A2000 slot adapter problem with sparkling on picture

[bmthr]

Hi,

I build two A2000 slot rgb to hdmi adapters, and on both I have a problem with sparklings on edge of icons.

Look at the video with problem:
https://youtu.be/2YZ0NK6CTBE

What I miss on this boards?
I build classic RGB to HDMI adapters V1 and V2, and never get any problem.

Thank you!

[LinuxJedi]

I have a rev 6 coming to me today with the same problem on this card. I'm going to be trying the capacitor fix and the VHC IC fix over the next few days.

[bmthr]

I put now 47pF capacitor between pin 9 and pin 11 of header connector and picture is perfect. Didn't change any IC all IC's are LVC

[LinuxJedi]

Oh nice! @c0pperdragon was suggesting a 200pF I think, glad that works. I'll try the VHC as well since I already have some coming.

[bmthr]

I try with 220pF, but not work, totally loosing a picture, with 10pF still have sparklings and with 47pF picture is perfect and fully normal

[LinuxJedi]

Took a closer look at the Amiga 2000 I was given with this fault. Turns out that it has a 8373R3 which is a Super Denise. Setting the jumper correctly fixes the image. So it turns out I wasn't reproducing the problem :/

[LinuxJedi]

Another thing that might cause noise is that the spare inputs in U5 were left floating, this is usually bad in CMOS ICs.

[Bloodmosher]

For me the magic combination seems to be:

• LVC chips
• Add the 47pf cap as described above
• Settings Menu->Overclock CPU: 40
• Settings Menu->Overclock Core: 170
• Sampling Menu->Sync Edge: Trailing with +ve PixClk

With the above I see a perfect picture on even my most finicky display (Predator X27) using A2000, A3000, and both Denise variants.

I use a heat sink on the Pi and have let it run for many hours without issue (so far!), with core temp reaching 41.16C at the moment.

[bmthr]

Great informations

[LinuxJedi]

@Bloodmosher did you have any glitching (not sparkling) before the setting change?

[Bloodmosher]

@Bloodmosher did you have any glitching (not sparkling) before the setting change?

My main setup is an A2000 with an Eyoyo LCD, and to be honest I never noticed any issues on it (it is also pretty small so things are just harder to see). I recently acquired a dead A3000 and while fixing I dropped one of the boards in and it had noticeable problems immediately. This led to a long drawn out evaluation and scrutinizing of pixels and I eventually found a way to consistently see issues even on the A2000 (with the Predator X27 being the absolute "best" at finding issues for some reason). As for glitching versus sparkling - I would say more glitching on the a3000 for sure. Maybe the daughterboard just increases the potential here.

[LinuxJedi]

Something else to note, I've tried an A2000 rev 4.4 today with the board using both R3 and R4 Super Denise. I get a really noisy picture unless I jumper the board to Denise. I have no idea why yet.

[Bloodmosher]

Something else to note, I've tried an A2000 rev 4.4 today with the board using both R3 and R4 Super Denise. I get a really noisy picture unless I jumper the board to Denise. I have no idea why yet.

I've got a 4.1 that I test regularly with... for Super Denise probably only R3 though. Nothing super obvious on this page that describes revisions (http://amiga.resource.cx/mod/a2000.html). Other than 4.4 getting a change for "cost reduction". :-)

Is this with or without the 47pf cap and the software config changes described above?

[LinuxJedi]

I've got a 4.1 that I test regularly with... for Super Denise probably only R3 though. Nothing super obvious on this page that describes revisions (http://amiga.resource.cx/mod/a2000.html). Other than 4.4 getting a change for "cost reduction". :-)

Yea, I've just gone over the schematics with another engineer, we can't see an obvious reason for it. This one has been through some repairs, so maybe something related to that.

Is this with or without the 47pf cap and the software config changes described above?

Bare bones, no cap, factory software settings.

[Bloodmosher]

I've got a 4.1 that I test regularly with... for Super Denise probably only R3 though. Nothing super obvious on this page that describes revisions (http://amiga.resource.cx/mod/a2000.html). Other than 4.4 getting a change for "cost reduction". :-)

Yea, I've just gone over the schematics with another engineer, we can't see an obvious reason for it. This one has been through some repairs, so maybe something related to that.

Is this with or without the 47pf cap and the software config changes described above?

Bare bones, no cap, factory software settings.

Worth trying out the 47pf cap and config changes; all the ones I make I just do both...

[LinuxJedi]

Worth trying out the 47pf cap and config changes; all the ones I make I just do both...

Or I just leave the jumper on Denise :D

As for the 47pF I also have a new design of board which arrived today which has an optional place for that (amongst other changes).

[Bloodmosher]

Worth trying out the 47pf cap and config changes; all the ones I make I just do both...

Or I just leave the jumper on Denise :D

As for the 47pF I also have a new design of board which arrived today which has an optional place for that (amongst other changes).

Nice! From the other thread it sounded like you were fishing for stuff that might impact picture quality and/or allow use of some other chip variants. Looking forward to hearing about your experience with the changes.

But also - just try a stock board with the 47pf and config changes (and make sure you have a heat sink). Might give a good point of reference- that solved all issues I have seen.

[LinuxJedi]

For those interested, after some testing I have upstreamed my rev 2 design to hoglet67/RGBtoHDMI#197. Amongst other things it has a place (C6) to add the optional 47pF capacitor and makes U5 completely optional for Super Denise.

[LinuxJedi]

Photo of rev 1, rev 2 bottom.

[bmthr]

Nice work..

[mamejay]

I had the same issue and only had a 27pf cap. This has fixed my issues

[Bloodmosher]

Do you think these changes solve the need to change the software config described above? Or is that still necessary? I can probably build one and try on my favorite test setups to find out...

[mamejay]

Do you think these changes solve the need to change the software config described above? Or is that still necessary? I can probably build one and try on my favorite test setups to find out...

Worked for me without the need to modify any OS settings

[Bloodmosher]

Do you think these changes solve the need to change the software config described above? Or is that still necessary? I can probably build one and try on my favorite test setups to find out...

Worked for me without the need to modify any OS settings

Just to clarify- in your case are you using LinuxJedi's board + a 27pf cap? Or the original board + 27pf cap?

[bmthr]

I never modify software on my V1 boards, just add 47pF, and on latest builds only use 74VHCxx IC's. I test all my builded adapters on Amiga 2000 board rev 6.2

I'll order LinuxJedi's boards and test it... I'm really curious!

[Bloodmosher]

Photo of rev 1, rev 2 bottom.

If leaving out U5 it will get confusing to have these in circulation, as in the picture where you have it set for Super Denise but yet U5 is missing. What happens when you do that? :)

The parts are so cheap it seems better to just build for all possible machines. Same for C6- it doesn't hurt to add and most likely solves display issues on some setups.

[LinuxJedi]

If leaving out U5 it will get confusing to have these in circulation, as in the picture where you have it set for Super Denise but yet U5 is missing. What happens when you do that? :)

The parts are so cheap it seems better to just build for all possible machines. Same for C6- it doesn't hurt to add and most likely solves display issues on some setups.

There are two possible paths for the clock for Super Denise, this is different from the v1 design. The first is C1 XOR C3 in U1, the other is C1 XOR C3 then inverted twice using XORs in U5 (this emulates the original v1 board behaviour which was C1 XOR C3 then inverted in U5 and inverted again in U1). I added a U5 place to this board in case anyone wants to use it (maybe adding the extra couple of gates affects timing), but it is pretty much redundant. The 0ohm in J2 decides which path is taken. This is all documented in the README.md in my commit too.

This photo is one of a series of test boards I made, some with C6, some without, some with U5, some without. All to test various hardware configurations. Nothing wrong with adding the capacitor, it didn't have any affect on the machine I tested it on positive or negative. YMMV, this is why I made it an option.

Also in reference to the other question, stock software settings.

[Bloodmosher]

If leaving out U5 it will get confusing to have these in circulation, as in the picture where you have it set for Super Denise but yet U5 is missing. What happens when you do that? :)
The parts are so cheap it seems better to just build for all possible machines. Same for C6- it doesn't hurt to add and most likely solves display issues on some setups.

There are two possible paths for the clock for Super Denise, this is different from the v1 design. The first is C1 XOR C3 in U1, the other is C1 XOR C3 then inverted twice using XORs in U5 (this emulates the original v1 board behaviour which was C1 XOR C3 then inverted in U5 and inverted again in U1). I added a U5 place to this board in case anyone wants to use it (maybe adding the extra couple of gates affects timing), but it is pretty much redundant. The 0ohm in J2 decides which path is taken. This is all documented in the README.md in my commit too.

This photo is one of a series of test boards I made, some with C6, some without, some with U5, some without. All to test various hardware configurations. Nothing wrong with adding the capacitor, it didn't have any affect on the machine I tested it on positive or negative. YMMV, this is why I made it an option.

Also in reference to the other question, stock software settings.

Thanks for clarifying. I'll try one out on my most finicky setup and see if it still needs the software tweaks.