laurent.artaud[AT]free.fr wrote:
...

> Note that (if I wasn't clear) a function generator is an electronic 
> device (I don't have any, it's quite expensive) used in laboratories to 
> test other electronic devices. It have nothing to do with any software 
> in the sense that the signal stays analogous all the way and is in no 
> way digitalized.
...

Please note that many modern function generators are indeed
software "driven". I.e. a uP or a DSP generates the waveform
which is then DA-converted and filtered.


-- 
Tor Olav
http://subcube.com
http://subcube.net

Post a reply to this message

> It certainly is a tricky area this. I know I have seen banding in
> images, and I know the solution is greater colour depth - but boy is
> it ever hard to come up with an example. Believe me I've tried. 
> 
> Now even though I've failed to come up with an image example does not
> mean it does not exist. A quick Googling will show that plenty of
> other people think they way I do. Mind a lot think otherwise too.

Yes, there are a lot of (24bit) images that show banding. Especially 
gray scale images are problematic: 256 shades of gray are simply not 
enough in many cases. This is a case when traditional B&W print still 
excels - especially when using medium or large format cameras. Color 
gradations simply look better. I don't think there are many printers (or 
video cards) that accept 48bit data - does anyone know?

Severi S

Post a reply to this message

laurent.artaud[AT]free.fr" <"laurent.artaud[AT]free.fr wrote:
> > I will have to try that one out when I get some sound editing
> > software installed (just reinstalled Windows).  What CD player did
> > you use to play back the audio CD?
> >
>
> I did not test it myself, sorry if I may have implied it unknowingly.
> Anyway, my explanation of the process was simplified for a text demo.
>
> Note that (if I wasn't clear) a function generator is an electronic
> device (I don't have any, it's quite expensive) used in laboratories
> to test other electronic devices. It have nothing to do with any
> software in the sense that the signal stays analogous all the way and
> is in no way digitalized.
>
> On the other end, using a software to generate the waveforms to create
> the CD is a far better way than trying to sample the function
> generator's output.

Indeed, I was thinking about digital generation of signals, sonic Ray
Tracing :-)

> Testing the 24/96 signal would be harder:
> 1) you have to find a software who can generate it ;
> 2) you have to play it.

Yeah, I have software and hardware to play 24/96 from my PC, it's just I
doubt my speakers could keep up [just looking], apparently they are
down -3dB at 22kHz, so who knows what they'll be down to at 48kHz.

Post a reply to this message

On Tue, 09 Mar 2004 22:37:18 +0000, scott wrote:

> laurent.artaud[AT]free.fr" <"laurent.artaud[AT]free.fr wrote:
>> > I will have to try that one out when I get some sound editing
>> > software installed (just reinstalled Windows).  What CD player did
>> > you use to play back the audio CD?
>> >
>>
>> I did not test it myself, sorry if I may have implied it unknowingly.
>> Anyway, my explanation of the process was simplified for a text demo.
>>
>> Note that (if I wasn't clear) a function generator is an electronic
>> device (I don't have any, it's quite expensive) used in laboratories
>> to test other electronic devices. It have nothing to do with any
>> software in the sense that the signal stays analogous all the way and
>> is in no way digitalized.
>>
>> On the other end, using a software to generate the waveforms to create
>> the CD is a far better way than trying to sample the function
>> generator's output.
> 
> Indeed, I was thinking about digital generation of signals, sonic Ray
> Tracing :-)
> 
>> Testing the 24/96 signal would be harder:
>> 1) you have to find a software who can generate it ;
>> 2) you have to play it.
> 
> Yeah, I have software and hardware to play 24/96 from my PC, it's just I
> doubt my speakers could keep up [just looking], apparently they are
> down -3dB at 22kHz, so who knows what they'll be down to at 48kHz.

That's not the point. It's not like 44khz gives perfect signal at anything
below 22khz, and nothing above it.

A 22khz wave form at 44khz looks like this:
 * * * * * * * * * * *
------------------------------
* * * * * * * * * * * *

Regardless of the actual shape of the wave.

Post a reply to this message

Tyler Eaves wrote:
> On Tue, 09 Mar 2004 22:37:18 +0000, scott wrote:
>
> > laurent.artaud[AT]free.fr" <"laurent.artaud[AT]free.fr wrote:
> > > > I will have to try that one out when I get some sound editing
> > > > software installed (just reinstalled Windows).  What CD player
> > > > did you use to play back the audio CD?
> > > >
> > >
> > > I did not test it myself, sorry if I may have implied it
> > > unknowingly. Anyway, my explanation of the process was simplified
> > > for a text demo.
> > >
> > > Note that (if I wasn't clear) a function generator is an
> > > electronic device (I don't have any, it's quite expensive) used
> > > in laboratories to test other electronic devices. It have nothing
> > > to do with any software in the sense that the signal stays
> > > analogous all the way and is in no way digitalized.
> > >
> > > On the other end, using a software to generate the waveforms to
> > > create the CD is a far better way than trying to sample the
> > > function generator's output.
> >
> > Indeed, I was thinking about digital generation of signals, sonic
> > Ray Tracing :-)
> >
> > > Testing the 24/96 signal would be harder:
> > > 1) you have to find a software who can generate it ;
> > > 2) you have to play it.
> >
> > Yeah, I have software and hardware to play 24/96 from my PC, it's
> > just I doubt my speakers could keep up [just looking], apparently
> > they are down -3dB at 22kHz, so who knows what they'll be down to
> > at 48kHz.
>
> That's not the point.

It is the point, I was saying that my speakers couldn't reproduce
frequencies higher than 22kHz accurately.

> It's not like 44khz gives perfect signal at
> anything below 22khz, and nothing above it.

Not perfect, but perfect in terms of most peoples' ears.  You *cannot*
sample a frequency higher than 22kHz at a 44kHz sampling rate - any analogue
to digital converter will filter higher frequencies out before it samples.

> A 22khz wave form at 44khz looks like this:
>  * * * * * * * * * * *
> ------------------------------
> * * * * * * * * * * * *
>
> Regardless of the actual shape of the wave.

That's not entirely true, the 22kHz must be put through an anti-alias filter
first before it is sampled.  And guess what?  That cuts off anything above
22kHz.  Hence, the actual analogue wave you are sampling will be exactly the
same whether it was a square wave, triangle wave, sine wave or whatever to
start with.

When you convert it back to analogue, it will come out as a 22kHz sine wave,
which is identical to what was _actually_ sampled.

Post a reply to this message

> 
> Please note that many modern function generators are indeed
> software "driven". I.e. a uP or a DSP generates the waveform
> which is then DA-converted and filtered.
> 
> 

I know.
But they are more expensive and for the test I was thinking about a full 
analog one would be enough.

-- 
Laurent ARTAUD (lau### [at] freefr)

Post a reply to this message

> Yeah, I have software and hardware to play 24/96 from my PC, it's just I
> doubt my speakers could keep up [just looking], apparently they are
> down -3dB at 22kHz, so who knows what they'll be down to at 48kHz.
> 

Well, if you have the software, try this:
- create a sin waveform at 20KHz and save it in a 16/44 format
- create a sin waveform at 20KHz and save it in a 24/96 format
- at 24/96, import the two audio files and compare them, the difference 
should be obvious (if not, try it with 22KHz waveforms).
- if you want to have fun, sub the 24/96 by the 16/44 and play the 
result (at 24/96). Depending of your ears and your hardware, you may be 
able to hear the difference.

Regards,

-- 
Laurent ARTAUD (lau### [at] freefr)

Post a reply to this message

laurent.artaud[AT]free.fr" <"laurent.artaud[AT]free.fr wrote:
>> Yeah, I have software and hardware to play 24/96 from my PC, it's
>> just I doubt my speakers could keep up [just looking], apparently
>> they are down -3dB at 22kHz, so who knows what they'll be down to at
>> 48kHz.
>>
>
> Well, if you have the software, try this:
> - create a sin waveform at 20KHz and save it in a 16/44 format
> - create a sin waveform at 20KHz and save it in a 24/96 format
> - at 24/96, import the two audio files and compare them, the
> difference should be obvious (if not, try it with 22KHz waveforms).

No.  There is no difference in the waveforms.  See my posts in
povray.binaries.misc.  Or download Adobe Audition demo version to try out
for yourself.

[1] This is a 20kHz sine wave saved and then loaded in 44.1 / 16-bit format
[2] This is file [1] converted to 96 / 32-bit format, then saved and loaded
[3] This is a 20kHz sine wave generated in 96/32 format, saved and loaded
[4] This is the difference between [2] and [3], computed in 32-bit format.

Please note that in [4] the scale goes from -32000 to 32000, so we are
seeing an error of between 1 and 2 per 32000, ie -84dB.  Theoretically there
should be no error, but I had choose a quality setting when I upsampled in
stage 2, so I'm sure that effects the result.

> - if you want to have fun, sub the 24/96 by the 16/44 and play the
> result (at 24/96). Depending of your ears and your hardware, you may
> be able to hear the difference.

I can't hear the -84dB signal on my machine at work, I'll try it at home but
I'd imagine I would need to have my amplifier up *far* higher than I do
normally.

Post a reply to this message

> 
> No.  There is no difference in the waveforms.  See my posts in
> povray.binaries.misc.  Or download Adobe Audition demo version to try out
> for yourself.

Silly me! Of course there is no difference: the filters keep only the 
fundamental, which is a sin at 20KHz! It's a triangular waveform, not a 
sin waveform you have to use.
ARGH!
I'm starting to loose myself!
I can't remember how the test was supposed to be done!
Was it 20KHz or 22KHz? Was it a triangular or a square waveform?

(I can't test it myself: I use Linux and have no sound editing softwares 
installed (I'll have to find one to try, but I will not be able to do 
that before this WE...))
(I can't find p.b.m...)

> 
> Please note that in [4] the scale goes from -32000 to 32000, so we are
> seeing an error of between 1 and 2 per 32000, ie -84dB.  Theoretically there
> should be no error, but I had choose a quality setting when I upsampled in
> stage 2, so I'm sure that effects the result.

Well, In my opinion, this error is just a rounding error during the 
conversions... It is not the one I was talking about.

> 
> I can't hear the -84dB signal on my machine at work, I'll try it at home but
> I'd imagine I would need to have my amplifier up *far* higher than I do
> normally.
> 

Well, if you can hear this signal, you have a really really really 
hi-end audio system. Most of the signal/noise ratios are around 70-80dB, 
so you might hear the amplifier's noise better that the signal.

Regards,

-- 
Laurent ARTAUD (lau### [at] freefr)

Post a reply to this message

laurent.artaud[AT]free.fr" <"laurent.artaud[AT]free.fr wrote:
>> No.  There is no difference in the waveforms.  See my posts in
>> povray.binaries.misc.  Or download Adobe Audition demo version to
>> try out for yourself.
>
> Silly me! Of course there is no difference: the filters keep only the
> fundamental, which is a sin at 20KHz! It's a triangular waveform, not
> a sin waveform you have to use.
> ARGH!
> I'm starting to loose myself!
> I can't remember how the test was supposed to be done!
> Was it 20KHz or 22KHz? Was it a triangular or a square waveform?

Assuming you filter at 48kHz for a 96kHz sampling rate, a 20kHz square wave
will appear as a 20kHz sine wave because the next harmonic in a square wave
is at 3 times the fundamental.  Can't remember what it is for a triangle
wave though.  Will test out this evening listening to them to see if I can
tell any difference :-)

>> Please note that in [4] the scale goes from -32000 to 32000, so we
>> are seeing an error of between 1 and 2 per 32000, ie -84dB.
>> Theoretically there should be no error, but I had choose a quality
>> setting when I upsampled in stage 2, so I'm sure that effects the
>> result.
>
> Well, In my opinion, this error is just a rounding error during the
> conversions... It is not the one I was talking about.

Yes, I think it is rounding error too.

>> I can't hear the -84dB signal on my machine at work, I'll try it at
>> home but I'd imagine I would need to have my amplifier up *far*
>> higher than I do normally.
>>
>
> Well, if you can hear this signal, you have a really really really
> hi-end audio system. Most of the signal/noise ratios are around
> 70-80dB, so you might hear the amplifier's noise better that the
> signal.


claimed specs:

Frequency response: 10Hz - 80kHz -3dB
SNR: 98dB

Unfortunately I don't think my speakers can keep up with that! (claimed
40-22kHz -3dB)

One thing that I do notice is that if I turn my amp up to full while
connected to my 2496 soundcard I hear no hiss or noise at all.  Switch it to
my SoundBlaster Live and wham! there is so much hiss, even when all outputs
are muted in windows ;-)

Post a reply to this message