Got it, and that Wikipedia chart looks super useful. I will pay attention to the image file size, and pick an EPROM that can handle that size, or larger with the same pin count package.

Using this info has now helped me identify why I'm having trouble figuring out in advance how to burn a U7 Bootflash EPROM to update the Sound OS on my LOTR before installing the LOTR LE 10.02 CPU code in preparation for adding a shaker. The file image I was intending to use to burn the U7 Bootflash EPROM is named BIOSv8.u8 and is 128KB in size. I got that file per this Pinside reference: https://pinside.com/pinball/forum/topic/how-to-install-shaker-motor-in-lotr-standard-edition/page/5#post-3080604. That must not be the correct file because a 128KB file would require a 1Mbit chip or larger, but a 1MB chip is not the correct pin count package (too big) to fit into the 28 pin socket at U7. So, I will start the search again for the correct image file for the U7 Bootflash EPROM, and when I find that then I can burn it to a 28 pin EPROM of the appropriate size (or larger including repeat copies to bulk up the file image).

Unless this is indeed the correct file, and there is a different reason why it shows at 128KB on my computer's hard drive?

How do the Checksums work for EPROMs? I just noticed that the checksum calculated by the GQ-4x4 for a given image file changes based on what Device you select. It varies for size of EPROM, and even more interesting for the same size EPROM the checksum differs between different EPROM manufacturers. For a particular file I have, when I select AMD AM27512 the GQ-4X4 software displays a checksum value of 00A68ED5, but when I select ST M27512 the GQ-4X4 software displays a checksum value 00FC9040. Why is that?

Hmm, it may be because I was erroneously looking at a file too large to fit on the target EPROM. Using another file, 64KB in size, I get exactly the same calculated checksum values when selecting either the AMD AM27512 EPROM as the Device, or the ST M27512 as the Device, which seems normal.

The checksum is usually the sum of all bytes you're programming.

If you're burning the same file to two EPROMs of the same size/bit orientation, the checksum will be the same.
So in your case the same file should show exactly the same checksum when burning an AMD AM27512 or a ST M27512 since these EPROMs are the same capacity and are equivalent.

That is good to know. I wish I had taken screen shots to show what I observed before. But looking again now using the same file as before I am indeed getting identical checksums from compatible EPROMS by different manufacturers. Not sure why I didn't see this when checking earlier, but I will chalk it up to "user error" of some kind. Thanks Quench!

After you change chip profile in GQ-4X reopen the image file.

If you open a 4k image while a 2k chip is selected and then select a 4k chip the software will truncate the last half of the image. So you must reopen the file again. I hope that makes sense.

That does make sense, and I did see the truncate message. I'll make it a habit to select the Device (chip) first, and then load the file.

Looking to do some ROM programming and going to pickup the GQ 4x4. I'm looking at EPROMs for Williams System 11 machines; they take the 27256 and 27128 chips. Looking on eBay there is different brands and models; mainly ST Micro and AMD seems to be the two major brands for this type. As long as it is 27256 or 27128, does the rest of the model number really matter? I'm seeing subtle differences in parts numbers past that point. Just want to make sure before I purchase. Thanks.

They could be different in terms of programming voltage etc. but for the most part you are ok. There is an eeprom version of the 256 that does not work IIRC.

What I do now is just buy winbond 27512 ee version and double/quadruple up the images.... way more convenient than using the UV lamp to erase. But I go through tons of burns so the burn-erase cycle is a PITA.

You could just get 256's and double them up as well, general rule is the older the eprom, the more variation in voltages, etc. with older ones being more stubborn.

Gotcha. So how does getting the 512’s and doubling or quadrupling save you from having to erase?

Do you see any issue with getting these?

ebay.com link: itm

ebay.com link: itm

The 256's aren't too badly priced. The 128's are overpriced.

The winbond 512's are electrically eraseable - the burner tells it 'erase first', then burn. It's more like an nvram that way. So no need to slap them in the UV eraser for 20-30 minutes after using (I write/test others software for these games so I need to erase/burn a lot of eproms).

I'll develop/test with the 512's, then when it's 'finalized', I'll burn to a 128/256 setup. I got 200 256's from a music instrument refurbisher a couple years ago, they were cheap, 200 for $110 or so. I had to scrape off a difficult to remove yamaha 'genuine' label on each of them and soak them in alcohol to get the glue residue off. My finger still twitch thinking of that.

Ok, so no issues with using an EEPROM whereas the originals were EPROM? I am going to be doing some software beta testing, so that may be very handy.

You might be thinking of the AT28C256 which has a slightly different pinout to 27256

SST27SF256 are a EEPROM version that looks pin compatible. Cheap out of China.

True.
Back in the day, I had a ROM-IT. The device was basically a computer controlled ram.
You downloaded the .bin file of the game to it and it plugged into the ROM socket.

So I should be Ok if I get the ones I linked above? In terms of that it’ll work in the machines (Bad Cats, Pin Bot). I checked the programmers list of supported chips and they are in there minus the last few digits.

Get the largest size EEPROM in the pin count package. It feels like a waste of money to stock 2764, 27128, 27256 and 27512 when a W27C512 can replace all of those and is probably the cheapest. There might be some rare cases where you need to pad the data in a certain way to use a 27C512 as a 27256 (supposedly some R-Dog sys 11, but not sure as the stock boards work fine with larger mem), but normally you can fill up the chip with repeated copies.

SST27SF256

has good endurance, 1000 cycles 100 years. But ive seen some flash memory I end up not buying because they are only guaranteed to hold memory for ten years after a burn, which is probably a conservative estimate, but worth looking at the datasheet and considering.

Data retention spec for early EPROMs like 2716 was 10 years too - the spec was usually at maximum ratings.
I haven't seen many of these 40 year old games with EPROMs suffering bit rot though.

The concern these days is the current crop of cheap programmers not burning cells with enough margin that will result in reduced data retention (data loss in years rather than decades).

The below article is taken from "Intels 1984 Memory Components Handbook" and discusses the move to intelligent/fast EPROM programming and dealing with verifying cells at different voltages to ensure good margins. May be an interesting read for some.

EPROM_Intel_Versatile_Algorithm_Equipment_Cut_EPROM_Programming_Time.pdf

I think I have never seen a pinball EPROM suffering from bit rot even though some are 40 years old. But some that I did burn myself with a home made programmer in 1980's, do suffer from occasional bit flips. Maybe that tells more from my programmer than the chips

This is probably my first EPROM project from 1985. A terminal from Intel data book application note, for my first homemade micro. The 2716 EPROM's are still OK!

term (resized).jpg

It happened with my Black Hole. I turned it on after a few months of no use and was suddenly getting the 7641 ROM error on the displays. The factory game 2716 EPROM had some bad bits. This was probably about ten years ago.

Impressive!

Might be the Black Hole was using EPROM's a bit earlier than Williams games. I think the first EPROMs I saw were in Sys 6 games, before that they used mask ROMs. Anyway, the BH EPROMs might have been quite early production 2716's so no wonder that with a "versatile algorithm" they might fail only after 40 years

As for my impressive terminal board... cover your eyes and be very careful looking at this rats nest at the back side... Prototyping boards were expensive then.

20221125_183517 (resized).jpg

I still have that chip, it was a Hitachi 2716. No date code but must be '79 or '80.

That is a work of art!

My side note:
I finally have a Williams/Bally DCS sound board EPROMs fail.
The game would boot and give one bong...
Through the years the "bong" would finally give six bongs to one bong... Finally stayed on six bongs.
I had to replace two EPROMs; all back to just one power up bong.
This is my first experience, with a personal game, that the EPROMs have failed.
Odd and fascinating at the same time.. I'm so glad pinball enthusiast people have made copies of the
SND software, I owe them a bit of gratitude.

Had the same thing on Pinbot sound ROM... Booted fine and worked for a couple minutes. Verified fine. Reburnt fine. Always same symptoms. Finally just tried another EPROM.... Bang, working great since then.

Probably just its time or heat related.

I just replaced a High Speed sound EPROM, failed checksum.

When i fixed original boards the PROMs on William's speech boards very often had bit rot. There would just a few a words garbled. Read back the chip and it was mostly good. Those are masked ROMs though if it matters. I can't really remember any other bit rot like issues on old EPROMs that where not a total failure of the EPROM and usually caused by over voltage or plugging the chip in "upside down".

I got back a GTB System 3 MPU where the six pin power connectors for the power board got swapped. Every EPROM in the game but the GPROM was damaged to a point it was totally dead. The MPU GPROM did not verify right but was close, I was able to erase and reburn it and it seemed fine.

My BSD occasionally gets a credit dot for ROM checksum error at power on. It always plays fine, I've ignored it for a couple years now.

My Sega Frankenstein had a few speach samples garbled but did not report any faults.
Cleaning pins and reseating did not help
Replaced the Rom and worked fine

The W27C512 will work in place of a 27xxx? I checked the pinout and it is slightly different on one or two of the address lines; everything else was the same I think. I’m looking to get some for testing/temporary use.

Yes, provided you replicate the data. The address lines do not change position.

Larger ROMs use pins which lower capacity DIP28 ROMs didn't, or programming signals were moved which don't affect normal read operations.

W27E512 are functionally the same as W27C512 and also the same price. The 'E' in the part number more clearly indicates them as EEPROM.

Here's a pin definition diagram I use for reference:

MPU-35a_27128ROM.png

Thanks for all the info. Trying to figure out now how I would double or quadruple the data on the GQ 4x4. I don’t see how you would do that.

The easiest way is in a command line with copy /b file + file filedouble or copy /b file + file + file + file filequad
Burner software itself you have to tell it to load at an offset which will change depending on the size you are burning for instance double up a 256 for burning on a 512 you load the file at offset zero (aka load it normally) then load the same file at offset +$8000 as that is the size of a 27256 EPROM

Awesome, well that is pretty easy! I figured so much with the offset and using the burner software… just didn’t know how’d you figure out the correct offset to use. Yeah I think the copy method is probably easiest. Thank you sir.

Anyone have a reliable source for 2732(a)'s? I have a batch I purchased off Ebay months ago and not a single one will program. Either with my newer burner or an ancient Pocket Programmer parallel port model. All of them fail around the 2% though a single one made it to 54%

Jameco has grab bags of eproms that you have to erase that are a good deal if they have any left.

You can also build an adapter to use other sizes as 2732.

If you are using the new 2732 "fakes" out there like the one below, you can usually get them to work if you set the burner for generic 2732 & programme @ 25V. Were your chips new/fakes or used? The fact you are getting a similar failure rate across all of them would seem like the case as that would not happen with a bunch of used unless all were compromised in the same way.

There are the fake new Hitachi's out there as well.

I just got another lot of used HN462716 eproms the other day, never had a problem with any lots I have bought & they are always erased when I get them.

https://www.ebay.com.au/itm/363837388333?hash=item54b665f62d:g:nY8AAMXQjwVQ~Uq5

https://www.ebay.com.au/itm/123502210036?epid=1111725623&hash=item1cc14e2bf4:g:kmkAAOSwM7lckKHh&amdata=enc%3AAQAIAAAA4PG9ZNov3OkZeugHtzWAtkp03WWAtTn5SjdyMHcazMp7Av%2F1qstjLaABWHXtM0BWYDiNH2R7r74GjFQncfgrxnzI8v6DswOVd7pkC2WsKxVDOe6rhmjK4Q8%2FSZpBK164FNNphOSa8gTxP92dQfK%2F7IUV9EUlyDA5Jcb%2FpwtKmc5f645HTmW20zfDknyU3s8o%2Fmwed%2F4tAQ1lcd5XULnBOivsj5ss2bm81Jj1pc0Oae7pFRrtNEd07Q%2FFBDm0oYiQQ%2BRm3rutEb2TfD1LxaV9kMTr2j5pvYCPVAOrTauTVWfE%7Ctkp%3ABFBMgPmGnfRi

IMG_4812 (resized).JPG

Probably 2 or so years ago there were counterfeits out of China of ST branded 2732A chips that were actually National Semiconductor 2732B chips based on the internal silicon die. These use a programming voltage of 12.5V, not 21V as printed on the chip. The below picture was sent to me from frunch of said part - looks the same as some of the ones Joydivision posted: I don't remember whether programming them at 21V blew them. Try programming them at 12.5V

2732_Remark.jpg

I had a note on my tube of these bootleg 2732s to programme set up as Generic 2732 at 25V - I thought andrew @Barakandi may have suggested this somewhere along the line? can't remember. So lower voltage not higher is recommended.

That looks like them. I tried all three voltages with the older burner and it made no difference.

Running a dump after shows 4-8 locations programmed and that's it. They erase fine but refuse to program further than the amount I indicated above.

Recycled some 2532's from Vector S&T, worked just fine.

These were new.

ebay.com link: itm

Don't trust the picture in ad, the cores are larger I think. I'll take a picture.

So the chinese M2732A chips need to be programmed at 12.5V to start with, if that doesn't work then try 21V & then 25V as you don't know really what they actually really are. Then some won't work regardless.

Used pulls are a much better option than these I think.

To be honest I have had a good success rate with these till now.

20231105_232554 (resized).jpg

20231105_232601 (resized).jpg

Sounds like you just got a bad batch from that seller.

You likely damaged them all programming at the remarked/rebranded voltage obviously being higher than what the chips really wanted.

Here's a web site showing some EPROM silicon die images - click on the 2732 link on the left frame. Does that National Semiconductor NM27C32BQ-150 die look familiar to the above ST remarked chips? Programming voltage is marked as 12.75V

http://brainstones.narod.ru/collection_uv_eprom/_uv_left.htm
http://brainstones.narod.ru/collection_uv_eprom/2732.htm

When you match the silicon die, the datasheets will tell what programming voltage and pulse width the chip needs.
Here's that genuine NS part picture for reference:

2732_ns_nmc27c32bq_150.jpg

Great info Quench! that's a great resource to definitively determine what chip you have & the programming voltage.

Agreed the silicon die matches the NS part. Also the other NS part NM27C32Q-45 is identical to the ST remarked eproms with the larger die that some have in the above pics..

I have looked very closely under magnification & the chips I have & they are not used, sanded & refinished chips, they are brand new for sure.

Would these be then newly manufactured chips based on the NS silicon die(s) & then marked ST to simply sell for more than an NS chip?

There used to be a great site (cpucollection.ca/2732.htm) that had silicon die pics of nearly every 2732 but sadly it died three years ago.

Well that one is kind interesting. The larger die matches the Eurotechnique ETC2732Q-45 chip. They either OEM'd the part from NS or vice-versa. At some point Eurotechnique became Thomson and later SGS merged with Thomson to become ST. So that one may be an origin from a former ST company. The ETC2732Q datasheet specifies it's a 25V program part, and the NS NM27C32 datasheet says the same. The Q suffix in the part number refers to ceramic DIP package.

Are you referring to the picture you posted above? They are definitely are all remarks. Considering there's different silicon die in the mix they cannot all really be manufactured on the same date code (also no way they were made in 2007) and with identical markings.

I doubt anyone is newly manufacturing 2732 EPROMs. If you're talking about other ST 2732 EPROMs that gives you this suspicion, post some pics so we can have a look.

I've had shit luck with any 24 pin EPROM lately. Years ago I the 2732 pulls from China where generally OK enough. Get bad batch every now and again. I bought from a few batches from different sellers and most of the 24pin erpoms like 2732 wouldn't take a burn.

Bigger the die, bigger the fail rate. At least it seems so... Even applies with the 27C160s I use. Whenever I peal a label off a 27C160 and see it is the giant sized window version, I go dammit. I don't think I've successfully re-used one of those yet. Rarely have an issue with the 28-32p EPROMs

Been meaning to try some AT28C16 salvages. Strange and unfortunate they didn't make AT28C32... skipped over to 64K

That would explain why those particular remarks worked with a 25v programme.

Yes, referring to the pics in post #83. I had another close look, here is a close up pic of one, yes remarks as you said, these are very good & do not look like the usual remarks where the top has been sanded, quite often a bit crooked too.

I think the whole chip maybe has been sandblasted (with maybe the glass covered) to look uniform then retinned/restamped. The top of the legs look sandblasted.

IMG_4825 (resized).JPG

Yeah, I had the same issue with the first 27C160 I had and every time I hit the program button I got a little further. Took something like 100 programs to reach the end with a successful write. In hindsight I should have tried bumping up the program voltage a little.

The small die 27C801 seem to program much easier than the larger die units.

They ran out of pins for the /WE signal to make a AT28C32.

Hmm, I always thought those grey silicon blobs were from factory. Looking at a bunch of local NOS ST M2732A a friend gave me, none of them have those blobs.

These were the last ST M2732A parts I got from China in August 2023.
https://www.aliexpress.com/item/1005004784226290.html
Their picture is of a remark, but I received genuine not remarked with original '88 date code parts with small silicon die - even still have the glue from the removed window label. Legs have been re-tinned. They don't have the grey silicon blobs. But I haven't tried them yet.
There is no guarantee an order placed today will be the same.

Recently I bought some 74HCT00 SMT parts from an AliExpress seller. They came as fake marked Philips chips but worked fine. I ordered more from the same seller and got genuine Philips parts with all sorts of date codes in the 90's. Then I ordered more and got the same as the first batch fake Philips marked parts.
I also ordered 74HCT00 from other sellers and received NXP 74HC00 parts (i.e. HC, not HCT)
It's a lottery what you get.

You saw the question & pic about the silicon before I edited it out! I realised the silicon blobs were of no importance & probably there from the factory or even added to the remarks.

As far as the bootleg 2732s... start at the lowest voltage and work your way up. If it takes an abnormally long time to program, probably need to go to the next voltage up.

https://www.marcospecialties.com/control/keywordsearch?SEARCH_STRING=2732a

Marco looks to be selling remarked chips.

The two with the small core would program at 21v but not at 12.5v.

The large core would not program AT ALL regardless of voltage.

Help line at Marco said they were sending more my way today and would look into the issue.

20231108_115327 (resized).jpg

The reality is it's probably difficult/expensive finding real NOS or used pulls in the US in decent quantity anymore. Guesstimate they went out of production in the 90's.

That wide core silicon die looks like an AMD 2732B chip. AMD 2732B are available through China like ST parts. I have some here that work ok, but NOT in a XGECU TL866ii+ programmer - it has a bug the manufacturer has ignored fixing.
Chinese clearly have no idea about these EPROMs. There is no such thing as a 2732A with 12.5V programming.
2732 = 25V programming voltage
2732A = 21V programming voltage
2732B = 12.5V programming voltage

2732_amd_am2732bdc.jpg