Amiga (Paula) vs Atari ST (Yamaha AY)

[Foebane]

Since we're on the subject of Paula/AY and samples, I have to ask:

Is it true that AY (the ST sound chip, nothing else) samples sound so fuzzy and hissy because they're actually only 4-bit (because of the 0-15 volume levels on said chip)? That just goes to show that whilst the AY can play samples, it was not really designed for them. I've heard similar hiss from POKEY samples, and I'm not sure, but doesn't the SID have 4-bit volume control? Correct me if I'm wrong.

[roondar]

This could be one reason yes. But there's multiple ways to play back samples, not all of them require changing the volume. It's also possible to use a square wave to play back samples (rapidly switching it on/off as needed) or use anything else that causes a 'click' (the famous old style SID samples made use of a 'bug' in the SID that caused an audible click when doing certain operations - Commodore fixed the bug in the newer SID chips and then those samples didn't work properly anymore). There's even a C64 sample method that uses the VIC-II to create the pulses needed IIRC. And I've even seen an Amiga example using the composite video output as an additional sample channel.

(I'd call the volume changing AM and the pulse changing FM, but this is probably just me making stuff up that sounds like it could be right :P)

I'm honestly not sure how the higher quality SID sample techniques work, though. Would be interesting to find out.

[chb]

Originally Posted by roondar

I'm honestly not sure how the higher quality SID sample techniques work, though. Would be interesting to find out.

There's quite a lot of information contained in this paper written by Mahoney:
https://livet.se/mahoney/c64-files/M...fvesson_v2.pdf

[junosix]

Here are some more examples of my Atari STE YM + DMA tracker that frank_b linked to: https://drive.google.com/drive/folde...PpptRadNFksUPV

It's 7 channels - the 3 YM channels, plus 4 channel 25kHz stereo multiplexed into 2 channels at 50kHz, meaning each of the 4 channels can occupy the full 8-bit amplitude rather than having to be shifted and combined. 3 of the sample channels are software-synth, based on the Konami SCC/PC Engine soundchips (there's a PDF of the manual in the link which explains things in more detail).

One of the problems with the STE is that the audio output stage isn't totally clean (almost all STEs suffer from CPU/bus whine on the left audio channel), and the volume of the YM and DMA chips aren't balanced (the Falcon cures both these issues and also has its own phenomenal sound capabilities). In my opinion the Amiga (with the low-pass filter disabled) always sounds cleaner for sample playback when playing the same samples. I love both computers and the styles of music they're each suited to, of course

[roondar]

I don't think anyone here will doubt that out of Atari ST(e), Amiga and Atari Falcon the Falcon is going to have the best sound. Commodore really ought to have included that DSP Dave Haynie wanted to add in the A3000+ (or at the very least ought to have offered a 16 bit version of Paula).

Ah well, I'm happy with what we have for most of these platforms. I kinda miss the times when every platform had its own 'sound'. Even though that is silly and sentimental of me to do

[Foebane]

Originally Posted by roondar

Ah well, I'm happy with what we have for most of these platforms. I kinda miss the times when every platform had its own 'sound'. Even though that is silly and sentimental of me to do

No it's not, don't be silly.

The same could be said for the graphics on the various home computers that came out in the 1980s that didn't 100% conform to the IBM PC standard, it was fun having platforms with their own distinct graphics and sound systems.

I remember having an argument in my college classroom against a die-hard PC nut (this was in the early 1990s) and defending the Amiga like a real zealot, and all the others did was scoff at, say, the price of the A4000 or how PCs did it better and cheaper. But I stood my ground. I only gave in to the PC craze in 1995 because Commodore had gone bust (thus, no more progress) and simply because PC games like Doom were promising a bright future (which they delivered: thankyouverymuch, FPS shooters of the late 1990s!)

I have moaned about the AY and SID, and praised Paula and POKEY, but that's a residue of the old schoolyard fights that most of us had over our systems, and the rivalries that remain to this day, despite emulation. I myself only met one other A8 user who I was happy to share games with, out of the entire school.

[pandy71]

Originally Posted by Foebane

Since we're on the subject of Paula/AY and samples, I have to ask:

Is it true that AY (the ST sound chip, nothing else) samples sound so fuzzy and hissy because they're actually only 4-bit (because of the 0-15 volume levels on said chip)? That just goes to show that whilst the AY can play samples, it was not really designed for them. I've heard similar hiss from POKEY samples, and I'm not sure, but doesn't the SID have 4-bit volume control? Correct me if I'm wrong.

Depends, signal volume/envelope control in AY is 4 bit (non linear - approx logarithmic), SID and POKEY are 4 bit linear AFAIR, ST YM is 5 bit (significant difference from AY). So noisy (high quantization noise) is unavoidable particularly for AY (where quantization scale is not linear). To get relatively high PCM quality both chips (SID and AY/YM) can exploit shape of generated wave (envelope) - linear ramp (sawtooth or triangle), SID count time, voltage linearly growing and at some point growing is interrupted but voltage stays - AFAIR this provide around 10 - 12 bit (internal DAC is 12 bit), AY *and ST YM) exploit different trick, they folding three linear ramps with around 3 bit each to form around 9 bit sample, also time consuming and critical.
Add to this sampling irregularity (phase error a.k.a. severe jitter).
So even HQ PCM reproduction is affected (this is easily audible, also software dependence produce some audible tones i assume associated to interrupts).
PCM playback on non DMA, non PCM architecture machines is always limited.

[Foebane]

OK, I'm confused now. What is the shorthand name for the ST sound chip? I gather I've been using the wrong name for years, now.

[Gorf]

Originally Posted by Foebane

OK, I'm confused now. What is the shorthand name for the ST sound chip? I gather I've been using the wrong name for years, now.

The shorthand name is "crap".


(sorry - I had to ....)

[idrougge]

Originally Posted by Foebane

OK, I'm confused now. What is the shorthand name for the ST sound chip? I gather I've been using the wrong name for years, now.

YM2149 or simply "YM".

People generally don't call the Yamaha FM chips "YM", but things like OPM, OPN, OPL etc.

[Foebane]

Originally Posted by idrougge

YM2149 or simply "YM".

People generally don't call the Yamaha FM chips "YM", but things like OPM, OPN, OPL etc.

Oh, great! I've been calling it AY!!! No wonder the confusion and people defending it!

[touko]

Originally Posted by gorf

the shorthand name is "crap".

:d
(sorry - i had to ....)

[pandy71]

Originally Posted by Foebane

Oh, great! I've been calling it AY!!! No wonder the confusion and people defending it!

Atari ST use YM2149 which is enhanced version of AY-3-8910 - so AY-3-8910 is worse than YM2149. Audio quality difference is meaningless and enhancements are mostly functional type.

Comparing with help of multiplatform games is is usually bad idea as in those times (but also today) game quality is limited by worst, common denominator platform i.e. 80's/90's by Atari ST/IBM PC (i.e. no HW acceleration offloading main CPU, architecture limitations etc.)

[Stefan Lindberg]

Originally Posted by Foebane

Yeah, when I said that the STE couldn't play back Amiga mods, I meant that it couldn't do so without modification. The STE DMA PCM audio chip is still TWO-channel, isn't it? So doesn't it have to do some processing to mix channels together for Amiga mod playback??

From what i learned so is the DMA of the STE only one channel... and then i mean "channel" as in as many different samples the sound system can play at the same time... like the Paula wich is a 4 channel soundchip and thus can play 4 different samples at the same time.
The STE only play one sample at the same time but the sample can be in stereo format, so to play 4-channel mods on an STE a lot of CPU power is needed just like on an ST... but you get better soundquality on the STE.

[rare_j]

Originally Posted by pandy71

Atari ST use YM2149 which is enhanced version of AY-3-8910 - so AY-3-8910 is worse than YM2149. Audio quality difference is meaningless and enhancements are mostly functional type.

So basically the same sound that a Speccy 128k produces. Charming in its own way for sure, I like a lot of Speccy tunes, but if I'd have upgraded from a Speccy to an ST to find it had exactly the same sound I would have been disappointed to say the least.

[Foebane]

Originally Posted by rare_j

So basically the same sound that a Speccy 128k produces. Charming in its own way for sure, I like a lot of Speccy tunes, but if I'd have upgraded from a Speccy to an ST to find it had exactly the same sound I would have been disappointed to say the least.

My Amiga-despising friend did just that.

[dlfrsilver]

Originally Posted by pandy71

Atari ST use YM2149 which is enhanced version of AY-3-8910 - so AY-3-8910 is worse than YM2149. Audio quality difference is meaningless and enhancements are mostly functional type.

Comparing with help of multiplatform games is is usually bad idea as in those times (but also today) game quality is limited by worst, common denominator platform i.e. 80's/90's by Atari ST/IBM PC (i.e. no HW acceleration offloading main CPU, architecture limitations etc.)

found on a great site :

"The General Instrument (later Microchip) AY-3-8910 and Yamaha YM2149F are very similar sound ICs. The YM2149F is essentally a Yamaha-branded AY-3-8910 compatible made under license."

[Bruce Abbott]

Originally Posted by dlfrsilver

found on a great site :

"The General Instrument (later Microchip) AY-3-8910 and Yamaha YM2149F are very similar sound ICs. The YM2149F is essentally a Yamaha-branded AY-3-8910 compatible made under license."

'Essentially' but with a few differences. The YM20149's clock frequency can be divided by 2 via a 'select' pin, and the D/A resolution was increased from 4 bits to 5 bits (so instead of 16 output levels there are 32).

To my ears at least the YM2149 sounds a little smoother than the AY, even when playing identical music (I tested both chips in my Aquarius Micro-expander, but have settled on the YM2149 because it sounds better).

Originally Posted by pandy71

signal volume/envelope control in AY is 4 bit (non linear - approx logarithmic)

The problem with the AY's D/A is not that it is non-linear, but that it's unsigned, so the steps are highly asymmetrical (the finest steps are scrunched up towards the bottom of the waveform, instead of in the center where they would be more useful). This could be corrected with an external circuit which provides polarity reversal with an extra bit, though it would probably be easier to just add a separate D/A converter.

I did some experiments using one of the AY-3-8910's I/O ports connected to an 8 bit D/A converter (initially just a resistor network like the Covox Speech Thing). Getting a stable playback frequency was tricky because the Aquarius has no interrupts or high frequency timer, and the CH376 USB chip is not buffered internally so I had to implement a software buffer with cycle-accurate timing.

[ Show youtube player ]

[pandy71]

Originally Posted by Bruce Abbott

The problem with the AY's D/A is not that it is non-linear, but that it's unsigned, so the steps are highly asymmetrical (the finest steps are scrunched up towards the bottom of the waveform, instead of in the center where they would be more useful). This could be corrected with an external circuit which provides polarity reversal with an extra bit, though it would probably be easier to just add a separate D/A converter.

Well... it looks to me quite OK - similar characteristics are common for example ulaw/Alaw https://en.wikipedia.org/wiki/%CE%9C-law_algorithm
Main difference is insufficient bit resolution (4 or 5) to cover wide dynamics range.

Originally Posted by Bruce Abbott

I did some experiments using one of the AY-3-8910's I/O ports connected to an 8 bit D/A converter (initially just a resistor network like the Covox Speech Thing). Getting a stable playback frequency was tricky because the Aquarius has no interrupts or high frequency timer, and the CH376 USB chip is not buffered internally so I had to implement a software buffer with cycle-accurate timing.

[ Show youtube player ]

Yep, FIFO and additional re-clocking (to form crude zero order hold re-sampler) should improve quality. Also it is recommended to use 1 value resistor and create second value by paralleling or by serial connection - on your video it looks like two different resistors values was used.
At some point decent results can be achieved. http://www.sonicillusions.co.uk/discrete_dac.htm

In fact i'm quite curious how AY sounds if supersonic noise or square wave is programmed and envelope is used.

[AmigaHope]

Originally Posted by pandy71

Well... it looks to me quite OK - similar characteristics are common for example ulaw/Alaw https://en.wikipedia.org/wiki/%CE%9C-law_algorithm
Main difference is insufficient bit resolution (4 or 5) to cover wide dynamics

What he's saying is that it's not nonlinear from the center of the waveform (i.e. signed, like a-law, mu-law, etc.), but rather that it's nonlinear from the bottom of the unsigned waveform, so that the most resolution is at one half of the extreme side of the waveform and the least is at the other extreme side of the waveform. It means you'd never be able to, for instance, make a good sine wave, instead it would have a nice sine shape on the bottom and a square shape on the top of the waveform.

Now I don't know if he's *correct* or not, but that's what I got from what he said.