip. For the task of decoupling, film capacitors should be better than ceramic capacitors. Electrolytic caps alone without something else is a big no-go, as they're way too slow.
Very good point! My knowledge of circuit design from that point of view is severely lacking, but I can see what you mean by that (and that's also why that's such a common feature in most 8-bit computers).
Very nice tip. Out of interest why are film caps better than ceramic for this as I've struggled to ever find good information on the difference, tho I was somewhat aware of electrolytic being slow before. I won't ask about tantalum as I can't cope with more than 3 types 😁
@@mogwaay Ceramic caps tend to perform worse at higher frequency. At the speeds needed for the DRAM row-refresh surge, they are usually far out of spec. Film capacitors and silver-mica keep within spec for faster operation.
One very common fault mode, since these chips are getting old, is that the cells get leaky and loose data prematurely. When you test some of the chips everything seems to be fine at first but you have to extend the time for the RAS signal to the maximum limit for refreshing a row. I had hundred of 16 kbit DRAMs (ITT 4017) loosing their contents if the RAS cycle time was at maximum. I even went so far to test the refresh time beyond the datasheet limit and see at which point data gets corrupted in order to see how much margin was left. So I sorted out the ones which were just above the maximum refresh time since they would have failed within just weeks after my repair. A lesson I learned after I repaired the first units with DRAM faults.
That's a really good point. It would be awesome to have a tester that can have that level of resolution and increase the delay until it fails and report that. That would be challenging to do with an Arduino, but maybe a faster microcontroller or using some tricks it could be possible. How did you manage to test that? Did you do a dedicated circuit, or did you test it in the computer itself?
@@NoelsRetroLab Since this was in the end 90s I used a 16 MHz 68332 system also known back then as "RaubKatCE". Hard to find an image today: docplayer.org/docs-images/91/105790635/images/77-0.jpg This one has loads of timers and would be quick enough to use the interrupt function for that. But I was cheap and lazy and polled timer counter registers just to hit the DRAM once the timer went over a certain value. It worked though. Funny thing with some DRAMs: When not refreshed they had all the bits low on even RAS addresses and high on odd RAS addresses. Or the other way round. That was weird. Never figured out why.
About Arduino protection: Atmega328p datasheet states that there are protection diodes on its inputs but the current through them should not be more than 1mA. That in fact means that puting a 10K resistor on each input should protect the Arduino from +12 and -5 volts.
All chips have that kind of protection, it's there for ESD purposes. 10k surely would protect against all the possible shorts, but might as well render the whole thing not working properly in the first place. Depends on the input impedance of Arduino and leakage current of retro memory chips. I've debugged a board at my first job years ago, which didn't work because of placing protection resistor on both ends of the trace - each was 3k3 iirc. Due to leakage currents all ligic signals were read as zero. Instead, I'd do what should have been done anyway - overcurrent protection on supply pins and external protection diodes on each logic trace.
It is a nice idea for a specific IC that works between two levels, but... the problem comes if I try to use the same system to work with differentes rates, and latencies.
You know, I tried to explain 0:48 in a recent video of mine and just ended up waffling for about 5 minutes and spending ages on diagrams that just made the whole thing even more confusing. I cut that whole section out in the end. And there you go and clearly and concisely explain the whole thing in less than a minute 😅
Haha, glad to hear that 😃 There's a lot more that can be said about that, but that seems like the minimum you need to know in order to actually use it.
I remember, as a teenager back in the early 80's working with a service technician servicing System 80's (TRS-80 clone), that these had 8x 4116 DRAM's onboard (socketed), and it pretty much became known that perhaps 90% of system failures were faulty DRAM. Therefore, it became standard practice that the first thing done was to just replace the 8x 4116 chips, and in 9 out of 10 cases, the computer was fixed! This also meant that many, still good 4116's, were thrown out. But, in the scheme of things, the time involved to fully troubleshoot, to identify the specific chip at fault, would have cost more than the super quick (and 90% effective) full "8x 4116 swap-out" repair. Crazy as that seems.
That makes sense, although depending how early in the 80s that was, that RAM could still be pretty expensive. But yeah, it does fail a lot, and some people even today recommend just swapping all the chips out when one of them fails. I don't tend to do that, but I can see the reasoning behind it.
@@NoelsRetroLab Yes, they were expensive chips. I even remember, as an electronics / microprocessor hobbyist back then, saving up, just to buy 1x 4116 chip to experiment with (in the days when we typically used 2114 1Kx4 SRAM's for our builds). So yes, I distinctly remember this proces and all of the 4116's being through out, simply because as a lowly paid teenager at the time it sure seemed absolutely crazy! ... and Yes, I was able (at a later time), to rescue some tubes of suspect 4116's from the rubbish and build my own circuit to test them. I still have a few tubes of these spare 4116's to this day. :)
Another great video from Noel. As Chris Jameson said, unbelievable timing. I'm working on a similar project to test an 8087 I bought for my IBM PC. Not easy though.
I was looking at your T-shirt thinking, "that looks familiar" and then the name "Sabre Wulf" popped into my head. So I looked it up but discovered it was a side-on game with a very similar looking character but not isometric. I also had Knight Lore in my head but for some reason thought the player character was a knight, but I looked it up anyway and discovered that is indeed the character from your T-shirt. Even weird, on wikipedia it says "The player, as Sabreman, has been bitten by the Sabre Wulf". So there is a link between the two games as they were both made by Ultimate Play the Game. I didn't own this game but a friend had it on his Amstrad 464 and I played it a bit at his house.
I have a spare working Spectrum 48K board with sockets for both RAM chips and fitted with an EPROM which dual boot to normal or diagnostic software. Selection is by pulling the high address line high or low on the EPROM. Brendon Alford's diagnostic software is unbeatable and open source.
Unbelievable timing with this video, I had just built my own DIY dram tester to check some 6464 chips on an EGA card I was working on, using a STM32 BluePill with Arduino and, I failed as it didn't find the faulty chips! I had to use my ZX SPECTRUM +2A, kinda like you suggested you could use a regular ZX 🤣. I was also sadended that this project wasn't Open as I had tried using it as a reference. I will be very interested to check your code, I had used someone else's as a starting point but clearly I'd gone wrong somewhere. Anyway good info as always and super timely, if you could make a video on whatever I find myself working on next week, thatd be great 😋
Haha, I'm glad we're so attuned that I just predict what you're going to need next 😃 (Hint: Start working on some VIC 20 stuff). My DRAM test code is no great shakes (and it's also based on another existing one), but it's certainly functional, so it's a nice start for your own code.
Despite the project not being open source, it's not that hard to design a pcb *inspired by* this pcb. Change pin usage on arduino, write own software, done. I also realised there is an easy way to test the chips for shorts as these board have analogue in and outputs beside the digital ones that are used for the read/write tests. I write software in python but that is easily translated to the code arduino uses.
perfect Noel...just what i needed to test my ram chips! one thing i have issues with that tester board is that theres no power switch.. id rather not be inserting and removing chips with the power connected...
That tester has originated from an Oric forum. One user posted about his quick & dirty Arduino based DRAM tester, the project grabbed some attention, some people found a few bugs in the code and corrected it. Later on that code was used (with or without mods) on some commercial products like the one you have there. I built my own testers in the form of UNO shields with homemade PCB's. One shield for 4116 and one for 4164-41256. Having them on 2 separate boards enabled single sided PCB design, thus homemade PCBs. I'll link the said forum page below, but I suspect you already know the story behind the kit you assembled. forum.defence-force.org/viewtopic.php?f=9&t=1699&sid=68af1e37deb5f6c5f030a930a514ff3e
Thanks, I didn't know about that project. I knew about one I used as starting point of the code in my home-made one, but I'm sure they're all pretty much the same. The main thing going for this commercial one is also testing 4116 memories.
@@NoelsRetroLab Being Oric, the original code is for 4164 chips but can easily be modified to do 4116's too. That's what I did, and I am sure Stephen Vickers did the same too. There are no IO pins left for driving a screen if you use actual Arduino pins. 9 pins for address bus A0-A8 (A8 is needed for 41256) + RAS +CAS +Din + Dout, +WE +2 pins for leds, takes up 16 pins. I used 2 other pins for 2 jumpers, one for 4164-41256 selection, one for "loop test" or "run once" selection, all that remained was D0 and D1, which are reserved for serial RTX. So I decided to use the PC screen with a serial terminal for all kinds of messages during the testing. For power, I feed the Arduino UNO from the adapter jack with exactly 12V, use the same 12V for 4116, I get the 5V from Arduino's on board regulator and generate the -5V with an ICL7660 from +5V. I believe Stephen Vickers generates all voltages from 9V input, which is a much safer approach for a commercial design like his.
You could try to cut one of the rubber feet in half with a sharp knife and stick it with the cut side to the edge of the PCB to prevent it from tipping.
I've bought the Steve Vickers tester, it works fine and I've stick a plastic below the PCB to avoid short circuit. Noel is so nice to contact him and wish him to Open Source (also the idea of OLED), for the honest word I like the idea but turn out not every Open Source project did go the right way, simply speaking I don't want to assemble the board myself because I'm not confident for my skill, and usually Open Source project just like a abandoned child, the parent are gone and no one improved it anymore, I really don't care he charge a higher price if I can get the product in a better quality and future improvement, sometimes money is a good way to attract developers did excellent work. If the developer can release product and make a living, then we have many new gadgets to play with.
Glad you liked it. Interesting about the board without SMDs. Why do you have both versions? Did you make one first and then changed to the other one? Also, we were wondering about the red LED. What's the logic about when it's on and when it turns off? (other than at the end to report an error). I wasn't able to figure it out. Thanks!!
@@NoelsRetroLab The SMDs where added to test my own skills on the early 4164 Tester PCBs (just for fun) As I developed the Power circuit part of a 4116 board, I just kept them on. Especially as I was only selling fully assembled PCBs then. Over time I merged the 4164 and the 4116 in a one board for all thingy. The code runs through 4 phases, and the red light goes out after it passes Phase 1. (Mainly just because of my lazy programming of the code!!!) Because I was mainly selling kits now, I designed a PCB without SMD, as alot of us Retro Geeks have failing eyesight for those tiny SMDs ;-) Also gave me a chance to do the PCB in Black, White and my favorite Orange!!
@@Codersys : If you ever bother to rework it again, using a third color of LED for "testing" instead of just keeping the other two lit would be desirable. Shouldn't require more pins either, so long as you can use Charlieplexing or something else similar to hang both green & red off of a single pin. And maybe some protection resistors and/or diodes between Arduino & RAMs, though that would have a much higher effect on the component cost.
Finally somebody who made this project! I am wondering if it' s a RAM problem or anything else with my ZX Spectrum +. Maybe I shouldn' t have been "playing" with that screwdriver on the pins of the ICs, around 3 decades ago...
@matskatsaba I have personally abandoned electronics, so I am not familiar with the latest trends, but I don' t think these thingies existed so many years ago, either.
They probably didn’t answer because I had already made it and it was open source. I made it with both the LEDs and an LCD then I made it with the exact OLED you showed and sold a few. Mine was only for the 4116 chip though.
You're getting pretty good at the animated graphics Noel. Good info as usual, the best part of using these 8Bit systems is that's is possible to observe and (sometimes) understand all of the parts and how they work. I've not tried to test a DRAM chip in place with an IC test clip with my backbit. Interesting
@@NoelsRetroLab I really like it. It’s expensive, but it you do a lot of repairs then it may be worth it. I’ve been able to shortcut many hours of troubleshooting with it.
Interesting video. I think having some over voltage protection could be good. What I would like to see is the Red LED gone. Use the green LED just as a power indicator. Then have two RGB LED's, one at the end of each socket, with flashing yellow = testing, solid green = good, and solid red = bad, you know, like simple traffic lights.
All you need in to replace the two LEDs with a combined red and green double LED in one 3-pin package. Then it will flash yellow and red while testing, be solid yellow any time both lights should be lit, and only go solid red in the fail state. No other modification should be necessary, just use a double LED. Be careful, there are double LEDs with only 2 pins that have them wired opposite each other, so you'd get yellow by sending AC. You don't want those in this application.
Some while ago I built a static ram tester for the 2111 RAM chips in my reproduction MK14 project, and I built it on a Arduino Shield proto-pcb so it could be plugged on top of one of my spare Arduino Unos (like a lot of people I have a number going spare!). This means I can use the Uno for other experiments when not using the RAM tester (i.e. most of the time!). I was also going to build a generic chip tester using an I2C I/O expander or 2 but that project is still on the "to do" list :) I like the idea of making a more protected device with an OLED screen, and I'd probably want to arrange for the Arduino to be able to turn the +12/-5/+5 voltage rails off when not testing to make inserting/removing the chips a bit safer. The problem with the Arduino inputs is that +12v on an input will likely kill the input, and the -5v although clamped by the input diodes on the Arduino will probably cause excessive current to flow and kill the input anyway. One way to protect the inputs would be to use a series resistor to limit the current and a 5v Zener to clamp the voltage to 5v, so that negative or over-voltage would not be a problem. This would require a lot of extra components, however, and maybe not worth the cost considering how cheap Arduinos are today.
Nice. Maybe it's time to pick up that project again 😃 Agreed that it should be voltage protected ideally. Even though Arduinos a cheap, you WILL be using it with known faulty DRAM, so it's just a matter of time. Also like not having any power until the test starts.
I think it is an evolution of the DRAMARDUINO which was pretty much open source. The firmware was pretty hit and miss. This seems more accurate but I still get false negatives / best of 3 or 5 runs on some chips. Would be good to know what the firmware is actually doing lol. Great vid.
The seller sends the parcel tracked and insured because bad buyers lie and say the package did not arrive and eBay forces a full refund. I don't ship untracked any more.
Thank you for your support @Nuts n Proud. You are absolutely right, also postage to USA cost £15.15 plus taxes from Ebay too. Most people are honest, but I have been ripped off by a guy in Finland once too 😢😢😢
Yes, that would be fantastic. You pop in the DRAM chip, you tell it it's a 4164 and it reports if it works or not AND the maximum speed at which it works. That'd be sweet!
If PWM is used for the clock signal, then the speed can be easily be set arbitrarily from 0 to 20 (or whatever the duino supports) MHz, with one or two commands.
@@NoelsRetroLab Hi Noel really interesting video on the tester. I was rather hoping you would mention testing other types of chips, forgive me as i am not technically literate but would it be possible to have testers that can test other chips like VIC-II, c64 logic chips, CIA chips, PLA chips SID chips as well as other chips from zx spectrum and Amstrad ect.. basically all 8 bit system platforms individual chips? Is this feasible or even possible to do? I know we can swap chips with known good chips, but i have always wondered why there are not chip testers beyond DRAM chips ect? Would really like to hear your thoughts on why in any future videos regarding chip testers?
I looked into it a bit and when you're trying to validate something works at 60ns timings (which even some of these old parts do!) you need to be able to sample in the 10s to 100s of MHz. Not impossible but definitely trickier -- I'm not sure if it's even possible with a low end STM to get that kind of timing precision on GPIOs. I was looking into building a tester with a cheap FPGA but it's a back burner project...
@@josephlunderville3195 yeah, I came to the same conclusion. I've thought it might be a good job for a pocket beagle, but that starts getting into the same price as a cheap FPGA.
First of all, thanks for this Video. I like that you have explained how to do checks yourself and showing code before showing a "prebuilt" device. I think you are right on voltage protection, I also thought about that. Arduinos cannot handle 12V on the I/Os. But then I thought protection will expand the circuit massively and make the whole thing more expensive. An Arduino is a cheap device and it may be cheaper to buy a new one. Personally, I don't check that many RAMs that it would justify the higher pricing.
Glad you enjoyed it. Yes, you're right. That's another way to see the whole protection thing: cost of replacement x chance it would be damaged. I'm working on a review of a chip tester taking a completely different approach, so it should be interesting. Cheers.
The 7660 is the voltage source of for the +12 and -5V regulators. -9V is from its output pin and the +18V is via the attached voltage doubler comprising of the 2 diodes and capacitors. The 7660 isn’t really needed because the equivalent can be generated by an AVR output pin and a few discretes.
I don't know if the kit uses Ardnuino code, but if so, guess it could be much faster with direct C programming of the registers. digitalWrite had a significant overhead the last time I used it.
Yes, I learned about that recently. I'm not sure if that matters given the speed limitations of the DRAM, but I haven't measured anything. Could be interesting to find out.
Yes, that's something I only learned relatively recently (but that tester code is pretty old so I never changed it). I see it makes a huge difference in speed.
i read somewhere once for higher than 5v voltages to go into an arduino or something there is a way to split it into a few parts and feed it into the aurdino and not damage it. for example 3x 4v parts. but i dont know anything beyond that i just remember reading it when someone was talking about higher voltages
@@NoelsRetroLab no problem at all. im not clued up on this sort of thing but what i remember is someone was trying to process some sort of data and the voltage was too high so this was kind of a solution to still process it. i think it was data logging something
I was thinking that to protect against overvoltage, you could maybe use a 5.6V zener diode? Maybe add a LED for visual indication. Not sure about negative voltage though. 🤔
Power jacks that use negative at center pind is not odd. Back in time is was normal to use center pind as negative. At some point in the '80 it change to be positive on centerpin. It might be due to the awareness of EMC.
A quick continuity test from vcc pin to the other pins via code prior to voltage being applied would be a good preliminary test. This would also solve the hot insertion issue. Of course it would need a design change with relays or similar. #rabbithole
The 7660 has been around for over 30 years. It’s just a capacitive charge pump which was later inbuilt into MAX232 serial transmitters. The S variant has a boost pin to select a faster operation frequency. One can also use a 555 with a few diodes and capacitors on its output as a multiplier to accomplish a similar result. Just remember that its output power is very limited.
9:55 Power supplies in the 80s and before were commonly center negative. 2000s and later are almost always center positive. 90s is a mix. There are exceptions with some old ones being CP and some newer ones being CN, but for the most part that guidance should usually be true.
Indeed! I knew I wasn't imagining it. As a kid in the '80s, the majority of my power supplies were center-negative. Center-positive was quite uncommon at the time. It was only when I read the comments of a TH-cam video last year that I realized that many people today regard a center-positive supply as virtually standard and don't really think to check polarity. That came as a shock to me.
Putting a DRAM chip that is so sensible in a powered socket? And I clearly remember that some chips (4116) have a special order to apply the necessary voltages on power-on (-5V to be applied first and powered down last and never to be positive when 12V is applied). I know of a Eastern European country Spectrum clone that was literally destroying his RAM due to the faulty power-on sequence in the power supply. It was upgraded to 64k chips that have 5V only power supply.
Technically, you are correct 100%, but I designed my own tester PCBs using the same code with that the kit seen in the video. During prototyping, I've included small 5V relays controlled by the Arduino to switch the power lines on and off between chip changes, but later on I totally omitted them in the final design because I wasn't able to damage even 1 chip out of tens of them with hot swapping. Somehow they take it very well.
You could create one that tests in-circuit by having it plug in to the 6502/Z80 socket. Would need to understand the machine a little for things like banking but that's not hard to do.
A few months ago Mark also did a review about this tester, so I bought one. Shipping from the UK since brexit is no fun at all.. it takes ages, still waiting. Hopefully it arrive soon. I hope to find a device once to test Amiga rams.
honestly i think this would be better as a shield that you plug ontop of an Arduino Uno/Mega (or even a RaspPi) instead of such a large/bulky PCB obviously only if you could fit everything of course. (like couldn't you switch between 4116 and 4164 voltages using some MOSFETs so you can save one socket? downside you would have to generate 12V and -5V from the single 5V input, plus side you can use USB to power it, making it much more convenient to use) also i would personally never ever ever solder on an IC or Module like that, i would always use sockets in case i screw up the soldering or the PCB is wrong in some way so that those components can be easily removed without tedious desoldering.
Arduino Nanos are so cheap, that it's nice to have them dedicated to this instead of having to pull out the Uno, hook it up to the PCB, and star testing. Totally agreed about the transistors to route the right voltages and keeping it down to a single socket though! And while you have that, you can also turn off voltages completely before the test starts, so there are no problems inserting the IC.
Real men write self-relocating assembly-code on the native-computer to identify the bad bit/chip-LOL! 🤣 Did this on an Osborne-1 to identify a bad 4116; the only component to ever fail (other than the switching power supply). Great video, glad to see people working/playing and enjoying the old school hardware! 👍
@@NoelsRetroLab Good Man, thanks for sharing the code. Nothing like the precision, elegance and predictability you can get with assembly. Did a lot of development on a Mostek Z-80 assembler that had a 56-pin card-cage with 4-cards making up the computer-part; the assembler was in 2708 eproms; this was hooked to an ASR-33 teletype with a paper-punch to store the programs. Those were the days!
nice video. to be honest though, imho, they've spent a lot of work designing the pcb etc, and as they are including the cost of the arduino nano in there as well i'm not surprised at the price, well priced imho. the cost of shipping is probably more the UK royal mail shipping to the EU now we divorced you :D -- that said... the other issue i have is with yuou saying it should be open source..... that's fine and in a way i agree with you, but on the other side, as i say, he has to pay to produce the pcb etc, and you can gaurantee that if he had open sourced it, then within a couple weeks it would already be on banggood and ali express within a month lol
_You can gaurantee that if he had open sourced it, then within a couple weeks it would already be on banggood and ali express within a month lol_ As if _not_ open sourcing it makes any difference, it just raises the bar to cloning slightly. If there is a perception of money to be made, it'll get cloned.
Great review. Thank you I am pretty sure the creator of the project has hundreds of hours into the project. Shipping is really expensive in the EU. Being in the US I cant even buy the kit.
Hi Noel, one way to shorten the through hole protrusions is to clamp the pcb and run a standard steel file flat over the board till the protrusions are flush withj the solder blobs after which you can make it really smooth (no burrs at all!) with a piece of scotch bright. This can be handy if one is not planning to use a case for a pcb, for whatever reason. And I may be able to help you with getting the display running, I have been experimenting with ST7735s and ST7789 driven displays in python on an RPi 4B, even wrote my own FX routines. It is still a work in progress. Kudos for another fine video and interesting topic, this is definitely one I will be building when my attic is finished :P Would a collaboration to make our own arduino ram tester with a display and short protection, be an idea?
Wouldn't that cause all sorts of stress on the solder joints? People already yell at me sometimes for cutting the leads after soldering. If I did that some people would have a heart attack 😃
@@NoelsRetroLab I have done a few times, I think people are way too careful with this stuff. If soldered properly it has only been heated a few seconds and is more than capable of handling a few strokes with a file. Afterwards, reflowing the joints is an option. Besides, the calculated cost of the pcb is 4 bucks, so what's the problem LOL. Of course I wouldn't do it on a retro board. Would you be interested in some technical information about those SPI displays?
Haha, it comes and goes depending on laundry rotation. It's gotten quite a bit of holes, so I think there's only one red left and it's about to go away 😃
I built one a month or so ago with an Arduino and breadboard (perf board). I found it passed bad chips. I modified the code (although I ain't worth a hoot in C++). It now found some weirdly bad chips. Between now and then, I had bought a DIY Chris ram tester. Fast forward to last week. I was having problems. The Chris tester showed good ram. I stopped and finally built the Retro Chip Tester Pro that I have had for a year. It found that RAM chip BAD. I found my breadboard arduino monstrosity and it found the ram bad. A few weeks prior to this, I had a RAM chip that failed my tester and passwed Chris's tester. I sent him a message without a response. IK opened up the Chris tester and about the only difference is that he doesn't use resistors between the arduino and the 4146 RAM pins. I don't know if it is the code he has loaded in his or the lack of resistors?
These are only basic tests if you are not checking write or read timings. So a good chip could easily still be bad. The easiest way is just put a zif socket in the computer and run a memory test.
Right. These are very basic tests, but will still catch 90% (or more) of the failing chips out there. It would be great to extend the tests to include timings though!
Interesting. I knew the TL866 tested many non-EPROM chips, but I didn't realize it also tested RAM. Let me check... Hmm... in the list of supported chips I have locally it doesn't like 4164 or 4116. Are they listed by a different number?
@@NoelsRetroLab My bad, I missed the point that 4116 is Dynamic ram, but minipro is only able to test static ram. I had a bunch of 3564 and 4364 and they tested perfectly fine as 6264, so this made me thinking that 4116 will do the same, until I realized that this is dram.
I wonder if something like SCR lockup could happen with the power always being applied to the ZIP socket like that. It might be possible for the -5V to make contact first or perhaps a bit after the other rails have connected. Could be an idea to have some MOSFETs on the power pins to the DRAM that supply power only when under test?
Yes, that's a good point. It would seem safer to only apply power when the test starts. And if when that's solved, then it makes it easier to route power to whatever pins and you can test multiple chips in a single socket. But obviously it complicates the design.
How about timing? Does it test timing requirements. I don't know how many speed grades are there for these old DRAM chips, but it is an important factor in a computer.
The problem with this kit is that one can by 4 sets DRAMs(x8pcs) in that price and replace them all! I just bought TESLA DRAM for 6,5 euros+4euros shipping from the Czech Republic and without customs taxes.
@@mal2ksc I don't believe that I won't be able to find 8 working chips out of 32 chips bought at the same price tag of that tester..... After all high rated sellers on eBay have their own testers.......
Oh, so I'm supposed to put my hand over my work when trimming thicker components? That will be good news for the cat. I’m not at the stage of testing any ram yet, I’m very much at the beginning of my journey with electronics. However, May I ask if you could program an ATmega328p rather than use a full Arduino Uni or Nano?
Your cat will love all the bits flying around! 😃 Yes, I'm sure you can use a microcontroller directly, it's just that the Arduino makes it easier to get going, but it's the same idea. You'll probably have more control over the timings skipping the Arduino if you're going to go that route.
Thanks! I think both LEDs turn on at startup, whether you reset it or power the board, but I'm not 100% sure. The weird bit is that the red one turns off half way through a test. We might be able to get Stephen (the designer) to pitch in and explain what the meaning of the LEDs is.
I wonder if using one of the cheap ARM modules like the 120MHz Adafruit ItsyBitsy M4 Express or one with an even higher clock speed would also allow RAM speed compliance testing.
I doubt we need anything that fast, but it would be easy to run through the calculations. It could be cool if they would report the true speed of the DRAM though (15ns vs 12ns or whatever).
Here's how to solve the shipping cost problem: get 4 like-minded enthusiasts to pool an order with. These kits are light and I'll bet you could ship 5 of them for about the same cost as one. That way, you split the shipping 5 ways and pay only 20% of the cost! :-)
Right. That's what I was afraid of. Someone else mentioned that the real damage is a current over 1mA, so that could be prevented with some resistances. Something worth considering for future versions I suppose.
Hehe 😃 Which IC pin cheat tool? The ones taped to the side of my workbench? Yes, those come in really handy since I still can't memorize the pin layout for the Z80 or even for measly 4164 DRAMs! (it's kind of random crazy).
@@NoelsRetroLab 7:08 That one! The one that fixes up the pins so you can place an IC relatively easily. The PPP hasn't gotten back to me (probably because they don't exist) but I'm pretty sure that's cheating :P
Thanks for demonstration So I ahave Amiga 2000 with green screen And all traces are good I can't test Kickstarter unfortunately or buy one So I will test ram But the ram in rev 6 (256×4) 44c256 So can I use same idea with Arduino mega or I just dreaming Please give me your idea Sorry for my poor English
That's the thing: We don't know because the code isn't open source! It would be a really cool thing to add that to a tester and then report the true RAM speed.
It would be nice for a bit more coding and layout to support other DRAM and SRAM chips. A bit more involved but would be very helpful for the retro community. Nice project nonetheless.
Right. The problem is not so much the extra coding, it's bringing the right voltages to the right Vcc pins. That's why this particular one has two sockets. But I agree, having one ultimate DRAM testing open source project would be great.
@@NoelsRetroLab the I/O pins can be ran through level shifters for dealing with other voltages and a few reed relays to make it work with a bunch more chips and layouts. it would be more involved but the possibilities are there and even if it cost a few times more it would be worth the effort and the investment.
Other possibilities with the pins for the nano. 1) Put the pins in the other way up - short end through the mother board. It doesn't matter of the pins are long on top of the nano. 2) Put them in the nano first, then not all the way through the main board by supporting the nano with some little thing (or even by moving the plastic down).
Wow awesome. Your channel has a unique ability to make fairly random videos that always happen to line up with my issues at the time. Thanks so much for the content!
I'm no expert, but other products manage to deal with all kinds of chips in the same socket, so to be limited to two types with no protection against wrong insertion, makes it less than ideal. I also don't think the sockets should be live while the user inserts the chip. It would probably be best to also not have to solder the Arduino to the board, but supply some female headers instead.
That's true, but to be fair, routing Vcc and ground safely to different pins isn't trivial. But I'll review other testers that do that, so we'll see how they compare. Agreed as well about not great about the power being live all the time on the socket.
I would even say just use a normal cpu with the bootloader and the program instead of a full arduino would work better and if a chip fails or get killed, that is socketed, it's no big deal, just burn another one. This project is for people that know how to do it and you could make the board even smaller.
These chips ain’t logic chips. The minipro can also test the PLA, some guy made a video about it on TH-cam, he evenmade a file available to load in your minipro software to do so yourself.
Self hacking, can you download the arduino software, and figure out how to add the oLED screen? (not to steal the software, but for personal use) Is it using all the data pins for the tester? Seems so flexible. Julian Ilett is doing similar with his cobbled together remote temp sensing, creating boards first for his controls, then for a nRF radio board, now combining them all into a single project board using one of the small qty PCBoard makers, and even having them put the SMD parts in place. I could easily see this done with this project. BTW, also, what's up with the hoarding of dead parts? Or now that you are able to test, will you recycle or trash them? I showed someone your 'collection' and told them to stop bugging me for mine... ;)
The real value of that tester is the PCB, not the software, and since adding the OLED screen would require modifying the PCB... I think I'm just better off designing one from scratch. It's not that hard. But I hear people are making an open source project, so I'm hopeful about that. As for the dead parts... they come in handy for stuff like this, or when I want to test how a computer with a faulty chip would behave. They don't take much space, so I'm fine keeping them 😃
Mostly because I'm shaky in analog electronics and power distribution. Someone else can probably do a much better job at addressing those things than I can. And it sounds there are some good open source projects out there already.
Yes, I"m sure 😃 I've known Leo for a while. Besides, Stephen, the seller, already replied here with comments. So unless it's some kind of long-term con, it's safe to say they're two different people.
Yeah, someone else mentioned and I checked, but it doesn't seem to support them (even though it supports tons of other chips). That would have been great. I wonder why they don't do DRAMs.
Mmm, I absolutely don't mind it not being open source, simply because people don't work for free all the time. But I would expect a more polished product indeed and at least replying to questions. Communication is then key.
I have recently tried to test DRAM in a chip tester, and I found that you get a lot of false failures if you don't have a good 100nF-400nF decoupling capacitor right across the power lines on the chip being tested. These chips draw a lot of current very quickly when you pull the /RAS signal, and if the decoupling is too far away and the current is delayed by stray inductance in the power-leads, you're bound to loose bits in a perfectly good chip. For the task of decoupling, film capacitors should be better than ceramic capacitors. Electrolytic caps alone without something else is a big no-go, as they're way too slow.
Very good point! My knowledge of circuit design from that point of view is severely lacking, but I can see what you mean by that (and that's also why that's such a common feature in most 8-bit computers).
@@NoelsRetroLab I've regularly seen the caps having been soldered right on top of the VCC and GND pins of the ICs, on retro boards. 90's boards too
Very nice tip. Out of interest why are film caps better than ceramic for this as I've struggled to ever find good information on the difference, tho I was somewhat aware of electrolytic being slow before. I won't ask about tantalum as I can't cope with more than 3 types 😁
@@mogwaay Ceramic caps tend to perform worse at higher frequency. At the speeds needed for the DRAM row-refresh surge, they are usually far out of spec. Film capacitors and silver-mica keep within spec for faster operation.
@@TheMovieCreator thanks for that, good to know 👍
One very common fault mode, since these chips are getting old, is that the cells get leaky and loose data prematurely. When you test some of the chips everything seems to be fine at first but you have to extend the time for the RAS signal to the maximum limit for refreshing a row. I had hundred of 16 kbit DRAMs (ITT 4017) loosing their contents if the RAS cycle time was at maximum. I even went so far to test the refresh time beyond the datasheet limit and see at which point data gets corrupted in order to see how much margin was left. So I sorted out the ones which were just above the maximum refresh time since they would have failed within just weeks after my repair. A lesson I learned after I repaired the first units with DRAM faults.
That's a really good point. It would be awesome to have a tester that can have that level of resolution and increase the delay until it fails and report that.
That would be challenging to do with an Arduino, but maybe a faster microcontroller or using some tricks it could be possible. How did you manage to test that? Did you do a dedicated circuit, or did you test it in the computer itself?
@@NoelsRetroLab Since this was in the end 90s I used a 16 MHz 68332 system also known back then as "RaubKatCE". Hard to find an image today: docplayer.org/docs-images/91/105790635/images/77-0.jpg This one has loads of timers and would be quick enough to use the interrupt function for that. But I was cheap and lazy and polled timer counter registers just to hit the DRAM once the timer went over a certain value. It worked though. Funny thing with some DRAMs: When not refreshed they had all the bits low on even RAS addresses and high on odd RAS addresses. Or the other way round. That was weird. Never figured out why.
Nice to see dedicated chip testers - especially when the diagnostic tools are not specific enough to identify a specific bad chip.
About Arduino protection: Atmega328p datasheet states that there are protection diodes on its inputs but the current through them should not be more than 1mA. That in fact means that puting a 10K resistor on each input should protect the Arduino from +12 and -5 volts.
Aha! That's really useful hard data. Thank you!!
All chips have that kind of protection, it's there for ESD purposes. 10k surely would protect against all the possible shorts, but might as well render the whole thing not working properly in the first place. Depends on the input impedance of Arduino and leakage current of retro memory chips. I've debugged a board at my first job years ago, which didn't work because of placing protection resistor on both ends of the trace - each was 3k3 iirc. Due to leakage currents all ligic signals were read as zero.
Instead, I'd do what should have been done anyway - overcurrent protection on supply pins and external protection diodes on each logic trace.
Don’t rely upon the ESD protection diodes as a replacement for ESD handling precautions. It’ll bite you in the proverbial when you least expect it.
It is a nice idea for a specific IC that works between two levels, but... the problem comes if I try to use the same system to work with differentes rates, and latencies.
You know, I tried to explain 0:48 in a recent video of mine and just ended up waffling for about 5 minutes and spending ages on diagrams that just made the whole thing even more confusing. I cut that whole section out in the end. And there you go and clearly and concisely explain the whole thing in less than a minute 😅
Haha, glad to hear that 😃 There's a lot more that can be said about that, but that seems like the minimum you need to know in order to actually use it.
I remember, as a teenager back in the early 80's working with a service technician servicing System 80's (TRS-80 clone), that these had 8x 4116 DRAM's onboard (socketed), and it pretty much became known that perhaps 90% of system failures were faulty DRAM. Therefore, it became standard practice that the first thing done was to just replace the 8x 4116 chips, and in 9 out of 10 cases, the computer was fixed! This also meant that many, still good 4116's, were thrown out. But, in the scheme of things, the time involved to fully troubleshoot, to identify the specific chip at fault, would have cost more than the super quick (and 90% effective) full "8x 4116 swap-out" repair. Crazy as that seems.
That makes sense, although depending how early in the 80s that was, that RAM could still be pretty expensive. But yeah, it does fail a lot, and some people even today recommend just swapping all the chips out when one of them fails. I don't tend to do that, but I can see the reasoning behind it.
@@NoelsRetroLab Yes, they were expensive chips. I even remember, as an electronics / microprocessor hobbyist back then, saving up, just to buy 1x 4116 chip to experiment with (in the days when we typically used 2114 1Kx4 SRAM's for our builds). So yes, I distinctly remember this proces and all of the 4116's being through out, simply because as a lowly paid teenager at the time it sure seemed absolutely crazy!
... and Yes, I was able (at a later time), to rescue some tubes of suspect 4116's from the rubbish and build my own circuit to test them. I still have a few tubes of these spare 4116's to this day. :)
Another great video from Noel. As Chris Jameson said, unbelievable timing. I'm working on a similar project to test an 8087 I bought for my IBM PC. Not easy though.
I was looking at your T-shirt thinking, "that looks familiar" and then the name "Sabre Wulf" popped into my head. So I looked it up but discovered it was a side-on game with a very similar looking character but not isometric. I also had Knight Lore in my head but for some reason thought the player character was a knight, but I looked it up anyway and discovered that is indeed the character from your T-shirt. Even weird, on wikipedia it says "The player, as Sabreman, has been bitten by the Sabre Wulf". So there is a link between the two games as they were both made by Ultimate Play the Game. I didn't own this game but a friend had it on his Amstrad 464 and I played it a bit at his house.
That's exactly right! Your brain made some good associations! 😃
I have a spare working Spectrum 48K board with sockets for both RAM chips and fitted with an EPROM which dual boot to normal or diagnostic software. Selection is by pulling the high address line high or low on the EPROM.
Brendon Alford's diagnostic software is unbeatable and open source.
Right. If you already have it, that's a fantastic solution.
Unbelievable timing with this video, I had just built my own DIY dram tester to check some 6464 chips on an EGA card I was working on, using a STM32 BluePill with Arduino and, I failed as it didn't find the faulty chips! I had to use my ZX SPECTRUM +2A, kinda like you suggested you could use a regular ZX 🤣. I was also sadended that this project wasn't Open as I had tried using it as a reference. I will be very interested to check your code, I had used someone else's as a starting point but clearly I'd gone wrong somewhere. Anyway good info as always and super timely, if you could make a video on whatever I find myself working on next week, thatd be great 😋
Haha, I'm glad we're so attuned that I just predict what you're going to need next 😃 (Hint: Start working on some VIC 20 stuff). My DRAM test code is no great shakes (and it's also based on another existing one), but it's certainly functional, so it's a nice start for your own code.
Despite the project not being open source, it's not that hard to design a pcb *inspired by* this pcb. Change pin usage on arduino, write own software, done. I also realised there is an easy way to test the chips for shorts as these board have analogue in and outputs beside the digital ones that are used for the read/write tests. I write software in python but that is easily translated to the code arduino uses.
perfect Noel...just what i needed to test my ram chips!
one thing i have issues with that tester board is that theres no power switch.. id rather not be inserting and removing chips with the power connected...
Good to hear! Good point about the power switch. That seems like a useful feature!
Yeah I agree about the power switch too.
Yes lol, I cringed every time
That tester has originated from an Oric forum. One user posted about his quick & dirty Arduino based DRAM tester, the project grabbed some attention, some people found a few bugs in the code and corrected it. Later on that code was used (with or without mods) on some commercial products like the one you have there. I built my own testers in the form of UNO shields with homemade PCB's. One shield for 4116 and one for 4164-41256. Having them on 2 separate boards enabled single sided PCB design, thus homemade PCBs. I'll link the said forum page below, but I suspect you already know the story behind the kit you assembled.
forum.defence-force.org/viewtopic.php?f=9&t=1699&sid=68af1e37deb5f6c5f030a930a514ff3e
Thanks, I didn't know about that project. I knew about one I used as starting point of the code in my home-made one, but I'm sure they're all pretty much the same. The main thing going for this commercial one is also testing 4116 memories.
@@NoelsRetroLab Being Oric, the original code is for 4164 chips but can easily be modified to do 4116's too. That's what I did, and I am sure Stephen Vickers did the same too. There are no IO pins left for driving a screen if you use actual Arduino pins. 9 pins for address bus A0-A8 (A8 is needed for 41256) + RAS +CAS +Din + Dout, +WE +2 pins for leds, takes up 16 pins. I used 2 other pins for 2 jumpers, one for 4164-41256 selection, one for "loop test" or "run once" selection, all that remained was D0 and D1, which are reserved for serial RTX. So I decided to use the PC screen with a serial terminal for all kinds of messages during the testing. For power, I feed the Arduino UNO from the adapter jack with exactly 12V, use the same 12V for 4116, I get the 5V from Arduino's on board regulator and generate the -5V with an ICL7660 from +5V. I believe Stephen Vickers generates all voltages from 9V input, which is a much safer approach for a commercial design like his.
I just use a Z80 as a DRAM tester because the Z80 is designed to work with DRAMs.
You could try to cut one of the rubber feet in half with a sharp knife and stick it with the cut side to the edge of the PCB to prevent it from tipping.
Good idea! I'll try that.
OH!!! That's what "ras" and "cas" are all about! You've just answered a question I asked 40 years ago! :)
I've bought the Steve Vickers tester, it works fine and I've stick a plastic below the PCB to avoid short circuit. Noel is so nice to contact him and wish him to Open Source (also the idea of OLED), for the honest word I like the idea but turn out not every Open Source project did go the right way, simply speaking I don't want to assemble the board myself because I'm not confident for my skill, and usually Open Source project just like a abandoned child, the parent are gone and no one improved it anymore, I really don't care he charge a higher price if I can get the product in a better quality and future improvement, sometimes money is a good way to attract developers did excellent work. If the developer can release product and make a living, then we have many new gadgets to play with.
Another great video, and thanks for the review of my PCB design. There is also a version with NO SMDs now too, for people that hate them things ;-)
Glad you liked it. Interesting about the board without SMDs. Why do you have both versions? Did you make one first and then changed to the other one?
Also, we were wondering about the red LED. What's the logic about when it's on and when it turns off? (other than at the end to report an error). I wasn't able to figure it out.
Thanks!!
@@NoelsRetroLab The SMDs where added to test my own skills on the early 4164 Tester PCBs (just for fun)
As I developed the Power circuit part of a 4116 board, I just kept them on. Especially as I was only selling fully assembled PCBs then.
Over time I merged the 4164 and the 4116 in a one board for all thingy. The code runs through 4 phases, and the red light goes out after it passes Phase 1.
(Mainly just because of my lazy programming of the code!!!)
Because I was mainly selling kits now, I designed a PCB without SMD, as alot of us Retro Geeks have failing eyesight for those tiny SMDs ;-)
Also gave me a chance to do the PCB in Black, White and my favorite Orange!!
@@Codersys : If you ever bother to rework it again, using a third color of LED for "testing" instead of just keeping the other two lit would be desirable. Shouldn't require more pins either, so long as you can use Charlieplexing or something else similar to hang both green & red off of a single pin.
And maybe some protection resistors and/or diodes between Arduino & RAMs, though that would have a much higher effect on the component cost.
Great video. More testers please. All kinds. Reminds me of the tube testing days at Radio Shack.
Next video coming up is a super-duper tester. Stay tuned 😃
Finally somebody who made this project! I am wondering if it' s a RAM problem or anything else with my ZX Spectrum +. Maybe I shouldn' t have been "playing" with that screwdriver on the pins of the ICs, around 3 decades ago...
@matskatsaba I have personally abandoned electronics, so I am not familiar with the latest trends, but I don' t think these thingies existed so many years ago, either.
Nice video as always!On ebay there is an italian seller with a version with OLED screen.
They probably didn’t answer because I had already made it and it was open source. I made it with both the LEDs and an LCD then I made it with the exact OLED you showed and sold a few. Mine was only for the 4116 chip though.
You're getting pretty good at the animated graphics Noel. Good info as usual, the best part of using these 8Bit systems is that's is possible to observe and (sometimes) understand all of the parts and how they work. I've not tried to test a DRAM chip in place with an IC test clip with my backbit. Interesting
Thank you! Glad you're enjoying them. I actually didn't know about the BackBit until now. Interesting!
@@NoelsRetroLab I really like it. It’s expensive, but it you do a lot of repairs then it may be worth it. I’ve been able to shortcut many hours of troubleshooting with it.
Nice, it should be open source of course! Socket the Nano to build in more robustness
@5:37 and that's the ideal time to thank our todays sponsor PcbWaaaa...! 😂
Interesting video. I think having some over voltage protection could be good. What I would like to see is the Red LED gone. Use the green LED just as a power indicator. Then have two RGB LED's, one at the end of each socket, with flashing yellow = testing, solid green = good, and solid red = bad, you know, like simple traffic lights.
Yeah, that would be pretty nice.
All you need in to replace the two LEDs with a combined red and green double LED in one 3-pin package. Then it will flash yellow and red while testing, be solid yellow any time both lights should be lit, and only go solid red in the fail state. No other modification should be necessary, just use a double LED. Be careful, there are double LEDs with only 2 pins that have them wired opposite each other, so you'd get yellow by sending AC. You don't want those in this application.
Wonderful analisys and review! Let's do it open source and better
Some while ago I built a static ram tester for the 2111 RAM chips in my reproduction MK14 project, and I built it on a Arduino Shield proto-pcb so it could be plugged on top of one of my spare Arduino Unos (like a lot of people I have a number going spare!). This means I can use the Uno for other experiments when not using the RAM tester (i.e. most of the time!). I was also going to build a generic chip tester using an I2C I/O expander or 2 but that project is still on the "to do" list :) I like the idea of making a more protected device with an OLED screen, and I'd probably want to arrange for the Arduino to be able to turn the +12/-5/+5 voltage rails off when not testing to make inserting/removing the chips a bit safer. The problem with the Arduino inputs is that +12v on an input will likely kill the input, and the -5v although clamped by the input diodes on the Arduino will probably cause excessive current to flow and kill the input anyway. One way to protect the inputs would be to use a series resistor to limit the current and a 5v Zener to clamp the voltage to 5v, so that negative or over-voltage would not be a problem. This would require a lot of extra components, however, and maybe not worth the cost considering how cheap Arduinos are today.
Nice. Maybe it's time to pick up that project again 😃 Agreed that it should be voltage protected ideally. Even though Arduinos a cheap, you WILL be using it with known faulty DRAM, so it's just a matter of time. Also like not having any power until the test starts.
I think it is an evolution of the DRAMARDUINO which was pretty much open source. The firmware was pretty hit and miss. This seems more accurate but I still get false negatives / best of 3 or 5 runs on some chips. Would be good to know what the firmware is actually doing lol. Great vid.
The seller sends the parcel tracked and insured because bad buyers lie and say the package did not arrive and eBay forces a full refund. I don't ship untracked any more.
Oh right, absolutely. For something about about that cost that makes a lot of sense.
Thank you for your support @Nuts n Proud. You are absolutely right, also postage to USA cost £15.15 plus taxes from Ebay too.
Most people are honest, but I have been ripped off by a guy in Finland once too 😢😢😢
I've been thinking of creating a similar project. I'd like to see support for testing at various speed-ratings.
Yes, that would be fantastic. You pop in the DRAM chip, you tell it it's a 4164 and it reports if it works or not AND the maximum speed at which it works. That'd be sweet!
If PWM is used for the clock signal, then the speed can be easily be set arbitrarily from 0 to 20 (or whatever the duino supports) MHz, with one or two commands.
@@NoelsRetroLab Hi Noel really interesting video on the tester. I was rather hoping you would mention testing other types of chips, forgive me as i am not technically literate but would it be possible to have testers that can test other chips like VIC-II, c64 logic chips, CIA chips, PLA chips SID chips as well as other chips from zx spectrum and Amstrad ect.. basically all 8 bit system platforms individual chips? Is this feasible or even possible to do? I know we can swap chips with known good chips, but i have always wondered why there are not chip testers beyond DRAM chips ect? Would really like to hear your thoughts on why in any future videos regarding chip testers?
I looked into it a bit and when you're trying to validate something works at 60ns timings (which even some of these old parts do!) you need to be able to sample in the 10s to 100s of MHz. Not impossible but definitely trickier -- I'm not sure if it's even possible with a low end STM to get that kind of timing precision on GPIOs. I was looking into building a tester with a cheap FPGA but it's a back burner project...
@@josephlunderville3195 yeah, I came to the same conclusion. I've thought it might be a good job for a pocket beagle, but that starts getting into the same price as a cheap FPGA.
First of all, thanks for this Video. I like that you have explained how to do checks yourself and showing code before showing a "prebuilt" device.
I think you are right on voltage protection, I also thought about that. Arduinos cannot handle 12V on the I/Os. But then I thought protection will expand the circuit massively and make the whole thing more expensive. An Arduino is a cheap device and it may be cheaper to buy a new one. Personally, I don't check that many RAMs that it would justify the higher pricing.
Glad you enjoyed it. Yes, you're right. That's another way to see the whole protection thing: cost of replacement x chance it would be damaged. I'm working on a review of a chip tester taking a completely different approach, so it should be interesting. Cheers.
@@NoelsRetroLab Looking forward to seeing it ;)
I think the commercial design can be improved in order to only use 1 zif socket. And also turn off the red led when necessary, as it bothers a little.
I'll have to see how it's getting -5v, 5v, and 12v out of 9v DC. I still have a lurking paranoia about the 19VAC-DC converter board on my old Ti99.
It looks like it's using some kind of voltage inverter AND a 7905 or something like that.
The 7660 is the voltage source of for the +12 and -5V regulators. -9V is from its output pin and the +18V is via the attached voltage doubler comprising of the 2 diodes and capacitors. The 7660 isn’t really needed because the equivalent can be generated by an AVR output pin and a few discretes.
I don't know if the kit uses Ardnuino code, but if so, guess it could be much faster with direct C programming of the registers. digitalWrite had a significant overhead the last time I used it.
Yes, I learned about that recently. I'm not sure if that matters given the speed limitations of the DRAM, but I haven't measured anything. Could be interesting to find out.
Replace those digitalReads and digitalWrites with direct port access and it will go a LOT quicker.
Yes, that's something I only learned relatively recently (but that tester code is pretty old so I never changed it). I see it makes a huge difference in speed.
i read somewhere once for higher than 5v voltages to go into an arduino or something there is a way to split it into a few parts and feed it into the aurdino and not damage it. for example 3x 4v parts. but i dont know anything beyond that i just remember reading it when someone was talking about higher voltages
Interesting! That's a good starting point to investigate that further. Thanks.
@@NoelsRetroLab no problem at all. im not clued up on this sort of thing but what i remember is someone was trying to process some sort of data and the voltage was too high so this was kind of a solution to still process it. i think it was data logging something
I was thinking that to protect against overvoltage, you could maybe use a 5.6V zener diode? Maybe add a LED for visual indication.
Not sure about negative voltage though. 🤔
Power jacks that use negative at center pind is not odd. Back in time is was normal to use center pind as negative. At some point in the '80 it change to be positive on centerpin. It might be due to the awareness of EMC.
16:30 add a slightly cut off foot in the middle under the pushbuttons, adapt it.
Good idea. I should try that.
@@NoelsRetroLab A similar trick can be used to get closer to the corner with the sticky outy bits in the way.
A quick continuity test from vcc pin to the other pins via code prior to voltage being applied would be a good preliminary test. This would also solve the hot insertion issue. Of course it would need a design change with relays or similar. #rabbithole
That IC, 7660S, is quite interesting for voltage conversion!
The 7660 has been around for over 30 years. It’s just a capacitive charge pump which was later inbuilt into MAX232 serial transmitters. The S variant has a boost pin to select a faster operation frequency. One can also use a 555 with a few diodes and capacitors on its output as a multiplier to accomplish a similar result. Just remember that its output power is very limited.
7660/555 also very handy for generating LCD display module voltages.
If one is clever, one can just use an AVR output pin as the charge pump oscillator!
A wire with a clip to assert reset might help it work in-circuit.
9:55 Power supplies in the 80s and before were commonly center negative. 2000s and later are almost always center positive. 90s is a mix. There are exceptions with some old ones being CP and some newer ones being CN, but for the most part that guidance should usually be true.
Indeed! I knew I wasn't imagining it. As a kid in the '80s, the majority of my power supplies were center-negative. Center-positive was quite uncommon at the time. It was only when I read the comments of a TH-cam video last year that I realized that many people today regard a center-positive supply as virtually standard and don't really think to check polarity. That came as a shock to me.
Putting a DRAM chip that is so sensible in a powered socket? And I clearly remember that some chips (4116) have a special order to apply the necessary voltages on power-on (-5V to be applied first and powered down last and never to be positive when 12V is applied).
I know of a Eastern European country Spectrum clone that was literally destroying his RAM due to the faulty power-on sequence in the power supply. It was upgraded to 64k chips that have 5V only power supply.
Technically, you are correct 100%, but I designed my own tester PCBs using the same code with that the kit seen in the video. During prototyping, I've included small 5V relays controlled by the Arduino to switch the power lines on and off between chip changes, but later on I totally omitted them in the final design because I wasn't able to damage even 1 chip out of tens of them with hot swapping. Somehow they take it very well.
You could create one that tests in-circuit by having it plug in to the 6502/Z80 socket. Would need to understand the machine a little for things like banking but that's not hard to do.
A few months ago Mark also did a review about this tester, so I bought one. Shipping from the UK since brexit is no fun at all.. it takes ages, still waiting. Hopefully it arrive soon. I hope to find a device once to test Amiga rams.
You could also add a checksum and or parity for the memory test.
honestly i think this would be better as a shield that you plug ontop of an Arduino Uno/Mega (or even a RaspPi) instead of such a large/bulky PCB
obviously only if you could fit everything of course. (like couldn't you switch between 4116 and 4164 voltages using some MOSFETs so you can save one socket? downside you would have to generate 12V and -5V from the single 5V input, plus side you can use USB to power it, making it much more convenient to use)
also i would personally never ever ever solder on an IC or Module like that, i would always use sockets in case i screw up the soldering or the PCB is wrong in some way so that those components can be easily removed without tedious desoldering.
Arduino Nanos are so cheap, that it's nice to have them dedicated to this instead of having to pull out the Uno, hook it up to the PCB, and star testing. Totally agreed about the transistors to route the right voltages and keeping it down to a single socket though! And while you have that, you can also turn off voltages completely before the test starts, so there are no problems inserting the IC.
Real men write self-relocating assembly-code on the native-computer to identify the bad bit/chip-LOL! 🤣
Did this on an Osborne-1 to identify a bad 4116; the only component to ever fail (other than the switching power supply).
Great video, glad to see people working/playing and enjoying the old school hardware! 👍
Oh but I did! Self-relocating, self-modifying and all 😃 github.com/llopis/amstrad-diagnostics Only drawback is that it only works on one platform.
@@NoelsRetroLab Good Man, thanks for sharing the code. Nothing like the precision, elegance and predictability you can get with assembly. Did a lot of development on a Mostek Z-80 assembler that had a 56-pin card-cage with 4-cards making up the computer-part; the assembler was in 2708 eproms; this was hooked to an ASR-33 teletype with a paper-punch to store the programs. Those were the days!
Royal Mail shipping from the UK, in 4 days with tracking and a signature, is £10. Even with packaging £18 is steep.
nice video. to be honest though, imho, they've spent a lot of work designing the pcb etc, and as they are including the cost of the arduino nano in there as well i'm not surprised at the price, well priced imho. the cost of shipping is probably more the UK royal mail shipping to the EU now we divorced you :D -- that said... the other issue i have is with yuou saying it should be open source..... that's fine and in a way i agree with you, but on the other side, as i say, he has to pay to produce the pcb etc, and you can gaurantee that if he had open sourced it, then within a couple weeks it would already be on banggood and ali express within a month lol
Thank you @Daytona FPV and yeah postage in UK is only £1.99 1st class post ;-)
_You can gaurantee that if he had open sourced it, then within a couple weeks it would already be on banggood and ali express within a month lol_
As if _not_ open sourcing it makes any difference, it just raises the bar to cloning slightly. If there is a perception of money to be made, it'll get cloned.
Thanks for the video =)
Great review. Thank you I am pretty sure the creator of the project has hundreds of hours into the project. Shipping is really expensive in the EU. Being in the US I cant even buy the kit.
Hi Noel, one way to shorten the through hole protrusions is to clamp the pcb and run a standard steel file flat over the board till the protrusions are flush withj the solder blobs after which you can make it really smooth (no burrs at all!) with a piece of scotch bright.
This can be handy if one is not planning to use a case for a pcb, for whatever reason.
And I may be able to help you with getting the display running, I have been experimenting with ST7735s and ST7789 driven displays in python on an RPi 4B, even wrote my own FX routines. It is still a work in progress.
Kudos for another fine video and interesting topic, this is definitely one I will be building when my attic is finished :P
Would a collaboration to make our own arduino ram tester with a display and short protection, be an idea?
Wouldn't that cause all sorts of stress on the solder joints? People already yell at me sometimes for cutting the leads after soldering. If I did that some people would have a heart attack 😃
@@NoelsRetroLab I have done a few times, I think people are way too careful with this stuff. If soldered properly it has only been heated a few seconds and is more than capable of handling a few strokes with a file. Afterwards, reflowing the joints is an option. Besides, the calculated cost of the pcb is 4 bucks, so what's the problem LOL. Of course I wouldn't do it on a retro board. Would you be interested in some technical information about those SPI displays?
Thanks
Great vid. Missed the red cloth though... 😋
Haha, it comes and goes depending on laundry rotation. It's gotten quite a bit of holes, so I think there's only one red left and it's about to go away 😃
As far as I know, all Arduino projects under Arduino license needs to be open source
I built one a month or so ago with an Arduino and breadboard (perf board). I found it passed bad chips. I modified the code (although I ain't worth a hoot in C++). It now found some weirdly bad chips. Between now and then, I had bought a DIY Chris ram tester. Fast forward to last week. I was having problems. The Chris tester showed good ram. I stopped and finally built the Retro Chip Tester Pro that I have had for a year. It found that RAM chip BAD. I found my breadboard arduino monstrosity and it found the ram bad.
A few weeks prior to this, I had a RAM chip that failed my tester and passwed Chris's tester. I sent him a message without a response.
IK opened up the Chris tester and about the only difference is that he doesn't use resistors between the arduino and the 4146 RAM pins. I don't know if it is the code he has loaded in his or the lack of resistors?
Maybe a few diodes would help with protection, that red light being on when testing looks annoying.
You could just get the MiniPro TL866CS for $80, it will test any ram chip and logic chip plus program about any chip.
Is that different from the regular TL866? Because mine doesn't do DRAM chips, at least not 4164 and 4116.
I have TL866II+ and it doesn't check these kind of DRAMs
These are only basic tests if you are not checking write or read timings. So a good chip could easily still be bad. The easiest way is just put a zif socket in the computer and run a memory test.
Right. These are very basic tests, but will still catch 90% (or more) of the failing chips out there. It would be great to extend the tests to include timings though!
Better buy MiniPro tl866 for that money. Despite being a pretty good programmer, it can test all kinds of RAMs and 74xx/4xxx logic.
I don't think the TL866 will check 4116 ram.
Interesting. I knew the TL866 tested many non-EPROM chips, but I didn't realize it also tested RAM. Let me check... Hmm... in the list of supported chips I have locally it doesn't like 4164 or 4116. Are they listed by a different number?
@@NoelsRetroLab My bad, I missed the point that 4116 is Dynamic ram, but minipro is only able to test static ram.
I had a bunch of 3564 and 4364 and they tested perfectly fine as 6264, so this made me thinking that 4116 will do the same, until I realized that this is dram.
I wonder if something like SCR lockup could happen with the power always being applied to the ZIP socket like that. It might be possible for the -5V to make contact first or perhaps a bit after the other rails have connected. Could be an idea to have some MOSFETs on the power pins to the DRAM that supply power only when under test?
Yes, that's a good point. It would seem safer to only apply power when the test starts. And if when that's solved, then it makes it easier to route power to whatever pins and you can test multiple chips in a single socket. But obviously it complicates the design.
Love the T-shirt
How about timing? Does it test timing requirements. I don't know how many speed grades are there for these old DRAM chips, but it is an important factor in a computer.
No, not at all. It would be great to make a tester that tells you the real response time though. Sounds like a fun project!
I'm in Canada and the shipping appears to be the same
The problem with this kit is that one can by 4 sets DRAMs(x8pcs) in that price and replace them all! I just bought TESLA DRAM for 6,5 euros+4euros shipping from the Czech Republic and without customs taxes.
Those can be bad too. If even one in ten is bad, you will be glad for the ability to test them before you install them.
@@mal2ksc I don't believe that I won't be able to find 8 working chips out of 32 chips bought at the same price tag of that tester.....
After all high rated sellers on eBay have their own testers.......
Hi! Can this project test the 3764 - 200ms? Thank you!!
Oh, so I'm supposed to put my hand over my work when trimming thicker components? That will be good news for the cat.
I’m not at the stage of testing any ram yet, I’m very much at the beginning of my journey with electronics. However, May I ask if you could program an ATmega328p rather than use a full Arduino Uni or Nano?
Your cat will love all the bits flying around! 😃 Yes, I'm sure you can use a microcontroller directly, it's just that the Arduino makes it easier to get going, but it's the same idea. You'll probably have more control over the timings skipping the Arduino if you're going to go that route.
Hard to believe such a basic project isn't OSHW.
I am looking for same kind of PLCC Dram chip tester on board. Suggestion let me know.
Great video as always! Did you try what happens to the LEDs if you reset the Arduino while the tester is powered?
Thanks! I think both LEDs turn on at startup, whether you reset it or power the board, but I'm not 100% sure. The weird bit is that the red one turns off half way through a test. We might be able to get Stephen (the designer) to pitch in and explain what the meaning of the LEDs is.
@@NoelsRetroLab Yep, red light goes out after phase 1 out of 4. (no reason, my bad programming ;-) )
I wonder if using one of the cheap ARM modules like the 120MHz Adafruit ItsyBitsy M4 Express or one with an even higher clock speed would also allow RAM speed compliance testing.
I doubt we need anything that fast, but it would be easy to run through the calculations. It could be cool if they would report the true speed of the DRAM though (15ns vs 12ns or whatever).
Good idea, it can be done with a cheaper 72MHz BluePull Stm32 for sure.
Lol. The intro ❤️ sad face :(
Here's how to solve the shipping cost problem: get 4 like-minded enthusiasts to pool an order with. These kits are light and I'll bet you could ship 5 of them for about the same cost as one. That way, you split the shipping 5 ways and pay only 20% of the cost! :-)
Great, if you're fortunate enough to know 4 local like-minded enthusiasts. :)
True. This is a bit of niche item that might be hard getting that many people into it though.
Any voltage beyond the range -0.5 V .. +5.5 V will damage the ports of Arduino...
Right. That's what I was afraid of. Someone else mentioned that the real damage is a current over 1mA, so that could be prevented with some resistances. Something worth considering for future versions I suppose.
@@NoelsRetroLab AVR can sink/source about 20mA/pin and 40 mA total but as other mentioned - power sequence, intermittent voltages, ESD...
Only the memory in the computer I'm using now makes me sad. Chips from micros don't give me any trouble (though I don't have any
Hehe 😃 Which IC pin cheat tool? The ones taped to the side of my workbench? Yes, those come in really handy since I still can't memorize the pin layout for the Z80 or even for measly 4164 DRAMs! (it's kind of random crazy).
@@NoelsRetroLab 7:08 That one! The one that fixes up the pins so you can place an IC relatively easily. The PPP hasn't gotten back to me (probably because they don't exist) but I'm pretty sure that's cheating :P
Thanks for demonstration
So I ahave Amiga 2000 with green screen
And all traces are good
I can't test Kickstarter unfortunately or buy one
So I will test ram
But the ram in rev 6 (256×4)
44c256
So can I use same idea with Arduino mega or I just dreaming
Please give me your idea
Sorry for my poor English
Do you not need current limit resistors on the LEDs?
can you update link - DRAM Tester Kit
Do any of these ebay DRAM tester projects check if the refresh cycles actually refresh or do they simply write and read back straight away?
That's the thing: We don't know because the code isn't open source! It would be a really cool thing to add that to a tester and then report the true RAM speed.
Where can I buy this DRAM Tester Kit? The link above does not work anymore... Please help.
It would be nice for a bit more coding and layout to support other DRAM and SRAM chips. A bit more involved but would be very helpful for the retro community. Nice project nonetheless.
Right. The problem is not so much the extra coding, it's bringing the right voltages to the right Vcc pins. That's why this particular one has two sockets. But I agree, having one ultimate DRAM testing open source project would be great.
@@NoelsRetroLab the I/O pins can be ran through level shifters for dealing with other voltages and a few reed relays to make it work with a bunch more chips and layouts. it would be more involved but the possibilities are there and even if it cost a few times more it would be worth the effort and the investment.
Yes, definitely. I'll review other testers that do that, so it should be interesting to compare.
@@NoelsRetroLab You should do a vid with the TL866 minipro to see how well it deals with SRAMS.
Other possibilities with the pins for the nano. 1) Put the pins in the other way up - short end through the mother board. It doesn't matter of the pins are long on top of the nano. 2) Put them in the nano first, then not all the way through the main board by supporting the nano with some little thing (or even by moving the plastic down).
4 diodes (bridge )after 9v jack and it doesnt matter if center is 9v or outer....
👍🍺
Wow awesome. Your channel has a unique ability to make fairly random videos that always happen to line up with my issues at the time. Thanks so much for the content!
This video didn't work on my TV. It did work on my PC though.
You probably have a failing DRAM chip in your TV. Better check them 😃 Just kidding. TH-cam can be fickle at times. Glad you managed to watch it.
Ok.... now do a 2101a static ram :)
I'm no expert, but other products manage to deal with all kinds of chips in the same socket, so to be limited to two types with no protection against wrong insertion, makes it less than ideal. I also don't think the sockets should be live while the user inserts the chip. It would probably be best to also not have to solder the Arduino to the board, but supply some female headers instead.
That's true, but to be fair, routing Vcc and ground safely to different pins isn't trivial. But I'll review other testers that do that, so we'll see how they compare. Agreed as well about not great about the power being live all the time on the socket.
I would even say just use a normal cpu with the bootloader and the program instead of a full arduino would work better and if a chip fails or get killed, that is socketed, it's no big deal, just burn another one.
This project is for people that know how to do it and you could make the board even smaller.
buy a minipro and get a eprom flasher and logic component tester as well
These chips ain’t logic chips. The minipro can also test the PLA, some guy made a video about it on TH-cam, he evenmade a file available to load in your minipro software to do so yourself.
Self hacking, can you download the arduino software, and figure out how to add the oLED screen? (not to steal the software, but for personal use)
Is it using all the data pins for the tester?
Seems so flexible. Julian Ilett is doing similar with his cobbled together remote temp sensing, creating boards first for his controls, then for a nRF radio board, now combining them all into a single project board using one of the small qty PCBoard makers, and even having them put the SMD parts in place. I could easily see this done with this project.
BTW, also, what's up with the hoarding of dead parts? Or now that you are able to test, will you recycle or trash them? I showed someone your 'collection' and told them to stop bugging me for mine... ;)
The real value of that tester is the PCB, not the software, and since adding the OLED screen would require modifying the PCB... I think I'm just better off designing one from scratch. It's not that hard. But I hear people are making an open source project, so I'm hopeful about that.
As for the dead parts... they come in handy for stuff like this, or when I want to test how a computer with a faulty chip would behave. They don't take much space, so I'm fine keeping them 😃
Why don't you design a PCB based off your breadboard? You'd have your own open source DRAM tester that way
Mostly because I'm shaky in analog electronics and power distribution. Someone else can probably do a much better job at addressing those things than I can. And it sounds there are some good open source projects out there already.
6:00 Are you sure that the viewer who bought it wasn't the eBay seller?
Yes, I"m sure 😃 I've known Leo for a while. Besides, Stephen, the seller, already replied here with comments. So unless it's some kind of long-term con, it's safe to say they're two different people.
@@NoelsRetroLab yep definitely :-)
It's a bit expensive. Great vid though.
Thank you! I think other testers are going to be more expensive, but they'll also offer more features. We'll see how things turn out at the end.
It's look okey but this tester will not test timmings. FPGA tester can check timmings etc.
Nice video! Do you make the chipmusic yourself? I like it :)
Thanks! Oh man, I wish I could make music like that. No, that's all the genius of McKlain (see links in description). He makes some wicked chiptunes!
Or you end up with TL866-II all in one device...
Ok, my bad, it doesn't support DRAMs
Yeah, someone else mentioned and I checked, but it doesn't seem to support them (even though it supports tons of other chips). That would have been great. I wonder why they don't do DRAMs.
I bet you can build a much better one considering your programming skills. Please do consider
Mmm, I absolutely don't mind it not being open source, simply because people don't work for free all the time. But I would expect a more polished product indeed and at least replying to questions. Communication is then key.
Nice video, buy a 3d printer to print a case
I'm so dying to get a 3D printer!!! But so far I don't have the space. Soon though. Soon 😃