[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[linrad] Re: Rate decoupling and sound card bandwidth in general

> Resuming on my previous email I would like to add, that the change
> in sampling rate is certainly not intended to change the generality
> of 'Linrad' but just a work-around for the fact that the output rate
> cannot be decoupled from the input rate for certain hardware.
Yes, understood - but how should Linrad know what the user wants?
Adding yet another parameter will increase confusion among new users
so I have to take the decisions once and for all. Before doing that
I would like to know for sure that this really is a problem;-)

> Surely
> the default option would be 48000/48000, but if that is unreasonable
> (with regard to unnecessary load on the CPU), then a new output rate
> would be available, with the appropriate 'Linrad' routines to get
> corresponding undistorted output. For many effective solutions it
> would save the purchase of a second sound card, though actually
> the cheapest possible CPU (maybe a Duron 1800 at present) would
> have no trouble in producing the extra cranking to provide solid
> 48000 kHz samples even with the present "legacy" calculations
> (interesting to see what it would mean for 192 kHz).
I think that anything you can buy today is much faster than necessary,
the problem only comes up on 486 and early Pentium computers.

On my old 266MHz Pentium MMX, the time consumed by all the baseband
processing is like this (the load from the actual writing to the
sound device and for the sound device itself is not included)

Rate    Load
6000     16%
12000    19%
24000    24%
44100    32%

Linrad only runs properly for a 6000 sampling rate with the WSE units.
To get the load from the output processing at higher rates I had
to deselect the second fft and remove the window from the first fft.

Running normally with 2x96 kHz bandwidth and the WSE units the load
is about 95% and going above 6000 on the output is a problem.

On the other hand, with an input of 2x48 kHz only the 266 MHz machine
should have no problem to run 48 kHz for the output.

All of the above is with one channel for the output.

> My knowledge of PC sound cards does not go far enough to know the
> general state of affairs with those, but since they are sound cards
> relative to human hearing, I have so far assumed that it would be
> more than fair that there would be a filter limiting the bandwidth
> to 20 kHz, even if over-sampling (up to 192 kHz) would be used.
No. The HIFI enthusiasts appreciate a perfect phase response.
adding a simple filter that would attenuate somewhat at say 40 kHz
would give a small phase shift at 16 kHz. A more complicated
filter would need inductors or op-amps and would degrade
performance or cost a lot of money.

The solution is that sampling is done at for example 64 times higher
frequencies, many MHz, and perfect phase response is then obtained
in a digital filter. Luckily for us, those filters are designed to
work well at the lowest sampling speed (8kHz) where 4kHz is sent
into the PC. When sampling at 96 kHz the same filter is used and
therefore the full bandwidth is obtained:-) I think the HIFI enthusiasts
value high bandwidth and since it is what comes out automatically
at minimum cost that is what we get:-)

> However, as the Delta-44 is used for a 96 kHz bandwidth in 'Linrad',
> this implies at least a 48 kHz input bandwidth. I would like to be
> informed whether this is natural for any card (or specifically for
> the Delta 44), or whether it is a user-inflicted modification on
> the Delta-44 circuit board.
As far as I know, full bandwidth is standard for all cards.

> It is good to hear about the new OSS-
> drivers allowing further pushing to 192 kHz, but does it also mean
> that the analog section of the Delta-44 board is ready to accept
> this higher input bandwidth? As such the RMAA 5.4 quality tests
> under Windows has shown 20 kHz bandwidth limiting for most cards.
192 kHz on a Delta44 using OSS is just silly. The spectrum does not
contain any strong signal outside the 96 kHz bandwidth. What comes
outside is aliases from the 96KHz bandwidth attenuated by the
filtering of the upsampling routine.

> http://audio.rightmark.org/products/rmaa.shtml

I think these results are completely irrelevant since the output
amplifier of the soundcard is included in the signal path. The
A/D converters are probably not influencing the results at all.

> My Delta-44 is still in the box, as I have only studied integrated
> audio solutions lately, in order to have a precursor for more
> serious upcoming 'Linrad' work (including using the WSE-units for
> 70 MHz cross-band reception; coincidentally being the 1st WSE-IF).
>                                           73, "Zaba" OH1ZAA/NNoY
> P.S. After re-reading your full reply I was not sure whether the
>       OSS 192 kHz settings were useful for CPU-load tests only, or
>       that they would allow real 192 kHz sampling. Also it was very
>       unclear to me what you mean by unhappy distortion under 4 kHz
>       (most of us are probably unhappy with distortion, whatsoever).
So now you know. 192 kHz is completely useless and not supported by
Linrad for a Delta44. You have to fool the Linrad setup routine by
manually editing the parameter file to see that there is a rate conversion
built into the OSS package.

I do not care about distortion above 4 kHz. I am an old man and I do
not hear much above such a high frequency;-)


Leif / SM5BSZ

This message is sent to you because you are subscribed to
  the mailing list <linrad@xxxxxxxxxxxxxxxxxxxxx>.
To unsubscribe, E-mail to: <linrad-off@xxxxxxxxxxxxxxxxxxxxx>
To switch to the DIGEST mode, E-mail to <linrad-digest@xxxxxxxxxxxxxxxxxxxxx>
To switch to the INDEX mode, E-mail to <linrad-index@xxxxxxxxxxxxxxxxxxxxx>
Send administrative queries to  <linrad-request@xxxxxxxxxxxxxxxxxxxxx>