Ferranti PC33

[moogway82]

Hello!

I'm new to this forum and I hope you don't mind me reaching out here. I'm helping my friend try to repair a Ferranti PC33 that appears to be very similar to the Ferranti PC31 which is shown in this forum thread from 2020:

https://www.vintage-radio.net/forum/showthread.php?t=165923

I'm trying to reverse engineer some bits of it to see if I can see where it's getting stuck - it's currently being halted (not READY) by the 8284 and I'm trying to figure out what signals control that using the 5150, Olivetti M24 (a 8086-based XT machine) schematics.

I would be really grateful if the user gridrunner who posted some photos on the old thread could add some more photos of their working board, esp just to check that the ULA chips and the PAL chips are in the right places. Also, if it was possible to scan more of the User Manual which is shown in that thread too would be such a big help.

We can find no information on the PC-33 anywhere and very little on the other Ferranti PCs of this era.

Many thanks in advance,
Cheers,
Chris.

[SiriusHardware]

Hello and welcome - I can't help you directly as gridrunner and others here probably can but during my last visit to the Manchester Museum Of Science and Industry some time ago I noticed that they had quite a strong affinity for all things Ferranti, no doubt because the company had such a large presence in the North West area.

I'v just had a brief scoot around on their website and they do have quite a bit of Ferranti related material including a PC31 (not PC33) - it may be worth contacting them to see if they hold any technical information - a service manual, for example, for your machine.

The Centre for Computing History (Computer Museum in Cambridge) also has a PC31 (Not PC33) in its collection and quite a lot of Ferranti related material so it may also be worth dropping a line to them to ask if they have any service info for the PC33. They have been very helpful to forum members here in the past.

[bobbyball]

I worked at Ferranti in Cheadle Heath at the time these PC's were current - we would have had one in our office, together with it's "EGA" monitor. Prior to being office based I did work in the calibration laboratory where these PC's were also serviced but didn't have much to do with that part. What Ferranti did do as a cost cutting measure, certainly in their earlier models (Advance 86 for example) is to not use gold plated contacts on the edge connectors. This caused umpteen problems needless to say. Not sure whether they "corrected" this in later models such as the PC33 but I suspect not as my colleagues who were servcing them did seem to have a lot of trouble.

[moogway82]

Hi SiriusHardware, thanks for the tips - I'd seen the CfCH page, it's what made me think of searching for PC31 stuff, which led me here, as the machines looked very similar from the outside at least - gridrunner's post showed me they were pretty much identical mainboards inside too! I will get in touch with them if I need more help as I've visited a couple of times, but it would be a bit tricky to visit at the moment. Didn't know about the Manchester Science Museum, good to see they have some bits too, will keep it in mind.

Hello, bobbyball, that's really interesting that you were actually there at Ferranti, it's sad that they folded like they did with that US arms company nonsense, what a shame for such a long running UK tech company... I don't think we have an issue with edge connectors on this machine - but the board is super weird - no solder mask, but traces mostly hidden in inner layers, very weird, makes it quite hard to reverse!

Thanks again for the feedback, this is a cool looking forum - I frequent stardot a bit, the Acorn computer forums and this looks like a nice community also, love these oldskool forums!

[SiriusHardware]

Yes, we are quite fond of stardot here too.

Gridrunner still seems to be around (last activity here a week or so ago) - you could try PMing them about this as they may be set up to be notified about PMs via email.

[ortek_service]

I wonder if one of those old Power-On Self-Test (POST) ISA-Cards may be useful?
- That outputted a failure code to a 2digit LED display and were quite-simple (Just 2 or 3 IC's / maybe a simple PAL device to do address-decoding).
You used to be able to pick these up quite cheap from China a few years ago, although probably no longer made by now, with ISA-slots long being consigned to history (as well as original PCI etc. replacements, with PCI-Express or even M2 now being the norm)
Some PC's also outputted the POST code on the Parallel-Port - Mainly Laptops, that lacked the ISA-slot?. And again, I've seen cheap POST cards for this.


That's assuming the BIOS firmware is running to start-with (maybe to initialise the 8284?) and the 8284 nRDY problem isn't preventing it from running.

I did wonder if this was hardware-compatible with other PC's, so you might be able to use the BIOS EPROM's from another similar PC, in case these were corrupt (maybe due to 'bit-rot' / charge-leakage over the years?) - I couldn't spot a checksum on their labels, and I presume there isn't an image of these to be found on the 'net so far? (In which case it wouldn't be a bad idea to read them out, if you have an suitable EPROM etc -Programmer and may be upload them somewhere - even here - (particular once they have been determined to be OK, and got it working) for posterity.

However, this PC is really slightly-odd / non-standard in that it is a mixture of 8 and 16bit, and falls between the original IBM PC (Model 5150) / XT (8bit model 5160 version - there was also an XT-286 model 5162) that used the 8088 @4.77MHz 8bit bus version of the 16bit bus 8086 and the IBM AT that used the enhanced instruction set 80286 @6MHz.
So it's more-like the long-running series of Amstrad x086 PC's that claimed to be '16bit', but couldn't run some software for other 16bit IBM's/Compatibles, that needed a 286.
I think HP etc. may have also done something similar.

It also only has 8bit slots, but that may be because some of the signals on the extra extension to the original 8bit slot required a 286 - In particular the upper address lines were only present on the 24bit (16MB) physical address space, compared to the 20bit (1MB) physical address space on the original 8086 & 8088.

I presume the original PSU is still OK / Power connectors are standard 6pin+6pin = 12pin 'AT' standard ? There is a Power Good signal on Pin1, that may be holding the Motherboard in reset, if not correct? And excess ripple on rails, due to failing capacitors could also be an issue + may need a minimum load for it to start-up properly, so might need an HDD / FDD's etc. attached (or a substitute load resistor, like AT's with the HDD required).

It may also be worth removing and reseating all the socket-ed IC's, in case a bad contact on one of these has developed over the years.

[moogway82]

Originally Posted by ortek_service

I wonder if one of those old Power-On Self-Test (POST) ISA-Cards may be useful?

My friend had been using it with a POST-card but it wasn't providing any extra information as the CPU is immediately halted - the "READY" signal from the 8284 is stuck LOW as its inputs are making it do so (nAEN1 & RDY1 not active). Also, I've had a quick look at the BIOS code in a dissasembler and I couldn't find any OUT 0x80 statements so I'm not sure it uses POST codes. An alternative XT BIOS is a good idea if we can get the CPU un-stuck...

Originally Posted by ortek_service

I presume the original PSU is still OK / Power connectors are standard 6pin+6pin = 12pin 'AT' standard ? There is a Power Good signal on Pin1, that may be holding the Motherboard in reset, if not correct?

It may also be worth removing and reseating all the socket-ed IC's, in case a bad contact on one of these has developed over the years.

My friend has been over the PSU's very carefully, he's a bit of a wizard at them, so they seem ok - good call on the Power Good signal, I must check that, however the system isn't stuck in RESET, that clears fine. He also dexo-it/reseated all the socketed chips, which I could do again, but I trust that's been OK.

Thanks very much for the suggestions, some really nice things to think about, just need to get some more time to look at it again, cheers!

I'll also try PM'ing, never like to do that first with a new forum as don't want to come across as some spammy weirdo (I'm just a regular weirdo )

[ortek_service]

Originally Posted by moogway82

>>
My friend had been using it with a POST-card but it wasn't providing any extra information as the CPU is immediately halted - the "READY" signal from the 8284 is stuck LOW as its inputs are making it do so (nAEN1 & RDY1 not active).
>>
>>

Unfortunately I've not had any previous dealings with the 8284.
So I've just been looking at some details on it: https://en.wikipedia.org/wiki/Intel_8284 (Plus the datasheet - attached here, for easier reference)

I hadn't realised that this was actually an oscillator / divider IC, so can't easily just remove (If it was socketed), if it was just a peripheral I/O IC, that wasn't needed for processor to boot, and force the READY output pin.

The datasheet isn't exactly too-clear about its READY logic, but it seems that both RDY1 & RDY2 are only used in 'Multiprocessor' systems.
- And I'm not sure if having a maths co-pro IC (Does the PC33 have this fitted) qualifies for that?
It does say to ground the nAENx pin to make it always active, if not using RDYx, but then (Test?) circuits are shown with nAEN1 tied-high (so inactive) and RDY1 not shown (So maybe then 'don't care' / that floats-high on TTL?).

It also seems that the state of the nASYNC pin determines whether both 'Ready' circuits are to be used (for synchronous, multi-processor circuits) or Asynchronous operation.So it may be worth seeing what logic-state is on that pin.
It may also be useful to do 'cold'(Unpowered) resistance measurements of all the pins on the 8284, to see if any are either permanently ground or to supply (although might be via a few k-ohms pull-up resistor). That would hopefully gives some more clues as to how it is being used in the PC33.

Not really also having delved too-deeply into the 8086 (or 8088), I have also discovered it also has a 'Minimum' and 'Maximum' (Used with 8087 etc Maths Co-Pro?) mode input pin. With some system block / connection diagrams (Although, unfortunately, these don't show where Rdy input(s) to the 8284 typically originate from), shown here: https://edurev.in/studytube/Minimum-and-Maximum-Mode-8086-Sy...a259c-5c41-4d19-a85c-8ddec2e7a9b7_t

And I've found this book, that does cover the 8284(A), as well as the 8086/8088 etc that might be useful, in understanding these systems: https://dn790005.ca.archive.org/0/items/TheIntelMicroprocess...Microprocessors%206th%20Edition.pdf
With section 8-5 around page 304-306 (PDF pages 318-320), having a lot more detail. It seems that READY output is an OR function of RDY1 & RDY2 circuits, so if either of these are valid, then you should get a READY output.

Figure 8-16 again depicts the internal structure of the 8284A. The bottom half of this dia¬
gram is the READY synchronization circuitry. At the leftmost side, the RDYl and AENl inputs
are ANDed, as are the RDY2 and AEN2 inputs. The outputs of the AND gates are then ORed to
generate the input to the one or two stages of synchronization. In order to obtain a logic 1 at the
inputs to the flip-flops, RDYl ANDed with AENl must be active or RDY2 ANDed with AEN2
must be active.

[ortek_service]

I also saw that in this example from the book, I previously mentioned, that RDY1 is derived from a wait-state circuit. And if all the links were missing / no DIP-Switches set, then it might never get a RDY1 signal. See attached diagram from page 306.

[ortek_service]

This thread contains quite a few links to schematics / suggestions of PC's that use the 8086: https://forum.vcfed.org/index.php?threads/8086-based-ibm-compatible-pc-schematics.76337/

I did see some info about (also having a place in Manchester) ICL Computers, maybe acquiring Ferranti Computing division. But although ICL produced 'Quattro' 8086 based computers, it seems they ran (Concurrent) CP/M(86) on these, and it wasn't until much later that they started producing (286 etc) PC-Compatible 'clones'.
However, like Ferranti, there's very-little info on these on the 'net so far.

[moogway82]

Hi Ortek_service,

Many thanks for all the information - yes the 8284 READY logic is a bit of a head scratcher - esp when you thinking in terms of creating a WAIT signal where everything in inverted! I think this Ferranti machine is fairly closely following the IBM 5150 in design and as such RDY1 is usually used by the DMA logic to request DMA control of the bus and 'halt' the CPU (not implemented using an actual HALT, just !READY and then externally switches off the buffers/transceivers). nAEN1 is then used by devices - and other sub-sytems - to insert Wait States, ie, DRAM timing, slow cards and chips and in this machines case - I think the 8/16bit Bus Convertor logic to give it time to do 2 bytes to make up a word transfer on the ISA bus. Last night I think I identified the signal that this uses to pass to the ULA to trigger nAEN1 to go LOW (it's been HIGH up till now), but I still need to map out more of that system to see why it's trying to add WAIT states when it shouldn't...

Interesting info on ICL taking on Ferranti - I looked it up but could only find information on a 1960's transfer of computing tech, which would all predate the personal computer revolution... I'm still sad that Ferranti folded, just been reading a little on their ULA technology and it was good stuff...

Cheers,
Chris.

[moogway82]

Originally Posted by ortek_service

This thread contains quite a few links to schematics / suggestions of PC's that use the 8086: https://forum.vcfed.org/index.php?threads/8086-based-ibm-compatible-pc-schematics.76337/

I missed this, great page I hadn't spotted on the VCFED forums, will be having a bit of a look over any of the 8086 based PC schematics - I've been using the Olivetti M24 as a main reference on how the 8-16bit Bus Conversion works, but more examples would be great!

[ortek_service]

Originally Posted by moogway82

Hi Ortek_service,

Many thanks for all the information - yes the 8284 READY logic is a bit of a head scratcher - esp when you thinking in terms of creating a WAIT signal where everything in inverted! I think this Ferranti machine is fairly closely following the IBM 5150 in design and as such RDY1 is usually used by the DMA logic to request DMA control of the bus and 'halt' the CPU (not implemented using an actual HALT, just !READY and then externally switches off the buffers/transceivers). nAEN1 is then used by devices - and other sub-sytems - to insert Wait States, ie, DRAM timing, slow cards and chips and in this machines case - I think the 8/16bit Bus Convertor logic to give it time to do 2 bytes to make up a word transfer on the ISA bus. Last night I think I identified the signal that this uses to pass to the ULA to trigger nAEN1 to go LOW (it's been HIGH up till now), but I still need to map out more of that system to see why it's trying to add WAIT states when it shouldn't...

Interesting info on ICL taking on Ferranti - I looked it up but could only find information on a 1960's transfer of computing tech, which would all predate the personal computer revolution... I'm still sad that Ferranti folded, just been reading a little on their ULA technology and it was good stuff...

Cheers,
Chris.

Yes, I think many of us first encountered Ferranti via their ULA's - most famously used by Sinclair in the ZX81 & Spectrum, but Acorn did also use a couple in the original Beeb (before switching to (CMOS? Cooler running) ones from VLSI, in later ones). Many of which have now been fully reverse-engineered, as there often wasn't much documentation released on these.

I hadn't realised that these Ferranti PC's also had ULA's in them (that it seems the 8086 Amstrad PC1512 also had), that makes schematics alone more-difficult to understand, without knowing the internal logic of the ULA(s). Although I do recall seeing a picture of a (Ferranti?) ISA video card used in one of their PC's that had at least four 40pin ULA's on it!

It does seem that the circuitry of an 8086-based PC is a bit more complicated than an 8088-based one, as the motherboard logic circuitry has to do some extra signals generation around the 16bit databus and 8bit ISA slots.

If the 8284 really does essentially OR (RDY1 & nAEN1) with (RDY2 & nAEN2), then I do wonder what the logic states are on RDY2 / nAEN2 inputs / what these connect to?

[MrSilenVid]

I also picked up a Ferranti PC-33 sometime in July last year. From a guy near the Lincoln area. Judging from the first photo, I believe your friend got his Ferranti from the same person.

I unfortunately have a similar issue with mine. The front LED stays lit orange, which means it hasn't posted and I get no video at all.

It would maybe be beneficial if we help each other out to fix these Ferranti machines?

Do let me know if you make any progress. I haven't taken a proper look yet to see exactly what the problem is on mine.

[moogway82]

Hi all,

Sorry for the lack of updates, but I have made some good progress in the meantime.

First we found the DMAC was faulty, it was asserting it's HRQ from boot without a DRQ signal present. So we swapped that and the RDY1 signal went high.

Then by enabling the 74F193 counter by pulling a signal to get the nLD=High, I identified that the counter, which provides the steps for the 8/16bit bus conversion process, was faulty, it's bit0 output was stuck high. We replaced this and the nAEN1 signal went low and the READY line was finally free to go High and we go proper activity from the CPU, finally we're running!

It still wasn't doing much, no beep or video but with a Logic Analyser on the CPU lines, I could see that the CPU was correctly going through bits of the BIOS code id disassembled, I could also see on those traces that bit7 of memory reads were always coming back high even when writes to the same addresses were low. Following up with oscilloscope on DRAM chip providing bit7 for those addresses (a 41256) I could see it always returned a high.

We then took Adrian's Digital basements friend's XTRAMTEST and split it into high/low bytes and my friend burned it on to 2 Eproms and as that will just blindly init all the basic chipset chips as soon as it boots without checking anything. It initialised video and I could test the RAM and basic chipset functionality. This was great news as I was worried the 8/16bit bus conversion that relies on PALs to work was broken, but it clearly works fine as CGA/MDA video would be a real stress of this. I think the original bios does more RAM tests first and fails over OR there may be some kind of service mode that the dip switches set as it does read the PPI very early on in the code and it was easy to check the ram on the LA as it was just in a loop writing a 1 to sequential bits of the 16bit databus to address 0x00000.

Current issue is the Timer is getting no clocks and I've traced that line to a mystery ULA (#23) up by the RAM. Timer CLKs should just be a halved PCLK from the 8284, usually just through a flip flop, so should be easy to replicate if the ULA is bust. Also I thought this ULA was generating RAS, CAS and Chip select lines but it's not. In fact apart from PLK in, Timer CLKs out, I can only find 1 more signal connected to anything else on the board! I've not had much time to poke more at it, but at present it looks like that ULA does precious little. I did get time to pull it out of the board, and the operation of the machine was the same as before - it could run the XTRAMTEST, just no Timer clocks, same errors/passes on the RAM chips - so it really isn't anything super critical.

I'd like to appeal again to gridrunner if they are about, I would love to see the user manual, and specifically the DIP switch settings as at the moment I'm assuming they are the same as the XT/5150 but there are some extra switches and id love to know what they do, it might help to free the original BIOS to get a bit further, I'd love to see it init video and say hello for the first time in decades...

I'm making a schematic with what I've found out so far, and I will publish that, and anything else I find out, like ROM dumps, to GitHub and let people know.

MrSilenVid: wow, that's very cool, thanks for reaching out. I'd love to see some pictures of your board, would be fascinating to know how similar/different it is and what issues you find - I wonder if you have working Timer (8253) CLK inputs?

Originally Posted by ortek_service

Ferranti PC's also had ULA's in them

By the looks of this board, Ferranti were drowning in their ULAs - I'm surprised they weren't installed in everything back in the day! I'm really interested to get to the bottom of this new mystery ULA (ends in #23) that seems to do nothing so far but halve the PCLK for the Timer chip which just needs a simple flip-flop - seems a bit of a waste of a 40-pin ULA chip! Only reason I can see to do that was if you had a massive stash just sitting there! Perhaps this chip did do something on previous models (tape input for the Advance86 for example?) and they only needed the PIT clk generation. But still seems like waste of PCB space at the least... I will keep poking at it as I'm very keen to know if this is the case..

Originally Posted by ortek_service

I do wonder what the logic states are on RDY2 / nAEN2 inputs / what these connect to?

RDY2/nAEN2 are just pulled to VCC/GND as per the 5150 schematics. In essence this is just a boring XT clone, but I think it's a bit more than that just because it's one of the few 8086 based machine with more discrete logic and it's a bit quirky (weird brown PCB, seems to use a PCB layer as a solder mask!) UK made thing by one of the biggest providers to the UK home micro scene...

Cheers,
Chris.

[ortek_service]

Thanks for the update - You've made a lot of good progress.

I wonder if a photo of the DIP-switch settings on the one MrSilenVid (or a Museum) has, may be helpful, if your's has been changed at some point and original settings forgotten?

Although, it would still be useful to know exactly what all the switches did!
- I recall once getting an IBM PC to work, by just trying various combinations and discovering you had to have the right setting for the installed memory for it to work! (It was only later, that I acquired a Babani book by Robert Penfold all about PC's internals, that detailed the settings and this was all in the mid 1990's before Internet was common / much content on there.)

[moogway82]

Originally Posted by ortek_service

Thanks for the update - You've made a lot of good progress.

I wonder if a photo of the DIP-switch settings on the one MrSilenVid (or a Museum) has, may be helpful, if your's has been changed at some point and original settings forgotten?

Although, it would still be useful to know exactly what all the switches did!
- I recall once getting an IBM PC to work, by just trying various combinations and discovering you had to have the right setting for the installed memory for it to work! (It was only later, that I acquired a Babani book by Robert Penfold all about PC's internals, that detailed the settings and this was all in the mid 1990's before Internet was common / much content on there.)

Thanks, I was really chuffed to get a video signal out of it at least. I'll exchange notes with MrSilenVid for sure. And yeah a bit of trial and error DIP Switch flicking might be in order!

[SiriusHardware]

Did you try PMing gridrunner? We have a saying on Tyneside 'Shy Bairns Get Nowt'.

Anyone who reads through the thread above can see you are seriously committed and well enough equipped to fix the machine, but it may just be that GR doesn't visit very often or read this section very often. Good work so far - I have to admit I dread coming across a ULA in anything as it can be a show stopper if you can't get another one and you can't replicate its original function.

That's why old machines like PETs are so lovely to work on - there are no custom ICs in them except the ROMs, and they are replicable / easy to substitute even now.

[moogway82]

Originally Posted by SiriusHardware

Did you try PMing gridrunner? We have a saying on Tyneside 'Shy Bairns Get Nowt'.
...
Good work so far - I have to admit I dread coming across a ULA in anything as it can be a show stopper if you can't get another one and you can't replicate its original function.

That's why old machines like PETs are so lovely to work on - there are no custom ICs in them except the ROMs, and they are replicable / easy to substitute even now.

Hia, yup I think youve sussed me out, I'm definately a bit of a shy bairn . Ta for the prod, Ive reached out via PM now. Don't want to just jump into a new community and start PM'ing asking for stuff, just bad show if you ask me - but happy to take a hint!

Yeah I'm a bit wary of the ULAs too, however I did manage to make my own Acorn Electron ULA from an FPGA (with a lot of help) so it is possible to replicate if needed. Also the PC/XT operations are very well understood - I'm pretty confident that ULA #22 is pretty much just Sheet 2 of the IBM 5150 made into a single chip - definitely replicable in a CPLD/FPGA. It's just killing me that for ULA #23, I can find a total of 3 signals joined to it! A 40pin ULA for just 3 signals is mad! Got a busy few days coming up so not going to get time to work on this, shame as I really want to try creating my own Timer Clock signal to get DRAM refresh going, swap the bad RAM chip and try the original BIOS again... patience, patience...

[SiriusHardware]

I did manage to make my own Acorn Electron ULA from an FPGA (with a lot of help) so it is possible to replicate if needed. Also the PC/XT operations are very well understood - I'm pretty confident that ULA #22 is pretty much just Sheet 2 of the IBM 5150 made into a single chip - definitely replicable in a CPLD/FPGA.

That's quite an impressive level to be at, not many people can do that sort of thing (Forum member 'Realtime' springs to mind).

What do we have to do to keep you here?