14.57731 bit audio

[nikosidis]

Here are the outputs from my A1200 in PAL.
Everything in one file but in the right order. 6 tracks.
Please listen and say who you think are most clear or sound the best.

https://files.fm/u/v2u22h9wm

It is little doubt that the first track sound most noisy.

Nr. 2 is the most clean. That is piano-half

[Thorham]

Originally Posted by nikosidis

Remember this is PAL and not productivity outputs.

I know, I'm testing in PAL only.

Originally Posted by orangespider

This is weird. I did manage to get it to play noisy on my system, but the noise is gone every time I do a full shut down, wait 10s and start.

If by shutdown you mean turn the system off, then I can't do that without resetting because my hard disk spins up too slowly. Shouldn't make a difference, right?

Originally Posted by orangespider

Maybe your system has a different default channel delay?

Let's hope not.

Originally Posted by orangespider

If it's not same on all systems, then I won't get away without figuring out a way to sync the channels.

How would you even do that without somehow measuring the output?

[orangespider]

Originally Posted by Thorham

How would you even do that without somehow measuring the output?

Well if I can force each channel to start at the tick I want it to, it would do the trick...

Now that nikosidis has uploaded the output from his Amiga, do you get the same noise or more when you play it?

If it's the same noise, then it's not a sync issue.

[Thorham]

Originally Posted by orangespider

If it's the same noise, then it's not a sync issue.

It's the same.

Could you perhaps use a different test song? I find the piano music too loud. I've uploaded a short 4 mb WAV to the Zone. It's called piano.wav and is also piano music (FF10-2 intro from FLAC). It makes the noise easier to hear.

[orangespider]

Originally Posted by Thorham

It's the same.

Could you perhaps use a different test song? I find the piano music too loud. I've uploaded a short 4 mb WAV to the Zone. It's called piano.wav and is also piano music (FF10-2 intro from FLAC). It makes the noise easier to hear.

Can you check the productivity version too?
http://s000.tinyupload.com/index.php...75100994351510

I don't hear any noise on this, so I wonder how does this sound at your place. Also it is possible we are not using the same piano base because I used compressor on the sample I had to make it louder, so it added some noise to the source.

edit: Where is the song you prefer for testing? I will convert that one so that we use the same source for comparison.

edit2: If the noise the same then it is great! I was afraid you're hearing something that doesn't show up here. So I don't have to bother with syncing after all. If I hear it you hear it too. Best news today.

[Thorham]

Originally Posted by orangespider

Can you check the productivity version too?

Sure, however, are you using standard productivity with VGAOnly? I'm not (no VGAOnly, and tweaked monitor file), so this may affect things.

Originally Posted by orangespider

Also it is possible we are not using the same piano base because I used compressor on the sample I had to make it louder, so it added some noise to the source.

Right, that's why the after checking your piano music and my FF10-2 one sound similarly loud.

Originally Posted by orangespider

edit: Where is the song you prefer for testing? I will convert that one so that we use the same source for comparison.

It's in the Zone and it's called piano.wav. Should be at the top of the list.

Originally Posted by orangespider

Best news today.

[orangespider]

Originally Posted by Thorham

It's in the Zone and it's called piano.wav. Should be at the top of the list.

I don't have access to the zone.

The productivity version expects multiscanroductivity screen mode (58 Hz refresh 29.29 Khz horizontal, otherwise you'll get some weird artifacts.)

Maybe nikosidis can record his output if you can't play it back that way.

[nikosidis]

This is the recording from productivity.

https://files.fm/u/peehc47f8

Is not that to noisy compared to the old version?

[Thorham]

Originally Posted by orangespider

I don't have access to the zone.

To get access see here: https://eab.abime.net/showthread.php?t=26201

Originally Posted by orangespider

The productivity version expects multiscan: productivity screen mode (58 Hz refresh 29.29 Khz horizontal, otherwise you'll get some weird artifacts.)

Yes, but are you also using the VGAOnly program and an unmodified monitor file? Also, are you using the Workbench 3.0 or 3.1 versions? I can test all this, but I need to know what you're using exactly.

[orangespider]

Originally Posted by nikosidis

This is the recording from productivity.

https://files.fm/u/peehc47f8

Is not that to noisy compared to the old version?

That is VERY noisy. Not at all how it sounds here....

[orangespider]

Originally Posted by Thorham

To get access see here: https://eab.abime.net/showthread.php?t=26201

Ok, when changing to productivity mode, I managed to completely break my converter and now it outputs extreme noise. I did download the piano.wav, so once I have my code functional, I will convert and upload. No clue what is going on right now but I can't even convert the old files anymore.

[orangespider]

http://s000.tinyupload.com/index.php...61459445079120

Pal version of Thorham's piano song. It is quite a bit noisy but it is what it is.

[Thorham]

Originally Posted by orangespider

Pal version of Thorham's piano song. It is quite a bit noisy but it is what it is.

Yeah, it's still noisy. What's the current problem?

[orangespider]

Originally Posted by Thorham

Yeah, it's still noisy. What's the current problem?

If I knew, I'd probably fix it.

Apparently something is wrong with my assumptions.

This is what I am working under:
1) audio channels start at cycle 0/4/8/12 from turning on the DMA
2) if you sync the channels into any other order, they will snap to the closest when they run out of the buffer (when they would trigger the interrupt if it is on) *
3) the volume counter has a period of 64 cycles that gets reset when both samples are played in a sample pair (at audxper 124, that means volume is reset at 248)
4) the volume isn't actually volume instead it plays short bursts at full volume when the counter is less than the specified volume
5) the DAC gives linear voltage depending on the byte value of the current sample

One or more of these assumptions appears to be wrong, I just don't know which one.

The snap seems to be working like this:
step 1: sync channel 3 to start at 74 instead of 12 at audxper 124
step 2: the buffer plays, and when it reaches the end, channel 3 snaps to position 12 again, but is now 1 full sample behind channel 0

[nikosidis]

From what I can hear the PAL versions is not bad compared to the old versions. It is the productivity version that is far from what it was.

[orangespider]

WinUAE disagrees with my take on the audio delay. If WinUAE is correct and I am wrong, then that explains the noise.

However the output from WinUAE doesn't make any sense to me.

I sent #$7f80 looped to channels 0 and 1 and #$807f looped to channels 2 and 3. I was expecting a larger delay between 0 and 3, but the output from WinUAE didn't do this. In fact the phase shift seems to be based on the audxper value and seems to be audxper / 4 for every single frequency. Also I would expect this to happen:

127 + 127 = 254
127 - 128 = -1
-128 - 128 = -256
-128 + 128 = -1

Instead the only values I am getting at negative ones. Anyone actually knows what is going on here?

Hmm.. I changed to 80807f7f / 7f7f8080 and now I am getting the results I was expecting. Getting really confusing here.

[nikosidis]

Originally Posted by orangespider

WinUAE disagrees with my take on the audio delay. If WinUAE is correct and I am wrong, then that explains the noise.

However the output from WinUAE doesn't make any sense to me.

I sent #$7f80 looped to channels 0 and 1 and #$807f looped to channels 2 and 3. I was expecting a larger delay between 0 and 3, but the output from WinUAE didn't do this. In fact the phase shift seems to be based on the audxper value and seems to be audxper / 4 for every single frequency. Also I would expect this to happen:

127 + 127 = 254
127 - 128 = -1
-128 - 128 = -256
-128 + 128 = -1

Instead the only values I am getting at negative ones. Anyone actually knows what is going on here?

Hmm.. I changed to 80807f7f / 7f7f8080 and now I am getting the results I was expecting. Getting really confusing here.

As WinUAE is not the real thing I'm pretty sure you found some flaws with the emulator.

[orangespider]

Originally Posted by nikosidis

As WinUAE is not the real thing I'm pretty sure you found some flaws with the emulator.

Well, unless we do a -lot- of measurements on the real machine, I can only measure WinUAE. As far as WinUAE goes, it blows my theory about the phase shift out of the water. Here are the measurements I did at different audxper values:

audxper: 128
expected 0-3: 3060
measured 0-3: 1580.08976738675
expected 1-2: 1020
measured 1-2: 564.11304187565

measured cycles 0-3: 6.2
measured cycles 1-2: 2.2



audxper: 256
expected 0-3: 1530
measured 0-3: 854.035477209028
expected 1-2: 510
measured 1-2: 346.057057333923

measured cycles 0-3: 6.7
measured cycles 1-2: 2.7



audxper: 512
expected 0-3: 768
measured 0-3: 491.021935422117
expected 1-2: 255
measured 1-2: 237.02317975879

measured cycles 0-3: 7.7
measured cycles 1-3: 3.7



audxper: 1024
expected 0-3: 384
measured 0-3: 309.51157156823
expected 1-2: 127.5
measured 1-2: 182.508343069696

measured cycles 0-3: 9.7
measured cycles 1-2: 5.7

So, if the real machine works anything like WinUAE, that would explain the noise I am getting. However I still have no idea where those delays are coming from and why aren't they at fixed cycles. Anyone knows what is going on there?

edit: From these measurements, it would seem that the phase delay in cycles is:
channels 0-3: 5.7 + audxper / 256
channels 1-2: 1.7 + audxper / 256

At least in WinUAE. I have absolutely no explanation why this would be true though.

[Toni Wilen]

I can check (not necessarily soon enough!) if channels are in sync with a scope if you create test executable that continuously plays repeating sine wave (or some other repeating waveform)

If no one else is interested in doing it.

UAE (or any other emulator/whatever) probably is not a good test case because audio is not emulated in 3.58MHz space with volume counter.

[orangespider]

Originally Posted by Toni Wilen

I can check (not necessarily soon enough!) if channels are in sync with a scope if you create test executable that continuously plays repeating sine wave (or some other repeating waveform)

If no one else is interested in doing it.

UAE (or any other emulator/whatever) probably is not a good test case because audio is not emulated in 3.58MHz space with volume counter.

Ah, I thought you did a cycle exact emulation of it. Well then I guess I'll have to measure on a real Amiga. Also it means whatever I do won't sound well in WinUAE later on. Guess I will have to torture nikosidis or if you can make more precise measurements it would be even better.

In a perfect scenario we should measure audxper at 128/256/512 and 1024 to see if the phase shift changes at all depending on audxper (it does in WinUAE). I have a repeating pattern beep that is perfect for this. If you're up for it, I can send you the 4 test case .exe files.