Fake chips from China

[argonn]

While I was getting materials for my computer, I stumbled upon some 4MHz Rockwell 65C02 chips, allegedly produced in 2014. Hmm. The same seller had NMOS Rockwell 6522 chips, also produced in 2014. Do I smell a fish? Or maybe I am wrong?

I ordered some: "5pcs" of R65C02P4. Here is what I got (they shipped very fast):

As you can see, every package is slightly different. Some have those two shallow holes near the ends, some do not. I tested them. Three chips appear to draw less power than other two (those act just like my old MOS6502AD). I tested them for BRA instruction: NMOS 6502 treats it as NOP with operand. Two chips appear to be real 65C02, three are 6502. One of them is "low power" though, drawing same amount as 65C02s do.



This one was also able to run at 6.667MHz, all the others (including my trusty MOS6502AD) were able to reach 5.333MHz. Tried searching for "CMOS 6502 without BRA", but you can imagine what are the results

Let's take a closer look at the chips now:





Bottoms are scratched, notches at the edge all look different, there are signs of sanding on tops, some corners are rounded too. Leads are shiny, not mat, as it is the case of new WDC parts. Also: MEXICO on top, KOREA / PHILIPPINES on the bottom. Right. Are those real chips? Yes. But not new, and labels on top are bogus.

https://www.youtube.com/watch?v=j9nasajve1o

[BigEd]

Not really fake, then, but (probably) recovered from old equipment and not very carefully screened and labelled - the sellers not too concerned about the difference between 6502 and 65C02.

[Martin A]

Unfortunately I've seen similar before, with the Z80 in that instance:

http://primrosebank.net/computers/mtx/p ... _fakes.htm

Sanded and re-marked NMOS Z80's being passed off as the 20mhz CMOS part - presumably because of the price premium on the newer/faster chip.

[BigDumbDinosaur]

There are companies who have ability to positively determine if a device is the real McCoy or a fake without doing anything untoward to the device in question. I use two parts liquidators for obtaining NOS items, such as the 53CF94 SCSI controller I use in my POC units. These liquidators contract with a test lab to verify the authenticity of the items in question. Although I pay a little more for the parts, I've never been burned with a fake.

As soon as you go through sources on eBay or in China all bets are off. I'm inclined to agree with Ed that these parts are likely pulls that have been rescreened, as these are not parts with sufficient value to warrant the expense of creating a working counterfeit. As always, caveat emptor!

[argonn]

There are some local trusted sellers here too. One even has original AMD 80386 chips from 1993, at the original retail price (they seem to ignore inflation) Got most chips for my project from his shop. Another is also liquidator, he sells a lot of Z80-related stuff, and quite a lot of 1970s to 2000s NOS electronics (got SCC2692 chips from him, around 1 USD per piece). Mouser is way too expensive, I ordered FPGA breakout board once, shipping was around the same as the board cost, and I had to pay tax too.

On ebay I found only one guy selling real new WDC chips in EU. He also sells original W65C22 chips. I don't think any NMOS chips are being produced today, probably even military moved past this technology, albeit they are known for being conservative.

[BigDumbDinosaur]

argonn wrote:

...got SCC2692 chips from him, around 1 USD per piece...

Check with your source some time and see if they can get you either the 26C92 or 28L92. These two have FIFOs on both transmitters and receivers, Un the case of the 28L92, a sizable performance boost can be gotten by configuring it to use 16-deep FIFOs—the interrupt load drops dramatically. The 2692 has no transmitter FIFO on the transmitter, which means it will interrupt after each datum is transmitted.

The other thing the 28L92 offers is fast bus timing, which allows it to easily keep up with a 14 MHz system without wait-stating.

Quote:

Mouser is way too expensive, I ordered FPGA breakout board once, shipping was around the same as the board cost, and I had to pay tax too.

Mouser's prices are okay, but they tend to use couriers such as FedEx and UPS for international orders, who charge big bucks to ship from the USA to other countries.

Quote:

On ebay I found only one guy selling real new WDC chips in EU. He also sells original W65C22 chips.

How do you know they are genuine WDC parts?

Quote:

I don't think any NMOS chips are being produced today, probably even military moved past this technology, albeit they are known for being conservative.

Most countries' armed forces are not that far behind the technology curve. There is absolutely no good reason to use NMOS parts, unless refurbishing an old piece of hardware and wanting to keep it as original as possible.

[argonn]

BigDumbDinosaur wrote:

How do you know they are genuine WDC parts?

I bought some, W65C22 is consuming less power than NMOS 6522, outputs are stiffer too. This guy is selling PLCC versions of all chips, most China sellers have only DIPs. It might not be definitive clue, but leads are mat, it looks like tin was deposited electrolytically. Every attempt to solder makes the coating shiny. Putting chip in a socket makes some markings too: in places it was touching metal leads are shiny. Just like blueberry bloom: once you touch it it's gone.
The price of the parts seem to be about right too.

[BigDumbDinosaur]

argonn wrote:

I bought some, W65C22 is consuming less power than NMOS 6522, outputs are stiffer too.

Do you have the W65C22N or W65C22S?

[argonn]

Markings:
W65C22S6TPG-14
SA1309A
A4V338.1
Taiwan

[BigDumbDinosaur]

argonn wrote:

Markings:
W65C22S6TPG-14
SA1309A
A4V338.1
Taiwan

You have the W65C22S, which has no current limiting on outputs, and a totem-poll IRQ output.

[argonn]

Added some diodes to IRQ lines. I hope that 1N4148 will be good enough, if not, there are still some JFETs around

[BigDumbDinosaur]

argonn wrote:

Added some diodes to IRQ lines. I hope that 1N4148 will be good enough, if not, there are still some JFETs around

The 1N4148's forward drop is too high to reliably pull the MPU's IRQB input down to where an interrupt will be recognized. This is where a Schottky diode should be used, such as the attached.

Another method if you have multiple 65C22s is to wire each one's IRQ output to an input on a multi-input AND gate, and let the AND gate drive IRQB. One of the gate's inputs should be reserved for a wired-OR interrupt circuit to take care of devices with open-drain IRQ outputs. The below illustration, copied from Garth Wilson's 6502 website, shows how to do it.

Mixed IRQ Circuit

However, I would use a 74AC11 or 74AC21, not a 74HC11.

diode_schottky_npx_bat85.pdf

Small-Signal Schottky Diode

(126.05 KiB) Downloaded 103 times

[argonn]

Roger, ordered BAT85 diodes, but I'll consider some faster logic anyway. I don't have HC/AC11 in my collection, will order some anyway, along with other stuff.
Using some programmable logic could make identifying IRQ sources simpler (reading one register), not to mention guaranteed delays. Just a thing to consider for the future, and there is a pretext to.

[BigDumbDinosaur]

argonn wrote:

Roger, ordered BAT85 diodes, but I'll consider some faster logic anyway. I don't have HC/AC11 in my collection, will order some anyway, along with other stuff.

74AC logic is generally three to five times faster than 74HC, plus can source and sink more current. The only thing you have to be mindful of when designing with 74AC are the extremely fast output rise and fall times.

Quote:

Using some programmable logic could make identifying IRQ sources simpler (reading one register), not to mention guaranteed delays. Just a thing to consider for the future, and there is a pretext to.

As the design gets more complex it becomes more difficult to implement all the desired features with discrete logic and not hit a performance wall due to cascaded propagation delays. A CPLD gives you much more predictable performance, as well as the ability to handle complex logic without having to use a washtub full of gates.

[jazz]

Those of you who have been around long enough may remember the issue of fake UARTS
They worked just fine - some of the time but the manufacturing costs were cut by only
including half the required buffer (if I remember correctly) These fake chinese chips
caused all manner of problems because they appeared to be correct and worked "some of the time"

If you are suspicious at all I'd say don't use them in anything serious.