What is the magic timing?

[RevivingRetro]

Maintainer here: I am hijacking the first post in this thread to put a note and avoid confusion.

The thread was originally about timing settings on some NTSC machines (in short: you might need to set lower timings such as 13/9 or 11/7 using either SKpicopatch before flashing the firmware, or SKpicotiming when the SKpico is already installed -- then everything works as it should).

Starting from August 3rd, the thread turned into "searching for a defect in a C128D" (the SKpico works well in those machines, as confirmed by extensive tests (see below) by @RevivingRetro, thanks!).

---

I have a dozen v2 DAC boards - all came assembled from PCBWay. I am using Amazon purchased Raspberry Pi Picos (vendor Vis Viva).
I have tried a multitude of different timing settings 13/9, 12/8, 6/6, 5/5 - using their own unique assemblies and picos and different C64 boards (250407 and 250425). While the 12/8 and 13/9 timings seem to fare much better than the default 15/11 - there is still sporadic no sound and access to the control panel via sys 54333. I have to reboot multiple times to get either to work.

Am I missing something here?

[SX64man]

#70 (reply in thread)

I found it can be a bit tricky to find the best setting on some C64s, especially if they are NTSC. If any of the machines have aftermarket non-MOS parts such as 15/25ns GAL or other PLAs or 6510/6526 substitutes, eeprom ROMS, new 70ns ram modules etc.., it gets even more challenging as the timing of these new hardware solutions is considerably faster than the original MOS parts. I have had decent luck on a few NTSC machines with 11/7/100, 13/9/100 but YMMV. I am reflashing f/w with the new values rather tha using the SKpicotiming util. _ I only use that to get it working and then reflash it with the same values. Probably overkill but...

[RevivingRetro]

Thank you for the reply. Interesting you had to use 100 for the padding. I used 20 and pass the harness test with no error.
I should have mentioned the machines are all NTSC.

I followed the thread you linked and it seems like its going to take a lot of experimenting. I don't want to give someone a machine with these types of issues though, so I'm going to try the simple 2040 and if that doesn't work out, I'll go back to the legacy.
These are only costing around $12usd in parts/build so it's not much an investment to walk away from. :)

[RevivingRetro]

@SX64man

Just FYI

All my machines tested are NTSC and all have OEM chips - none aftermarket.

I tried 11/7/20 and that seem to be more consistent (thanks for that config).
The 20 offset still works for me not showing the SID as bad in diag.

I'm now wondering if maybe different VIC-II versions might make a difference as well.
I'll have to check that out when I get a chance :)

[SX64man]

@RevivingRetro
Phi2 (the first parameter) is easily viewable on 6510/39 but not sure where to get the readbus value for the best troubleshooting of exactly what does the second parameter change. A typical scope views of phi2 (yellow), and a5 (purple) is below for two machines with working
good audio, menu, etc..
Unfortunately the non-working scope version is not really any different using those same inspection points.

[SX64man]

On passing the diags - I noticed something interesting. The diag always passes for me when the firmware is first loaded and the config is in default mono mode. However when I SYS in there and change it to my preferred stereo setup combo (see below) then that's where it no longer passes and I must reflash with the "100" value 3rd parameter to get the diag to pass again.

[SX64man]

This is the hardware setup I have been using on the 5 machines. Rather than getting a5/a8 directly from the 6510 - I get them at the closest vias carrying the signals. Here is a 250407 NTSC board. I have also tested getting a5/a8 at the 6510 and it made no difference in behaviour in terms of audio or menu corruption/lockout etc.

..

[RevivingRetro]

I modified my settings to yours (btw - notice the sporadic menu glitch that pops up in the middle of the screen when you hover over the POT setting) :)

I'm still using 20 as the offset and I still don't get the BAD U18 error - I tested it on 250407 and 250425
But I also see I'm using v2 DAC SIDKick (no need for the external DAC board) and I'm on v211

Using your timings of 11/7 seemed to have provided relief from the sporadic SYS54301 issues though

[SX64man]

Excellent good to hear! I think the difference with the diag is probably due to the long cables from the user port to the joystick ports on my harness. Looks like yours is a more local design at each port and probably has less loss to the joystick ports.

[SX64man]

I also did some more testing using the SKpicotiming.prg and it was interesting that for most of my NTSC machines, 11/7,12/8, or 13/9 seemed to be the sweet spot - because as soon as it got to 14/10, boom the dreaded unstable message popped right up. I think I will suggest to FRNTC's documentation that for NTSC the default might be safer as one of the 3 values above rather than 15/11. I'm curious if other NTSC users have the same experience.

[RevivingRetro]

I think the difference with the diag is probably due to the long cables from the user port to the joystick ports on my harness. Looks like yours is a more local design at each port and probably has less loss to the joystick ports.

That would seem to make sense! :)

[frntc]

Great that you figured out values that work. For the signals: Ax comes from the CPU and is typically less of an issue -- more interestind is the CS. And yes, in general the clock frequency is slightly higher for NTSC and thus smaller values are likely to work better. The ideal paddle value mostly depends on some capacitors on the C64 board, parts which do not get better over decades...

What sporadic glitch when hovering over pot-values do you mean? The "pico"-sprites move according to the paddle values, is it that?

[RevivingRetro]

What sporadic glitch when hovering over pot-values do you mean? The "pico"-sprites move according to the paddle values, is it that?

yes, the "pico"-sprites move according to the paddle values

and thank you for the clarification re: signal/capacitor values

BTW - Your SIDKick has been a great SID alternative. Loved it since the first iteration!
I have an immense amount of C64 and SX-64s (close to 200 units total). All restored/working. I'm planning to start selling them and wanted to provide customers an affordable option by choosing SIDKick over original SID. It would help reduce the cost by $20-$25. So getting this dialed in was pretty important. So far I've ordered 20 of each version from PCBWay. So, thank you for introducing the SIDKick!! :)

[SX64man]

Thanks for the explanation. Absolutely. great work! I'm going to do some scoping of CS and see if it will be possible to dial in the timing values for the individual machines. Of my 5 NTSC 250407s, 3 are using 11/7/100, 1 using 13/9/100 and the last one is a Bwack 407 reproduction mb and is using 10/6/100. I have had these 4 vintage machines repaired for a while but no SIDs to put in as the original ones had all died. So your solution is perfect, in particular because it can do stereo very easily.

On an unrelated subject, maybe should be on separate thread. Is it possible to tweak the 6581 hi pass filter so that less artifacting is present or is this a limitation of the hardware not being able to sample the higher frequencies cleanly. It's pretty apparent as a fuzziness sound in 6581 mode using SidTester filter test. In 8580 mode it seems better. So perhaps it's something buried in the original reSid algorithms. it's not a big deal but that is the reason I am using dual 8580s with digi in the config.
awesome work!

[SX64man]

Did a few more tests today - I wanted to test settings on a modern NTSC C64 (https://github.com/bwack/C64-250407-Replica-KiCad) board with a few mods and modern replacement ICs. I'm happy to report that this 11/7/100 setting produces good stereo audio and passes the diag tests on a modern mb with new caps, components, JCIAs, Gal PLA, TEBL modulator replacement customized for DAC (no internal audio), 6510 with Phiagra buffer, and motherboard taps for a5/a8.

[SX64man]

scope tests on differences between 11/7/100 and 13/9/100 don't appear significant to establish a clear difference in timing shift. Differences, if any, are very subtle.

Measured at SKPico SID 6 (phi2) and 8 (CS)