Which Amiga computer technically has the best audio quality?

[stx2199]

The A1200 sound output is far louder than the A500
so you will get better audio quality

[TuRRIcaNEd]

Originally Posted by Calabazam

There are differences in filters, hence the difference between A500 sound and A1200 sound that some can perceive.

That may be the case, however was it not essentially almost always the case that game and demoscene coders would disable the hardware (Butterworth) filter in order to achieve the best mid- to high-end clarity?

They may be differences in how caps behave and type of caps.
Old damaged caps will alter the sound. Tantalum caps as i have in an A600 are theoretically possibly altering the linearity of the sound output.

Cards on the table - I am at most an amateur, self-taught audio engineer - but I feel confident enough to state that (presuming the caps are functioning as they are supposed to) one might just be able to pick up an audible difference in that respect if you have the ears of a bat.

A very sensitive bat.

If anyone can actually hear a difference, i guess it is just a matter of preferences to say which is the best.

Correct. The joy of audio is that everyone's preference is subjective.

For decades now, there has been an ongoing debate amongst C64 aficionados as to which revision of SID sounds "best". What's very important to understand (aside from "best" being subjective) is that SID, unlike Paula, generated its tones internally, and due to the nature of chip manufacturing processes, literally no two SID wafers - even of the same revision - were likely to sound exactly alike, if one were to go deep enough into that rabbit hole.

Paula does not generate her own tones - she literally takes data from Chip RAM, is able to process that data through many functions that are useful in terms of audio and musical reproduction, and feeds the processed data to a DAC setup which converts the signals to an analogue audio signal pumped out of the stereo phono sockets. The only part of that system that might affect the final output is the character of the DACs - and the whole point of a DAC is that they're supposed to be generic and predictable.

Case in point - the Sega Mega Drive/Genesis used a YM2612 FM sound chip which in basic terms is a 6-channel version of the 8-channel YM2151 used on Sega's 1980s arcade games. However, the YM2612 was designed to be a low-cost component and - unlike the YM2151 which required an external DAC - had a basic DAC on the chip die itself, which is why FM music on a real Mega Drive tends to sound somewhat tinny compared to the arcade equivalent (a shortcoming that no longer exists as far as emulators are concerned).

Back in 1991 I rescued my family's old Marantz hi-fi amp and speakers from the garage, set it up in my bedroom and hooked it up to my A500 (which made everything sound *epic* for that time... ). Two years later when I got my A1200 for Christmas, I naturally swapped it out and could perceive zero audible difference.

Many years later when I started trying to get my head around audio engineering and music production, I went full-nerd and spent a day comparing the output from my "real" A1200 to that of the modern emulators, and observed that the analogue output from the A1200 had a very slight, but (just about) perceptible bias towards the low-end and mid-range - meaning that OG Paula + DAC has slightly more low-end "oomph" and mid "presence" compared to the raw dump from WinUAE.

So what does that mean? In all honesty, not all that much. The likelihood of any noticeable variance in audio clarity between different iterations of Amiga hardware is beyond negligible. If your intent is to stick your Amiga's audio output through a hi-fi for a more immersive experience it doesn't matter in the slightest which model you use, and if your intent is to capture Amiga audio output to use in your own musical productions, chances are you're going to run it through EQ settings tailored to your own tastes anyway, so the same applies.

[TCD]

Originally Posted by TuRRIcaNEd

A very sensitive bat.

It would be interesting in a 'blind test' to see if people can even hear the difference between real hardware and emulation.

[TuRRIcaNEd]

Originally Posted by TCD

It would be interesting in a 'blind test' to see if people can even hear the difference between real hardware and emulation.

Back in the day (we're talking about 1997-ish), having reluctantly switched to Wintel for Uni, I had a feeling that the output of Fellow/WinUAE sounded a bit "flat" compared to the real thing - but I had no idea how to express it or understand why that might be. This is why I did a nerdy A/B test many years later.

[Pyromania]

Sound chip in the Vampire V4 is quite advanced and it blows Paula away.

[TuRRIcaNEd]

Originally Posted by Pyromania

Sound chip in the Vampire V4 is quite advanced and it blows Paula away.

On paper, of course; that said it's hardly surprising given we're talking 2020 FPGA vs. 1985 NMOS, no? What remains to be seen is whether those capabilities are ever used.

Anyways, as far as I can tell that wasn't the aim of the original question, which seemed to be whether any of the "classic" Amiga models had inherently superior audio fidelity despite every single one of them using Paula.

[pandy71]

Originally Posted by sokolovic

A600 were cheaper, smaller, easily expandable and have the same soundchip than every Amiga.
Seems logical to me to use them.

Nope, there are different Paula revisions and later (PLCC) are better - probably CSG improved on all those years also Paula layout...

Originally Posted by Henryk Richter Amiga Paula vs. System Theory

It’s worth noting that the ECS revision of Paula got significant
improvements wrt. DAC accuracy. The ECS Paula can be found in SMD Amigas as standard
configuration (8364R7PL). A500+ usually also has the ECS Paula equipped (8364R7PD). By
my experience, the ECS Paula is less commonly found in A3000 sytems. If you can get a hold of
an ECS Paula for A500/2000/3000, I’d recommend to install it.

Secondly Paula is not everything - there is many factors involved in audio quality - for example virtual ground circuitry used in Amiga to allow Paula DAC's emit/sink current that is later converted to voltage, also filters are different. PCB layout is important, decoupling is important.
Seem CD32 was designed around same Paula as A600 but overall audio signal path is simply better.

So despite Paula didn't changed (we know that this is untrue as mentioned earlier) since 1984/85 there are differences sometimes more important, sometimes less so you can have Amiga that has higher audio quality.
In half of 80's high quality audio from DAC's was just at the beginning of popularity, 5 years later at the beginning of 90's this was quite common knowledge and this know how is visible in Amiga circuitry evolution.

[chip]

Paula rulez !

[TuRRIcaNEd]

Originally Posted by pandy71

sometimes less so you can have Amiga that has higher audio quality

I hate to be a pain in the bum, but is it OK to ask you to explain how?

[pandy71]

Originally Posted by TuRRIcaNEd

I hate to be a pain in the bum, but is it OK to ask you to explain how?

Many small things, for example how power supply traces are laid on PCB, sometimes they can deliver cleaner power to analog circuitry, sometimes less, if there is no proper grounding then this will affect audio quality, analog signal lines laid close to noisy signal, replacing resistors in power supply with more efficient ferrite beads - there is no audiophile voodoo here - nowadays this is standard practices and you can easily find whole chapters in for example IC datasheet how to properly design PCB layout, how to place components, what should be avoided.
And important remark - this knowledge evolved in decades - i bet if in 1984 such knowledge exist with same level of "popularity" as today then overall audio circuitry will be better since first design - CD32 is last Amiga, also it was designed with at least cooperation with application engineers dedicated to CD audio DAC implementation so such knowledge propagate down to Paula circuitry.

Some topic is currently going on this how much A600 can be improved:

https://eab.abime.net/showthread.php?t=86880

also this thread can be interesting: https://eab.abime.net/showthread.php?t=109051

It can be nice to see if someone offer those retro PCB's for various projects with changes addressing all those small things - this not ruin legacy heritage of Paula...

[8bitbubsy]

Originally Posted by Karlos

The A1200 has a much brighter sound as the cutoff filter is tuned much higher.

This. It sounds way less filtered than all the other Amigas!
Every other Amiga has a fixed low-pass whose cutoff is around 4-5kHz (depending on model). It really does remove the high fidelity from the sound.

[chb]

Originally Posted by TuRRIcaNEd

For decades now, there has been an ongoing debate amongst C64 aficionados as to which revision of SID sounds "best". What's very important to understand (aside from "best" being subjective) is that SID, unlike Paula, generated its tones internally, and due to the nature of chip manufacturing processes, literally no two SID wafers - even of the same revision - were likely to sound exactly alike, if one were to go deep enough into that rabbit hole.

Paula does not generate her own tones - she literally takes data from Chip RAM, is able to process that data through many functions that are useful in terms of audio and musical reproduction, and feeds the processed data to a DAC setup which converts the signals to an analogue audio signal pumped out of the stereo phono sockets. The only part of that system that might affect the final output is the character of the DACs - and the whole point of a DAC is that they're supposed to be generic and predictable.

A bit off topic: I think there's a misunderstanding how the SIDE works. It indeed features, in contrast to Paula, oscillators and noise generators to create sounds - but they work in the digital domain, their output is fed through a DAC (one per channel, there's actually a lot of DACs in the SID for different tasks...). Now, those exact DACs are partly responsible for the unique sound of the SID, more precisely the older 6581, because they are quite non-linear (due to manufacturing limitations) and introduce distortions.


But the issue that's responsible for most variations between SID chips is filter variation; AFAIR they use a FET on the chip as a voltage-controlled resistor (voltage provided by a DAC, ofc), and that method is very sensitive to manufacturing tolerances and even operating temperature. If one disables the filter, SIDs sound more similar across different revisions. They also fixed that behavior with the 8580...

[CCCP alert]

Originally Posted by TuRRIcaNEd

Correct. The joy of audio is that everyone's preference is subjective.

Not when talking about max/min frequency range, distortion, humm hiss. I am talking about Amstrad vs Kenwood tape deck 'audiophile' style tests on the output of a reference signal not stuff like bass bias/missing mid-range or 'warmth' of output sound etc which is different and personal sure.

To be fair, after adjusting the volume control on a TV/monitor you're not going to hear much difference really, or even a budget midi-hifi etc, stuff you would only notice on a decent amp and decent large hi-fi speakers I suspect.

My friend had an A500 with a slightly better Kenwood hifi with proper Dolby Pro Logic and better speakers etc than mine and I couldn't really tell the difference from my machine using a slightly less sophisticated Pioneer at home 10 minutes drive away. Probably have to test them side by side

Would be interesting to compare 1000/2000/3000/4000 as these were expensive machines. The 1000 motherboard has a lot going on with the most components and that riser to the daughterboard.

[TuRRIcaNEd]

Originally Posted by chb

Now, those exact DACs are partly responsible for the unique sound of the SID, more precisely the older 6581, because they are quite non-linear (due to manufacturing limitations) and introduce distortions.
...
...most variations between SID chips [are the result of] filter variation; AFAIR they use a FET on the chip as a voltage-controlled resistor (voltage provided by a DAC, ofc), and that method is very sensitive to manufacturing tolerances and even operating temperature

Correct. You've gone into more detail than I could manage to remember precisely, but based on what I've read over time the filter's sensitivity to manufacturing tolerances was the main factor behind no two SIDs sounding exactly the same.

Originally Posted by CCCP alert

Not when talking about max/min frequency range, distortion, humm hiss. I am talking about Amstrad vs Kenwood tape deck 'audiophile' style tests on the output of a reference signal not stuff like bass bias/missing mid-range or 'warmth' of output sound etc which is different and personal sure.

I hear you - though this goes back to what chb was saying - for better or worse Paula remained unchanged from the 1985 A1000 to the 1993 A4000T; and all of Paula's audio processing happened in the digital domain. The DAC conversion to analogue output (which is where "max/min frequency range, distortion, humm hiss" can become an issue) always happened at a single point between the chip's output traces on the PCB and the stereo RCA sockets on the back of every Amiga machine. With the hardware lowpass filter disabled, there was very little opportunity for the audio output to be affected by manufacturing/component variance.

On a bit of a tangent (though sort-of relevant), about 16 years ago I borrowed a DAT deck from a friend of mine so I could export my old indie band's recordings from DAT to my PC. Being an older DAT deck, the SPDIF output used was of the RCA plug type as opposed to TOSLink - and the cable he gave me had all mod cons - gold plated on both ends with the cable itself being almost as thick as a baby's arm... But SPDIF is a digital protocol; the signal is either on or off - which means that technically the signal transmission quality should be the same regardless of whether one's using an expensive or cheap cable.

To be fair, after adjusting the volume control on a TV/monitor you're not going to hear much difference really, or even a budget midi-hifi etc, stuff you would only notice on a decent amp and decent large hi-fi speakers I suspect.

Right - and as I said above, I'd been running my A500 through a fairly decent Marantz hi-fi amp and set of speakers since about 1991. When I swapped it out for my A1200 at Christmas 1993 I could make out no discernible difference at all.

My friend had an A500 with a slightly better Kenwood hifi with proper Dolby Pro Logic and better speakers etc than mine and I couldn't really tell the difference from my machine using a slightly less sophisticated Pioneer at home 10 minutes drive away. Probably have to test them side by side

Dolby Pro Logic was a technology for decoding 5.1 home cinema audio from soundtracks encoded with Dolby Surround ( https://en.wikipedia.org/wiki/Dolby_Pro_Logic ). Amiga audio output was always standard stereo (2.0) and there was never any software for the Amiga which encoded the output as Dolby Surround - having Pro Logic on the amp would have made absolutely no difference whatsoever. As long as your hi-fi and speakers were of decent quality, there's no reason they should have sounded much different.

Again, if you'll forgive me another tangent, as an amateur musician and semi-pro recording and live audio engineer I'm very particular when it comes to knowing and understanding my equipment (I've been fortunate to have been taught by several people who are professionals as good as it's possible to find). One truism that comes up time and time again is that there are an awful lot of BS and snake-oil merchants operating in the "audiophile" market. One particularly egregious example was a vendor selling audio cables with gold/platinum-plated connectors claiming that the specific weave and metallurgy in their product resulted in "optimal electron alignment" (or words to that effect) making their cables better than any competitor. I think they were charging something like £75/ft. For a cable.

And it's utter *gash*.

Plating the connector with a better conducting metal may theoretically improve conductivity, but in practice it doesn't make a blind bit of difference because the electronics of the hi-fi components and circuitry within the speakers are designed to operate in exactly the same manner regardless of any extra conductivity. Furthermore, while thicker cables with better insulation are theoretically capable of handling a greater load, the actual load they're carrying will be the same regardless of which cable one uses.

Case in point, shortly after I moved last year I bought a second-hand B&O hi-fi setup* from eBay for my living room. Despite being manufactured in 1982, it had recently been checked and reconditioned and it still sounds absolutely immense.

It's an all-in-one unit, so all the circuitry between phono/tape/radio/aux is internal. However the socket and plugs to the speakers are of exactly the same type as those used on the Amstrad (a brand you rightly used as an example of dubious quality) stereo I had as a kid.

Yes, you're reading that right - a B&O unit was using the same kind of socket and speaker connections as an Amstrad of a similar vintage. In other words, audio quality/fidelity comes pretty much entirely from the signal coming out of the amp. The socket and cable types don't make a blind bit of difference.

Would be interesting to compare 1000/2000/3000/4000 as these were expensive machines. The 1000 motherboard has a lot going on with the most components and that riser to the daughterboard.

But as I said above, the silicon used for Paula - along with the DAC and output architecture - was the same from the A1000 to the A4000T. The process for manufacturing the motherboards (and the DAC components being used) was the same regardless of whether the motherboard was going into a machine intended for the home or professional market.

As someone who dreamed of writing music since I was tiny - and subsequently being useless trying to learn 'cello and piano in the "normal" way, but ended up finding my feet learning to write music on trackers in my early teens - I hope you can understand that I go back a very long way when it comes to Amiga music and audio stuff. The point is that based on my understanding, in a purely technical sense there's absolutely no reason any Amiga (regardless of model) should sound perceptibly different from any other.

If ever there was a limit on Amiga audio fidelity (over and above Paula's being 8-bit and inherent frequency constraints), it was arguably a matter of RAM usage.

In order to demonstrate that claim, check out DJ MeTune's conversion of Axel F (which takes up 432k) :

[ Show youtube player ]

[* - I blame my Dad for getting me started on the B&O rabbit hole... Apropos of nothing, the Marantz setup I butchered and hooked up to my Amigas was originally his purchase in the early '80s. He'd long since buggered off, so I felt no guilt nicking it for my own purposes. ]

[AmigaHope]

It should be noted that while Paula is more or less unchanged between revisions, and the systems varied in the analog circuitry that affected the sound after it came out of Paula, the core limitation in overall audio quality came from Paula's DMA access window, which was tied to horizontal scan rate. Thus, Paula coupled with OCS Denise had an inbuilt limitation to its sampling rate (about 28Khz), achieving higher sample rates required that Paula be paired with a Super Denise or Lisa which could operate at horizontal scan rates higher than 15Khz.

Of course on older machines you'd have to bypass the baked-in low-pass filter (the hardwired one, not the one tied to the power LED) to even hear the higher frequencies, and early A1000s didn't even have the LED-tied low pass filter disable.