• Please review our updated Terms and Rules here

The curse of fake parts out of China.

I have done de-capping at work to ensure that ICs purchased through brokers have been what was stated on the plastic. And compared the silicon to known good devices (parts taken off a donor board that we were remanufacturing).

All were good apart from an Intel 8080 clock IC. The silicon was completely different with no manufacturer markings whatsoever.

However, we only use it as an oscillator, and not for the internal reset and ready synchronisation etc. As a result, it was pretty easy to 100% test the functionality of every device and everything was good - despite it not being a 'real' Intel device.

Dave
 
daver2 said:
I have done de-capping at work to ensure that ICs purchased through brokers have been what was stated on the plastic. And compared the silicon to known good devices (parts taken off a donor board that we were remanufacturing).

All were good apart from an Intel 8080 clock IC. The silicon was completely different with no manufacturer markings whatsoever.

However, we only use it as an oscillator, and not for the internal reset and ready synchronisation etc. As a result, it was pretty easy to 100% test the functionality of every device and everything was good - despite it not being a 'real' Intel device.

Dave
Click to expand...
I like the way you remanufacture boards that have vintage parts on them. You must also regard these things as very trustworthy and worth the effort, that is if they are the genuine parts to re-make the clone boards.
 
Yes, this is for work...

I generally go to a supplier that has the silicon wafers still from Intel. A number of 'vintage' devices I used were encapsulated from the silicon wafers.

I have genuine Intel 2716 EPROMs encapsulated in ceramic with 2016 date codes on them! All tested to the original Intel specifications...

We had a fault with one of the Intel 8274 dual serial devices. When I returned it, they tested it on the machine and it passed. Being a sceptic, I asked for the test results, which were forthcoming. Intel have their own internal test vectors for the silicon that exceeds anything that have ever been published in the data sheets or books!

Other devices they have in stock have traceable paperwork back to the original component manufacturer.

Dave
 
Had a weird issue in a project using the previously-mentioned aliexpress AS6C4008 static RAM chips where a nearby memory cell kept seeing a value change. I was not pushing the envelope on speed this time; it was a very slow retro project. I spent a while debugging my code to try to find every conceivably way I could possibly have stuffed a value into the wrong memory cell, but came up lacking. Eventually I swapped it for a mouser AS6C4008 and the problem went away. This is the second set of problems I've encountered with those SRAM chips. It takes the economy out of it when you spend an hour debugging something that turns out to be a bad IC.

Now I need to either toss all the damn things (I think I have over a hundred -- have some big projects planned), or build a tester to test them all.

Scott
 
smbaker said:
Had a weird issue in a project using the previously-mentioned aliexpress AS6C4008 static RAM chips where a nearby memory cell kept seeing a value change. I was not pushing the envelope on speed this time; it was a very slow retro project. I spent a while debugging my code to try to find every conceivably way I could possibly have stuffed a value into the wrong memory cell, but came up lacking. Eventually I swapped it for a mouser AS6C4008 and the problem went away. This is the second set of problems I've encountered with those SRAM chips. It takes the economy out of it when you spend an hour debugging something that turns out to be a bad IC.

Now I need to either toss all the damn things (I think I have over a hundred -- have some big projects planned), or build a tester to test them all.

Scott
Click to expand...

I recently had problems with UVeproms misbehaving in a Character Generator application.

I have figured out playing with Eproms now this is the task that shows up marginal defects, when they are asked to operate at a speed faster than they can (even if it is within spec).

I am going to build a test jig for them to drive them to their max limits speed wise, with a VFO time-base driving an address generator and find out when the read output data corruption occurs.

It is not sufficient to test them with a read in a programmer like the GQ-4x, the reads are too slow and can all look just fine, that is, until you put that ROM into more speed demanding circuits.

I suspect that a lot of recycled ROMs with varying forms of borderline internal damage will conk out at lower speeds than they should. Of course the really bad ones fail even on a "slow test"

Throw those IC's in the bin, before they land in the lap of some unfortunate punter.
 
daver2 said:
I have done de-capping at work to ensure that ICs purchased through brokers have been what was stated on the plastic. And compared the silicon to known good devices (parts taken off a donor board that we were remanufacturing).

All were good apart from an Intel 8080 clock IC. The silicon was completely different with no manufacturer markings whatsoever.

However, we only use it as an oscillator, and not for the internal reset and ready synchronisation etc. As a result, it was pretty easy to 100% test the functionality of every device and everything was good - despite it not being a 'real' Intel device.

Dave
Click to expand...
Any chance you have a picture of the fake Intel 8080 clock IC(s)?
I assume you are talking 8224 here?

Bill
 
Yes, it was a Siemens SAB8224P.

The Curve Tracing testing of the package pins gave identical results between a known good device and the samples:

1707212792671.png

A de-cap of a known good device indicates markings similar to:

1707212874230.png

This is a die photograph of the samples (note no text anywhere):

1707212954246.png

We use an 8 MHz crystal and only use the OSC and THETA_2_TTL outputs - so the devices are 100% testable in the product as it rolls off the production line. All of the parts worked when tested at the end of the production line and we have had no problems with any of the parts during burn-in and use.

The other (different) Siemens parts we procured (8289 and 8287) did have identifiable silicon die markings within the packages.

Dave
 
Last edited:
One other thing I have noticed, a subtle detail, but probably important , because the Devil is in the detail.

When a lot of IC's come directly from a manufacturer, and have never seen any use, or been fitted to a pcb or to a typical DIL socket, the angle of the pins, after the bend after they exit the IC body, are splayed too far apart to be fitted to a DIL socket, or pcb holes that are suited to a DIL IC.

They require bending or forming to get the right spacing between the pins.

Recently I received a shipment of NOS Fujitsu 2732A-20 Uveproms from Poland. Right away I noticed the original looking pins with the wide spacing (as the part came from the factory). Also the Gold plated die base and a very large die inside.

They are perfectly normal and test and program with no errors and run in a high speed character generator application.

So I have no doubt in my mind that they are the genuine vintage NOS Fujitsu article, not any kind of re-cycled part.
 
Wide pin angle is not necessarily an indication of a new part. Chinese sellers can easily duplicate that.
 
Yeah, I reckon if they're willing to laser-clean and re-label a part, it's not exactly a lot of extra work to have a pin straightening die that sets the pins to the right, "new" angle.
 
smbaker said:
Had a weird issue in a project using the previously-mentioned aliexpress AS6C4008 static RAM chips where a nearby memory cell kept seeing a value change. I was not pushing the envelope on speed this time; it was a very slow retro project. I spent a while debugging my code to try to find every conceivably way I could possibly have stuffed a value into the wrong memory cell, but came up lacking. Eventually I swapped it for a mouser AS6C4008 and the problem went away. This is the second set of problems I've encountered with those SRAM chips. It takes the economy out of it when you spend an hour debugging something that turns out to be a bad IC.

Now I need to either toss all the damn things (I think I have over a hundred -- have some big projects planned), or build a tester to test them all.

Scott
Click to expand...
Testing is key. Too many things in the retro community have to be sourced from places where you can't count on the authenticity of the product.

BTW, thanks for all the awesome designs you've shared over the years. I purchased one of your ETA-3400 boards a while back and look forward to building it.
 
I am reminded of this document which was presented at a NEPP Electronics Technology Workshop, after a 2009 story about NASA getting counterfeit parts. There is a lot of good information in here with photos regarding blacktopping / remarking, die examination, counterfeiting trends (up to 2010, when it was written), and worst offenders like IC Brokers.


If you REALLY want to get into the weeds, here's a couple hundred pages (just some light evening reading, right?) on counterfeit component avoidance:

 
cjs said:
Yeah, I reckon if they're willing to laser-clean and re-label a part, it's not exactly a lot of extra work to have a pin straightening die that sets the pins to the right, "new" angle.
Click to expand...
Though the appearance of the pin surface helps. Most of the fakes I have seen, the pins have been re-plated or have tooling marks or marks from previous socket insertions. When they are new, there is more often just a very thin uniform coat of Tin plating, devoid of any tooling marks, this feature, in conjunction with pins that have never been formed to fit a pcb or socket is a very good sign the chip has never seen any use.
 
Amardeep-AC9MF said:
Testing is key. Too many things in the retro community have to be sourced from places where you can't count on the authenticity of the product.

BTW, thanks for all the awesome designs you've shared over the years. I purchased one of your ETA-3400 boards a while back and look forward to building it.
Click to expand...
Thanks. That ETA-3400 board in particular can be really finnicky about what RAM is used on it (even when it is name brand). In the github repo I believe I have a list of part numbers that I've tried.

Scott
 
skate323k137 said:
This is covered a bit in the first link, fake test results / certification
Click to expand...

I once saw an interesting documentary about fake Helicopter parts. A company ordered a new main rotor from a trusted supplier, had all the conformance certificates and was a traceable part, yet, somehow along the way, it had been switched with what would be the genuine part (the old bait and switch). Everybody in the supply chain pleaded innocence and due diligence. The part was a used and vandalized component that had multiple holes drilled in it before it was sent to the dump. Somebody found it, filled those holes with car body filler, and painted it so beautifully it looked like the top of Elton John's personal piano.

No matter how things are "covered" in documentation and procedures, the crooks will still find a way to get them in the chain and fool everybody, so you need to be prepared for that.
 
You must log in or register to reply here.
Back
Top