Follow-up to fake or real chips? I think I know the answer now.

Follow-up to my previous videos about possible fake chips from AliExpress. I think I found the answer to the question on what was up with these chips!
Part 1: • Basement Rant: Did I b...
Part 3: • My first PCB! A Tandy ...
Part 4: • Viewer mail: All-in-on...
Video on the jr-IDE installation in my IBM PCjr
• IBM PCjr Part 4: It's ...
Jonard Tools EX-2 Chip Extractor:
www.jonard.com...
On the chips I bought:
Alliance AS6C4008
-55PCN 1635
I246219J0152G

[@vant4888]

Adrian, you have oscilloscope on the bench - record SRAM access timings on both machines. It is obvious that china chips are used and remarked ones and that PCj must have slower memory access than Tandy.

[@hakemon]

That's precisely what I was thinking. I was just talking to myself saying "I bet the RAM is too slow."

[@williammentink]

@@hakemon I've heard of that from China. Take a slower part, grind off the old silk screen, then put a new silk screen on for a faster and more expensive part. Not necessarily a fake SRAM, but one with a different speed rating.

[@Alexis_du_60]

I remember one of my DDR2 RAM sticks would start failing (data corruption) if I set the timings at a certain value (5-5-5), however if I changed them to 6-6-6 the errors would somehow go away.

[@Okurka.]

@Mack I bet they are 150 ns ICs, they run at 6.66 Mhz.

[@MrKeebs]

This is a great suggestion, please give this a try Adrian! :)

[@PhilXavierSierraJones]

Ah, black-topped chip. They grind off the top, apply a black dye to make it look like it's intact and apply a new etch or new print. Very common for fake chips. They might be chips from lesser-known manufacturers.

[@colindevaux4476]

had chips like this.under the scope you could see the old print and different mfg date

[@TheMadmagik]

Chips have been top sprayes and the legs ground and tinned . Which is probably why one fell off Common practice over there. They are not always fake but they are not new they are pulls from scrap units cleaned up and rebadged. Sometimes work but sometimes dont , they dont even bother testing them!

[@azzajohnson2123]

Yep measuring the thickness of the chip accurately would be the give away.

[@chrisw1462]

Yep.. they do that so the chips that leave the factory look good, and the count is right. Then they end up at, for lack of a better term, a chip chop shop that mods the nomenclature.

[@samuellourenco1050]

Why would they grind of the top? To hide the original manufacturer and part number? Doesn't make sense.

[@PeterCCamilleri]

The Tandy runs at about 7.2MHz while the PCjr runs at a slower 4.77MHz. The speed is probably the difference,

[@15743_Hertz]

Yup, you got floor-sweepings. The paint on the chip is called black-topping. Paint is used to cover up the original markings. The substandard chips probably failed one of the many quality tests that the manufacturer does to their product with the offending chip tossed in a bin for disposal. Whether the substandard chips were sold, picked out of the garbage or whatever doesn't matter. I went through this with some hard to find power transistors. After about 5 different batches, I gave up and settled for a quality local source close equivalent and ended up swapping the base and collector to get my transmitter to work. I wasted many weeks because of unscrupulous vendors.

[@15743_Hertz]

@@adriansdigitalbasement Maybe a bad batch from a subcontractor? The possibilities while finite, are still a large subset.

[@GGigabiteM]

@@adriansdigitalbasement Another common scam is relabeling lower end parts as higher end parts. This is a big problem in the vintage computing community for high demand parts like fast CPUs and FPUs. Sometimes scammers get away with it because the lower tier part will run at the higher speed, but may have a higher heat output and draw excessive power. But this doesn't always past muster, scammers will relabel different parts with lesser features as higher end parts, like a 68LC040 to a 68040 (the LC part lacking an FPU.)

[@Okurka.]

@@adriansdigitalbasement They are 150ns SRAM ICs.

[@vylbird8014]

I had the problem with diodes. Turned into a project to build a low-cost diode tester that can measure Vfd-vs-I. It's basically a curve tracer, except without the graphical display.

[@olegkostoglotov8800]

You would think that they would use flat black enamel paint rather then something that is akin to cheap ink , but I guess paint cost's too much?

[@bennyturbo]

Haha, the excitment when the computer booted with your homemade card... That's what I crave! That's the beauty of this hobby! well done :) Subbed!

[@SianaGearz]

The surface on the pins and one of the pins just breaking off indicates that the China chips are not new chips, they are salvaged from discarded old electronics, and the top has been cosmetically resurfaced and the pins were resurfaced with paste and hot air rather than chemically like at the factory. They might be of a different spec or even from a different manufacturer or might have degraded in-use. I wonder what is underneath if you go for some more powerful solvents, maybe acetone. Probably nothing particularly interesting really, just a sanded epoxy surface. Sometimes ordering from China is appropriate, but often you'll get stuff like this.

[@PauloSilva-ll4vs]

I agree with you, it is a salvaged chips.

[@SianaGearz]

​@@HankScorpio64 Aliexpress is responsible for reimbursing the customer, not the seller. Then they can take care of enforcement with respect to the seller domestically, they are after all a mainland company. It shouldn't be your problem whether the seller is a fly-by-night operation. Yeah sometimes you need a chip, and sometimes a used chip is the only one you can get, so you don't have much choice, and the local sellers all just resell the China stuff with no extra QC but with markup, so it makes sense to just reach closer to the source. Other times you need something cheap, quality be damned, and other times you need a Chinese domestic part and know the Chinese seller and trust them, that happens too. But thing is Alliance Semiconductor specialises on ongoing production of obsolete type memory ICs for the slow industries that need to maintain 30 year old products, you get their fresh chips via normal suppliers.

[@MagesGuild]

Precisely.

[@jaymartinmobile]

Back in the original PC and AT days there were often issues with the speed of the RAM chips that would cause weird bus contentions and errors like this. I often had people purchase 150ns RAM from us for a "turbo XT" and then would complain when it wouldn't boot when set at 10Mhz back in the day. I would wonder if maybe the fake chips are actually much slower but pin compatible RAMs that were removed from older computer boards. Maybe these are like 120 or 150ns RAM not the 55ns your true Alliance RAMs are. This would also make them work in a slower PC but just a few Mhz can make them not reliable. One way to prove this might be to spray freeze-spray on the suspicious chip and boot to see if the error goes away until the chip heats back up. Usually when one of these chips is super cold they will work much faster but as soon as they heat up they degrade to their native speed, whatever that was before it was painted over.

[@cheater00]

that might be true. try putting them in a freezer in a ziploc for 15 minutes, and putting them in the pc right away.

[@thomasguilder9288]

Or use freezer spray while booting the computer

[@lrblanton]

*

[@misterhat5823]

Remarked chips of a lower speed grade seems quite likely.

[@RS-ls7mm]

Fake RAM chips are not new. I remember running into the same problem 40 years ago when RAM chips were crazy expensive and some vendors tried to sell remarked chips.

[@PuyoPuyoMan]

I actually wrote a technical report in college about these counterfeit chips, lemme tell you after doing the research for that you'll never catch me buying anything from China. As for these it's common to take rejected or lower spec chips, sand off the markings and sell them as fully functioning or higher spec. The soldered looking leads could also be a sign those chips were taken from some other device that's been through who knows what and then etched with a newer date code. Well I could go on, but well in short always be really careful about your supply chain when buying components, because counterfeits are bad.

[@inferior2884]

Therefore, I have a bad 74LS173...

[@AureliusR]

I'd love to read your report -- was it published anywhere?

[@PuyoPuyoMan]

@@AureliusR Nah, it was just a report for a technical writing class. The research is pretty interesting, but trust me, it wasn't very good lol I can point you to a couple of the sources I used though, they're still fairly interesting: "Counterfeit Integrated Circuits: A Rising Threat in the Global Semiconductor Supply Chain" and "Response to counterfeit integrated circuit components in the supply chain" (parts 1 and 2) are both what I used for starting points.

[@thedungeondelver]

2 days 2 AB videos? You spoil us, sir!

[@tHaH4x0r]

A dead give away that they are probably recycled chips is that they have tinning on their legs. They desolder the parts, sand off the top, repaint it and mark them with the same batch codes. However, removing tin from the legs is too much effort so instead they re-tin all the legs. Not a lot of actual manufacturers do this with brand new chips. Usually these chips aren't 'bad' persé, but they can be marketed as a higher speed grade that they actually are, and of course can be damaged by their relatively rough treatment in removal from the board and heat cycling.

[@bradkamrath]

This was a ridiculously entertaining video, Adrian. Great job, man!

[@williammanganaro9070]

I am in agreement with others here. It looks like the "Fake" chips have slower access times than whats marked. It works in the PC Jr because the access time in the fake chip is sufficient enough for the computer to work (PC Jr = Slower memory cycle times = gives the fake chip more time to output and input data). Good video !! Fun to watch !

[@Hqbwheicjebw]

I do a TON of old hardware restoration as a hobby (I am a late 80’s child) and I have never even thought of fake chips, heard of them but never came across them. Time to pull out the old scope and also getting inside these old pieces of silicone.

[@Alexis_du_60]

I'm tempted to say that it might be that the Tandies seems to be more aggressive on the memory timings, thereby stressing the chip to the point of making it fail, while the PCjr would be more lenient on timings.

[@JohnJones-oy3md]

I would love to see that memory decapped. "The die don't lie."

[@MoraFermi]

It might be a "remanufactured" chip. China is notorious for pulling chips out of old scrapped hardware and reselling them as "new".

[@jaybrooks1098]

Me too.. The only reason to remove a old logo and number is to trade it for a more lucrative model. It’s probably a cheap knockoff thats been rebranded. Crudely.

[@jaybrooks1098]

And de lidding might give you the silicon build number.

[@Patchuchan]

@@MoraFermi They also would sell parts as spec higher then they originally were.

[@FennecTECH]

Mora Fermi i mean. should be easy if IPA is eating it away. Just toss it in a bowl or cup of it

[@KaelumYodi]

There could be any number of differences that are causing what you are seeing. The 2 most common would be 1) the speed of the chips is actually slower than the real chips, 2) the impedance of certain pins may be lower than the real chips and thus overloading the bus. The reason that the 74245 buffers are used is to ensure that the bus isn’t overloaded, which is absolutely required when stacking boards. It wouldn’t take very much time to add a 74245 to your board, and ensure that it is only actively enabled when the memory is being addressed. Also, USE A GROUNDING STRAP! You should be grounded (through a 1M resistor) whenever handling any IC or circuit board. You have no idea how much static damage that you are doing to those chips. You’ll never feel a small static charge under 20,000v, but it will blow/burn holes into the silicon. Environments where the relative humidity is below 10% or above 90%, are the most dangerous. Just learn the basics for anti-static protections, if your going to be doing this stuff. This is why I’ll never buy components from an electronics store.

[@zincmann]

The old adage applies here "You get what you pay for" And "If it's too good to be true, it probably is"

[@olegkostoglotov8800]

Except when it comes to Chicom export businesses, you may get nothing of what you pay for.

[@timrb]

Reclaimed chips from scrap, reprinted. May be same brand or different. See if you can scrub them with acetone.

[@krisbleakley9455]

I also agree these are salvaged chips, they're most likely not 55ns as marked. Another example is eproms. I know for a fact that ST discontinued eproms in 2011/12, if you got eproms with a date code after that be weary. A 100ns eprom, may in fact be a 120ns or 150ns chip remarked, same applies for sram and flash memory. This is not new it's been going on for years. It's not limited to memory either, I see YM3812 chips with the latest date codes and it probably hasn't been produced for 25 years.

[@mtbevins]

Thanks for the great video. I love the process.... It is fun when it does not work for you (as with me often) and watch you work through it. Troubleshooting can be frustrating but so worth it when you finally corner the problem.

Is it possible that our "Chinese friends" actually re-labelled a slower rated chip as a faster rated version? And the PCjr may work with slower RAMs as well, so it's the reason it's not a problem there. Just guessing here ...

[@alanhightower976]

The Jr also only performs patterning on 64K blocks at a time and the pattern is only a typical AA/55 sequence. So if there is any aliasing in the fake chips (eg if they are 128K for example), the less sophisticated memory tests will still pass.

[@vant4888]

Adrian can you use that oscilloscope and record memory access timings on both machines ?

[@vant4888]

Is it possible to temporarily slow down Tandy using signal generator instead of quartz ?

@@adriansdigitalbasement Not with SRAM, but I have similar experience with the Z80 CPUs (in fact, not just me, on some Hungarian Z80 computer forums many users describe the same problem): Chinese "fast" ones (let's say 8MHz and above) often does not work on any clock faster than 6MHz. My theory, that they re-label slower chips as faster. So they're not totally fake chips, they are Z80 (just not the speed grade they claim) ... But still they can earn more by asking for more money because it's a faster version. And they also lower the chance they will be caught on cheating, because many people actually does not push the parts even to their supposed limits, so they can see, "it's fine". With a totally fake (ie, even not a Z80 at all ...), it's still higher chance for sure!!, that customers will complain ... About the "slow 512K RAM" sounds strange, well, indeed, you're right. I have some 70ns one from Mouser, but I haven't even seen anything slower than that in that capacity. But who knows! Everything is possible as you said, maybe they can get some faulty chips from some factory which are sorted out on some automated tests, that they have problems but seems to work on a much slower access time what it should do. So then now they try to sell garbage ones, basically :-O

[@SimonZerafa]

@@adriansdigitalbasement What's the speed of the memory included with the Tandy, on the motherboard? How does this compare with the specs for the 512Kb chips, both the Chinese ones and the others? 😊

[@flyguille]

I know what is happening... I fall the same in a MSX mod. All your chips are ok, but different batch and maybe different factory, but not fake. The problem is, not at digital level, but at analog level of the digital signal, when the address bus has different loads.. A0 A1 A2 has more circuit load than A3 A4 ...etc., that happens because some adresses lines often has to go to more places than others addresses lines, by example, for all those I/Os chips which has 8 internal registers which the cpu must address, but the other addresses lines are not reaching all those chips, and those only go to a multi-purpose chip selector.... etc., that makes the ADDRESSES lines to be unbalanced with the load across all addresses lines. Everything is fine within the original load/noise ecosystem.... but then, you start adding mods, and place more loads. And there is when just a tiny more load makes a "more unsensitive" chip to not catch up the signal correctly. This kind of trouble if the load passes the threshold always, will make problems very consistently, not ramdon problem, because the source is too consistently which is different loads across the addresses lines.

[@laharl2k]

flyguille Sadly registered sram chips dont exist.

[@pixelflow]

Can we just convince Behringer (makers of clones of old CEM synth on a chip ic's put into their Roland/Korg synth clones) they should get into the retro-pc remake business :)

[@turpialito]

Probably the most underrated comment on this thread :-D

[@Cybernetic_Systems]

Nice work Adrian, those chips must not be the same spec as marked. The tinned legs is a classic sign they are chip pulls and re-furbed.

[@twocvbloke]

I'd guess at them being "Sinclair Grade", as in, something's wrong with them in part, but work up to a certain extent so can be used for some things but not others... :P

[@Everett-xe3eg]

Thats every chip manufacturer ever. Especially CPU die OEMS

[@trinidad17]

It's a possibility but what's more like it is that they are rebranded clones in order to improve sales. If they can clone MCU and CPUs what stopping them from making SRAM clones and some of their clones even have more features and run better. Said that, rebranding is obviously not a nice move from them. Adrian should check the LCSC website and try some compatible SRAM modules from there, they sell from Chinese manufacturers, but not rebranded and stuff that actually works fine.

[@TheTurnipKing]

Possibly clones that aren't quite up to spec?

[@AiOinc1]

@@Everett-xe3eg It actually used to be fairly rare. If a part was defective it was recycled or trashed, never resold. Nowadays, they sell them as lower model parts in many cases, which is why they have such high product yields now. Their production has improved, sure, but for the most part their defective chips can go on to become 100ns or 150ns parts instead of the rated ones. Also, Intel was one of the first companies to do this.

[@xxycom8963]

“Sinclair Grade” would not have worked in PC Jr. The Sinclair memories were the ones where the upper page of memory will not work.

[@Lightning666]

The chips cut with pliers from the board and then chinese solder legs on it again. Thats why legs are shiny and thick. They do it with MOSFETs too.

[@watermalone3841]

I Just Discovered Your Channel Last Night! I Watched the *"LEFT FOR DEAD"* C64 Video Series, and Became an Instant Fan of Your Work! You can Literally Feel the Love You Put into Your Craft as the Videos Play, and it's Incredibly Inspiring! Plus I get to Learn about so Much Unique Information and the Backstories Behind the Tech You're Working On! Who Needs a College Student Loan Debt when You Teach N00BS like Me how Shit Works and How NOT to do Home DIY, All for Free?! I'm 32 and Still Burn My Toast in the Morning! People Like Me *NEED* People Like You on God's Green Earth!! Much Love!

[@trinidad17]

Wear noise cancelling headphones so I block the hum in my office, then decide to watch this video and listen to someone else's background hum.

[@GameTechRefuge]

Hi Adrian. Was just watching LGR, discussing fake Yamaha chips. He recommended checking out a video from plgDavid here on you tube. Well worth checking both LGR and plgDavids videos on the subject. I'm frankly a little shocked at the process behind the reconditioning of these fake chips. All the best from Ireland.

[@adriansdigitalbasement]

Thanks yeah I just saw LGRs video on my feed (not watched yet) and I have plgDavid's video on my watch later list to go back to. Fascinating stuff honestly.

[@GameTechRefuge]

@@adriansdigitalbasement I'm afraid to go through my EPROM collection :(

[@HeyBirt]

They are likely much slower RAM chips that were remarked to make them more saleable. A friend of mine bought an entire roll of transistors form a Chinese broker as he was in a pinch. He looked at them under a microscope before building thousands of PCBs with them and thought they looked OK. He called the actual semiconductor manufacturer and found out the markings on the transistor were bogus. The seller has sanded down and remarked a whole roll of inexpensive transistors so of course they do this with chips they can get a few bucks out of too. Unless there is no option I'll buy from Mouser, Digikey, etc.

[@BertGrink]

As my handyman father used to say, it can be expensive to buy cheap; even though he was specifically referring to hand tools, the idea holds up in any area, methinks.

[@BertGrink]

@Lassi Kinnunen Well, that's true, but my father's point was that if you buy cheap tools, there's a greater risk that they may break, thus forcing you to buy a replacement tool; on the other hand if you buy quality from the beginning, you can have a tool that lasts a lifetime.

[@DerMartexus]

God damned! The new intro is sooo awesome! So that sounds like good news. :) Can you do some more rubbing on the Chip? Maybe some sort of original name appears.

[@DerMartexus]

@@adriansdigitalbasement I think so, too. That sucks so much.

[@AndyHullMcPenguin]

The suspect chip may be slower, but using the 74 series buffer may be sufficient to allow you to get away with using them (it will have the effect of cleaning up the signals). In other words, the fakes are probably fully functional but lower spec. chips, most likely pulled from used equipment (hence the nice shiny soldered legs). A quick rub down with 400 grit to remove the old markings, a brush with fat black sharpie pen (or more likely a Chinese sharpie clone), and a whizz under the laser etcher and a yingyang 100ns clone ram chip becomes a known brand 10ns chip + $$$ profit. The other trick used by some of these suppliers is to add in 10% of random scrap chips with the correct number of pins overprinted with a valid part number, to a bandolier or two of say 10000 genuine chips. This boosts your profit by 10%, assuming your customer doesn't spot them and puts the failure rate down to their own processing. That could turn each $1,000 profit in to $1,100 for next to no effort, and it cleans up your scrappage rate of already dodgy parts, turning them in to money. This is organised crime at work. Big money is being made with these ripoffs.

[@r.d.machinery3749]

I saw a video of a process used to pull chips off old circuit boards in China. They dumped the boards on top of a roaring fire, waited for the solder to melt and then banged the boards up and down on a slab of concrete to release the chips. :D

[@martinda7446]

I'm glad Vic Reeves is doing consumer computer vids these days.

[@incrediblySmart]

Every time. It reminds me of a time of just hanging out with a friend experimenting and figuring things out.

[@kneehighspy]

digging that intro 👍

[@kneehighspy]

tone167 miss the old 80’s computer chronicles.

[@melvinolson8381]

@Steven Hawkins please upload!!!

[@kneehighspy]

Steven Hawkins hate for you to go to all that trouble. gonna check out archive.org. i did watch a bunch of episodes here on yt.

[@kneehighspy]

Steven Hawkins your efforst are much appreciated!

[@mattparker9726]

7:18 Adrian, try actually USING the chip in a real-world RAM intensive operation. See what happens.

[@jrmcferren]

As I just noted in my top level comment, if the CPU clocks are different on the Tandys Vs the PCJr, these chips may operate reliably

[@CommodoreFan64]

@@jrmcferren that's what I was thinking as well.

[@Okurka.]

@@jrmcferren Reliably at a lower speed. I hope he gets the refund.

[@coreyclarke6929]

oh man do i feel your pain.... i have been dealing with a batch of fake eeprom for the last couple weeks, i was starting to loose my mind over it, out of 10, only 3 worked, and 4 of the bad ones actually went up in smoke during erasing! the chips i got were not quite as obvious as the ones you got, but there were some giveaways.

[@robertturner4913]

I stick the SRAMs like this in an old TOP853 programmer and run the SRAM test. It works pretty good for such a cheap, now discontinued, programmer. I would not install potentially fake chips in a machine like that - depending on what the chip really is, if its power pins are in the right place, there could be smoke and burned traces. I have smoked several fake HD6309E's from AliExpress - almost everyone I've bought was fake.

[@jackconvex9524]

This would drive me absolutely insane with frustration

[@nickrog6759]

Looks like the ALLIANSE Re-silked screened China chips are genuine BUT slower MHz models than the ones you need . Late to the Party as the solution to the Mystery has already been solved down below in the Excellent well behaved comments Below 😀. I had to ask a similar questions 20 times in 20 different ways before even coming close to a correct conclusion in the day . I am now Subscribed for the Whole package - Adrian's Facinating Content and the Helpful & Polite Commenters - that's a rare 1st for me on KZbin !

[@francoisleveille409]

There is this 'fine' gentleman on various Commodore forums selling VIC-20 and C16 RAM expansions and they ALL use Chinese SRAM chips he found on eBay. I told him not to do that he attacked me with a flurry of insults. My comments to him were 'moderated' on the VIC-20 Denial forum.

[@brendanrandle]

would be cool to see a jig made up to test the speed of these chips

[@andygozzo72]

yeah, i'd say they're not 'fakes' as such, they're purporting to be ram chips...and they are (although may not be of alliance manufacture, so in that way they'd be 'fakes'/counterfeit) ... but most likely slower types remarked as faster, like with a lot of the cheap '20mhz' z80s... maybe old used 'recovered' parts...

[@laharl2k]

Its called an arduino

[@Kaxlon]

I bought a bunch of SST EEPROMs from China a few years back. They are correct size, but not the same speed. Sometimes clearing works, sometimes not. I treat them as OTP now. And had zero complaints from users. Maybe you can check timings as suggested below?

[@gremfive4246]

So you proved what we already knew, the PCJr is superior to the Tandy. Excellent video....

[@brianv2871]

Said no-one ever. Hahaha

[@BlackEpyon]

No, the PC-Jr is inferior to the Tandy is most aspects, which is why the PC-Jr tanked shortly after release, but the Tandy 1000 line, which was initially based on the PC-Jr, took off. Tandy learned from IBM's mistakes, fixed them, and brought the price down to where ordinary people could afford it. I haven't looked at the decoding logic on the Jr-IDE board, but I'm familiar with one of the guys who designed it (he goes by "eeguru" on the vintage computer forums). The PC-Jr is completely wonky in how it managees it's RAM, but the Tandy 1000 is quite simple. The Tandy shoves all of the expansion RAM straight to the beginning of the address, reading the first 384KB, then the onboard RAM controlled by Big Blue, which includes the video mask. The PC-Jr has 64KB base RAM, then the video mask, then whatever expansion RAM is addressed after that (if my memory serves). You can plug a (working) AS6C4008 straight into the expansion bus on the Tandy using A19 as the chip enable, and it will work without any decoding logic, but we haven't determined yet on the forums if overlap might cause issues. The thread is here, if you want to look: www.vcfed.org/forum/showthread.php?67544-I-wish-to-create-a-new-DMA-RAM-expansion-card-for-the-Tandy-1000-line&p=576980#post576980

[@Psychlist1972]

I know Mouser is expensive for small numbers of parts, but I never buy ICs or other important things from Ali or Ebay. Especially when it comes to vintage parts, you're likely to get either fakes, or more often, rejects and pulls that are being sold as good/new. Those rejects and pulls may work on one computer and not another, depending upon tolerances. In your friend's shot showing Digikey and China, the giveaway is not just the bad printing, but if you look at the circular indent, you can see it has been coated just like the top. In the genuine part, the circular indent (die pin release, I believe), is clean and flat, not coated. There's a good site out there that covers these types of things to look for, but if I post a link it'll get marked as spam. The re-topping process does involve a black paint-like coating. Also, almost all ICs are made in China today. It's not an indictment of China, it's an indictment of non-reputable sellers who do not keep a verfiable supply chain. (NOte that Alliance is fabless -- they probably get their dies made in China as well) Real Q: Why order from random sources again after having been burnt before? Have you learned the lesson, now? :D Edit: Ok. At the end you answer that question :)

[@Psychlist1972]

BTW, that chip puller is a plastic, erm, Chinese copy of a metal part designed elsewhere :)

[@Arcadecomponentscom]

Common problem. Take slower chips and chips from different manufacturers, remove the top, and etch on new markings. If they are painted on, the paint rubs off easily. If they are lasered on they will use wrong fonts or logo versions that don't match the date codes on the chips. Legs will be retinned in a solder pot which is why you'll often see thicker solder on the wide tops of the legs and leg grunge where the leg bends and meets the body of the chip. If you buy them by the tubes you can often see the chips all have the same date codes, logos, etc, but will have different mold markings and body shapes where they came from different batches and even different manufacturers.

[@burnte]

They’re lowered spec’ed chips that have been remarked as better chips. That’s why they work in a less strenuous application but fail in tighter tolerance applications.

[@sx64man]

Adrian, I feel your pain, I''m pulling out my hair with some 2114s for the Vic20 that I got from China and 40 of the 80 are reporting failed in the ram expansion... same red foam BTW, we probably got them in the same place on AliExpress... something is very weird as they almost work... I suspect timing issues or failed tolerances on the China chips. Put in a few 2114s that came out of C64s and no issues... great video.

[@turpialito]

Did you check the pins on the motherboard slots? They may be dirty or bent out of contact, making the pin float. Get some light in them and use a magnifying glass, followed by a blast of compressed air and contact cleaner, perhaps?

[@mprebbz]

You are lucky enough to have digikey in your country - I honestly can’t see why you would take the risk of buying from random sellers in China. Interesting videos though. Your channel is added to my subscriber list

[@robertturner4913]

The fakes I've got - they use solvent to remove the original markings. Then they apply a black "paint". Then they remark the chip. For example, using this method, they might sell hitachi 120nS chips as Alliance 70nS chips. Since the SRAM pinout is a jedec standard they can sell that one chip as any of a number of different brands and speeds of sram, etc. It is an atrocious practice. Many are also removed from printed circuit boards which is why the pins are tinned with solder. The access timings are likely the reason for your problems.

[@Adanos_ger]

Even though this isn't a fresh video anymore: I tend to buy things from aliexpress that are more disposable. Like LEDs, Resistors, Transistors, e.g. If I need ICs I tend to buy them from known good sellers locally. You still save a lot of money, but are on the safe side basically.

[@jp-um2fr]

This should upset a few, sorry. I am English, old and almost housebound.50 years ago my hobby was valve amps. Rather than watch that dreadful TV I decided to take up electronics as a hobby again. Things have changed just a bit. 3 years on and many linear power supples getting more and more complex I now have quite a nice store of bits. It's all from China. I could never afford the real thing and I don't think much is made in the UK now. I underpower everything, fans, thermal cutouts, etc, etc all over the place. All I can say is what I have found, very few problems. Most of my transformers, heatsinks, etc. have come from old Monacor amps. 8A at 36V is not a toy. Mind you 245V right across the chest brought back memories. No fool like an old fool. In the early days I paid over £30 with postage for 4 x LM317HVT, my one and only order from the U.S. All I can say is thank you China, you have stopped me going mad.

[@buckykattnj]

Check It!... that's a trip down memory lane, no pun intended. Used to use that almost daily from roughly 1988 to 1993, IIRC. Up until I got a job at an OS/2 shop doing DB2 and telephony on brand new, high end PCs that didn't need hardware diagnosis so most of my time-tested old DOS go-to tools fell into disuse.

[@bobblum5973]

You mentioned that your card didn't have an LS245 (74LS245), do you mean not at all? If you don't have one, then the "fake" memory may be loading the bus more than the "real" ones, causing trouble when it gets accessed. Either that or just timing of the signals. As for the buffer, I've seen problems with different manufacturers of the same part, or using 74xx instead of 74LSxx ones. The easiest answer is what you found: avoid the cheapest price if quality or specifications might be in question (you often only get what you pay for). Thanks for the good video, keep up the good work!

[@superspedboy0076]

im assuming that the tolerances made with the china chips are wider, so they could have less exact QA. example if a resistor is 8k oms, the china could be 7,950 oms, instead of straight 8k. however PCjr could also have wider tolerances so IBM could work with more sub-par parts. Personally I would use an oscilloscope and map it from the good chips and the china chips. that would show how the chip is made and the difference between the 2 and point to you why thats happening. love your vids Adrian as always man!

[@Gerardus1970]

They are possibly genuine or from other manufacturers but certainly recovered chips, hot oiled to make the pins look new and smooth, sanded, painted and remarked (sometimes to reflect later/better specced/or more desirable batches). It's rampant. They sell them as new, but they are not.

[@piconano]

The reason maybe that the timing in the PCJr is not critical to this IC. You put it in a faster bus, and it acts up. See if cold spray makes it stop working in the PCJr.

[@battra92]

Definitely worth buying from good sources. The Capacitor plague and gutter oil should have taught us all a good lesson.

[@UpLateGeek]

If you look at the output, it's only the first nibble that's incorrect, the rest is correct. And it's consistent in how it's incorrect: A=E, B=F, 8=C, 9=D. Let's look at those values in binary: A=E: 1010=1110 B=F: 1011=1111 8=C: 1000=1100 9=D: 1001=1101 As you can see, it's the second most significant bit that's being flipped from a 0 to a 1. Butt! It's only affecting the first byte, why not the second? It's an 8-bit bus, so the surely second byte is just another read from the same data lines on the same chip, so we should be seeing the same problem in the second byte, right? My guess is that the second read is soon enough afterwards that whatever circuit controlling the output of that bit is charged enough to hold it for long enough to read valid the second time. And I guess it's working in the PCjr either because of the bus transceiver on the expansion card presenting a lower load, so the pin can stabilise in time, or the slower access time is enough to do the same. Either way, the chips are fake. But I'd still be interested to see a comparison on the oscilloscope of that pin to a working one. (Also, still haven't got around to testing that SRAM I got for my 486 boards yet. Probably doesn't help that I'm spending my time staring at datasheets for other people's chips!)

[@Jpetersson]

I wish I knew as much about electronics as Adrian. I know way more than the average person though...

I wonder if adding the transceiver IC to your board would get that suspicious memory chips working. It could be that those IC's have some kind of issues bringing the data bus to high impedance or that they cannot source/pull enough current... Joust guessing...

[@Mosfet510]

I had that happen once (that I know of lol) with a pair of LM338's. They were supposed to be 1.25-32v@5A. What they were doing over in counterfeit land was packaging LM317's inside the TO-3 case. It kept maxing out at 1.5A. The store gladly replaced them and found out they had been had as well.

[@danweecc]

It's SRAM so maybe they've relabelled a slower speed RAM as a faster one and the timing issues are causing access issues. Check the timing between the read logic edge and the data appearing on the bus, comparing between the good and bad chips.

[@garydunken7934]

You get those RAM errors due to timing problems. Use a scope or logic analyser to further dig in.

[@ArixZajicek]

LS245 might only be needed for DRAM and not SRAM? If it's a dedicated refresh circuit it'll have to refresh the RAM without it affecting the bus lines. Maybe they're like, genuine rejects that someone picked from the dumpsters and pawned off in Shenzhen? Crazy behavior but glad you have a sure solution now!!

[@jrmcferren]

The 74LS245 is only a bus transceiver with 3 state outputs. Basically these allow the switching of a device on and off the bus. It wouldn't matter if he was interfacing with SRAM, DRAM, ROM, or even a UART, the IC would be the same.

[@ArixZajicek]

@@jrmcferren Right, but the SRAM and DRAM chips are usually have tri-state outputs on their own. In theory you wouldn't need it for the DRAM if you wanted to refresh from the CPU, but if they have their own refresh circuitry, they'll need to be detached from the bus for that, hence the transceivers, but of course SRAM won't need refreshed either by the CPU or any support circuitry so there's no need for a transceiver between it and the main but. That said, it's just a thought, since my theory would require that the address lines be able to be detached from the address bus too :P

[@KonradZielinski]

The problem is that at least some of the time the local chips are just as questionable. I got a bunch of bad 74HC chips from a local retailer, went to another store in the same chain, got new chips plugged them into the same circuit and it worked.

[@hqqns]

Hi Adrian, I'm not sure if it's the recording but your 1084 looks like it could do with some focusing - easily doable by adjusting a pot on the flyback.

[@lyfandeth]

When those RAM chips were made, all of them were milspec tested for speed, and tagged by speed. Once the quota for each speed was filled, the rest of the production batch (which could be top grade, but that's money wasted on testing) was sold off as minimum speed, untested and cheap. Some wholesaler might have bulk ordered the untested chips, remarked them, and sold them. That's where extra paint or odd markings would come from. Or, they might be RFE, details unknown.

[@SeanGrimes]

I'm glad you got it sorted! From either the fake or lowbinned chips! Please please please provide the specs for you memory "+" board. My HX will be really happy!

[@Gerardus1970]

They are most probably not fake chips, they are used, recovered chips which are then remarked - often to a faster speed. The texture on the top is different because they sand the tops back before remarking the chips. You can tell because the two circles on top of the chip are a different depth to genuine. The China chips are shallower. They also put the chips in a hot oil bath to clean up the solder on the pins, smoothing it.

[@Codeaholic1]

You probably need a buffer between the memory and the bus

[@wishusknight3009]

Makes sense. The PCJr has a more sophisticated ram controller circuit.

[@Codeaholic1]

@smbakeryt had a similar problem on his storage/ram expansion card for his XT

[@timmooney7528]

The chip may be a close to spec cousin re-labelled or a QA reject that could benefit from a buffer.

[@Dukefazon]

6:30 - the fake chip was like "We've been compromised! Quick, bite the cyanide capsule! They'll never get me alive!"

[@1985stout]

6:48 To turn on that using the "suspicious" chip scares me.

[@hugovangalen]

Me too! I'd try the good one in something that doesn't work. I think Adrian loves living on the edge. :)

[@BigCar2]

I would guess the tolerances used or operating specifications for the Chinese chips are different from the Digikey chips. So they'll work for some uses, but not for others.

[@gusbert]

Definitely subtle differences in the SRAM access timings, perhaps made worse by the omission of the bus transceiver on your card. All it takes is a few tens of ns to make things go bad :-(

[@charlesjmouse]

Late to the party: FWLIW: I tend to agree with the sentiments in the comments. The Chinese chips are probably 'genuine' for a given value of 'genuine' but are likely: -Salvaged parts -Slower versions relabeled -Marginal / test-failed parts relabeled I've got to admit to being cheap and so tend to buy China-sourced IC's anyway, but take a very close look at the pictures and any feedback, and don't buy form sellers who don't show pictures of the actual items for sale. This doesn't fully protect me of course but as I mostly build in to my designs the option to add some margin I usually get away with it... the times when I don't it's usually because I'm dealing with a retro-system that was built on the cheap anyway, and sometimes where I've not followed the IC conventions used in more 'official' upgrades - this is the one that catches me out most often as I try to keep chip count as low as possible and I sometimes run in to "so why did they include that IC?", delete that, and then find out it was included to paper over the cracks in the OEM design. (I'm also in the habit of freely publishing my designs for others to use as they see fit as I very rarely make up boards for others / sale - too much hassle and as a result of issues like this combined with being a single person who doesn't have the time / resources to extensively test stuff. It's frustrating to get unhappy feedback from people complaining about hobbyist boards that have run afoul of edge cases, even though I only ever use known genuine parts in other people's boards) I hope that helps. I just stumbled across your channel, rather enjoying it. :-)

[@ropersonline]

You having installed the Jr-IDE into the Tecmar jrWave housing initially confused the heck out of me. I know you mentioned it briefly at 3:08, but it still confused me because I didn't immediately get that the old sidecar and modern innards were totally different products.

[@PauloSilva-ll4vs]

Good job man, your job seems to be a great one, working with these kind of jewel.

[@peteregan9750]

they are probably scrap production - may have a slight error that only occurs during ceartian circumstances

[@Okurka.]

Nah, they're just 150 ns chips marked as 55 ns.

[@Plan-C]

Iffy socket or connection that changes when you shove chips in and out maybe?

[@percivul1786]

What about QEMM? It's a really old mem management proggy I remember from the early 90's and whatnot. It could access and check himem as well as exmem (if I recall correctly). You might have to check out some less than reputable locations to find it, but I think it MAY do what you're asking for.

[@colonelbarker]

I think it's probably worth including an octal bus on your board if you're able.

[@Xaltar_]

Norton Utilities for MS-DOS, I vaguely remember using that to test high RAM in the 90s. Back before norton software basically turned into mallware :P

[@pow9606]

Yea. the word malware didn't exist in my vocab during the dos days. No internet, modems, routers etc. Maybe bnc networking. Zip files were a must for transfers then.

[@Xaltar_]

@@pow9606 Haha yeah pkunzip :)

[@crowbarviking3890]

When i asked my father about the Tandi (german engineer in the 80s), he always complained about them being flimsy. So i would not rule out that the Tandi is too.. picky.. when it comes to signal quality etc. Especially when compared to an IBM... btw. the black paint can come from a rework session. I would not bet on that these chips are really new. Most likely they are "refurbished", in chinese that means, borderline OK and added new paint.

[@MrCodyswanson]

Well done sir, thanks for keeping us updated!

[@steveg5122]

The 1000 may be more....sensitive to timings than the IBM PCJR. The chips as others have said are probably genuine just...old ones. it could be just slightly off-spec enough.

[@snooks5607]

Adrian, when getting new chips from Digikey try buying one that's slower, 150ns or something, and see if it reproduces the error the china chips had?

[@jeffburrell7648]

I agree with v ant, you probably received sub-par, rebranded chips from Ali Express. Years ago we had this problem in one our products - we could not get parts from our normal distributor and had to go to the gray market. The parts we received did not match their data sheet specs and it caused no end of trouble trying to figure out what the problem was. Obviously, we never went gray market again.

[@PaleozoicPCs]

That solder-shine on the pins of the AliExpress chip is a dead giveaway they're salvaged. The V-20 CPU I bought off eBay to upgrade my EX looks the same. I only paid $2.50 for it and it works so I haven't had any complaints, but I have wondered if the silkscreening that claims a 16mhz rating might be suspect. It's still really mysterious why even a used chip would behave like that if they're actually AS6C4008s; the 55ns rating is about four times faster than they need to be for the Tandy's memory cycle time. Did they make them in slower grades? They could be 628512s in disguise, but I think the slowest speed grade of those is 100ms, which should still be fast enough.

[@wanjockey]

Great, well thought out video. Very good information to have. Appreciate all of your videos.

[@FirebayRefurb]

I've learned the lesson hard way too, and only order from China if Digikey, Mouser or Farnell can't provide me chips 🙂

[@michaelgraff6978]

On your original board, you did not have bypass caps. Please try inserting them on your test board, and see what happens. I suspect this may help; the PC Jr board you installed the chips in have these bypass caps. Also, not having the bus transceiver could easily explain this. Timing matters, as does good power.

[@rickhunt3183]

I don't have a very clear view of your board, but I see no decoupling capacitors. The problem you're having could be due to fan out. The octal bus transceiver does double duty to isolate the bus and provide increased drive. Are your chips on the custom board in tristate mode when the cpu is not accessing it?, or just not enabled? I would look in the area of fan out for the problem. Just my thoughts from taking a fast peek at the board. I haven't been able to clearly examine your work so I could be totally wrong. Cpu fan out is 1.

[@MrRadioAct]

Two thoughts: -That black paint looks like laser marking spray (www.trotec-materials.com/news/laser-marking-products). I would bet they are using a cheap CO2 (k50?) laser that can not engrave the epoxy package directly therefore the paint and blury print. -I had a run in with gray market components through work a few years back and after hand testing 1000 pcs, I came to the conclusion that what we had bought must have been fab rejects that were scooped from the bin and resold as genuine. Many of the parts simply didn't work, but interestingly many sort of "worked", just not to spec. The vendor didn't even blink when we sent back 600 pcs as defective, he had a ton more to replace them.

[@TR-mx1wk]

I'd like to suggest a "possible" reasoning that the new chips may not work within your constructed board, and the same reasoning that the old ones will. There is a term in manufacturing called the "Fan-out Capability" of each pin. Since you've not used the LS245 (Buss transceiver), in your circuit, you may be surpassing the capability of the Buss Fan-out. My suggestion, add the LS245 to your circuit and try again. ThanQ...

[@TR-mx1wk]

My reasoning why the old ones work and new ones don't in this case, is that "possibly" the new ones were manufactured with a "lessor" specification by China, than the old ones had for their "Fan-Out" capability.

[@dhpbear2]

8:00 - I noticed it's counting up SLOWER than before (6:55) since you swapped the 'upper' chip! I'm guessing the 'fake' parts are TOO slow for your design?

[@MechaFenris]

Now THAT is a head-scratchingly weird problem. I hate when those sorts of glitches come up. Troubleshooting nightmare. :)