2708 EPROM Power Sequencing

[ChrisOddy]

Average - I just used a fluke current clamp. I'll need to dig out my Tektronix current probe to look at the peak.

If I use the same modules that I used on my Acorn PROM Programmer power supply (http://www.theoddys.com/acorn/Replic...mer_board.html) then there should be plenty in hand.

IR0512SA 5 to 12V @ 250mA
IR1215S 12 to +/-15V @ 100mA

As far as I can tell from the data sheets the Vpp will draw large peaks during programming, up to 40mA, should be OK with the 100mA unit and with a reservoir capacitor(s) to smooth things out.

[G6Tanuki]

It's worth looking at the different spec-sheets from different manufacturers: the 2708s I always specced were TMS2708JL from Texas Instruments.

Back in the 70s and 80s it was not unusual for competing manufacturers to 'improve' an original part to get one over on the original-part's designers/manufacturers and attract designers to specify their improved parts rather than those by the first-to-market manufacturer, if only because the improved parts made circuit-design simpler [meaning faster-to--prototype and cheaper too, when it went from design to production]

I remember certain chips - [tri-state bidirectional bus-drivers come to mind] - were always seen as 'fragile' by some designers who shied away from using them: they'd clearly only had experience of first-generation/pre-release versions - some of which did indeed have strangely-narrow power-sequencing requirements.

[ortek_service]

Originally Posted by ortek_service

>>
>>
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.

I later remembered that Chris was initially using a 2716 for the firmware EPROM on his re-make one, so probably only had +5V supply / no path for +12V / -5V to flow anywhere with no 2708's present.
And unlikely that there would have been any latch-up issues, without delays to Vcc pins of any IC's caused by series resistance and large capacitors etc.

[ortek_service]

Originally Posted by 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.

Yes, I recall some previous 2732 / 2532 etc? EPROM's you gave me to check had quite a few duff ones (as well as ones Leigh had bought, I was programming for him). So there does seem to be quite a lot of bad ones from China etc, even if not the often counterfeit / remarked IC's.

Were those currents measured with a 2716 being used for the Softy's firmware EPROM, or have you now got an original 2708 fitted / measured them on Derek's one?

[ortek_service]

Originally Posted by ChrisOddy

Average - I just used a fluke current clamp. I'll need to dig out my Tektronix current probe to look at the peak.

If I use the same modules that I used on my Acorn PROM Programmer power supply (http://www.theoddys.com/acorn/Replic...mer_board.html) then there should be plenty in hand.

IR0512SA 5 to 12V @ 250mA
IR1215S 12 to +/-15V @ 100mA

As far as I can tell from the data sheets the Vpp will draw large peaks during programming, up to 40mA, should be OK with the 100mA unit and with a reservoir capacitor(s) to smooth things out.

Yes, it should only be a pulse of a few ms / ten's of ms from the Vpp supply, when programming, with the others not much different from reading.
- Unlike the first 1701/1702 ones, with complicated polarity reversals for programming and > 0.5A currents that meant needing multiple programming loops of a short pulse in each location to prevent overheating so took 20mins just to do 256Bytes. Although they were shortly replaced by A-suffix versions that drew a bit less and allowed much faster programming.

[ChrisOddy]

I've now measured the currents with my current probe and yes they are all pretty static except, as expected Vpp, though the peak is only 10mA.

Measurements were made with the firmware in a 2708 now that I can program them !

So far no problems with any damage to chips, I've been switching the -5V supply on first and last, fingers crossed.

Chris

[circuitryboy]

Reincarnating that board so rapidly is a superb achievement. Could we have a 'completed' picture?
I was thinking: if the '8060 in the original Softy is deceased, then it's very likely the character PROM was hit as well. (DM74S287s seem to be plentiful on the 'e' site.)
And thank you for listing the firmware. I've been looking at that. Particularly the (patented?) "TRANSWIFT" code.
You've probably already corrected that error at 01A9 (hand assembled code!) - should be 9CCF. SC/MP takes any non-opcode as a NOP, but μcycles can vary.

[ChrisOddy]

No strangely the character PROM was OK and the 8154 but the RAMs and a couple of logic chips are playing up. I haven't touched Derek's board for a few weeks as I've been concentrating on the replica but will have another look at it now I have I learnt so much about how the SOFTY ticks.

I would like to fully disassemble and comment the code - I've made a start but the 8060 instruction set is so different from any other processor I'm familiar with, the Xchange instructions in particular I find challenging to follow.

The only thing remaining on the replica is to add a modulator - I need to test it with an original UM1111 and also the more widely available UM1233, I'll need to dig out one of those old analogue TV's !

Once its all done then I'll put all the info on the website, unfortunately as the weather has started to improve the garden now starts to take over and slow things down a bit.

Chris

[ortek_service]

Originally Posted by ChrisOddy

>>
>>
I would like to fully disassemble and comment the code - I've made a start but the 8060 instruction set is so different from any other processor I'm familiar with, the Xchange instructions in particular I find challenging to follow.

The only thing remaining on the replica is to add a modulator - I need to test it with an original UM1111 and also the more widely available UM1233, I'll need to dig out one of those old analogue TV's !
>>
>>
Chris

Well I can't be as bad as PIC16 etc. strange mnemonics, that even though there's not many, I had to keep consulting the programming manual when hand disassembling / commenting some code once, having been used to more standard 68xx/65xx / Z80 & 80xx etc. + a bit of ARM.
Although I've yet to look at the register-less entirely stack based Fairchild F8.

AFAIK, Digital TV's will still receive analogue-TV transmissions (or at least the ones I've used, > 10years old). Although if they only tune to exact channel frequencies, and lack fine tuning, then you might need to tune the modulator to the TV (Ch36 etc). And if TV's now lack AFC, because they are PLL'd to a crystal-reference, then they may not cope too well with an unstable RF-Modulators transmit frequency (Particularly the 'crude' UM1111 1-transistor oscillator with no output load-isolation buffer?)

[SiriusHardware]

Have to agree on PICs, at least coming at them from an assembly language point of view - calling registers and memory locations 'files', constant bank switching and on early devices at least, woe betide you if a lookup table or a subroutine fell across a page boundary. I only really started to use PICs properly when I finally grudgingly learned a little bit of 'C', which hides all that low level PIC nonsense from you. (Cue volcanic eruption from Karen, if she were here...)

But at least the PIC did have a proper subroutine call / return mechanism, whereas all you can do on the SC/MP is swap the PC with one of the 16-bit pointers. That's the thing I really miss most on the SC/MP.

Back on topic, did Chris eventually just go for an 'analogue' method of power rail sequencing - a relay powered from the -5V to switch in the other rails?

[ChrisOddy]

I have just gone for a simple relay switched by the -5V rail - circuits nearly finished but the gardening and 86m of fence to erect are coming first whilst the weather is good !
Chris

[ortek_service]

Yes, the disassembly I first did on some PIC-code, that just to pulsed an O/P in a loop involved quite a bit of bank-switching to access the right locations in the initial setup. But a bit easier, if this is auto-generated for you by a PIC C-compiler library, although trying to follow disassembly of what a C-Compiler has produced is even more fun.
I remember that many said a PIC's architecture wasn't well-suited to efficient C-compiling (probably also the case for many uC's with few registers) / some early PIC C-Compilers weren't too good / rather buggy.

Coming to C relatively late, after BASIC, Assembly language & Modula-2 etc, I'm not a fan of it's rather cryptic syntax, that's not always obvious what is going on if you don't use it often - Most assembler can be much easier, with clear instructions. And there's now probably fewer people who now produce PC-Programs using C (where many of us started using it in the days of DOS - if not into Unix, before Linux, that this was mostly written-in), with many moving onto a least C++ if not C# and Java etc. - maybe Python+Qt etc

And even with uC's (particularly embedded systems), many C-compilers also support quite a bit of C++ (even if just the clear syntax options and not the Object-Oriented stuff). So can make using C (often referred to as more-portable assembler) a bit more easier to follow.


Back to Softy, I thought Chris was still just manually individually-switching the supplies at present, as a precaution, as hasn't quite finished a design to do it automatically.
It does seem that it's still a bit unproven that supply sequencing is required - particularly in view that only IC's on the main +5V rail have failed on Derek's one. And with so few others still out there, it's difficult to find out how reliable they've been long-term.

Multiple simultaneous IC failure does seem a bit suspicious, although I have found this not too long ago on several computer, that had been previously working OK until sudden failure, like with a Commodore VIC'20 & C64 etc (Although many MOS technology devices don't have a great long-term reliability record with later C16/+4 special. CPU version notoriously-bad)
And a ZX Spectrum 48K & 128K - where many of the DRAM's had failed, even though PSU rails were still OK / many single-rail ones had failed.

[SiriusHardware]

I have often described C as a 'write-only' language, that is, generally only readable, if at all, by the person who wrote it. .

When I (attempt) to write C I usually try to write it so it performs one clear action per line - even if it generates code four times as long it is worth it in terms of my being able to understand my own code at any time afterwards. I tend to prefer not to use 'C' shorthand, such as

variable++;

Preferring instead

variable=(variable+1);

If the 'C' compiler is any good it will generate the same code from either version, and I know which one is the more 'English' and easy to read.

One thing 'C' does have going for it is the use of clear markers for the beginning and end of code blocks, whereas Python uses the crazy method of indentation to define code blocks. It's incredibly easy to mess up the existing indentation when trying to insert another level of indentation - the code can end up so badly indented that no amount of panel beating will straighten it out again. Whenever I make a major change to working Python code I work on a copy of the working code, never the original code.

You may never get to the bottom of what took out all those devices in that original Softy - possibly misconnection of the PSU at some point in its history, you may unfortunately never know.

If I owned an original PSU and original Softy I might be tempted to insert a 'safe start' circuit between them on its own PCB - along the lines of what Chris is doing. I'd probably put high wattage 'just above rail voltage' zeners on each rail as well, in case of short circuit (input-to-output) failure of any of the linear regulators.

[ChrisOddy]

Here's my schematic for a SOFTY 1 power supply. Just finishing off the layout, the PCB will fit in an RS enclosure part no. 192-0783.
The mains PSU and XP power DC-DC switchers are avaliable from RS.

[SiriusHardware]

May I suggest a back-EMF suppression diode across the relay coil RL 1.1? You might be surprised by the size of the reverse-voltage 'kick' generated by a DC electromagnet when its power is removed and the magnetic field around it collapses.

[ChrisOddy]

This relay has one built-in !

[SiriusHardware]

I know some do, I did not know if that one did. Anyone who builds your design may just use any 5VDC relay they can lay hands on, so it may be worth pointing out in the build notes or on the diagram itself that the relay you have used / specified does have one.

[ChrisOddy]

Finished.

[Orakle42]

The "must release" voltage on the relay is only 0.5v, is this OK? ( in the worst case it means the -5v could be only -0.5 before the relay opens). The capacitors C8, C3 & C5 have yet to discharge....just a thought.

Orakle42

[ChrisOddy]

I've assumed that the significant problem is on initial power up which the relay should fix. When powering down there might be some energy remaining in capacitors but the risk of causing a problem is low especially as the +5V rail is heavily loaded by the SOFTY and the +12 and +26 rails will discharge quickly through R1/3 even if not being used ?