Page 20 of 28

Re: A new generation of PLA for the C64 is now available.

Posted: Fri May 19, 2017 11:37 am
by janilaa
eslapion wrote: Sat May 06, 2017 1:01 am Just a word of warning concerning a PLA sub selling on eBay that also uses the XC9536XL: Me and Fredric have examined this product and we were curious as to why it would carry the 5ns version of that chip - PLAnkton carries the 10ns version.

Considering the speed of a C64 and that most other PLA substitutes have problems because they are actually too fast, there is absolutely no logical reason to use a faster chip that also turns out to be considerably more expensive - PLAnkton uses the 10ns version of the XC9536XL, it is the slowest and cheapest member of that family of chips and only it has the necessary characteristics to faithfully replicate the slow N82S100 on which the Commodore PLA was based.

Upon closer inspection, we have determined that the reason why the 5ns version was chosen is because it is in fact Chinese made counterfeits. Chinese clones of the XC9536XL are all marked as 5ns but they have all sorts of different response speeds. Also, on these counterfeit ICs the slew rate limiter as well as the low power mode don't operate properly. Of course, they are considerably cheaper than all genuine Xilinx XC9536XL, no matter what their rated speed may be.
Hi,

I'm not happy about interfering to a thread where you you are trying to sell your product, but I think I have right to defend my product.

You are just guessing why I have used 5ns chip. The real reason is that by using 5ns chip I have a lot more possibilities adjusting the timing. If you would have measured the timing, you would have noticed that J-PLA doesn't have 5ns propagation delay, but instead about 25-30ns on every output, like the original 906114-01. Adjusting the delay can be made also on VHDL program, not just with an external RC-circuit.

I bought first CPLD's from China, but that doesn't mean they are counterfeits. I believe that they are genuine, but perhaps marked bad on the factory, sold to a dealer, who tested those again and found some to be working. Or maybe something else... impossible to know. Also possibly as good and genuine as the one's bought from authorized dealer. Anyway, I have tested those to have exactly same characterictics as the the one's bought from authorized dealer. Also the propagation delay is 5ns on every IC (as marked on top). All features, including slew rate limiter and low voltage modes work exactly as genuine, that's why I believe these to be genuine. I have also run very long test-routines on each J-PLA before selling them. But just to avoid any possible problems I decided to buy CPLD's about month ago from Digikey as well. And yes, these are the same 5ns versions also. So, all the latest I have sold and every item in the future will have CPLD bought from authorized distributor.

Re: A new generation of PLA for the C64 is now available.

Posted: Fri May 19, 2017 11:40 am
by janilaa
e5frog wrote: Mon May 15, 2017 9:23 am In the next reply Jani himself (maker of J-PLA) confirmed that he has used counterfeit chips.
I did not confirm using counterfeit chips. I confirmed buying chips from China.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 7:04 am
by eslapion
janilaa wrote: Fri May 19, 2017 11:37 am Hi,

I'm not happy about interfering to a thread where you you are trying to sell your product, but I think I have right to defend my product.

You are just guessing why I have used 5ns chip. The real reason is that by using 5ns chip I have a lot more possibilities adjusting the timing. If you would have measured the timing, you would have noticed that J-PLA doesn't have 5ns propagation delay, but instead about 25-30ns on every output, like the original 906114-01. Adjusting the delay can be made also on VHDL program, not just with an external RC-circuit.

I bought first CPLD's from China, but that doesn't mean they are counterfeits. I believe that they are genuine, but perhaps marked bad on the factory, sold to a dealer, who tested those again and found some to be working. Or maybe something else... impossible to know. Also possibly as good and genuine as the one's bought from authorized dealer. Anyway, I have tested those to have exactly same characterictics as the the one's bought from authorized dealer. Also the propagation delay is 5ns on every IC (as marked on top). All features, including slew rate limiter and low voltage modes work exactly as genuine, that's why I believe these to be genuine. I have also run very long test-routines on each J-PLA before selling them. But just to avoid any possible problems I decided to buy CPLD's about month ago from Digikey as well. And yes, these are the same 5ns versions also. So, all the latest I have sold and every item in the future will have CPLD bought from authorized distributor.
There are many details to answer here but the main giveaway sign that a XC9536XL (or XC9536) is a counterfeit circuit is the type of plastic used for the package, the second one is the type of ink (or etching, actually) used on the plastic package.

While genuine Xilinx ICs have a low contrast ink with the information etched on the surface of the packaging, fake chips have a high contrast ink with no etching at all. The surface of the plastic is also very smooth, slick and even shiny on fake ICs.

Finally, virtually all Chinese sourced XC9536XL have exactly the same date code on them, something completely nonsensical. Every time I ordered a batch of ICs from Digikey, they had a different and fairly recent manufacturing date code. The last one, 1701, indicated the chips were made in the 1st week of 2017.

As for the technical details you provide, I am perfectly aware of the slew rate limiting capability built into the XC95xxXL family of circuits as we use it on PLAnkton. We elected to use the 10ns version because the datasheet specifically indicate some of the slower response speeds are only available on the 10ns version of this IC.

See: http://www.xilinx.com/support/documenta ... /ds058.pdf
On page 5 see Slew Rate Limited Delay and, most importantly, Internal Low Power Logic Delay (PLAnkton operates on the low power mode of the XC9536XL)

Also, the figures you give of 25-30ns latency on every output are false.

Both myself and Thomas 'Skoe' Giesel have carefully measured the response speeds of various generations of MOS 906114-01. Commdore 1st used the N82S100 then later created it's own versions of the PLA and called them the 7700 and the 8700. I had instruments which were more accurate than those used by Skoe (500MS/s vs 100MS/s).

The N82S100 had a latency of 38ns on all outputs but later ICs were considerably faster with only the CASRAM output being delayed. The faster 8700 had a response speed of about 16ns on all outputs except the CASRAM which had a latency of 24ns - this is consistent with HMOS technology speed with an added buffer on the CASRAM line.

PLAnkton can have a CASRAM latency of 24 or 40ns and it can be set by a jumper line on the underside of the module. Achieving this requires the usage of an RC filter. If you had implemented a delay of 25-30ns on all outputs then 8 different RC filters would have needed to be installed on your product and it carries no such things. Doing something like that serves no purpose as only a CASRAM delay improves compatibility and using a faster 5ns version of the CPLD with the intent of achieving a latency of 25-30ns is purely nonsensical.

Using a slew rate limiter on all outputs only prevents ringing on the output lines.

Because of the way it is configured, PLAnkton's signals are strictly indistinguishable from those of a MOS 906114-01 type 8700 on both a 500MS/s logic analyser and a 100MHz oscilloscope. yet it performs the same job with 1/5th the power of the genuine chip.
janilaa wrote: Fri May 19, 2017 11:40 am I did not confirm using counterfeit chips. I confirmed buying chips from China.
There are no authorized distributors for Xilinx in China.

If you ever manage to get something other than the 5ns version from a Chinese source then maybe, just maybe it isn't a fake IC.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 10:23 am
by janilaa
eslapion wrote: Sat May 20, 2017 7:04 am There are many details to answer here but the main giveaway sign that a XC9536XL (or XC9536) is a counterfeit circuit is the type of plastic used for the package, the second one is the type of ink (or etching, actually) used on the plastic package.

While genuine Xilinx ICs have a low contrast ink with the information etched on the surface of the packaging, fake chips have a high contrast ink with no etching at all. The surface of the plastic is also very smooth, slick and even shiny on fake ICs.
The one's I bought from China have etching, so this proves them genuine?
Only first few of my PLA's had CPLD's bought from China. I have used CPLD's from Digikey for a while and will use these in future,
so I think we can stop this junk about counterfeits.
As for the technical details you provide, I am perfectly aware of the slew rate limiting capability built into the XC95xxXL family of circuits as we use it on PLAnkton. We elected to use the 10ns version because the datasheet specifically indicate some of the slower response speeds are only available on the 10ns version of this IC.
Yes, the slew rate limiting goes proportionally with the IC's response speed. 5ns version has 3ns additionsl delay, and 10ns has 4.5ns additional delay.
What was your point? What were you trying to prove?
Also, the figures you give of 25-30ns latency on every output are false.
No they are not. Have you measured it?
We are speaking of latency from Input to Ouput, just to be sure.

In my current version (1.10), seven of the outputs are about 25ns.
In this version I added about 5ns more delay to #CASRAM, for added compatibility with old boards.
I did test the first version (1.00), which had same ~25ns delay on all outputs, including #CASRAM to work
pefectly with old boards too (which originally use slower 82S100). But there might be
a board out there requiring larger delay, so I added about 5ns delay to #CASRAM.
That's what I can do with the 5ns version! Can you do the same with yours?

Just in case some board needs even more delay I added also RC-delay (like the one found on newer C64 boards) option on 1.10
version, but that's not enabled by default. Most of the boards in use are newer boards anyway,
so this 25-30ns version is more suitable (and closer to original) for them.
PLAnkton can have a CASRAM latency of 24 or 40ns and it can be set by a jumper line on the underside of the module. Achieving this requires the usage of an RC filter. If you had implemented a delay of 25-30ns on all outputs then 8 different RC filters would have needed to be installed on your product and it carries no such things. Doing something like that serves no purpose as only a CASRAM delay improves compatibility and using a faster 5ns version of the CPLD with the intent of achieving a latency of 25-30ns is purely nonsensical.
Here you prove that PLAnkton does not have correct delay on other outputs than #CASRAM!
I did measure PLAnkton having about 10ns delay on all other outputs than #CASRAM.
PLAnkton is advertised having all outputs within 2ns from the original. Why in my measurements
all except #CASRAM have about 10ns delay?
Did I receive mangled version or are all the PLAnktons like this?

Just believe, using RC-delay is not needed. Things can be done inside CPLD, even if you didn't do
it like that. OK, I'm using about all the macrocells to achieve it, so delay cannot be raised more, but
hey, it's not needed!
There are no authorized distributors for Xilinx in China.
If you ever manage to get something other than the 5ns version from a Chinese source then maybe, just maybe it isn't a fake IC.
Again a false information. There are authorized Xilinx distributors in China, Avnet and Comtech.
I would have linked distributor list from Xilinx's site, but not sure if it's allowed on this forum.
You can check it there.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 1:32 pm
by eslapion
janilaa wrote: Sat May 20, 2017 10:23 am
Also, the figures you give of 25-30ns latency on every output are false.
No they are not. Have you measured it?
We are speaking of latency from Input to Ouput, just to be sure.
I haven't measured IT, I have measured THEM since there are at least 6 different versions of the PLA which were provided with various Commodore 64 and 64c. Every single one of them had a different response speed.
In my current version (1.10), seven of the outputs are about 25ns.
In this version I added about 5ns more delay to #CASRAM, for added compatibility with old boards.
I did test the first version (1.00), which had same ~25ns delay on all outputs, including #CASRAM to work
pefectly with old boards too (which originally use slower 82S100). But there might be
a board out there requiring larger delay, so I added about 5ns delay to #CASRAM.
That's what I can do with the 5ns version! Can you do the same with yours?

Just in case some board needs even more delay I added also RC-delay (like the one found on newer C64 boards) option on 1.10
version, but that's not enabled by default. Most of the boards in use are newer boards anyway,
so this 25-30ns version is more suitable (and closer to original) for them.
PLAnkton can have a CASRAM latency of 24 or 40ns and it can be set by a jumper line on the underside of the module. Achieving this requires the usage of an RC filter. If you had implemented a delay of 25-30ns on all outputs then 8 different RC filters would have needed to be installed on your product and it carries no such things. Doing something like that serves no purpose as only a CASRAM delay improves compatibility and using a faster 5ns version of the CPLD with the intent of achieving a latency of 25-30ns is purely nonsensical.
Here you prove that PLAnkton does not have correct delay on other outputs than #CASRAM!
I did measure PLAnkton having about 10ns delay on all other outputs than #CASRAM.
PLAnkton is advertised having all outputs within 2ns from the original. Why in my measurements
all except #CASRAM have about 10ns delay?
Did I receive mangled version or are all the PLAnktons like this?
PLAnkton offers EXACTLY the same response speed as the MOS 906114-01 type 8700 rev. 2

What sampling resolution and logic threshold level did you use to get 10ns ? I usually get 14-16ns which is the same as the 8700 rev.2.

You answer as if all C64 PLA had exactly the same response speed which is absolutely not the case. We elected to replicate one which works on all C64 AND makes it easiest for us to achieve the desired result with this type of CPLD.

Only one C64 PLA had a response speed of 25-30ns on all outputs (measured 28ns really) and it is the Synertek PLS100N. The N82S100 was slower, all 7700 and 8700 PLAs were faster but this caused no compatibility problems.
Just believe, using RC-delay is not needed. Things can be done inside CPLD, even if you didn't do
it like that. OK, I'm using about all the macrocells to achieve it, so delay cannot be raised more, but
hey, it's not needed!
Using macrocells involves using some sort of clock and will provide a response in time steps which are likely asynchronous to the C64 instead of a fixed delay.

Just like genuine Commdore PLA, PLAnkton uses strictly combinatorial logic to have static response speeds instead of macrocells which are clock based.
There are no authorized distributors for Xilinx in China.
If you ever manage to get something other than the 5ns version from a Chinese source then maybe, just maybe it isn't a fake IC.
Again a false information. There are authorized Xilinx distributors in China, Avnet and Comtech.
I would have linked distributor list from Xilinx's site, but not sure if it's allowed on this forum.
You can check it there.
I will check on that but I am quite sure AVNET is U.S. based. They may have offices in China.

Added edit:
AVNET has sales representatives in China and Hong Kong but their actual headquarters are in Arizona USA and primary sales locations are Phoeniz - Arizona, Diegem - Belgium, Poing - Germany.

Added edit:
After checking on eBay, all the XC9536XL ICs I could find on eBay which were rated for 10ns were either PLCC (discontinued) or much more expensive than the price on Digikey. All the reasonably priced XC9536XL with VQFP package on eBay were rated for 5ns and at a much lower price than Digikey - they were priced lower than the 10ns equivalent which makes absolutely no sense.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 2:25 pm
by janilaa
eslapion wrote: Sat May 20, 2017 1:32 pm I haven't measured IT, I have measured THEM since there are at least 6 different versions of the PLA which were provided with various Commodore 64 and 64c.
Ok, I may have misunderstood. I thought you were refering to J-PLA, not real C64 PLA's.
What sampling resolution and logic threshold level did you use to get 10ns ? I usually get 14-16ns which is the same as 8700 rev.2.

You answer as if all C64 PLA had exactly the same response speed which is absolutely not the case. We elected to replicate one which works on all C64 AND makes it easiest for us to achieve the desired result with this type of CPLD.

Only one C64 PLA had a response speed of 25-30ns on all outputs (measured 28ns) really and it is the Synertek PLS100N. The N82S100 was slower, all 7700 and 8700 PLAs were faster but this caused no compatibility problems.
I admit that I don't have measuring equipment as accurate as you and my measurements might have some error. I'm using logic analyzer with 200MS/s sampling rate. Low level threshold is 0.8V and High 1.6V.

I am aware of different timings on different C64 PLA's.

Right now I have 906114-01's dated between years 86-88 on my table, which I assume to be 8700 by the date and they all show about 25ns propagation delay constantly on all outputs, except #CASRAM which has about 30ns when measuring with my equipment.
PLAnkton however shows about 10ns on all other outputs, and about 35ns on #CASRAM.
Small variations of 5ns to either side sometimes because of resolution.

It is possible, that the 906114-01's I have are not 8700, even if they should be by date, or my measurements are are wrong somehow, but
this is the case. When comparing late 906114-01 outputs to PLAnkton, I get about 15ns difference on outputs other than #CASRAM.
On my J-PLA the results on the "other outputs" show mostly no difference in timing when compared to late 906114-01's.
Using macrocells involves using some sort of clock and will provide a response in time steps which are likely asynchronous to the C64 instead of a fixed delay.
Clock is not needed.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 2:35 pm
by eslapion
janilaa wrote: Sat May 20, 2017 2:25 pm Right now I have 906114-01's dated between years 86-88 on my table, which I assume to be 8700 by the date and they all show about 25ns propagation delay constantly on all outputs, except #CASRAM which has about 30ns when measuring with my equipment.
The PLS100N is usually dated 86-88 and was mostly distributed in Europe.
Using macrocells involves using some sort of clock and will provide a response in time steps which are likely asynchronous to the C64 instead of a fixed delay.
Clock is not needed.
Then I suggest you check the exact amount of power used by your product.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 2:42 pm
by janilaa
eslapion wrote: Sat May 20, 2017 1:32 pm After checking on eBay, all the XC9536XL ICs I could find on eBay which were rated for 10ns were either PLCC (discontinued) or much more expensive than the price on Digikey. All the reasonably priced XC9536XL with VQFP package on eBay were rated for 5ns and at a much lower price than Digikey - they were priced lower than the 10ns equivalent which makes absolutely no sense.
Yes, the 5ns versions are bit more expensive, but the price difference doesn't matter much.
I originally bought 5ns versions and made the logic for them, so I am willing to pay the 0.50$ extra
on chip so that I don't have to change VHDL and make all the tests again. Most of the
price of the final product is work anyway, material prices are very low.

It is also possible that the chips I bought from China were this first time genuine and good, but
if I would have ordered them more, nobody knows what would I have received.

Originally I ordered chips from China, because I made few PLA's for my self for my own C64's.
Then when I got some extra pieces and everything seemed to be fine, chips seemed to be working
correctly etc. I decided to sell extra pieces on the ebay.
After these were sold, I thought why not make some more as I have plenty of PCB's already and
decided to order CPLD's from Digikey to avoid any problems.

Don't worry, I dont have time to make these at the rate you have been selling PLAnkton. Isn't little
competition just healthy?

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 2:48 pm
by eslapion
Look at item 191704575881 here: http://www.ebay.ca/itm/10-PCS-NEW-XC953 ... 1704575881

At 12.77$US for a pack of 10, this is about half the price you would pay for the same chip (5ns) on Digikey and they do not offer a price break on larger quantities. On the photo is the 10ns version so this is obviously not the product you're going to receive if you place an order.

The 5ns version of this chip is supposed to be 32% more expensive than the 10ns version. The only reason you end up paying less for this faster version is because this Chinese seller is selling counterfeit circuits.

I elected to replicate the characteristics of the 8700 rev.2 because it was the profile which provided the easiest approach to making a faithful PLA with the cheapest genuine XC9536XL I could find - it turns out using the less expensive chip provides the best results.

I knew the vast majority of XC9536XL and older XC9536 ICs sold on eBay are fake because I have seen in the past people use the non-XL purchased on eBay to make PLAs for themselves or for resale and could notice the differences.
Here you prove that PLAnkton does not have correct delay on other outputs than #CASRAM!
You had another type of PLA in your hands and you replicated it's profile which is somewhat different but it does the job just as well. This being said, you can't say mine is wrong and yours is good.

I am fairly certain the CPLD on the J-PLA unit we got is a counterfeit one but I never said your product doesn't do what it is supposed to do; perform the job of a C64 PLA reliably. It does.

Re: A new generation of PLA for the C64 is now available.

Posted: Sat May 20, 2017 3:31 pm
by janilaa
eslapion wrote: Sat May 20, 2017 2:48 pm You had another type of PLA in your hands and you replicated it's profile which is somewhat different but it does the job just as well. This being said, you can't say mine is wrong and yours is good.

I am fairly certain the CPLD on the unit we got is a counterfeit one but I never said your product doesn't do what it is supposed to do; perform the job of a C64 PLA reliably. It does.
I didn't either say that U17 plankton wasn't good. You have put a lot of work on it and it has proven to be excellent replacement. And a lot of C64's have been saved with it, which is great!