Digitally Remastered 8 bit released

[meynaf]

Originally Posted by demolition

It can be proven mathematically that 44.1kHz can provide a perfect reproduction of all frequencies from DC to 20 kHz.

Does not mean nothing else can creep in, like components of higher frequencies.

Originally Posted by demolition

Since the human hearing does not go beyond this range, 44.1 kHz should be enough for perfect reproduction. The only issue here is that it is impossible to make infinitely steep low-pass filters. In the early days, the filters were very poorly made and allowed high frequencies to fold down into the audible range, known as aliasing. This sounds very bad. Today modern DSP and oversampling techniques means that these filters can be made a lot better and in practice can push aliasing products down below -100dBFS in the audible range so they are no longer a problem. This is a bit harder with 44.1kHz than with 48kHz.

There are still harmonics to be heard, or aliasing noise, or quantization noise, or whatever it's called. I supposed it would be harmonics, which could well be wrong. However something can be heard that makes 48Khz sound slightly better than 44.1, that's all. The theory behind, i've no idea of. I'm just stating a fact.

[demolition]

Originally Posted by meynaf

Does not mean nothing else can creep in, like components of higher frequencies.

Sure, if the anti-aliasing filter is not good enough, then frequencies above Fs/2 will mirror down below Fs/2. This sounds horrible, particularly since they are inharmonic. When sampling @ 48 kHz 16 bit, then the anti-aliasing filter should ideally have a 0 dB gain @ 20 kHz and less than -96 dB gain @ 24 kHz. This is quite a steep roll-off but doable today if done right. Cheap A/D converters will not do it right, and then you will get a better result with 96 kHz since that will loosen up the restrictions for the filter significantly.

Just because cheap converters perform bad, doesn't mean that 48 kHz can't be good enough.

Quantization noise relates to the bit depth of the sampled signal. With 8 bit it is quite audible with most sounds, but with 16 bit you will need to turn the volume up very loud to be able to hear it, and then you will still only be able to hear it during very soft parts of the music, so effectively it is not an issue with most music. In cinemas where you have a large dynamic range, it makes sense to go above than 16 bit.

[Pat the Cat]

Originally Posted by demolition

Just because cheap converters perform bad, doesn't mean that 48 kHz can't be good enough.

Even a 48KHz sampler isn't going to keep track with a 56K Amiga sample.

I'll go over the main points of this thread;-

1) The examples are not clipped. You cannot clip when transferring a mod to a Wav with Sonique. You CAN clip if you over compress, unless you tell the digital compressor software not to.

2) The reason for the digital delay is the way that the human mind interprets audio. If a sound has out of phase left and right elements, played through speakers, people can give it a direction.

3) Neither these examples nor any Prodigy track is truly "distorted". Instead, they have had their decibel levels raised - in this case, to 130 or so. So of course they sound distorted unless you turn your levels down.

These examples and Prodigy's run the razor's edge. One more step either way, they would be losing fidelity. Instead, they soak up and use all the available bandwidth.

That's why they are "Always outnumbered. Never outgunned."

It's also possibly why they don't bother about arguing that their works are clipped and distorted. They know that they aren't, and if other people mistakenly think that, then it's somebody else's mistake, not theirs.

The difference between the Prodigy and myself is - I'm not trying to sell a record. I'm not trying to promote an album, do a tour, arrange a gig, or even sell a T-shirt.

I'm saying this is how you can make original Amiga audio sound - meaty, dynamic, powerful, and attractive to 21st century ears.

If somebody else wants to do this, design, build, test and sell hardware, fine. I can't claim Intellectual Property over maths. I certainly can't claim IP over a work based on a Public Domain release (unlike Akira, who seems to think that's legit).

[Daedalus]

Originally Posted by Pat the Cat

3) Neither these examples nor any Prodigy track is truly "distorted". Instead, they have had their decibel levels raised - in this case, to 130 or so. So of course they sound distorted unless you turn your levels down.

I'm confused by this. Are you saying that changing the volume of my playback equipment will affect whether it sounds distorted or not? The Prodigy's music sounds the same regardless of what volume I play it at - turning it down sounds just as distorted, but at a lower volume. That's because that's the effect they intended. But you seem to be saying (and did so earlier in the thread too) that the distortion people are hearing is because they have it turned up too loud. If that's what you're trying to say, that's nonsense of the highest order.

If I have to turn my equipment down to hear your music correctly then you're doing it wrong.

[demolition]

Originally Posted by Pat the Cat

Even a 48KHz sampler isn't going to keep track with a 56K Amiga sample.

Sure it is. 48 kHz can perfectly capture all the frequencies up to 20 kHz which covers the human hearing. It will not accurately capture frequencies above 20 kHz (and nothing above 24 kHz), but only your dog will care about that.

Prodigy uses distortion to a large degree, like many others as a musical tool. As Daedalus writes, this has nothing to do with the volume you play back at unless you have really bad speakers. If the distortion changes with volume, your system is non-linear and this is bad. Of course any speaker will distort when it becomes loud enough, but it should not happen within normal listening levels.

When you write '130 or so', what do you mean? Decibel? Because Decibel is not a unit, it only tells on a log scale how large something is relative to something else.. On a digital signal, 0 dBFS is full scale and you cannot go any higher. Most music will have a level of between -20 to -10 dBFS RMS. The higher, the more compressed it generally is.

Listen to this track and tell me it is not distorted:
[ Show youtube player ]
They distorted several of the instruments on purpose and it sounds great. It is one of my favorite Prodigy tracks.

[Thorham]

Just use whatever sounds good and be done with it

[Anakirob]

Originally Posted by Pat the Cat

The Amiga has two ways of doing Audio - Paula and the CPU. Paula limits you to 28,867 samples per second playback on NTSC, a little higher on PAL (which is why a tiny minotiry of Amiga audio doesn't play on NTSC systems)

Really? I was always under the impression that NTSC allowed a slightly higher sample rate than PAL, having something to do with the screen refresh rate (although I know this is not the same thing as horizontal refresh rate which allows 56KHz under certain screenmodes), at least in the table of notes the rates are slightly higher under NTSC than PAL. Can someone clear this up?

[meynaf]

Originally Posted by Anakirob

Really? I was always under the impression that NTSC allowed a slightly higher sample rate than PAL, having something to do with the screen refresh rate (although I know this is not the same thing as horizontal refresh rate which allows 56KHz under certain screenmodes), at least in the table of notes the rates are slightly higher under NTSC than PAL. Can someone clear this up?

You're right, it's NTSC that can go the highest.
IIRC frequency limit is 28604 for PAL, 28867 for NTSC (period 124 in both cases).

[Anakirob]

Cool, thanks for that. I have another question about Paula when playing the two uppermost notes in ProTracker, which are both above the maximum frequency range of the Paula. For those of you who don't know, they play ok, but with glitches, presumably from skipping the odd bit of data.

Is playing samples at rates higher than the Paula can actually output a feature of the Paula, or is there some kind of coding trick used in module replayers to allow these notes?

[meynaf]

Originally Posted by Anakirob

Is playing samples at rates higher than the Paula can actually output a feature of the Paula, or is there some kind of coding trick used in module replayers to allow these notes?

They play with occasionnal glitches in most module replayers, too. The only way to play them safe is by some kind of downsampling.

[Anakirob]

So does that mean module players and trackers DO do some coding tricks to prevent some kind of crash which would occur if one were to attempt to feed an illegal value to Paula?

[meynaf]

Originally Posted by Anakirob

So does that mean module players and trackers DO do some coding tricks to prevent some kind of crash which would occur if one were to attempt to feed an illegal value to Paula?

No, they really feed illegal values to Paula. No crash can occur from this, Paula will just eventually output bogus sounds.

[robinsonb5]

Originally Posted by Anakirob

Is playing samples at rates higher than the Paula can actually output a feature of the Paula, or is there some kind of coding trick used in module replayers to allow these notes?

It's just a side-effect of how Paula works. The sample rate is set by telling Paula how many clocks to wait between emitting samples, and there's nothing stopping you from setting that sample rate very high. However, if Paula is ready to play a new sample before it's arrived, the previous word of data is repeated, which causes the glitches you talked about.

The limit is in how quickly the DMA system can feed fresh data to Paula. The DMA system is in sync with the video display, and only one word of data per channel can be fetched per scanline. That's why the maximum sample rate varies slightly between PAL and NTSC systems. It's also why with 31KHz video modes those glitches disappear - because the audio DMA slots are now roughly twice as frequent.

[NorthWay]

Originally Posted by robinsonb5

It's also why with 31KHz video modes those glitches disappear

OT but: Could you hack up a special display mode to do some crazy high frequency audio playback? Happily without an active display output so you don't fry your monitor.

[Thorham]

Originally Posted by NorthWay

OT but: Could you hack up a special display mode to do some crazy high frequency audio playback? Happily without an active display output so you don't fry your monitor.

I wouldn't be surprised if there's a DMA limit. I've tried 56khz using Hippo Player, and there were glitches. With the CPU the limit is much higher, and perhaps the Copper can do it too.

[daxb]

"AddAudioModes dblscan" from AHI packet should do the trick.

The DBLSCANoption does not have anything to do with the audio mode
list.  If specified, it will open and then immediately close a
native, double-scan screen.  On some systems using a graphic card,
this will enable >28 kHz sample frequencies with the native audio.
You need an appropriate monitor driver in DEVS:Monitorsto make
it work.

[Pat the Cat]

Originally Posted by meynaf

You're right, it's NTSC that can go the highest.
IIRC frequency limit is 28604 for PAL, 28867 for NTSC (period 124 in both cases).

First edition HRM lists minimum period values as 123 PAL, 124 NTSC. Measured in clock ticks. That gives a total maximum playback rate of 28.867 KHz on BOTH kinds of Amiga.

CBM didn't ever write that part of the HRM of course, Hi-Toro/Amiga Inc did. I have not checked the later HRM edition, so it could be a typo, or perhaps CBM just didn't like the idea of their American customers having a slightly suspect sample rate. It might explain Akira's claim that the released version is different in pitch to the original. Not sure about that, have no idea what tools they used or their source of the audio, but in terms of playback on a PAL system it is a match, for frequency response.

As Sonique is French, with a SECAM background, the writer probably took account of the different playback rates and adjusted accordingly. That's my guess, perhaps it only works with PAL sourced mods after all.

I think I muddied the waters when I said that Paula was limited. Other posters are more accurate here, it is the DMA via chip RAM that causes the limits, and while you could use the CPU to write a custom copper list for the next vertical blank so that the copper writes the values into the audio DACs on time, it would probably be very glitchy unless you got it absolutely perfect every time. That's on the Amiga, of course.

Amiga 8 bit Sampling software capable of capturing audio at that rate shut down all activity while sampling, and started up again when the user had clicked a mouse button or similar to indicate the sample recording be stopped and the system returned to "normal" again.

[Pat the Cat]

Originally Posted by Daedalus

I'm confused by this. Are you saying that changing the volume of my playback equipment will affect whether it sounds distorted or not? The Prodigy's music sounds the same regardless of what volume I play it at - turning it down sounds just as distorted, but at a lower volume. That's because that's the effect they intended. But you seem to be saying (and did so earlier in the thread too) that the distortion people are hearing is because they have it turned up too loud. If that's what you're trying to say, that's nonsense of the highest order.

If I have to turn my equipment down to hear your music correctly then you're doing it wrong.

This is not a release of "my music". If people want some easy listening over the internet, there are plenty of available examples.

This is a technology demonstrator for developers to study, in terms of what happens if you use purely digital, software techniques on Amiga sound data. With the desired goal of a hardware add on to enhance Amiga audio.

If I thought I could do that on my own, with a completely designed solution, controls, limiters, filters etc etc I would have just done it on my own. it would have saved an awful lot of goalkeeping.

[Thorham]

Originally Posted by daxb

"AddAudioModes dblscan" from AHI packet should do the trick.

Or use HippoPlayer or DeliTracker. They support >28khz without AHI.

[Pat the Cat]

Been a long time since I played one... IIRC correctly, there are minimum requirements for playing fast samples. Accelerator and fast RAM, if you want to do so from Workbench. You can do so on any Amiga technically, but you can't do much else at the same time, on a chip RAM only / 68K machine.

Take this with a pinch of salt, I only played very briefly with an early version of Hippo. a long long time ago.