HCR Micron Plus EPROM Programmer

[hoglet]

hopefully all goes well fingers crossed.

Thanks!

Here are some photos of it fully built up:

So, does it work? Yes it worked first time, which was a bit of a surprise as all the ICs came from my spares bin and were untested.

I was going to record a video of the software running, but it won't look any different the existing videos:
- Burning EPROMs with the BBC Micro (Least Significant Byte)
- Burning an EPROM using the HCR Micron Plus 64 programmer for the BBC Micro (RetrolitUK)

So far I have only programmed 27128 (21V) and 27128A (12.5V) and read a 2532.

But I've checked that VPP appears in the right place for each of the ROM types.

It seems a really nice programmer, and I can see myself using it as my main programmer for Beeb sized EPROMs.

Some observations:
- current consumption is only about 400mA (@9VDC) and even after an hour nothing is particularly hot.
- VPP is spot on (12.5V/21.0V/25.0V), due to it being adjustable with trim-pots
- Under load, VCC to the EPROM is slightly on the low side (4.8V and 5.8V rather than 5.0V and 6.0V), because of voltage drops through the switching transitor, relays and traces. This could be addressed by bumping the Zener from 3.6/4.7V to 3.9/5.1V. I won't do this unless this actually causes a issue. This is probably the case with the original programmer, as this part of the circuit is unchanged.

The only anomoly I have found is, in the inactive state where the socket is "powered down", pins 20 and 22 (OE and CE) are at 4.8V.
These pins are driven directly from the latch, so this could be addressed with a software fix. It's still a bit of a surprise, because the design includes buffers to isolate the address and data lines when inactive.

I wonder if someone with a real programmer could make some voltage measurements:
- start the software (v1.40 or v1.41) and select a 27128 device
- measure the voltage on each of the socket pins wrt pin 14 (GND)

This would confirm this not due to a subtle mistake on my part....

Now I just need to find a suitable case for it.

Dave

[KenLowe]

Nice .

I think I might need to get one of these. Even if it's just to program those pesky 25v 2532 EPROMs for my Atom .

[BeebMaster]

Definitely I would like one as well, even if it's just the PCB and parts list.

[vanekp]

Very nice that it works 1st time that's always a big bonus when you had to basically build it from photos, I use to have one of these programmers I am sorry I did not take it with me when I moved over seas as with all my BBC stuff, I just use a TL866II+ these days works for most things.

[hoglet]

I think I might need to get one of these. Even if it's just to program those pesky 25v 2532 EPROMs for my Atom .

So, I just tried to program a TMS2532A and it was less than successful.

For the record, this was the IC:

This started off with AtoMMC 3.0A in it, which had, many years ago, been working in a Atom. I was able to read this in the programmer and see data. I erased it, blank checked it, and then attempted to program it at 21V. I picked 21V because that's what my Needham EMP-20 uses for a TMS2532A, which is how I previously programmed it.

That's when things started to go wrong....

A couple of seconds into programming, the relays clicked and the green "socket active" LED when off. The programming software continued to run, so this was obviously a "glitch" rather than a planned action.

I let the programming pass complete, and the verifiy failed on byte 0.

I then tried to read the ROM into the buffer, so I could see how far the programming had got, and that proved not to be possible. As soon as the software tried to apply power to the socket, it would "glitch" and the socket would power off.

I tried a second (different) TMS2532A (containing ECONET 3.50) and was still able to read this.

So it definitely looks like the first TMS2532A has become damaged.

I put the EPROM in a breadboard and applied power. Current draw was 50mA, which is about right. I then connected the pin 21 (VPP) to 5V, and the current draw went up to 200mA and the PSU current limited.

So it looks like the VPP pin has somehow blown.

Now, this could be one of several things:
- the EPROM was on the way out anyway, and just died of "natural causes"
- the EPROM was damaged by an error in the programming algorithm
- the EPROM should have been programmed at 12.5V
- the EPROM was a fake (the die is suspiciously small, and the date code is 8732, which is quite late)

Now, as this was previously working in an Atom, it must be some kind of 2532, rather than the much more common 2732. I think it came from Charlie Robson when I bought my AtoMMC.

I also have a small tube of EPROMS marked as TMS2532A that are definitely remarked 2732s, so I then tried to program one of these (at 25V), and that worked. As far as I know, the programming algorithm is the same, it's just the pinout that is different.

I'm really not sure what to make of this.

I'll do a couple of things now:
1. Try to locate a datasheet for the TMS2532A circa 1987 and to confirm the programming voltage.
2. Scope the socket during programming to see if anything seems incorrect.

Thoughts anyone?

Dave

[BigEd]

Would it be safe to always start by trying at lower Vpp? Start at 12 and only go to 21 if it fails to program.

[hoglet]

Would it be safe to always start by trying at lower Vpp? Start at 12 and only go to 21 if it fails to program.

If you are not sure, then in general this would be a safer approach.

But I was pretty sure that 21V was the correct voltage, as that's what the Needham EMP-20 uses for a 2532A, and I have programmed these successfully before in that programmer.

Dave

[BeebMaster]

I can't seem to find a datasheet for 2532A, only 2532 which says the programming voltage is 25V.

Apart from (I know this is crucial!) the slightly different pinout, are these the same as 2732? Is it that the Atom ROM sockets are wired for 2532 rather than 2732?

[hoglet]

I can't seem to find a datasheet for 2532A, only 2532 which says the programming voltage is 25V.

Searching on archive.org it was for sale from 1987 to 1992.

But I can't find a data sheet either, not is it mentioned in the TI MOS Memories Databook of 1989.

Weird...

I've found a few references to it being a 21V device, and none to it being a 12.5V device.

So I think something must have "gone wrong" during programming.

Apart from (I know this is crucial!) the slightly different pinout, are these the same as 2732? Is it that the Atom ROM sockets are wired for 2532 rather than 2732?

That's correct.

Dave

[daveejhitchins]

Dave - are you able to read the chips identity? You might be able to do this in your Needham's? That is if it ever had one!

Dave H.

[hoglet]

Dave - are you able to read the chips identity? You might be able to do this in your Needham's? That is if it ever had one!

I don't thinks EPROMs of this type/era supported the manufacturer's identification code (via 12V applied to A9). I think it's only supported by the (CMOS) 27C64 and later.

My plan is to ask Revaldinho to look at the die under his microscope (he doesn't know this yet....). There is a photo of a genuine TMS2532A here.

I've just looked at the VCC/VPP sequencing on the scope during programming, and there might be an issue with the 2532 algorithm...

Here's what the 27128 looks like (scale is 20ms/div):

The top trace is VCC, the bottom trace is VPP. You can see VPP starts off at 5V, then rises slowly to 21V over about 50ms.

Here's what the 2532 looks like (scale is 2us/div):

You can see VPP comes to 21V very quickly, and is 21V for about 2us before VCC reaches 5V.

The 2532 datasheet page 4 says says "VCC must be applied before or at the same time as VPP". Maybe it being applied 2us early has caused some kind of latch up. I'm a bit sceptical that this is the cause, but it's a possibility.

The other possibility is that I programmed it at 25V rather than 21V (though I did this slowly and carefully, so I don't think I messed up).

Dave

[BigEd]

I wonder if you'd see anything by measuring resistance from Vpp to ground, on a good vs bad part? Or perhaps testing it as a diode??

It does somewhat look like a no-longer-connected-on-the-inside pin.

[MartinB]

@Dave - There’s an interesting (maybe) thread here on 2532A programming if you haven’t already seen it….

https://forums.arcade-museum.com/thread ... ms.455368/

[hoglet]

I wonder if you'd see anything by measuring resistance from Vpp to ground, on a good vs bad part? Or perhaps testing it as a diode??

It does somewhat look like a no-longer-connected-on-the-inside pin.

It's actually the shorted, rather than open, as a lot of current flows into VPP when you try to raise it to 5V now. But not a direct short - the voltage has to get above 1.2V for current to start flowing. So it's kind of like a pair of diodes down to GND.

Dave

[hoglet]

@Dave - There’s an interesting (maybe) thread here on 2532A programming if you haven’t already seen it….

https://forums.arcade-museum.com/thread ... ms.455368/

Thanks Martin, that's an interesting read.

It's actually the same thread I came across earlier that had the die photos.

I do now think my "late" TMS2532A was genuine; it looks identical to the genuine die shot in that thread (with a tiny die), and it was successfully programmed as a TMS2532A (21V) in a Needham EMP-20.

Now, here's an interesting scope shot, showing the start of the programming sequence for a 27128:

The top trace is VPP (pin 1), the lower trace is PGM (pin 27).

It looks like the software starts programming while VPP is still ramping up. You can see it really struggles to program the first byte, compared to the subsequent bytes. It takes 80ms to program that first byte, compared to 8ms for the subsequent bytes. It only succeeded because it's using an intelligent algorithm.

I can't imagine this is deliberate.

I can't currently explain the slow rise of VPP from 5V to 21V. There is a 4.7uF capacitor on the ADJ pin of the LM317. But my back of the envelope calculation suggests the time constant should be more like a a few ms, not 40ms. I will double-check CPC sent me the right capacitor.

For reference, the schematic is here

Anyway, this doesn't explain the blow TMS2532A....

Dave

[Mince]

I wonder if someone with a real programmer could make some voltage measurements:
- start the software (v1.40 or v1.41) and select a 27128 device
- measure the voltage on each of the socket pins wrt pin 14 (GND)

This would confirm this not due to a subtle mistake on my part....

I can do that — do you just mean instantaneous voltages with a multimeter, or over time with an oscilloscope? Also, any particular pins (such as the ones you were querying), or all of them?

Will take me until later in the week, though, as I have The Australian Pink Floyd tomorrow and then a baseball committee meeting on Tuesday!

Now I just need to find a suitable case for it.

I think it needs to be brown else it won't work properly!

As an aside, I've not got any 2532s and have never tried programming one in it (although I've programmed 2732s successfully... and unsuccessfully by accidentally picking 2532 mode, for that matter).

[hoglet]

I can do that — do you just mean instantaneous voltages with a multimeter, or over time with an oscilloscope? Also, any particular pins (such as the ones you were querying), or all of them?

Whan I posted this I was just interested in the instantaneous voltages (i.e. with a multimeter) of all the pins when the green LED is off, and no EPROM present.

But if you are up for it, a scope plot of VPP (pin 1) and PGM (pin 27) @20ms/division for the first 200ms of programming an 27128 (with the EPROM present). I used an 28-pin DIP IC test clip to attach to the pins. I'm mainly interested in VPP, so if you can only easily get to that pin, then that's fine. Just don't accidentally short it to pin 2! Trigger on the rising edge of VPP passing 2V. I'm interested in the rise time of VPP, see the scope shot in the previous post.

Dave

[hoglet]

I can't currently explain the slow rise of VPP from 5V to 21V. There is a 4.7uF capacitor on the ADJ pin of the LM317. But my back of the envelope calculation suggests the time constant should be more like a a few ms, not 40ms. I will double-check CPC sent me the right capacitor.

I've thought about this some more, and I can explain the slow rise now...

Here's a fragment of the schematic:

(Assume R10/R11 are effectively disconnected)

I had naively expected when the 5V clamp (ZD3) is released that the VPP rise would correspond to the RC time-constant of 220R (R13) x 4.7uF (C16), which is about 1ms

But that's completely wrong...

The LM317 maintains 1.25V across Vadj and Vout, and as these pins are connected by a 220R resistor that means that about 5.7mA will flow through the resistor. This acts like a constant current source.

When the clamp is released, the intial voltage on C16 is 3.6V. However, as there is a 4K (R12+RV1) in parallel, then 1mA of the 5.7mA will flow through this resistor, leaving 4.7mA to charge the capacitor.

So how quickly does this charge initially?

Q = I * t = C * V

dV/dt = I / C = 4.7mA / 4.7uF = 1.00 V/ms

So it's going to take at 17ms to rise from 3.6V to 21V

But even that is wrong, because as VPP rises, then so does the current flowing through the resistor (R12+RV1), meaning over time less and less of the 5.7mA will be left to charge the capacitor.

So it's going to take significantly longer than 16ms to get close to the final value, which is exaectly what I saw on the scope:

It actually takes ~50ms to get "pretty close".

If Mince does make the same measurement, I expect he will get the same result. The programming software really should wait at least 100ms for VPP to stablize before it starts programming.

Dave

[hoglet]

The other dangling issue is why did the TMS2532A die at the start of programming?

My best hypothesis is it's because in the 2732/2716/2564/2532 programming code, the power supply sequencing doesn't meet the data sheet requirement that VCC must be stable at 5V before VPP applied.

Looking through my Atom spares, I found one more identical looking TMS2532A from the same batch.

So I had a go at updating the software to properly sequence the power supplies:
https://github.com/hoglet67/HCRMicronPl ... t/e955d989

This is the with the original 1.41 version:

The top trace is VCC, the bottom one is VPP, which is clearly the faster of the two to rise.

And this is with my modified 1.42 version:

The top trace is VCC, the middle one is VPP, the bottom one is nPGM.

You can see VPP now come up in several steps:
- 0V (before VCC is enabled)
- 2V (after VCC is enabled, the VPP pin acts just like a TTL input, and there is a 3K3 pulldown)
- 5V (after VPP is enabled, the VPP regulator is initially clamped at 5V)
- rising slowly from 5V to 21V (ater the clamp is removed)

When programmng starts, VPP has been stable at 21V for about 100ms.

With this updated code I've successfully programmed my one geniune TMS2532A several times now without any issues.

In an ideal world I would now revert to the original 1.41 software and prove (beyond reasonable doubt) that it was indeed the power supply sequencing that caused the demise of the first EPROM. But I'm not going to do that, as I only have one of these EPROMs remaining. I'm comfortable the new code is an improvement.

Dave

[hoglet]

And for completeness, here are the before and after shots programming a 27128:

Version 1.41:

Top trace is VPP, bottom trace is nPGM.

Version 1.42:

Top trace is VCC, middle trace is VPP, bottom trace is nPGM.

It stuggles to program the first byte, as VPP is still rising, taking 80ms to complete it.

Programming doesn't start now until well after VPP is stable, and all bytes program in ~8ms.

Hopefully there won't be any more issues now.

Dabe

[Super_7b]

Hi Dave,

This thread has been silent for a while now, so I'm wondering if you've been doing any more work on the EPROM Programmer?

Looking at the forum, you have loads of other stuff active, so I can well imagine this is on the back burner, considering that it seemed to be working correctly in the latest posting.

I am keen to build one myself, but if you think there may be more to do I will hold off.

Do you intend to have any boards made and sell them or should I look at downloading the files and sourcing my own board(s) from someone such as JLCPCB?

BR

Mick

[hoglet]

This thread has been silent for a while now, so I'm wondering if you've been doing any more work on the EPROM Programmer?

Looking at the forum, you have loads of other stuff active, so I can well imagine this is on the back burner, considering that it seemed to be working correctly in the latest posting.

I am keen to build one myself, but if you think there may be more to do I will hold off.

Do you intend to have any boards made and sell them or should I look at downloading the files and sourcing my own board(s) from someone such as JLCPCB?

I'm not planning any more work on this - as far as I known everything is working and I'm happy with the current PCB.

I do have a few spare boards which I'd be willing to sell for £15 inc UK postage (which will help cover my initial outlay and fund future projects).

Or feel free to source your own.

Dave

[Super_7b]

Hi Dave,

Thanks for the prompt reply. I've PM'd you to purchase a board.

BR

Mick

[hoglet]

Thanks for the prompt reply. I've PM'd you to purchase a board.

Your board is on it's way.

I've added a BOM to the wiki, including links for some of the more unusual parts:
https://github.com/hoglet67/HCRMicronPlus/wiki/BOM

I've not listed suppliers for everything because I had quite a lot of the bits in stock already.

Do ask if you are struggling to source anything.

Dave

[Super_7b]

Hi Dave,

Just started gathering parts to see what I have and what I need to buy, checking against the BOM.

I noticed in the BOM that it mentions "HM6464LP 2 U12,13", but in the schematics/silkscreen they are listed as HM6264LP. I assume this is a typo in the BOM.

I have some HM6264LP-70 chips, so I suppose these are ok as there is no specifed speed in the schematics.

BR

Mick

[hoglet]

I noticed in the BOM that it mentions "HM6464LP 2 U12,13", but in the schematics/silkscreen they are listed as HM6264LP. I assume this is a typo in the BOM.

FIxed now.

Let me know if you spot any more.

I have some HM6264LP-70 chips, so I suppose these are ok as there is no specifed speed in the schematics.

Pretty much any speed should work.

Dave

[Super_7b]

Hi Dave,

The board is now completed and I tested it with no ICs (except the MC34063AP power conversion chip) to ensure the voltages were sane (they were).

I then installed all the ICs and retested. All looked fine with no hot components and no release of the magic smoke. I was able to check the 25V line and tweak it to the nominal value, but I guess the 21V and 12.5V lines would need some action from the logic to select them before I could adjust them. How did you achieve this?

I also need to get an image of the control software to load to my SWRAM. In the Github repository I can see the source for the corrected version. What assembler toolchain do I need to use to create the binary? I have access to Windows 10, but also a couple of flavours of Linux (Gentoo on a real machine and Ubuntu on VirtualBox). I'd be happier installing the correct tools on Linux, if possible.

Once I have it all working I shall look out for a suitable case.

Thanks

Mick

[hoglet]

I was able to check the 25V line and tweak it to the nominal value, but I guess the 21V and 12.5V lines would need some action from the logic to select them before I could adjust them. How did you achieve this?

Here's a procedure to adjust all three voltages:

1. Unplug U7 (74LS374) and U11 (7406)

2. Connect a volt meter across R4 (on the 30V range) and power on. Initially the meter should read 0V. If it doesn't, something is wrong.

3. Jumper-wire pin 19 of U7 to GND and adjust RV1 for a reading of 25V.

4. Now add a second jumper-wire connecting pin 4 of U11 to GND and adjust RV2 for a reading of 21V.

5. Move the second jumper so it's connecting pin 6 of U11 to GND and adjust RV3 for a reading of 12.5V.

6. Power down, remove the two jumper wires and re-insert U7 and U11.

I also need to get an image of the control software to load to my SWRAM. In the Github repository I can see the source for the corrected version. What assembler toolchain do I need to use to create the binary? I have access to Windows 10, but also a couple of flavours of Linux (Gentoo on a real machine and Ubuntu on VirtualBox). I'd be happier installing the correct tools on Linux, if possible.

It uses BeebASM which can be built for Linux.

Let me know if you need a binary.

Dave

[Super_7b]

Hi Dave,

Once again, many thanks.

I successfully installed BeebASM on a Gentoo machine and was able to assemble the v1.42 source file to a BBC .ssd file. This was read into the target Model B from its Gotek and it responds correctly to *EPROM and all the menus are as expected.

I carried out the tests you specified for the programming voltages and they were all quite near straight out of the box. I did the OCD thing and tweaked them as near as I could to the specification.

I can't actually test programming an EPROM in anger until my ribbon cable arrives. I had the plugs on hand, so I simply ordered a 1 metre length of cable. Would it be risky to use the full length?

BR

Mick

[hoglet]

I can't actually test programming an EPROM in anger until my ribbon cable arrives. I had the plugs on hand, so I simply ordered a 1 metre length of cable. Would it be risky to use the full length?

The HCR EPROM programmer doesn't include termination resistors, so I would try to keep the cable fairly short. The one I use is 60cm long.

Dave