Queries on AGA Copper and 14-bit Audio

[paul1981]

I wouldn't get to obsessive with this. Signal to noise ratio does not tell you anything about the quality of a signal anyway, it only tells you the amount of noise. I suppose the real aim here is to try and ascertain whether there can be 14-bits of resolution. Whether your PC sound card is up to such a measurement I seriously doubt.

[paul1981]

Regarding it being unfair to compare Paula 14-bit at 44.1KHz with a cd, well it does a pretty good job I say. If you listen to it you'd agree too. No, it's not as good, due to NO oversampling, but it's bloody good - trust me.

[pyksy]

Originally Posted by pandy71

It is not so different - different is in Amiga 1000 - there is high order lowpass filter there. Every generation of Amiga has simplified filtering.



I have test signals already - generated by AP software for 16 bit 32000 then resampled to 28604 in SoX - there are 3 sets of signals, 8 bit, 16 bit and 14 bit on 16 bit (16 bit audio files, volume reduced by 4 then increased by 4). All files are stereo. 8 bit are 8svx, remain are wav. In total there is over 150MB - need to check how to compress them lossless and need to upload somewhere so anyone interested can download them and play on Amiga - I assume that PC with free software may be used as audio analyzer.

And you can compare it as we comparing analog signal.
Listening tests... anyone may propose own method for subjective testing - if you have ideas on this please share it - thank you.

The filters in a1200 and a4000 do differ from older Amiga models. See https://bel.fi/alankila/modguide/interpolate.txt for more information.

Sent from my LG-H815 using Tapatalk

[TroyWilkins]

I'm sure there are people here with the appropriate equipment to do this, so how about someone connect their Amigas sound output directly to a recording device, perhaps the mic in on their PCs sound card, and record the Amigas sound output while playing back various sounds, using different sound output configurations?

Then these sounds could be examined in a sound editor, and we could all have a listen to the captured sound ourselves?

Heck, we could even be really scientific about it, and have the captured sounds not labelled as to what they are, and see if people can tell by ear, this way we could even use the original sound at say 16-bit 44.1 kHz as a control? That sound converted to 8-bit 22050 Hz, converted to say 14-bit 22050 Hz, and the original simply converted to 16-bit 22050 Hz. That gives us 4 samples to listen to and compare, which is not an unreasonable amount in my opinion, and if it's short, say 15 seconds of audio, it should be enough to make a fair comparison.

Of course, the original sound would need some low, medium and high frequency sounds, as well as some time with no sound being played, but if the original sound sample is picked well, this could really sort out the argument as much as possible. Sure, our opinions of the sounds would be subjective, but loading them into a sound editor would enable direct comparisons, would it not?

Heck, I could even do this with my A600 and my crappy PC, although I'd have to use the onboard sound on the 970A-D3P, but it would provide at least something we could draw conclusions from. I don't have the appropriate stereo RCA to stereo 3.5mm cable, but I could easily order one from ebay.

So, what do we think of this idea? Any merit, or am I barking up the wrong tree?

[pandy71]

Originally Posted by paul1981

I wouldn't get to obsessive with this. Signal to noise ratio does not tell you anything about the quality of a signal anyway, it only tells you the amount of noise. I suppose the real aim here is to try and ascertain whether there can be 14-bits of resolution. Whether your PC sound card is up to such a measurement I seriously doubt.

There is nothing obsessive at least not on my side - curiosity and rationality.
I like Amiga, music but i reject all audiophile hype and those magical HDMI cables that dramatically improve picture quality.
And if Amiga is 8 bit but capable to do 14 bit why then precisely described PC audio input with 2 - 10 bits more accurate electronics can't be used for perform measurements?
Most modern audio PC cards have more than 80...90dB quality - as engineer i think it is sufficient to perform such tests in home.
My goal is to use state of the art audio analyzer and do this in a way accurate as possible but anyone may do same things in home.
And distortions are always distortions even if some people likes them more.
Knowing that average Amiga has something like 65.43dB 'quality' will suddenly not kill pleasure of listening music and game fx on Amiga?
I really miss 8364R4 or R5 datasheet with such things specified clearly by Commodore.

Originally Posted by paul1981

Regarding it being unfair to compare Paula 14-bit at 44.1KHz with a cd, well it does a pretty good job I say. If you listen to it you'd agree too. No, it's not as good, due to NO oversampling, but it's bloody good - trust me.

And that's why i don't understand this rejection of idea to test - for me knowing that it is not 86 but perhaps 71 or 64dB will not kill any pleasure from Amiga sound.
And there is large group of people preferring non oversampled CD - they believe that oversampling killing CD audio... so called NOS DAC... they wish to pay hundreds and thousands of Euro for sometimes very ubiquitous constructions. They even building own math theory http://www.sakurasystems.com/article...pling-DAC.html

Originally Posted by pyksy

The filters in a1200 and a4000 do differ from older Amiga models. See https://bel.fi/alankila/modguide/interpolate.txt for more information.

Thx, i know this and i use also this http://www.amigawiki.de/doku.php?id=...ice:schematics so overall circuitry can be modeled in SPICE.

Originally Posted by TroyWilkins

I'm sure there are people here with the appropriate equipment to do this, so how about someone connect their Amigas sound output directly to a recording device, perhaps the mic in on their PCs sound card, and record the Amigas sound output while playing back various sounds, using different sound output configurations?

Nope, MIC input in PC is very bad idea:
- it is MIC input - very high gain means lot of noise,
- frequently is not linear but has non linear characteristic to deal with wide range of MIC signals (there is dynamic compressor there)
- it is MONO input
- it provide power supply for microphones as such it may have phantom voltage on pins - it should not kill Amiga but anyway will disturb quality.

You can use for audio measurement any LINE Input on PC - it should be quite OK. If you dob in Audio card quality you can use external player source (CD, DVD or decent portable player from reputable vendor with defined parameters).

Originally Posted by TroyWilkins

Then these sounds could be examined in a sound editor, and we could all have a listen to the captured sound ourselves?

There is dedicated software for this:
http://audio.rightmark.org/index_new.shtml
https://www.trueaudio.com/index.htm
And other very similar - not a problem.

Originally Posted by TroyWilkins

Heck, we could even be really scientific about it, and have the captured sounds not labelled as to what they are, and see if people can tell by ear, this way we could even use the original sound at say 16-bit 44.1 kHz as a control? That sound converted to 8-bit 22050 Hz, converted to say 14-bit 22050 Hz, and the original simply converted to 16-bit 22050 Hz. That gives us 4 samples to listen to and compare, which is not an unreasonable amount in my opinion, and if it's short, say 15 seconds of audio, it should be enough to make a fair comparison.

So, what do we think of this idea? Any merit, or am I barking up the wrong tree?[/QUOTE]

https://en.wikipedia.org/wiki/Codec_listening_test
http://www.head-fi.org/t/486598/test...aims-and-myths

I will try to do objective part (but need to wait for V500 or revive my A3000).

[meynaf]

Aside of pure academic purposes, where's the point in measuring sound quality when the ear can't actually hear the difference ?

[TroyWilkins]

Originally Posted by pandy71

Nope, MIC input in PC is very bad idea:
- it is MIC input - very high gain means lot of noise,
- frequently is not linear but has non linear characteristic to deal with wide range of MIC signals (there is dynamic compressor there)
- it is MONO input
- it provide power supply for microphones as such it may have phantom voltage on pins - it should not kill Amiga but anyway will disturb quality.

You can use for audio measurement any LINE Input on PC - it should be quite OK. If you dob in Audio card quality you can use external player source (CD, DVD or decent portable player from reputable vendor with defined parameters).




There is dedicated software for this:
http://audio.rightmark.org/index_new.shtml
https://www.trueaudio.com/index.htm
And other very similar - not a problem.

https://en.wikipedia.org/wiki/Codec_listening_test
http://www.head-fi.org/t/486598/test...aims-and-myths

I will try to do objective part (but need to wait for V500 or revive my A3000).

Ahh, yes, good point. As I said, I could easily get the correct cable (may even have one here in one of the boxes of various cables) and could record directly from my A600. Shall I go looking for a suitable sound to playback on my Amiga, or does anyone have something suitable in mind?

[pandy71]

Originally Posted by meynaf

Aside of pure academic purposes, where's the point in measuring sound quality when the ear can't actually hear the difference ?

One of my procrastinated projects - use original chips and hook them to the FPGA, from one side this allow to analyze chips in some form of logic analyzer and at second it will allow for example to nullify CHIP RAM limits (CHIP RAM will be simply dual port memory so with CPU access without classic limitations).

As you know Jens made something similar http://www.amigahistory.plus.com/clone-a.html goal is create duplicate of chips in non commercial way (i assume Clone A will be never revealed by Jens and with new development on SAGA it may be a possibility that Clone A will be lost effort).

At some point i may wish to re-implement Paula and i need to understand also all DAC quirks.

[meynaf]

Originally Posted by pandy71

At some point i may wish to re-implement Paula and i need to understand also all DAC quirks.

Re-implement... for something identical or enhanced ?

[paul1981]

Originally Posted by meynaf

Aside of pure academic purposes, where's the point in measuring sound quality when the ear can't actually hear the difference ?

I see this from the other side totally. What's the point in measuring the difference in sound quality when it is obvious to the ear? If people can't hear the difference between Paula 14-bit @ 44.1KHz and a cd player or sound card playing the same audio data then there is either something wrong with their ears, their brain, or their playback equipment. That's not to say it doesn't sound good, because it is rather good. What I thought this thread was about is, how many bits are truly presented with Paula 14-bit playback. If anyone can measure it then good for them, but it's not going to make Paula sound better even if someone managed to squeeze out 16 true bits out of it.

[meynaf]

Originally Posted by paul1981

I see this from the other side totally. What's the point in measuring the difference in sound quality when it is obvious to the ear? If people can't hear the difference between Paula 14-bit @ 44.1KHz and a cd player or sound card playing the same audio data then there is either something wrong with their ears, their brain, or their playback equipment. That's not to say it doesn't sound good, because it is rather good. What I thought this thread was about is, how many bits are truly presented with Paula 14-bit playback. If anyone can measure it then good for them, but it's not going to make Paula sound better even if someone managed to squeeze out 16 true bits out of it.

Yeah, i can hear the difference. It sounds better with Paula 14-bit

[paul1981]

Originally Posted by meynaf

Yeah, i can hear the difference. It sounds better with Paula 14-bit

It certainly does. Paula 14-bit at 44.1KHz is very impressive when compared with the usual 8-bit sound at 28KHz.

[Thorham]

Originally Posted by paul1981

If people can't hear the difference between Paula 14-bit @ 44.1KHz and a cd player or sound card playing the same audio data then there is either something wrong with their ears, their brain, or their playback equipment.

It depends on the music. For some types of music the difference will be negligible, for some other types it will be large.

[NorthWay]

BTW, a while ago I was visiting a site with frequency tests etc and they had a comparison between a 16-bit sample and a properly processed 8-bit version of it (I think the frequency was high for both - probably 44KHz) and to my great surprise I had trouble hearing the difference with my Porta Pro connected to my work laptop. I don't know if that says more about the gear and my ears, but I had expected this to be a walkover and clear as day.

[Thorham]

Originally Posted by NorthWay

BTW, a while ago I was visiting a site with frequency tests etc and they had a comparison between a 16-bit sample and a properly processed 8-bit version of it (I think the frequency was high for both - probably 44KHz) and to my great surprise I had trouble hearing the difference with my Porta Pro connected to my work laptop. I don't know if that says more about the gear and my ears, but I had expected this to be a walkover and clear as day.

Again, it depends on the music. Loud music with limited dynamic range will often sound just fine in 8bit. Now try music with soft passages, like classical music...

[meynaf]

Originally Posted by Thorham

Again, it depends on the music. Loud music with limited dynamic range will often sound just fine in 8bit. Now try music with soft passages, like classical music...

That seems obvious. If we play some music which is f.e. at 1/8th of max peak, we drop from 16 to 13 bits - not really a big deal. However if we start with 8 bit, we end at 5. Not good.
8 bit can sound good, but only if it's fully used.

[Thorham]

Originally Posted by meynaf

That seems obvious.

Maybe not everyone realizes this fact. If the site NorthWay mentioned is the one I'm thinking of, then the example music chosen for the 16bit vs 8bit test is just about the worst possible. The site actually explains this, though, so it was done on purpose. They also have a proper version based on speech which becomes softer and softer, and they have dithered and noise shaped versions, too (where the 8bit speech can steel be understood at lower levels then plain 8bit). Quite interesting.

[pandy71]

Originally Posted by meynaf

Re-implement... for something identical or enhanced ?

Enhanced but at the same time full binary compatibility.

Originally Posted by meynaf

That seems obvious. If we play some music which is f.e. at 1/8th of max peak, we drop from 16 to 13 bits - not really a big deal. However if we start with 8 bit, we end at 5. Not good.
8 bit can sound good, but only if it's fully used.

This is not entirely true... as i wrote before - 4 times oversampled 8 bit (i.e. 176.4kHz) can provide higher quality than SACD.
Even in case of Amiga 8 bit at 56kHz with proper dithering and noise shaping will be close or better than CD - issue is that in normal conditions Amiga CPu is not capable to properly process signal due overall low power (perhaps V500 can do such thing especially if some additional HW acceleration can be added to ISA).
There is large number of studies about perceived noise-shaping - most important paper from LIPSHITZ, VANDERKOOY, WANNAMAKER:
http://www.ece.rochester.edu/courses...shitz_1991.pdf
http://mirlab.org/conference_papers/...fs/0300413.pdf
http://wiki.hydrogenaud.io/index.php...=Noise_shaping

I provided some time ago SoX scripts to try 8 bit with noiseshaping at Amiga and this is still not optimal as it will not use Amiga 56kHz (most of NS filters is designed or for 44.1 or for 48kHz) overall results will be better than 14 bit mode at a half of data (no need to push 16 bit at 44kHz).

Suboptimal yet better than 14 bit (more than 86dB dynamic in most interesting area - at 45473Hz and DMA only).

[meynaf]

Originally Posted by pandy71

Enhanced but at the same time full binary compatibility.

Then you don't need to copy the old behavior. A good sound quality is enough.

Originally Posted by pandy71

This is not entirely true... as i wrote before - 4 times oversampled 8 bit (i.e. 176.4kHz) can provide higher quality than SACD.
Even in case of Amiga 8 bit at 56kHz with proper dithering and noise shaping will be close or better than CD - issue is that in normal conditions Amiga CPu is not capable to properly process signal due overall low power (perhaps V500 can do such thing especially if some additional HW acceleration can be added to ISA).
There is large number of studies about perceived noise-shaping - most important paper from LIPSHITZ, VANDERKOOY, WANNAMAKER:
http://www.ece.rochester.edu/courses...shitz_1991.pdf
http://mirlab.org/conference_papers/...fs/0300413.pdf
http://wiki.hydrogenaud.io/index.php...=Noise_shaping

My comparison was intended for same replay freq.

[pandy71]

Originally Posted by meynaf

Then you don't need to copy the old behavior. A good sound quality is enough.

Of course it need to be binary compatible - if i don't care then simply i can do things completely different.

Originally Posted by meynaf

My comparison was intended for same replay freq.

If sample rate is higher than 44kHz then you can consider this as same replay freq.

Our hearing is not uniformly distributed across frequencies.
https://en.wikipedia.org/wiki/Absolu...old_of_hearing
https://en.wikipedia.org/wiki/Equal-loudness_contour
Our ears don't need same number of bits for different frequencies.