note too that, in a lot of cases, it seems like we're moving away from modeling and towards using a gajillion samples of the original instrument (the garritan orchestra series, ni's akoustik piano & elektrik piano, the bfd series, etc)... there seems to be a real shift towards quantity...

Jeremy wrote:


as far as synthetic instruments verses real instruments (the electric piano and the v-drums), your run down doesn't apply.

Yes and no, I think. I mean, I'm definitely totally embarassed by howmuch crap I wrote earlier about stuff that's totally well known, but I don't think it was completely off the subject. Because digital's so clearly quantized, people quickly come to the conclusion that it can never accurately represent a sound (that was said at least two or three times in this thread). But it's worth remembering that electrical analogs are also grainy, and at some point the analog breaks down. It actually is possible to make a digital signal that's smoother than an equivalent analog signal, the bumpiness just has a slightly different character. That's was my main point, and it applies to sounds coming out of digital synths as well as digital recorders.

The modelling is obviously a whole different issue, much more germaine I'm sure, but I personally don't really have anything to say about it. For what it's worth, I will say that Pspice will FAIRLY reliably tell me whether a circuit will work, but it does a pretty so-so job at predicting finer issues.

ned clayton wrote:

Jeremy wrote:


as far as synthetic instruments verses real instruments (the electric piano and the v-drums), your run down doesn't apply.

Yes and no, I think. I mean, I'm definitely totally embarassed by howmuch crap I wrote earlier about stuff that's totally well known, but I don't think it was completely off the subject. Because digital's so clearly quantized, people quickly come to the conclusion that it can never accurately represent a sound (that was said at least two or three times in this thread). But it's worth remembering that electrical analogs are also grainy, and at some point the analog breaks down. It actually is possible to make a digital signal that's smoother than an equivalent analog signal, the bumpiness just has a slightly different character. That's was my main point, and it applies to sounds coming out of digital synths as well as digital recorders.

The modelling is obviously a whole different issue, much more germaine I'm sure, but I personally don't really have anything to say about it. For what it's worth, I will say that Pspice will FAIRLY reliably tell me whether a circuit will work, but it does a pretty so-so job at predicting finer issues.

i actually read your entire post. it was definitely relevant and probably more informative than and of the typical 'look guys, use what you have bladdy bladdy bladdy' or the " gear doesn't matter, talent does bladdy bladdy bladdy" etc. type posts that come along with this conversation.

you'll have to forgive me, i've been in/read at least 17 analog verses digital threads on this forum, so i'm sort of typing from a 'didn't we go over this already?' stand point.

if i didn't have to wake up in 4 hours, i'd kerble all ya'll.

jeremy

Rodabod wrote:A big topic!

Digital is, and always will be an approximation of its analogue equivalent. It is inherently finite in its range of levels and will therefore never be exactly the same. Theoretically, the differences are very snall (especially with increased bit depths / rates), but still remains different. The big difference in sound between different a/d converters also adds to this difference.

I hate to argue, but the above is wrong. Nyqust proved it differently.
I'm not arguing what sounds better/worse analog/digital, but what you say is not correct.

Pohlman's "Prinicples of Digital Audio" chapters 2, 3 and 4 should cover it nicely.

ned clayton wrote:Totally. In fact the individual magnetic domains in tape are much much longer with respect to the wavelengths of sound being recorded than digital samples are, so if you had tape where the domains were perfectly lined up you'd hear a really grainy sound similar to low sample rate audio.

Huh? 20KHz is approx 1.7cm. The domains are about 20 um (I have no idea how to make that "micro" character show up, so I hope the u suffices)

some interesting, related links:

http://www.ndt-ed.org/EducationResource ... omains.htm
http://www.ee.washington.edu/conselec/C ... e/95x1.htm
The Complete Handbook of Magnetic Recording by Finn Jorgenson is also a good read on this topic.

goosman wrote:

Rodabod wrote:A big topic!

Digital is, and always will be an approximation of its analogue equivalent. It is inherently finite in its range of levels and will therefore never be exactly the same. Theoretically, the differences are very snall (especially with increased bit depths / rates), but still remains different. The big difference in sound between different a/d converters also adds to this difference.

I hate to argue, but the above is wrong.

You'll need to explain. Otherwise I could just respond, "you're wrong" too.

Nyquist.... You're not confusing sample rate, are you?

For a given digital system (ok, most digital audio systems), we allocate a fixed number of bits to represent analogue levels. Therefore, for a given system, the range of levels is finite. The resulting difference is not correlated with the audio, so is generally considered noise, but does have a nasty impact on the sound at lower bit-rates.

And as I said before, the qualities of A/Ds obviously have a big impact.

Rodabod wrote:
You'll need to explain. Otherwise I could just respond, "you're wrong" too.

Sorry, you're right, I should have said more, but I wanted to think about it some more and got distracted, so I just pressed post....

Rodabod wrote:
Nyquist.... You're not confusing sample rate, are you?

Well, I'm not confusing it, but the rate has a lot, actually everything, to do with it.

Rodabod wrote:
For a given digital system (ok, most digital audio systems), we allocate a fixed number of bits to represent analogue levels. Therefore, for a given system, the range of levels is finite. The resulting difference is not correlated with the audio, so is generally considered noise, but does have a nasty impact on the sound at lower bit-rates.

And as I said before, the qualities of A/Ds obviously have a big impact.

A signal that is bandlimited is constrained in terms of how fast it can change and therefore how much detail it can convey in between discrete moments of time. The sampling theorem means that the discrete samples are a complete representation of the signal if the bandwidth is less than half the sampling rate, which is referred to as the Nyquist frequency

The above is from http://en.wikipedia.org/wiki/Nyquist-Sh ... ng_theorem

Which goes into more detail about how a properly bandlimited signal can be completely reproduced.

If you're not too keen on trusting Wikipedia's data, you can find another explanation at http://www.lavryengineering.com/documen ... Theory.pdf

Dan Lavry is a very trusted name in the business of A/D/A converters.

From that pdf:
"Initial intuitive reaction may cause one to think that we do not have enough X's to be able to replot the original wave (red) with all of its details. That intuitive reaction is wrong. The key here is fact that the wave form is band limited. For a given bandwidth, the number of samples (X's)
need only to exceed twice the bandwidth in order to be able to retrieve the complete waveform, including any value between the sample times. Let us see how it is done."

And then he goes on to explain it.

It's pretty heavy stuff, so it's unfortunately not easily explained, which is why I think a lot of people don't get it quite right.

Again, I'm not saying that it sounds good, bad, etc. but given a properly designed system that conforms to the known limits of the Nyquist-Shannon Theorem you will get out exactly what was put in.

This is not new stuff, the core dates back to the 20's and 30's and is quite well understood now. Most A/D and D/A systems used in music today (I'll qualify it a bit and say above a certain price point) are properly designed.

Mathematically a signal sampled at 40 kHz has all the original information of frequencies up to 20 kHz. Unfortunately this same process of sampling *adds* bogus information above 20 kHz.

So in the real world of engineering even if you've done a perfect job of A/D conversion, the D/A conversion will have to include a final filter that lets all the information below 20 kHz through, and blocks all of the information above 20 kHz.

There are numerous strategies for dealing with this problem, but the need for this "brickwall" filter is a big part of the problem going from theory to practice in terms of sampling theory.

goosman wrote:

Rodabod wrote:
Nyquist.... You're not confusing sample rate, are you?

Well, I'm not confusing it, but the rate has a lot, actually everything, to do with it.

Ok, without trying to sound patronising, I think you have perhaps missed my point regarding bit depth / quantising. Sample rate is most definitely not "everything to do with it".

Example - try recording at any sample rate you like, say 48KHz, 96KHz, etc. and record using an 8-bit bit depth. How does that sound? Good? Bad? Metallic? Grainy? The reason - not to do with sample rate!

Sample rate only affects bandwidth as far as we are concerned, and yes, in theory, we can reproduce frequencies up to half of the sample rate. That's cool.

But!....

galanter wrote:Mathematically a signal sampled at 40 kHz has all the original information of frequencies up to 20 kHz. Unfortunately this same process of sampling *adds* bogus information above 20 kHz.

So in the real world of engineering even if you've done a perfect job of A/D conversion, the D/A conversion will have to include a final filter that lets all the information below 20 kHz through, and blocks all of the information above 20 kHz.

There are numerous strategies for dealing with this problem, but the need for this "brickwall" filter is a big part of the problem going from theory to practice in terms of sampling theory.

I had actually forgotten to mention this (although this thread started a while back now...) and it is a very good point.

Whilst everything is generally good in theory, we still can't implement our filters perfectly as we would like.

Again, better deisgned A/Ds sound better!

galanter wrote:There are numerous strategies for dealing with this problem, but the need for this "brickwall" filter is a big part of the problem going from theory to practice in terms of sampling theory.

While this may be true, you still get a perfectly reproduced signal out of the system. This is a different argument than the one Rodabod was making.

And just to keep the terms straight, it's the Nyquist-Shannon Theorem.
A theorem being a mathmatically provable statement, rather than a theory which could be viewed as an unproven opinion of how things may (or may not) work. I.e., we can prove what Nyquist and Shannon proposed to be true.

And to cleanse my palate: I like rock music made with analog methods.