Queries on AGA Copper and 14-bit Audio

[Thorham]

Originally Posted by roondar

Is it really that relevant wether or not an Amiga can compete with modern PC audio hardware which is 25 years newer and leans heavily on technologies which where either not available at all or not available for any reasonable cost back in 1985?

Partially, but what's most relevant is the way it sounds with various types of music. My A1200 seems to produce good quality for certain types of music compared to peecees, so that's good enough for me.

Originally Posted by roondar

If we're going to compare DAC performance, it should be done fairly - compare it to other DAC's from the 1980's, available for similar prices as Paula (i.e. a few dollars max). Comparing the Amiga against high end audio stuff is hardly fair (and outright silly when doing so against stuff from 2016!) and doing so does make it seem like the goal is to make the Amiga look bad (IMHO anyway).

That's not entirely true. Paula supports several megahertz (!!!) sample rates (audio DMA doesn't, of course). Since there doesn't seem to be any software which uses this, Paulas true capabilities are unknown. Asking how this holds up to current good quality sound card DACs seems entirely reasonable, and I expect that it should do fairly well.

[roondar]

Originally Posted by Thorham

Partially, but what's most relevant is the way it sounds with various types of music. My A1200 seems to produce good quality for certain types of music compared to peecees, so that's good enough for me.

Oh, I agree it sounds nice enough, but that's not what is being discussed - the discussion has focused on how many bits of resolution are being offered and the level of signal to noise inherent in Amiga audio. Both of which, I think, are not really all that likely to be better in early PC cards.

However, the thread doesn't compare how Paula sounds or even how it does compared to hardware available at the time, but compares things like SNR and resolution of Paula to current 24 bit/192KHz sound cards or even studio equipment and I don't really feel that's a fair comparisom.

Originally Posted by Thorham

That's not entirely true. Paula supports several megahertz (!!!) sample rates (audio DMA doesn't, of course). Since there doesn't seem to be any software which uses this, Paulas true capabilities are unknown. Asking how this holds up to current good quality sound card DACs seems entirely reasonable, and I expect that it should do fairly well.

I'm not so sure it will, because there is more to audio quality than sample rates.

And the discussion in the thread seem to be mostly about the 'other bits' (SNR, levels of distortion, etc) - where Paula does not score too well because technology these days lets you create a cheap DAC with outputs offering (compared to 80's consumer technology) insanely good values for these things.

[jarp]

All this fighting and nobody RMAA'd their Amigas? Well here you go, just for pure fun, RMAA test results played from Amiga 600 rev 1 with Furia 020 accelerator, recorded by PC and Asus Xonar STX. I used multiscan screenmode and play16 to play RMAA test signal.

8-bit (play16 verbosity=1 fast=1 filter=off output=paula8):



14-bit (play16 verbosity=1 fast=1 filter=off output=paula14):



There is clear difference in 8-bit and 14-bit output, RMAA shows it but also it can be easily heard when listening music. Led filter on / off also produces huge differences - led filter off of course being much better. High noise level can be easily heard as can be high distortion. Amiga 600 has hard filter which cannot be turned off and that makes frequency response crap. I would love to try this with Amiga 1200 which does not have it or some day de-solder the filter from the Amiga 600. Perhaps Amiga 1200 has much better noise floor also but rev 1 Amiga 600 definitely has A LOT of noise.

But there seems to be measurement errors also - RMAA says that dynamic range is 55 dB for 8-bit output but it's impossible since 8 bits allows only 48 dB, no? Theoretical limit for 14-bit would be 84 dB and RMAA says 70 dB so that is possible.

Well in any case this was just for fun. Even though RMAA measurements are poor I was hugely impressed by the quality of sound when I played back some acapella music. I was fully able to enjoy it. I mean damn, that sound chip was designed for 1985 computer and it STILL can do results which are good enough for average home audio equipment.

[pandy71]

Originally Posted by jarp

All this fighting and nobody RMAA'd their Amigas? Well here you go, just for pure fun, RMAA test results played from Amiga 600 rev 1 with Furia 020 accelerator, recorded by PC and Asus Xonar STX. I used multiscan screenmode and play16 to play RMAA test signal.

But there seems to be measurement errors also - RMAA says that dynamic range is 55 dB for 8-bit output but it's impossible since 8 bits allows only 48 dB, no? Theoretical limit for 14-bit would be 84 dB and RMAA says 70 dB so that is possible.

Well in any case this was just for fun. Even though RMAA measurements are poor I was hugely impressed by the quality of sound when I played back some acapella music. I was fully able to enjoy it. I mean damn, that sound chip was designed for 1985 computer and it STILL can do results which are good enough for average home audio equipment.

Thank you for this - hope it is clear why i ask for this.
Not sure if play16 do things correctly some numbers are worse than expected.
I converted RMAA signal to more proper Amiga sample rates.
They can be downloaded from http://wikisend.com/download/311120/8364_Test_RMAA.7z
There are two sample rates (period=124 and period 74), 8 bit files are 8svx and there are two types of them - plain without dither and noiseshaping and second with dither (HP TPDF) and for period=74 also noiseshaping (improved-e-weighted SoX http://sox.sourceforge.net/SoX/NoiseShaping ).
Sadly SoX has implemented limited noiseshaping (very narrow selection of sample rates as such for lower frequency noiseshaping is disabled - ffmpeg somehow support wider range of sample rate combinations).
I think that some bad results may be related to distortions related to incorrect (lack of) samplerate conversion.
Dynamics better than expected may be related to limited oversampling and maybe some errors in measurement procedure (due bad or not existing) sample rate conversion.
I will try to repeat this measurement with APx555 .

And at the end - i share same impression that Amiga audio is not bad at all if we remember that it is 8 bit, designed in half of 80's previous century and there is a lot of thing that can be improved (on software but also on HW/analog signal path side) - even slightly worse than expected numbers definitely not nullify joy and fun from Amiga music.

Originally Posted by roondar

@audio discussion:
Is it really that relevant wether or not an Amiga can compete with modern PC audio hardware which is 25 years newer and leans heavily on technologies which where either not available at all or not available for any reasonable cost back in 1985?

Of course we have better audio these days. And no, that is not at all relevant when looking at the Amiga audio out quality. It'd be much more interesting to know how it compares to audio produced back in the day by PC's and such and much less in current day high end studio performance.

The Amiga is not a high end audio device. Then again, neither was a PC back in the 90's (I'm pretty sure Soundblasters where actually quite a bit worse in the actually produced signal to noise ratio department until the AWE32*), so again why is this relevant at all?
*) as opposed to the stats on the box, which never held to any serious scrutiny.



If we're going to compare DAC performance, it should be done fairly - compare it to other DAC's from the 1980's, available for similar prices as Paula (i.e. a few dollars max). Comparing the Amiga against high end audio stuff is hardly fair (and outright silly when doing so against stuff from 2016!) and doing so does make it seem like the goal is to make the Amiga look bad (IMHO anyway).

Can't left this without explanation as it is quite clear that some of us see my question as inappropriate and assume in advance that i hate Amiga and my goal is to prove that Amiga audio is total crap.
At first there is no known official specification for Amiga audio seen as a music\sound\audio source - fair - ok - it was not the main purpose of Amiga.
At second there is not known official Paula specification covering DAC's aspects i.e. there is not known electrical specification of chip (weird as Commodore for sure defined some values and at least from time to time verified them).
I was curious about real audio signal parameters as it allow to describe system capabilities (i.e. approach to improve overall audio Amiga capabilities) side to this it may trig new directions for for example software development.
That's all - for sure it was not to compare DVD Audio where 24 bit, 192kHz audio is reference - however i proposed to use well known high quality source (as for example CD\DVD\Media player etc.) with defined sources to calibrate simple measurement equipment like PC audio Line Input so Amiga can be objectively measured and in objective way also all potential improvement can be verified - exactly same as jarp made http://eab.abime.net/showpost.php?p=...&postcount=123 .
Simple enough, nobody died, some values are not as good as expected so this need to be analyzed if this is real Amiga (Paula + rest of signal circuit) limitation or this incompatibility between signals and Amiga HW uniqueness (PC is fixed sample rate, Amiga is variable sample rate).

Side to this i'm not surprised about silence after jarp message - subjectively we love Amiga even if love is blind.

[jarp]

Btw. tried your RMAA files and results were extremely... wild. Frequency response was a roller coaster etc. Files do sound different also than mine when I listen them. We must have different versions of the RMAA or something.

I also "upgraded" my test gear from my Asus Xonar STX to an old E-MU 0404 USB. Mainly due to it's excellent integrated pre-amp (STX has better specs otherwise) - Amiga puts out quite a low voltage without a preamp and the one I used in my earlier measurements was not the best (~90dB noise floor). Repeated the tests but pretty much nothing changed (which was expected of course since my preamp was better than Amiga's outputs in any case).

[pandy71]

Originally Posted by jarp

Btw. tried your RMAA files and results were extremely... wild. Frequency response was a roller coaster etc. Files do sound different also than mine when I listen them. We must have different versions of the RMAA or something.

They may sound different as frequencies will be correct (Amiga is not capable to play 44100 or 48000 as such frequency can't be set in Paula), other issues may be related to dithering and noiseshaping (but there is no way to play level -60dBFS on 8 bit sample).
I've downloaded latest RMAA and saved wav's, later wav's was converted with SoX to something more suitable for Amiga.

SoX script (PC commandline) bellow:

@setlocal

@rem 58146 57208 56300 55420 54568 53741 52939 52160 51404 50670 49956 49262 48588 47931 47292 46670 46064 45473 44897 44336 43789

@set sr=47931
@set /A pre=%sr%/4
@echo %sr%
@echo %pre%
@rem lipshitz, f-weighted, modified-e-weighted, improved-e-weighted, gesemann, shibata, low-shibata, high-shibata

@set noiseshp=improved-e-weighted

@echo %noiseshp% 

@sox --multi-threaded --buffer 131072 -S -V -D "%1" "%~n1.sox" rate -v -s -a -I %sr% stats -b 8
@sox --multi-threaded --buffer 131072 -S -V -D "%~n1.sox" "%~n1_ECS_%sr%.8svx" dither -f %noiseshp% -p 8 stats -b 8

@set xx=640
@set yy=294

@set /a w=%xx%-144
@set /a h=%yy%-78

@sox --multi-threaded --buffer 131072 -S -V -D "%~n1.sox" -n spectrogram -z 86 -q 12 -x %w% -y %h% -o "%~n1_sox.png"
@sox --multi-threaded --buffer 131072 -S -V -D "%~n1_ECS_%sr%.8svx" -n spectrogram -z 86 -q 12 -x %w% -y %h% -o "%~n1_ECS_%sr%.8svx.png"

@del -q "%~n1.sox"
@endlocal
@pause

Originally Posted by jarp

I also "upgraded" my test gear from my Asus Xonar STX to an old E-MU 0404 USB. Mainly due to it's excellent integrated pre-amp (STX has better specs otherwise) - Amiga puts out quite a low voltage without a preamp and the one I used in my earlier measurements was not the best (~90dB noise floor). Repeated the tests but pretty much nothing changed (which was expected of course since my preamp was better than Amiga's outputs in any case).

I think about buying Behringer UMC202HD - it support sample rate up to 192kHz@24 bit so should be sufficient for this task.