Sunday, 6 April 2025

Building a 1581 Disk Drive Part 1

This is the first of several posts from my Patreon covering the build and testing of a 1581 clone, this first one from February 2024.

As part of another project I am working on, I have been testing on a variety of Commodore disk drives, and it occurred to me there was one I was missing.

I had tested 1540, 1541, 1541-II, 1570 and 1571, but I needed something that would sit between a 3½" disk and an IEC cable.

Commodore made such a thing, the 1581. Real ones of those are silly money these days. I have considered designing a replacement a few times, no doubt to be called a Mini 1581, but have always hit the problems of two out of production parts on the controller board, and the floppy drive itself.

The design is relatively simple, a 6502 processor, 8K RAM, 16K ROM, that part is easy, I can do that, but there is also an 8520 IO chip driving the IEC bus and a WD1770 disk controller.

The 8520 is out of production, and although it is a variant of the 6526, which is based on the 6522, and the 6522 is still available, the differences were too many to easily swap them without changing the firmware and making it incompatible. The WD1770 could be replaced by a microcontroller, possibly also including the 8520, but you still have the issue of a real disk drive to connect that lot to. These were standard models, all also long out of production.

I could replace the disk drive with an SD card, but then you pretty much have an SD2IEC anyway.

There are various 1581 clone designs floating around the internet. Someone who has built a couple of 1581 clones is Tim of Tim's Retro Corner. He produced two videos documenting the build and the steps required to get things working.

Tim is on my discord (see https://www.patreon.com/tynemouthsoftware), so I asked him how things were going with the 1581 drives, and which PCB design he would recommend.

He kindly offered to send me a set of the PCBs he used, and they arrived a few days later. The smaller PCB is an adapter board for PC drives, which I may or may not need.

Thanks Tim.

Thim.

This version is designed to fit into a 1581 case, or rather a 3D printed replica. I don't think at this point I am planning to do that, but we will see.

I started sorting out the parts required, which included a Commodore style power switch. I had ordered several versions of those when trying to find the best one for the Mini VIC.

The rocker on this red one was too big the VIC20 case, but will be fine here. Even if I go for a 3D printed case, I can just enlarge the switch cutout.

Mostly complete, the only bits I didn't have to hand were the ferrite beads and the EMI filters.

I did need to order a 4 pin DIN as although I am sure I have a bag of those somewhere, but couldn't find any.

Ah, that's not right. The ones Digi Key supplied seem to be this weird type with a different footprint.

Plan B.

One of the other drives I had tested was an Oceanic 118 / Excellerator Plus (repair post last month), and this had an interesting power connector.

Outwardly, it is a 5 pin DIN socket, shown here with the pin part of the plug installed so I could triple check the pinout (for reasons which will be clear if you have read the repair post).

However, the PCB appears to have been designed for a 4 pin DIN, the middle pin is missing.

You can see the pins are labelled 1,2,3,4, and not 1,4,2,5,3 as it would be on a 5 pin DIN.

I suspect they may have planned to use a 4 pin DIN, but changed to a 5 pin to avoid being too much of a blatant rip off of a 1541-II.

The standard sockets are arranged with compatible pinouts so, if you cut off the middle pin, you can solder a 5 pin DIN into a 4 pin DIN footprint.

So I did.

I added a link between the two OV pins, as only one was wired up on the PCB, not sure if that is the same as the 1581 would have had. The 0V rails on my power supply are not commoned, and are on separate wires, so I added a link on the back of the board to be sure.

That means I can use the same power cables I made for the Excellerator plus drive. (I quadruple checked the 5V and 12V rails were not transposed)

The other end was a 4 pin XLR that plugs into a 5V and 12V bench power supply I made for exactly this sort of thing. (looking back 13 years ago, I did take photos of the build but apparently didn't write up a blog post. Maybe a little late to do that now, but you never know. Watch this space.)

I didn't have any suitable 4 core cable to hand, so I braided my own. Works quite nicely.

Almost ready, just a couple of choices to make for some of the chips. The board is designed to take either a modern WDC W65C02S, or a vintage 6502. There doesn't seem to be any consideration given to the different drive levels and clock outputs on the W65C02, but all these boards seem to support both chips, so it must be OK?

It can also take a 6264 or 62256 RAM chip, although only 8K will be used with either chip.

The final choice is a WD1770 or WD1772. The later chip has extra features which are not used, but the chip can be used with a jumper change, as was done on later revisions of the 1581 itself.

I chose a WD1770, a W65C02S and a 6264, as those were the ones I had most easily available at the time.

There was one error with the PCB design to fix, the capacitor C5 is not wired up to the 0V rail, so it not doing much. I fitted it anyway so it didn't look odd.

Rather that wire that up, I went for wiring another 100nF capacitor directly to the 5V pin of the 8250 which it was meant to be decoupling the power supply for. The other end went to the nearest 0V point.

Almost time to fire it up, but I don't have a case, so I need some activity LEDs.

My first attempt was to use two surface mount LEDs.

It would have worked better with smaller LEDs, but the only ones I had to hand were 1206. 0805 would probably have been perfect.

A pair of standard 3mm LEDs worked a lot better.

Perfect, time to power on.

Red power LED, good sign. The green LED came on, and went off. Also a good sign.

Then it came back on, and then off and on and off again.

Then a gap.

Then three flashes again. And repeat.

Ah OK, this is an error signal, common on these drives. Time to check the manual.

Three flashes is bad RAM?

Odd, the RAM chip itself is fine, brand new and also tested externally, and I can't see any other problems. It also passes the zero page test every time, just failing on one of the other 31 pages.

I decided to try a 62256 instead, I soldered the blob jumper JP2. This connects 5V to pin 1 of the chip. On the 6264 this is NC, so I don't think it would be any problem to have 5V there all the time then you could fit either type of chip.

With the jumper set and the 62256 installed I tried again, but again got three flashes.

OK, time to test a theory. I had been worried before about the W65C02 option, with the timing and potentially the drive levels, maybe I should test an original 6502?

Again there are jumpers, but before I went up to the workshop to change yet more solder blobs, I thought I would double check if I could get away without changing them, just to test it out.

JP3 attaches 5V to pin 26. For the W65C02S, this is BE, Bus Enable, and needs to be pulled high (ideally with a 3K3 resistor, but direct 5V is OK I suppose). On the 6502 this is NC, so again I don't think it would be a problem to permanently tie it to 5V, specially if it were through a resistor.

The other jumper involved was JP1. This supplies 0V to pin 1. On the original 6502 this was a second ground pin. Frustratingly, the W65C02S has this an an output, Vector Pull, low when the interrupt vector is being read. Would have been nice if this had have been an input. Low to indicate "act like a real 6502", so it would act as a direct replacement when installed in a socket designed for a 6502. When pulled high, it could enable all the new features for systems designed to use them.

I tried a little test, I grabbed a few 6502 chips and stuck the meter into horseshoe mode and tested resistance between the two ground pins, pins 1 and 21, and on MOS chips, this was a few ohms. On Rockwell chips, it was about half an ohm.

Either way, I thought it would be good enough to test a 6502 with the jumpers set for a W6502S (but you couldn't do that the other way around, never plug in a W65C02S with pin 1 grounded, it is an output on that chip).

Powered on, and a single flash then off.

Sorted.

I guess that explains why all the photos I have seen of 1581 clones have used old 6502s and not the modern WDC chips.

I changed the jumpers anyway to be safe and also put back the 6264, and everything is still OK, single flash at power on.

OK, power on to a single flash of the green LED. Lets connect up the IEC cables and give it a go.

I am using a C64 with an Epyx Fastload Reloaded cartridge which has a DOS wedge, to make it easier.

Nice, that looks promising.

I started with a drive freshly extracted from a working Amiga 500. That should work fine, then I can work on finding a suitable drive to keep with it.

Time to try to format a disk.

If you have seen Tim's videos (and if not, why not? go and watch them after you have finished reading this), you will recognise this error.

74 Drive not ready. OK here we go.

I was hoping to have this working in a single post (and I may go back and edit it into a single post, depending on how long it takes - edit no I didn't).

So far I am seven drives in and the best I have had is one which started to format then failed.

I have tried Amiga, PC, A3000 and ST drives, hoping I would find one that just worked, either direct or with the PC adapter. But no luck.

Tim seemed to have good results with a Samsung SFD321B floppy drives, so I have ordered one of those (which should have been here last week....).

I have also dug out a very corroded A3000 board with a VL1772 on, in case makes any difference and if it has not lost too many pins.

Update from the future

The 1581 clone is now working well, with several further posts on Patreon if you don't want to wait until next week.

Adverts

All items on my Tindie store ship from the UK, so most of the world should be unaffected by recent tariff changes, however US orders may be subject to this and handling fees etc. Please be aware of this before ordering.

There are a selection of other repair and upgrade parts for various machines, many of which can be seen on this PET 2001 board.

There are also Minstrel 2 and Minstrel 3 kits and accessories available.

You can now get a Minstrel 2 kit for $200. Same as the 1980 pricing for the ZX80 kit on which it is based.

Or if you want ZX81 compatibility, you can get a Minstrel 3 kit for $200

Posted by at
Labels: , ,

Sunday, 30 March 2025

PET Composite Video Out - New Kit Version

I have a new kit listed on my Tindie store. PET Composite Video Out.

This replaces the previous version (left) which was only available pre-assembled.

With the new version, I wanted to simplify things a little. The old one had various jumper blocks to provide buffered or buffered and inverted versions of all the signals, to both the video output and a second output for a monitor. I don't think anyone ever used any of those.

It also had the "seemed like a good idea at the time" pre-made links on the jumper blocks, so it would work as is, or you could cut the track and fit the jumpers. When I used that on the jumpers on the Minstrel, lots of people skipped over that in the instructions, fitted the jumpers over the pre-made links, then wondered by the jumpers didn't work.

The new version bypasses all of that and contains only the bits that you need.

It is available in kit form if you want to built it yourself.

Its quite a simple through hole kit, so you should be OK if you know which end of the soldering iron to hold and have made sure you throw all of your lead-free solder in the bin before you start.

Or you can still order it assembled if you prefer.

It plugs into the monitor connector of a PET 2001 or 2001N.

Anything with the 9" monitor should work, 2001/2001N/3008/3016/3032/4016/4032.

It is not suitable for 40xx or 80xx PET/CBM models with 12" monitors and 6545/6845 CRTC chips. Those generate different frequencies and have different polarity signals. And no, it will not work with a SuperPET (since someone always asks with everything I do - maybe if someone could send me a SuperPET I could make things for it?)

These earlier PETs all used something very similar to NTSC (yes, even in the UK and other PAL territories).

16.625ms frames, give a 60.150Hz frame rate, close enough.

The vertical sync shown on the bottom trace, is the correct polarity, active low, and around the right pulse duration. That's a good start.

The video at the top is active low, so needs to be inverted so that the white pixels are high.

The horizontal sync in the middle is the wrong polarity, active high. This is easy enough to fix, but where things get tricky is the pulse starts too early, is way too long, and ends well into the visible video data.

Because of the timing of this horizontal sync pulse, the previous version had the video shifted slightly to the right on the screen. This was usually not a problem, and could usually be adjusted by the horizontal position control of your monitor if it bothered you.

It is only noticeable on my monitor if I turn the brightness right up on a screen full of text (which you can easily generate if you run the board without a CPU).

I have built various versions of PET Video Out over the years.

Most of which have been based on a vintage circuit from Commodore PET Users Club of England – Newsletter issues 1 and 2, page 9.

Note the version you normally see of this has this capacitor marked as 2200mf, this is incorrect, and should be 2200pf, or 2.2nf as we would normally write these days. I have corrected the version above.

That was my starting point, with only a minor change of adding the 100Ω resistor to drop the output level slightly.

To get around the issue with the horizontal sync pulse, an RC circuit (R1/C4) is used to generate a short pulse at the start of the PET's hsync pulse.

The bottom trace shows the rising edge at the start of the PET's horizontal sync pulse.

The middle trace shows the output of the RC circuit.

The top trace shows the horizontal sync pulse that is generated for the time the middle pulse is over the logic threshold. It this case, around 3µs.

This is then combined with the vertical sync by a NOR gate to generate an active low composite sync. This shows the start / end of the frame with the vertical sync pulse.

The capacitor / diode arrangement (C5/D1) turns this composite sync into negative pulses on the output. That is combined with the inverted video signal to give a composite video signal. It fits reasonably well to the expected -0.3V sync and +0.7V white level.

This new horizontal sync pulse is still too early, so it causes the video to be offset to the right slightly.

To fix that, I need to generate an initial delay pulse and use that to trigger the actual HSync pulse.

I wondered if I could do that without adding too many parts, but I only have a spare NOR gate?

I have used the spare NOR gate and a new RC circuit to create a the middle pulse in this capture, one that can be fed to the existing circuitry to generate a delayed horizontal sync pulse.

Well, that looks good to me, the HSync pulse is now delayed from the original and closer to the start of the video.

Ah no, that's too much. I need to change the resistor value.

Nope, still too much.

Ah, there we go. Perfect.

The delay pulse only turned out to be about 2µs, but it made an important difference to the signal. I also adjusted the other resistor in the second RC circuit, so the HSync pulse is now closer to the 4.7µs in the spec.

I have shown both versions of the circuit together so you can see the extra stage.

I have tried to get that all on one capture (cheating 5 channels out of a two channel scope).

This shows the signals progress from bottom to top.

  1. The PETs horizontal sync pulse too early and too long (17µs)
  2. The high-pass filtered version of A.
  3. The gated and inverted version of B (3µs delay).
  4. The high-pass filtered version of C.
  5. The gated and inverted version of D (4µs sync).

The horizontal sync that is actually used is the version marked E. This has been delayed and shortened with two RC filters and two inverters.

The final composite video output looks quite good. Since the border is black, there is no need for a separate back porch signal.

All sorted.

The board has it's own 5V regulator to reduce noise from the digital logic on the board. It runs from the unregulated 9V DC supply.

A good place to tap that is the polarity protection diode, a single large diode near the power connector. The end with the stripe is 9V DC.

You could also tap that elsewhere if it is more convenient. Anything from about 7V up to 25V should be fine.

The more I look at it, the more I can't help seeing the Servo Robot from Wheel in Space.....



Adverts

The PET Video out is available assembled or in kit form.

The price of the assembled version is the same as before, the kit version is cheaper.

There are a selection of other repair and upgrade parts for various machines listed on my Tindie store, many of which can be seen on this PET 2001 board.

Be the envy of all your friends with one of the limited edition run of blue PET Diagnostics modules.

Patreon

You can support me via Patreon, and get access to advance previews of development logs on new projects and behind the scenes updates. These are often in more detail than I can fit in here. This was actually three Patreon posts rolled into one, so a lot got cut out. This also includes access to my Patreon only Discord server for even more regular updates.

Posted by at
Labels: ,
Subscribe to: Posts (Atom)