Pokeymax v4 bring up thread

[foft]

2 hours ago, woj said:

So what's next?

What shouldn't be next is getting over-confident and taking your newfound micro soldering skills to your Mac Mini🤪. Part of the reason I wanted to hand solder a pokeymax was to practice before doing the SSD upgrade on it. I ripped a bunch of pads, ooopsy. I wonder if a Mac mini case could be made to fit an EclaireXL in nicely? 🤣

 

https://forums.macrumors.com/threads/apple-silicon-soldered-ssd-upgrade-thread.2417822/post-33146571

[woj]

11 hours ago, foft said:

What shouldn't be next is getting over-confident and taking your newfound micro soldering skills to your Mac Mini🤪.

Ouch! I have a fair collection of destroyed equipment from my "experiments", but that tops it. This is the reason why I strongly refused my son's requests to do a Nintendo Switch mod of some sort that he said he would need to unlock something somewhere to do I don't know what, I did not even look it up to see what it's all about, just said that I am not confident enough, no money for a new Switch if I break it, and that I do not have a microscope (I still don't, I did the PokeyMaxes with a macro mode on the phone and a magnifying glass). He offered to buy me one, if I would only agree to do it. He is 8 years old...

 

In any case, a good YT channel to mostly understand one's limits and practice required, rather than anything else, is NorthridgeFix, I am sure you know it, but if not I highly recommend it. 

[flashjazzcat]

As I was reading that other thread, I was already thinking 'not enough heat'. I was hot-airing an encoder chip (just a DIL, obviously) off an A600 yesterday and I thought I'd try a lower heat setting and it was torture. Cranked it back up to 400 degrees and all was well. Mac boards - being multi-layered - sink an enormous amount of heat. I'm still surprised the pads came off, though; perhaps the first nand pads essentially ended up being slow cooked?

 

That's a tough lesson, anyway. But it hopefully only happens once.

 

PS: Dosdude is great. I love his videos - constantly surprising what he manages to accomplish (I nearly said 'what he pulls off' there, but thought better of it).

 

Edited May 19 by flashjazzcat

[woj]

The problem I have is that my heat gun is from the lower shelf and the temperature figure is only a rough indication. Having said that, the absolute minimum to do anything on a RoHS board is always 380 degrees C, but that only worked well for single 0603 parts. Going too high though is also a no no, everything starts to burn, parts are flying off, etc. 

[woj]

On 5/17/2024 at 9:53 PM, foft said:

There is some example software here:http://www.64kib.com/pokeymax_files/software/

I finally managed to listen to the RSI Megademo Wasteland track! It took a bit of fighting with the COVOX addressing and such, but I succeeded with NeoTracker 1.8. The Inertia Player from this ATR of your was not usuable - SDX said 'not a binary' and upon the attempt to extract the exe file Altirra said the file system on that ATR is damaged and refused to do anything about it. I also noticed that some MOD files are not loaded correctly by NeoTracker (most notably the Airwolf track shipped with Inertia). 

 

What I am now searching for is the Amiga RSI Megademo loading screen tune, I'd really like to hear how it sounds on an Atari. Problem is I do not know the name of it, and my searching so far has failed me (I found other tunes from the same demo, but not this one). This one: 

 

[woj]

So I also exhibited profound stupidity, I decided to correct the soldering on the first Pokey of the two resistor networks and the 0.1uF cap for the opamp with the ugly blob that I left. Mostly because of my OCD. That turned into a three hour ugly  battle using the iron and the heat gun. After the first correction I got a buzzing noise on the stereo channels and no reflowing or braid supported solder sucking was helping. Multimeter suggested a short between two legs of the resistor network (not knowing which one as they are connected 1-1), but also otherwise giving weird readings on the whole output circuit. In the end I had to take them off fully, to discover there was indeed a deeply hidden bridge under the 20k one, clean up, use some paste and solder on two new ones (still had left one of each) with a heat gun. Not surprisingly, I also lost one of the unconnected lonely pads, let it rest in peace whereever it is (probably somewhere waiting to make a surprise short circuit one day again). That was not the end, iron and solder was still needed after this, and I had to purposely build a bridge on the ground side of the 20k resistor network, because for some reason some of the pins were not getting contact. I finally won, and apologies for a long post, but I really had to vent this out.

 

But that's not all. I think this is the weakest link on the board and perhaps it could or should be fixed. You @foft had a problem there, I had a huge one (own stupidity triggered), and on my second Pokey the resistor networks were also dancing and required attention, not to mention that my dry run soldering also created a bridge exactly there. I am thinking either the soldering pad footprint is too narrow for these (pads sticking to the inside down under the chip, rather than the outside, the bridge formed under there and there was no way to take it out) and should be widened. Or, since only three out of four resitor network channels are used, go with three single 0402 resistors. Brief look at the board design suggests it might be possible.

[foft]

Pleased you finally won. Do you have some good flux, this makes such a difference to reworking?

 

We could certainly try a version with 12 components instead of these 4. i.e. 3x10K, 3x20K, 3x18nF and 3x560. Alternatively I could try fitting the next size up of the arrays.

 

I should have time later this week to have a play in Eagle.

[woj]

The flux is supposed to be good, but I do not have a comparison really. Part of the problem yesterday was me being really tired, it was supposed to be 5 minute hack, I was not prepared to sit through this for so long. Tired enough at some point not to notice that my multimeter was on a wrong setting giving me bullshit readings. But I just could not postpone it, had to get it fixed.

 

BTW, the capacitor network size and soldering is perfectly fine.

Edited May 21 by woj

[woj]

Both my computers are wired up for PokeyMax and working, and with a little bit of fiddling around with the U1MB plugin code, I also managed to squeeze in all the plugins I need into my other computer. So now I am left with 4 original Pokeys, several stereo boards, and also other Atari chips, essentially enough to build a NUC or XEL (I hope this is not going where I think it might be going).

 

Also, I think I got to the bottom of this:

 

On 5/17/2024 at 10:31 PM, woj said:

however, the non-official PokeyMax update tool did it, and verified, but both processes ended up with progress 101% and "Error reading file" message

 

It seems there is a bug (I reported it on GitHub) on the end of file check for DOSes other than MyDOS. 

[woj]

On 5/22/2024 at 9:29 AM, woj said:

It seems there is a bug (I reported it on GitHub) on the end of file check for DOSes other than MyDOS. 

 

So, with help of @MADRAFi this was resolved (thanks again!), however, the bugfixing loop we went through tripped over the fact that the core.bin file I produced on my own computer was damaged. Today I looked into this, and this:

 

dd if=${TYPE}.pofswap bs=1 skip=2216 count=32768 of=init2.bin
dd if=${TYPE}.pofswap bs=1 skip=118952 count=202752 of=core2.bin

 

is the source of the problem, the offset from which you copy the data cannot be hardcoded, in case of my POF file produced by Quartus 23.1 it is 3 bytes later (I am currently testing the exact offset). (Surely, now that I know what the problem is I can adjust the offsets to my Quartus version, and for you it works on your setup, and currently we are probably the only two people actually building the core, nevertheless it is good to know about it for the future ;)). 

[foft]

Oh good to know. I’ve upgraded a few times without that changing. I guess now you can see my inventive use of spare core flash memory 😀. I think perhaps there is a map file with offsets so it could theoretically look it up.

Edited May 26 by foft

[woj]

16 minutes ago, foft said:

Oh good to know. I’ve upgraded a few times without that changing. I guess now you can see my inventive use of spare core flash memory 😀. I think perhaps there is a map file with offsets so it could theoretically look it up.

So I am looking into this at this very moment, it seems all my POF files are shifted by 3 bytes because of the "std" in the version name of Quartus. This essentially messes up the whole build script, the modify_pof scripts also break as they update some SID data in place, but probably at a wrong offset. So probably even the POF file that I produced and programmed at some point is not good either...

 

EDIT: so yes, this is exactly it, all meaningful payload is shifted by 3 bytes because of "std" in the header. When I corrected modify_pof to seek at 3 bytes later, and modified makeflash to first create a temporary pof with 3 bytes skipped (even before swapbits) and then work on that I got a byte-to-byte identical core.bin as yours. I have not looked into the OpenOCD created files though. 

Edited May 26 by woj

[woj]

On 5/5/2024 at 5:08 PM, foft said:

First though I've been playing with the SPDIF output. I used a ~420ohm resistor and a diode to connect it to a DAC and that works.

I am at the stage of being ready to try this, now I found this, I thought I can just connect the SPDIF output directly to my digital amplifier, but clearly I should be careful, so how exactly does it go? Resistor and diode in series on the signal line? What diode and which direction? 

[foft]

It’s a 3.3v output and spdif expects more like 0.5v. So while your amp might be ok with 3.3v it’d be safer to drop it down.
 

I just used a random diode forward biased to get ~0.7v. I guess you could choose a special one with 0.5v drop.

 

I got my output pcbs btw. Toslink works fine using the plt133. Unfortunately I forgot the milling layer so can’t fit my 3.5mm socket, gah!

 

Anyway since I had to reorder I took the opportunity to design a stencil for reballing the fpga without taping off all the unused holes!

 

BTW if anyone is in or near Switzerland I’ll be showing the pokeymax v1-4 at Ouf Party this weekend. Along with something else very exciting!

[woj]

5 minutes ago, foft said:

It’s a 3.3v output and spdif expects more like 0.5v. So while your amp might be ok with 3.3v it’d be safer to drop it down.

I did my homework in the meantime, I was just not able to figure out if it is 3.3V or 2.5V. So essentially a voltage divider circuit like the one attached should do the trick, just I need to make sure the resistance values are OK to go from 3.3V down to 0.5-0.6V required by the standard (if I understood the wiki page correctly). This morning I found 1206 390ohm ones in the drawer, 220 and 100 barrels, and a bag of 0.1uF caps, so I should be able to glue something together and check with the oscilloscope that I got it right before I fry my amplifier. Next step is indeed a PCB, I need one to get this out and also to replace this stereo signal cable of mine that turned out sort of bulky...

 

There are also those 74HCT based circuits around for the TTL to Coax conversion, but for that I'd need power I guess, that would be a bit more of a contraption given that there are no power pins on the Pokey Max...

 

 

[woj]

So, that circuit that I quoted works just fine, I got hold of 300, 330, 360, and 390 resistors, the 390 one seemed to give the most up to the spec range, but honestly the spectrum on the oscilloscope was very spiky on the low/high transitions edges, so I was a bit unsure. Nevertheless, my receiver took it without complaints.

 

I did run into a probably non-issue, the mono detect was not working, but I have to check that again, because I forgot to check the settings after flashing the SPDIF core, and apparently those get reset, on top of this the U1MB plugin does its own thing, etc. etc. Needs one more less rushy session.

 

Sounds like time for designing my custom audio breakout board for all my inner cabling.

 

 

[foft]

I just ordered v2 of my toslink/audio output boards. I forgot the milling layer on v1, doh! Anyway they worked fine though so need some core fixes for spdif, I think something is clipping, overflowing or something like that.

 

I also found this interesting bga rework technique. Looks great for damaged soldermask cases but neat idea. No idea how much they cost though. https://www.instructables.com/Reworkng-a-BGA-Using-a-Stay-in-Place-Stencil/?amp_page=true

[woj]

1 hour ago, foft said:

I just ordered v2 of my toslink/audio output boards. I forgot the milling layer on v1, doh!

Pictures?

[woj]

10 hours ago, foft said:

I think something is clipping, overflowing or something like that.

I am too deaf to notice anything, to me it plays fine, I also do not have my ear trained for Pokey. Anyways, my mono detect issue was even simpler than I thought, the moment you play something stereo PokeyMax will go stereo and stay there until power down. Makes perfect sense, duh!

[foft]

It is meant to switch back to mono, if it doesn’t then I messed something up

[foft]

Just going to leave this here, finally it works, not perfect but getting better!

 

 

[Dinadan67]

Pokey, Antic, GTIA and Sally as FPGA? Wow, almost a full cyborg 🤖

[+scorpio_ny]

On 6/1/2024 at 3:35 PM, foft said:

Just going to leave this here, finally it works, not perfect but getting better!

 

 

This is fantastic @foft! This means there will be alternative options down the road for replacement hardware.

[foft]

10 minutes ago, scorpio_ny said:

This is fantastic @foft! This means there will be alternative options down the road for replacement hardware.

Just to be clear the antic is the fantastic work of @ijor as he describes here:

He kindly let me demo the FxAntic at the Ouf Party along with my other chip replacements, that I showed a few years back in 2018 at Silly Venture.

[+scorpio_ny]

7 hours ago, foft said:

Just to be clear the antic is the fantastic work of @ijor as he describes here:

He kindly let me demo the FxAntic at the Ouf Party along with my other chip replacements, that I showed a few years back in 2018 at Silly Venture.

Thank you very much for your hard work @ijor! It is appreciated!