2708 EPROM Power Sequencing

[ChrisOddy]

I'm currently testing my replica SOFTY 1 Programmer. Derek from Bygonebytes has kindly lent me his original board but that has a few blown IC's on it including the 8060 and I've been trying to establish why ?

The Softy 1 needs +12V and -5V in addition to +5V to power the 2708 EPROMs (firmware and the device being programmed).

I have my replica board running but on +5V only (with a 2716 for firmware), I havn't yet tested programming.

During my testing (of Dereks board) I've been careful to apply the +5V supply first followed by +12 and -5V my assumption being that keeping everything biased with +5V would prevent +12V or -5V from damaging anything ?

I've had a look through several 2708 data sheets to see if there is any info on power sequencing. The Intel data sheet states (in a small note) that -5V must be the first and last power supply applied, other data sheets don't actually state that but do reference all maximum supply ratings to VBB (-5V).

Dereks Softy is powered by a simple linear power supply, I assume it came with the Softy ? It doesn't look like there is any circuitry to sequence the supplies.

I havn't found any info in the Softy 1 manual either that suggest any precautions to be taken with power supplies ?

I've found several designs for 2708 programmers on the web but I've seen no specific attempt at power sequencing or mention that it is required ?

Presumably the 'apply -5V first and last' is to protect the 2708, my concern is how to protect everything else on the board !

thoughts ?
Chris

[Orakle42]

Just a thought, In my experience it is always worth ensuring that reverse voltage protection diodes are fitted across each power line in multi line systems to protect against different rise and fall times and reverse voltage scenarios. It is particularly important if different supply sources are used when testing. This may not be the cause of your failures but it will certainly eliminate possible future problems.

Orakle42

[ortek_service]

If the 2708 is the only device that uses the extra +12V / -5V rails, then I would expect that to fail first, before any strange parasitic-powering causes <0V / > +5V to get fed to anything else before the +5V rail is stabilised.

There could be a latch-up issues, where voltages are present on pins and powering IC's, before their main +5V supply is present, but would have to be from +12V rail.
I once had a +5V-powered EPROM, causing a parallel CMOS 9V? PLL IC to latch-up and get hot, as it had an RC supply filter causing a lag in this.
I resolved this, by putting series resistors between EPROM data outputs and PLL inputs, to limit current below that which could cause latch-up.

Putting reverse-polarity shunt diodes (or maybe some moderate-power zeners, just above rail voltage) on the rails, might help prevent incorrect voltages getting onto these.
But this maybe a bit over-cautious - like power-supply sequencing - that doesn't seem to be common on equipment from that era. So I presume they got away with not requiring it, otherwise there would have been a lot of failures.

I wonder how many Softy-1's are still working / failed after short periods originally, which could have pointed to a design problem.
They wouldn't have had this issue with the 'replacement' and sold for much-longer Softy-2 model as it was single-rail only.

[Mark1960]

The older equipment would be almost certain to be using linear power supplies, with simple transformer, rectifier, bucket capacitor and voltage regulator. The -5v supply would be much lower current than +5v, so would use a smaller bucket capacitor, which then allows the -5v faster turn on time than +5v.

Bucket capacitor for 12v could also be chosen to slow the turn on time for 12v,
Though this also slows the turn off time.

[buggies]

A friend's Nascom 1 board freshly built had its eproms killed by not having the negative supply sequenced properly.

[SiriusHardware]

This is actually starting to sound like a job for a microcontroller, (to switch the rails on and off in a precisely timed sequence). The microcontroller would need to have a well reservoired supply so that it could switch the output rails off in the correct sequence even in the event of mains input failure (the assumption would be that all the rails would have enough reservoir capacitance to keep them going for a few milliseconds, but on detection of mains failure the microcontroller would switch them off sooner than that, in a precise timed sequence).

[circuitryboy]

Checking my '2708' data found 3 manufacturers with "Vbb on first/off last" annotation - AMD, Intel, Motorola.
Is the PSU, pictured with the Softy, supplied by Dataman? And is the EPROM one of the damaged ICs?

[ortek_service]

Originally Posted by circuitryboy

Is the PSU, pictured with the Softy, supplied by Dataman? And is the EPROM one of the damaged ICs?

Chris hadn't been sent the PSU board - which is believed to be the original Softy (It seems Dataman only took over producing these from the Softy-2, which wasn't 3-rail anymore) designed / supplied one, and Derek had once had it working OK. But there's a possibility this has a fault on it with one or more of its supply-rail outputs.

The 2708 EPROM had survived OK, despite it appearing t be the most critical on having all 3 supplies present - Luckily, as Chris had originally just been sent that, to readout as it seems no-one had put a copy on the 'net until now.

[ortek_service]

Originally Posted by buggies

A friend's Nascom 1 board freshly built had its eproms killed by not having the negative supply sequenced properly.

It does seem that 4116 DRAM's were quite sensitive to missing supplies, as they often died in ZX Spectrums often with the DC-DC converter also faulty.
But the DC-DC converter wasn't too reliable and they did mod. it's design a bit over the various PCB-issues, to improve it / prevent incorrect voltages being fed out if it stopped oscillating.
And it's possible that a 4116 drawing too much current had killed the DC-DC converter circuit (Although I don't recall finding any dead 4116's with shorted supply inputs)

[ortek_service]

Originally Posted by SiriusHardware

This is actually starting to sound like a job for a microcontroller, (to switch the rails on and off in a precisely timed sequence). The microcontroller would need to have a well reservoired supply so that it could switch the output rails off in the correct sequence even in the event of mains input failure (the assumption would be that all the rails would have enough reservoir capacitance to keep them going for a few milliseconds, but on detection of mains failure the microcontroller would switch them off sooner than that, in a precise timed sequence).

Yes, once you've got a uC in a system, then it's common to use it to control all the switched-supplies for an orderly start-up / shutdown / can also provide test-modes & some self-checking / measuring of these.
But it does seem a bit overkill for something a bit retro like this, making it rather less original, that presumably was fairly reliable originally (although with so few still around / maybe sold originally, would rely on finding people who used them a lot back then to comment on issues they had, in case there was a major design-flaw)

[marty_ell]

Originally Posted by ChrisOddy

I've had a look through several 2708 data sheets to see if there is any info on power sequencing. The Intel data sheet states (in a small note) that -5V must be the first and last power supply applied, other data sheets don't actually state that but do reference all maximum supply ratings to VBB (-5V).

Presumably the 'apply -5V first and last' is to protect the 2708, my concern is how to protect everything else on the board !

thoughts ?
Chris

Bitter experience from the '70-'80s quickly taught me that early EPROMs/DRAMs/micros demand their -5/-12V rails to be powered up and close to or at their rated V before the +5V and particularly +12V get above a V or so (and this was often not specified in the manufacturers datasheets of the time). Otherwise there is a serious risk of latchup and thermal runaway to destruction. So Intel's comment on power-up is absolutely right.

But it is not so important that the -V holds up until the +V rails disappear during power-down; a properly running chip is not likely to latchup and if it did, it would only be for a few 10's of ms.

Originally Posted by SiriusHardware

This is actually starting to sound like a job for a microcontroller, (to switch the rails on and off in a precisely timed sequence). The microcontroller would need to have a well reservoired supply so that it could switch the output rails off in the correct sequence even in the event of mains input failure (the assumption would be that all the rails would have enough reservoir capacitance to keep them going for a few milliseconds, but on detection of mains failure the microcontroller would switch them off sooner than that, in a precise timed sequence).

If you are concerned only about the -V rail, I suggest a 555 timer will do the job. Just set it to switch FETs or close a relay to connect the +V rails after the time required for -V to stabilise. As 'belt and braces', assuming the -V rail is lightly loaded (usually is in old micro designs) you can fit a big reservoir on the -V rail which will hold it up until long after the +V rails have decayed.

[ChrisOddy]

My thoughts are to do something along those lines. From a 5V supply use a small switcher to generate -5V, use this to power a small relay that then switches on the +5V which also powers +12V and +26V switchers. I reckon the relay delay will be sufficient to allow the -5V to settle and I can add a diode and capacitor so that it stays up last after power is removed.

I'm just having a go at programming a 2708 on the SOFTY, once thats working I can do some current measurements and work out the details. I bought 5 used 2708, so far two are faulty!

[G6Tanuki]

I designed/built/marketed stuff [battery-powered digital pig-weighing scales / ultrasonic back-fat-measuring probes for cattle-markets/abbatoirs] in the early-80s using 2708s along with a 6502 microprocesor and 4116 RAM] and never needed to do anything specific about powering-up the rails in a particular order.

[SiriusHardware]

I think it's fair to say that at the time all the components were common or garden and easy to find if they ever needed to be replaced. That's probably why you never had trouble with them.

Many of those same components now are expensive, hard to get museum pieces and the need to avoid doing anything which might conceivably blow them up is exponentially greater than it would have been back in the day, so I can understand Chris not wanting to leave it to chance.

[g4wim_tim]

Just had a rummage in my 'junk draw' and found I have a number of old 27xx series eeproms of varying sizes - didn't see any 2708's though.

They were all good and have been kept in anti static foam.

I'm unlikely to use them, pm me with your requirements and I'll see if I can help out.

Regards Tim

[g4wim_tim]

Originally Posted by G6Tanuki

I designed/built/marketed stuff [battery-powered digital pig-weighing scales / ultrasonic back-fat-measuring probes for cattle-markets/abbatoirs] in the early-80s using 2708s along with a 6502 microprocesor and 4116 RAM] and never needed to do anything specific about powering-up the rails in a particular order.

Agreed, I even built myself a very crude programmer and even now nearly 40 years later the chip still works !

Regards Tim

[ortek_service]

Originally Posted by SiriusHardware

I think it's fair to say that at the time all the components were common or garden and easy to find if they ever needed to be replaced. That's probably why you never had trouble with them.

Many of those same components now are expensive, hard to get museum pieces and the need to avoid doing anything which might conceivably blow them up is exponentially greater than it would have been back in the day, so I can understand Chris not wanting to leave it to chance.

Although, rather strangely, the 3-rail 2708 survived. Whereas it was some of the single-rail devices like the harder/more expensive to obtain INS8060 CPU that had failed on Derek's one for reasons yet unknown.
So might be worth putting a 5V6 / 6V2 moderate power zener across the +5V supply, to prevent it getting lifted by other supplies, as 7805's etc only source.

But may still need to look at possible latch-up at power-on causing some devices to overheat & fail - as someone else said they'd found.
Although this might be fairly easy to identify, with a lot of power-cycling testing in case it's intermittent (hopefully finding out before CPU gets damaged!)
However, Chris has told me of suspected-issues with the reset circuitry / originally had what appeared to be the Firmware EPROM not programmed correctly, that could have been masking any power-on intermittent latch-up.

[ortek_service]

Originally Posted by ChrisOddy

>>
>>

I'm just having a go at programming a 2708 on the SOFTY, once thats working I can do some current measurements and work out the details. I bought 5 used 2708, so far two are faulty!

I presume some of the remaining 3 are OK! Or are they all duff?
What were the fault symptoms with the faulty ones?
- Is it possible that the Softy isn't quite working properly yet, so making these appear faulty, unless you've tested these on your other programmers with the DIY 2708-to-2716 readout-adaptor you'd previous used for original readout.

I also have the odd 2708 of unknown status, that I acquired amongst other used EPROM's I'd bought cheap from rallies over the years.
And another 5off on a rather vintage Fairchild F8 chipset Qume printer-controller? card, that it would be interesting to dump contents of these.

[ChrisOddy]

Well out of five 2708s two seem to work - a slightly better yield than I have got from recent purchases of PROMs from China (0 out of 10 that turned up).

One locks up the bus, one doesn't program and one programs and seems to verify but doesn't run reliably.

I have now measured the currents so can design a power supply:
+5V 330mA / 400mA
+12V 27mA / 60mA
-5V 22mA / 45mA
+26V - / 9mA

First figure is with no EPROM in programming socket, second is programming. All measurements made with current clamp meter.

[Mark1960]

Is that peak or average current? You don’t want the peak current to be triggering current limit and cutting out one of the supplies.