10 years. GDEMU project is now 10 years old, although the idea itself (including early prototypes) is a bit older still. Is there going to be another version? I sure hope so but every time I think I’m finally going to sit and make a new prototype something happens. Capital S something.
Details are frankly not that interesting, the point is I now have more work and a lot of it is rather new to me so the last few months were challenging. At times it felt like RPG where every skill has to be unlocked first to be used but it’s getting better now. I might even try to sell some ODEs soon.
Speaking of ODEs – the few Wizards that I promised to people who ordered but didn’t get one are almost ready but I simply had no time to do any testing. I’ll get to that in a week or two. I will have some Rheas and Phoebes as well, these are almost done. However starting this year I’ll have to raise the prices – everything is much more expensive now. Electricity, parts, shipping, living in general. I’ll add 15 EUR to each price, that should mostly cover the gap for now. The Wizards on the waiting list will go at current price, it would be unfair to accept the order and then raise the price before delivery.
Being busy somehow made me stop playing games, except on some weekends, so for casual amusement I decided to code another emulator. I’m creating AGI (early Sierra adventure game interpreter) for FM Towns machines. It started as collection of Python scripts to decode the game resources but eventually I wanted to write some C and 386 assembly code so this is how far I got:
It’s actually a composite image of several lower-level routines run on PC VGA (because it’s easier to develop on PC), the logic interpreter is nowhere near complete enough to run the game this far. But I got this far so I’ll keep tinkering with it. I’ve already tested a lot of ideas on Towns hardware, including Tandy sound emulation via OPN2 FM synth – it’s nothing like the SN76489 but at least offers 3 voices for music, instead of 1 like with PC speaker.
Here’s something different for the Xmass / New Year holiday week. I might have mentioned this before – I like vacuum tube stuff, I’ve repaired some radios and B/W TVs, and now I’m building my own radios for fun. AM radio stations might be disappearing but I can always make a small local transmitter to fake some signal.
This is one of the prototypes I’ve been working on lately. It’s just the back end – audio frequency amplifier for the speaker, actually works pretty decently considering the small output transformer (fully rewound due to open primary) and battery-powered tubes 1S5T and 3S4T. The “T” stands for Tungsram, these are modified 1S5 and 3S4 that require only half the heating current (so 1/4 the power) with pretty much identical specs except the 3S4T which is some 20% weaker. The problem here is 1S5T that’s pretty much dead (down to 20% emissions) but still works (with perhaps higher distortions) due to it’s unusual operating point at about 20V of anode/g2 voltage, at which the current is very low anyway.
So I was looking for another 1S5T, cheap since I have some NOS ones but I wanted something for the experiments that I wouldn’t miss if it got accidently damaged or killed. Used, cheap tubes are usually duds but I don’t need much so I’m prepared for it, it’s still 3 or 4 pieces for the price of one NOS. One seller had some other cheap tubes in stock and it so happend that I wanted to buy some for other experiments – and you don’t say no to 5 vaccum tubes for 30$ shipping included. Especially old mid to late 1930 tubes.
As you can see these are well used and require some minor repairs:
Glass is loose in the bakelite socket, needs some glue – this is normal for these tubes at this age
Soldering of the tube wires to the contacts at the bottom is very oxidized and questionable, already had some issues so it’s usually best to just resolder everything (which is a PITA due to said oxidation, old solder won’t melt and new one doesn’t stick)
The outside coat of paint is not just for show, below the red paint is a layer of copper that acts as an EM shield, this is connected to one of the contacts via a wire at the very bottom of the glass part – this is where the paint usually cracks and separates, breaking the connection
The blackened EF6 (that’s not dirt, or if it is I can’t get it off without risking further paint damage) needs the shield reconnected, possibly in a way that other EF6 has been fixed. I’m planning on using some silver-based conductive glue under the extra wire layer to make it more reliable.
Well the 1S5T I got was more than tired, it’s no longer usable – that happens. Frankly these directly heated battery tubes don’t live very long due to limited amout of oxide material that can be put on the thin cathode wire. But the other tubes… First lets look more closely at some of them:
Pre-owned by Wehrmacht and… Luftwaffe perhaps? Funnily enough the clean EF6 is older, marked 10-1938. I’m not buying those for any sort of collection so I don’t much care for these marks but I was not expecting this. What I also wasn’t expecting is the results of electrical testing – I’ve built a small test setup with couple of DC PSUs and 4 digital meters to measure currents and transconductance, at characteristic point given in the datasheets. This provides much more information than a simple emission test that most tube testers do.
So the EF6 have 2.37mA (the Wehrmacht one) and 2.70mA of anode current of the expected 3.0mA when the tube was new. Clearly tired but far from dead. Transconductance S stands at 1.6mA/V and 1.7mA/V of the expected 1.8, that is very usable. The EH2 ones stand at 3.76mA and 3.40mA of expected 4.2mA but at reduced anode voltage (tested at 100V instead of 250V) so it should be better than that. S is 1.32mA/V and 1.37mA/V of expected 1.4 – that’s great! My method might not be entirely correct (S is measured by delta between two DC points instead of AC injection) but I’ve tested quite a few tubes already, NOS ones too, so I know this method produces results very close to datasheet values – so it can’t be too far off.
I guess I’m building another radio based on these tubes then, I’m going to need something like EBF2 for the second IF stage and detector diodes, plus EL3N for the output tube. I got a set of period-correct IF transfomer cans that were meant to work with these (or even earlier) tubes. Winter is a great time to work on such projects – it’s cold outside and the tubes are nice and warm…
Sorry for long absence, I was very busy. I have some Rheas for sale, I will accept orders Saturday 2023-10-28 at 21:00 CEST (9pm for the 24-hr impaired folks). Please note there will be daylight saving time change a few hours later (at 03:00 on 29th) so make sure you don’t confuse CEST with CET.
I should have a small batch of DocBrowns and perhaps a few Wizards ready soon as well. And speaking of FM Towns devices, I’ve run quite a few tests on my Fresh and noticed that even on Towns machines that are 486 equipped from factory (not CPU swapped) most games simply do not work properly with cache enabled. This is mostly due to certain library files (that handle everything – graphics, sound, input) being designed for much slower 386. Which means that gray tower Towns with 486DLC are even less likely to work properly without extensive per-game patching – and there’s different versions of these files so that would be a lot of work to find and work around all the problematic code.
Rhea orders are now open closed.
UPDATE: DocBrown orders will open on Saturday 2023-11-18 at 20:00 CET.
DocBrown orders are open closed. Once I compile the list (1-2 days) I will send out confirmation emails.
UPDATE 2: Wizard orders will open on Saturday 2023-12-02 at 20:00 CET. Usually I don’t ship things in December but this is a small batch, shouldn’t be too bad. Which also means there might be more orders than I have stock, in that case some might not be shipped until early 2024 (February most likely). We’ll see.
I have some DocBrowns almost ready so I will open orders this Saturday (2023-04-15) at 20:00 CEST.
All Phoebes were shipped but I got one Type-2 still here – usually I’d just leave it for the next batch but this one is already packed, somebody corrected their order to a different type and I don’t want to unwrap it. If anyone is interested – leave a comment.
Rheas are shipped so I will open orders for Wizards this Saturday (2023-02-04) at 20:00 CET.
Orders are now closed. Please give me a day or two to process it all and I will send out confirmation emails. Though I can already tell I have a problem, the stock I prepared was calculated based on the previous batch and it looks like this time there were more orders. I’ll let you know the details in the email but some of you might have to wait for your Wizard.
UPDATE: I should now have enough Wizards to cover all the orders – shipping will resume next week.
UPDATE 2: Phoebe orders will open this Saturday (2023-03-25) at 20:00 CET. This is also a smaller batch than usual so I might have to either delay or cancel some orders if there’s too many.
I will open orders for Rheas this Saturday (2023-01-07) at 20:00 CET.
In other news, remember this device?
No, this is not another photo of it. This is a screen capture of an emulator running on my PC. Why emulate a multimeter you might ask – well I did explain before that it has an option for external interface but I don’t have it, and adding it would require updating the FW which is a set of 4 old EPROM chips in a very fragile PCB. So I’d rather make a new PCB with new part and the I/O module integrated into it, and then swap the whole logic PCB. Since I don’t want to use 8080 with it’s weird 12V logic and support chips I’d have to switch to 8085 (I actually have one), Z80 (have several) or just do it on a modern ARM Cortex MCU, preferably one with FPU so that I can have HW support for floating-point values (not really necessary seeing how 2Mcyc/s 8080 can deal with it in software, if only just).
So I’ve decided to emulate the 8080 to see how much work it would be to try and port that emulation to ARM in case I didn’t want to run a completly new code but rather emulate old FW but on modern hardware. To speed things up I decided to borrow Z80 core I made for Makaron and dumb it down to support only 8080 instructions (for emulation speed and to make sure I don’t accidently run Z80 code). This also helped me understand the FW much better, I now know how the A/D conversion works and how all the calibration data is used to make sure the device has the accuracy it claims.
Now the funny thing is I found a couple of nasty bugs in the Z80 turned 8080 core, makes me wonder just how many NAOMI games didn’t work right because of them. I might back-port the fixes some day just to check that.
And one more interesting bit, remember that Varta battery this multimeter used to supply the calibration NVRAM when power was off? The one made in Western Germany? Well turns out it was the original battery from the factory, I’ve replaced it recently and here it is desoldered with date code now visible:
October 1988, and it still holds nominal charge. Although these usually have pretty flat voltage curve right until they die, so it was high time to swap it anyway.
Sorry for the long silence, I was very busy and after that I caught a pretty nasty stomach bug that made me miserable for over 2 weeks. Then I decided not to open orders because it would mean shipping in December and that’s a bad idea for small vendors like me. Too many things get lost or damaged in shipping because everyone rushes to buy gifts for Xmas. Next ordering window will therefore be open in January 2023.
If you emailed me in the last 2 months or so and I never got back to you, please email me again. I try to answer all mail but I don’t always have time to reply right away (or I need to check something) and, if I’m busy enough, I can then forget to reply at all.
Speaking of Xmas, a nearby business was cleaning their storage rooms and throwing junk away. So I scored an older PC, with monitor, for free. It was dirty and pretty yellowish but dirt can be cleaned and I don’t really mind the old plastic look. The important thing is nothing was cracked or broken, and the thing actually works too. Here’s a few photos (after the cleaning):
First the mobo, it’s a 286 system with Intel CPU (8 MHz, with 12 MHz turbo mode) with 1 MiB of RAM. I don’t see any way to expand the RAM other than EMS ISA card (which I don’t have and don’t really need). RAM chips are mostly 80ns so no point in trying to go 0WS on this board, it’s unlikely to be stable. There is a co-processor from AMD, and it’s rated for 10 MHz but the mobo has it clocked at 4.77 MHz and that can’t be changed without modding. NiCd battery did corrode and was removed but haven’t managed to do significant damage to the copper just yet – no repairs were needed.
This is a Hercules clone, a full-length ISA card by the way. Works fine.
MFM controller for HDD and floppy drives, a bit shorter then the Hercules. This PC has an extremly simple Award BIOS that doesn’t even support IDE HDDs. All you can select is one of the 40 predefined HDD profiles for MFM/RLL or use a SCSI card with it’s own BIOS.
Speaking of HDD, there is an ST-251 with a whooping 40 MiBs of magnetic estate. And not only it spins just fine, there are no sector defects – even the ones from the sticker ended up between sectors and are not affecting the operation. HDD surface was tested with SpinRite and, after some 12 hours, got a clean bill of health.
This is a pretty big 5.25″ drive so on the photo I’ve added a bit younger SCSI ST12550N for comparison – and by modern standards that one is also a big and heavy drive. BTW that SCSI HDD has SGI firmware in it and reports less available space than a standard ST12550N would. My guess is SGI wanted more reliability and opted for more hidden sectors for remapping any bad ones. Not that is has many of those but man, that disk is loud and requires good airflow or else it wil get rather hot. I’ve picked it for the photo because those two drives sure take their time to spin up (and down).
Here’s the monitor. It’s another monochrome CRT except amber this time, and the tube is pretty strong too. Obviously needed a full clean both outside and inside (I’m not afraid to remove the tube, it’s safe if you handle it properly) but it was mostly extremly dusty, no tobacco residue to clean this time. Which is nice for a change. This monitor is frankly even better then the Philips I’ve just shown, it can better correct for image sharpness in the corners and it accepts both 50Hz and 60Hz signals – though the green phosphor has longer glow time, which on one hand causes some ghosting in games but makes 50Hz text mode display flicker way less on the other, good for the eyes if you need to work with one.
Also this monitor can natively accept CGA RGBI signal, without needing a card with CGA-on-MDA emulation like what the ATI offers. Obviously there are no colors but shades (intensity steps) instead. Coupled with its ability to properly lock to CGA refresh rates (this monitor uses a dedicated chip for vertical deflection rather than a bunch of transistors) it gives you the option to use either MDA, Hercules or CGA card to drive it. Neat. And here is a sample of what the CGA colors-as-shades look like, I’d say it actually looks better then the terrible cyan/magenta palette):
I also scored some other free stuff, old SCSI HDD (78 MiB in 5.25″ package), 5.25″ double density floppy drive, some 3.5″ floppies, 15″ SVGA CRT monitor (tube is not virgin but still decent) and various other bits and pieces. If there is interest I can make some more photos. All this stuff means I don’t have to buy myself anything for Xmas, there’s already enough junk to play with as is 🙂
I’m almost done shipping Rheas, and Wizard testing is slowly coming along. In other news – I got notified that shipping to Australia is not possible at this time. Not sure why, I assume it’s COVID-19 related? Kinda forgot about that one but, as colder weather is slowly coming to northern hemisphere, we might all get a rude reminder very soon.
Speaking of colder wheather, I got myself a new space heater – obviously a used one, quite a lot too, but it was cheap. Didn’t need any repairs but I took it apart and cleaned it properly. Replaced one capacitor I found to be a bit tired. The rest didn’t look too bad so I kept them all original.
The hue of the green is off in the photos, in reality it’s not “radioactive green” but a darker shade (as in, more green than green-cyan). Eyes perceive it differently than camera sensor, I didn’t bother trying to rebalance it. Strictly speaking this is a monochrome monitor with TTL inputs for video and intensity signal – so at best you can have 4 brightness levels including black. Obviously a PC Hercules card in graphics mode can only do 2 levels, on or off, you can do “shades” by dithering pixels since the resolution is pretty high – 720×348.
That being said I have ATI Graphics Solution card that can drive a mono (MDA/Hercules) or color (CGA) monitor, and best of all, it can emulate CGA on mono monitor, and MDA on CGA monitor – including CGA composite mode except without making the pixels fuzzy like an NTSC TV would. Haven’t actually looked at the signals with a scope but I suspect they are modulated per pixel to get 8 different brightness levels. This does have some limitations but works in general. Spot the differences in “CGA” screenshots below:
Working on older electronics is fun. These days if something breaks you might as well throw it away, it’s just a bunch of chips barely visible to the naked eye, and potted/glued so can’t even be opened. I guess I’m getting old because, while I sure enjoy my CPUs and GPUs and super-duper HD flat panel displays, I still miss the old days…
I’m almost done with DocBrown queue – I will have a few units left so if you missed the ordering window last time but already have Marty and can provide the serial number, please write in comments. If you don’t have a Marty (yet) just wait for another ordering slot later this year.
I will open orders for Rheas next Saturday (2022-08-20) at 20:00 CEST. I have some Wizards too but those need to pass testing first, and that is a lot of work for me. I hope to have them ready soon.
Speaking of testing, every now and then I get asked about ODE for the desktop Towns models. Well, somehow I’ve managed to misplace the prototype PCB. I damaged some chips, had to replace them, eventually damaged the PCB too, then it got lost somewhere. I have no idea where I put it…
So I made another out of Wizard prototype board with some extra hardware and wires. It kinda works now but is not very stable so I’ll probably need to design and order a better PCB. This should help with signal integrity – frankly though the Fresh machine is the hardest yet to get the ODE working. Possibly because it’s original drive is x2 speed capable and the chips expect different timings. I still hope to resolve these issues with software tricks to keep the hardware simple and reliable. Here’s a photo of what the current prototype looks like:
UPDATE: Rhea orders are now open closed. Give me a few days to process the orders. As for DocBrowns, I did say I have some left but not many. If you didn’t get an email from me then you’ll need to wait for the next batch to order. I expect it will take at least 6-8 weeks to make and test a new batch.
DocBrowns are finally here, orders will open this Saturday (2022-05-14) at 20:00 CEST. While sending confirmation emails for Rheas I’ve noticed that my email provider is again being blocked/bullied by some sites, no doubt to force me to switch to more “reliable” service. At least this time I got notified of the block, and the offenders are hanmail.net and bellsouth.net. Keep that in mind if you use those because I ask for Marty serial numbers and if I’m blocked you will not be able to see and answer my request.
And now time for some random photos. I got a new (well, new to me) toy for some 50 EUR shipped:
It’s an M1T 380 multimeter from Metra. In theory it offers 6 and 1/2 digits in VOLTS/AMPS when using rather slow (~2s) HI.RES mode, in reality I might have to mod it to provide better temperature stability (noise doesn’t seem to be a problem, not as far as I can tell – I’d need a modern and calibrated meter of this class to compare with), but the good news is some people already tried that with good results.
As you can see it was made in Czechoslovakia, behind iron curtain. So while the date says December 1988, the inside resembles very late ’70 or early ’80 western electronics. Back in the day we had to work with what was available. And that became very apparent once the meter arrived – there was no power cord and it seems Czechs/Slovaks back then used their own standard of plugs called Typ 5813 – that thing is still made out of bakelite. Wasn’t easy to hunt that down and I didn’t want to replace it with modern IEC socket if I could help it – but I’ve managed to order two NOS plugs from Velmo.cz and I made my own cable.
This is a 380.0 model, meaning no external interfaces, something I intend to change. I’d like to have a modern USB, or at least RS232 port (especially since it would offer full control over the instrument). I have a manual and full schematics – it’s all in Czech (which I don’t speak, not to mention all the technical terms) but I get enough of it not to be completly lost.
Seems to be in working order, after some cleaning and fuse replacement. One of the tantalum capacitors died and shorted after a few power-ups, that can happen with the old “teardrop” types. It was a bit annoying the cap in question was a part of the input chopper amp module, it’s a separate black box (no, really, see photos below) on A/D converter board that I had to desolder to even get inside. I’ve seen this entire box being replaced with a single modern OpAmp, might just do that too eventually.
The digital part is based on 8080 clone. Yup, not even 8085 (or Z80, which would be already considered somewhat obsolete in ’88). It’s probably because it’s a clone and was manufactured locally in many eastern bloc states, so much easier to source. There’s 1024 bytes of static NMOS RAM, plus more expensive CMOS RAM (256×4) to store the calibration data. This NVRAM setup uses 3V lithium battery to provide power when the meter is off, and for that you need CMOS tech if the battery is to have any reasonable life time. And get this, the battery on the PCB was manufactured in Western Germany (so even accounting for shelf time, it was installed there sometime in mid-90’s) and still measures over 3V. Yes, it will be replaced eventually.
The ROM is 8KiB (4x 2716 EPROM) and pretty much full. I’ve built a cable to interface with this board and dump the ROM (and calibration data), I’m slowly progressing through the disassembly. I opted to use Z80 mnemonics since I don’t really speak original Intel 8080. And the code, to fit into this space, has some hacks in it, and some bugs too, which doesn’t make it any easier.
The idea is to add the external interface board based on original schematics of that add-on module. There is a dump of the 2KiB external ROM of extra 8080 code for the RS232 board but I’m not sure it will work properly with my FW. I’ve compared it to the v3.6 which is known to work and there some patches in places where the REMOTE function takes over, so that might be a clue I need a FW update first. And removing these 2716 chips is not something I’m looking forward to, these PCBs are extremly fragile. Super easy to rip vias or traces.
So perhaps I should create my own CPU board instead with the RS232 interface already on it, I could fit all of it by using Z80 in place of 8080 (with a few tweaks to the timing-critical code) and modern RAM/EEPROM (or Flash, or even just EPROM) chips. This also saves me from having to work around another Czech-specific socket/plug combo that is unobtainium (that green one near PCB edge) and have just one PCB and some wires to the plug on the back of the unit.
Or, and this is something I’m seriously considering now, do away with Z80, put a modern Cortex-M MCU there. Run either my own code (after I figure out the complete protocol to control A/D board), or the 8080 code in emulator, or a mix of both. I could probably fit all of that in 5V compatible M0 part + MAX232 for the serial port. Or perhaps use M4F with hardware FPU, a few more chips to translate 3V3 to 5V, and code a few more functions like sliding window filter for example.
So anyway, as you can see it works but somebody messed with calibration and it was way off (this should be showing 10.00000V). I’ve already re-calibrated the DC ranges for VOLTS and AMPS, as well as OHMS – and now I’m testing my new toy for temperature and short-term drifts.
UPDATE: DocBrown orders are closed. I will process the orders and send out confirmation emails in 1-2 days, though I can already tell there’s a lot. I probably don’t have enough stock at the moment. I will have more but that will take 2 months or so, I’m planning on doing Wizards next and that takes a lot of my time.
GDEMU 