Audacity Forum

RustyL wrote:If Audacity is set to a project rate of 44100, I found out that my Sound Blaster card actually captures at 48k and internally converts to 44.1. In so doing, the clips have been converted to a value less than Audacity would display with red bars. Changing to 48000 project (capture) fixed my problem.

It is often helpful to set the Audacity project rate to a rate supported by your sound device.

Resampling properly done should not change the peak amplitude of the waveform.

Gale

I'm not resampling when I encounter the problem. My projects were set up to capture at 14.1 thinking that if I'm going to burn CD's I should uses the same rate and that it would result in smaller file size. The SB card's ADC evidently is locked to it's highest sampling rate but then internally, SB converts its output to the requested value such as 14100. No user controls are available to internally control ADC rate. Little do you know that the card is buzzing along at its highest rate, internally converting to the requested sample rate, and passing along rounding errors that may have occured. Perhaps they intentionally round to prevent divide by zero. The rounding is downward resulting in a clipped value of less than a linear maximum +/-1.000.

Here is data of a wave form clipped 10 times, Audacity doesn't redline it. The first two and last two samples are not part of the flat ;

sample-data5.txt 1 channel (mono)
Sample Rate: 44100 Hz. Sample values on linear scale.
Length processed: 14 samples 0.00032 seconds.
0.88046
0.90048
0.93143
0.93546
0.93515
0.93613
0.93661
0.93750
0.93829
0.93765
0.93866
0.93036
0.91104
0.90112


Maybe someone with more knowledge can explain and or verify this anomaly with SB and possibly other cards and formulate an FAQ response. I wasn't able to solve this problem until I searched SB and other places on the web.

Rusty

RustyL wrote:My projects were set up to capture at 14.1 thinking that if I'm going to burn CD's I should uses the same rate and that it would result in smaller file size. The SB card's ADC evidently is locked to it's highest sampling rate but then internally, SB converts its output to the requested value such as 14100.

Presumably 14.1 and 14100 are typos for 44.1 kHz and 44100 Hz?

Gale

RustyL wrote:I'm not resampling when I encounter the problem.

But you are saying SoundBlaster is resampling, correct?

RustyL wrote:'Here is data of a wave form clipped 10 times, Audacity doesn't redline it. The first two and last two samples are not part of the flat ;

sample-data5.txt 1 channel (mono)
Sample Rate: 44100 Hz. Sample values on linear scale.
Length processed: 14 samples 0.00032 seconds.
0.88046
0.90048
0.93143
0.93546
0.93515
0.93613
0.93661
0.93750
0.93829
0.93765
0.93866
0.93036
0.91104
0.90112

Where does that waveform come from and what are the original values? Are you trying to record a full-scale or clipped tone playing in Audacity?

RustyL wrote:Maybe someone with more knowledge can explain and or verify this anomaly with SB and possibly other cards and formulate an FAQ response. I wasn't able to solve this problem until I searched SB and other places on the web.

It's not a "frequently" asked question, and not an Audacity problem if I understand you, though it's interesting if true.

What bit-depth are you recording at? What version of Windows are you using? What version of Audacity are you using? What host are you choosing in Audacity's Device Toolbar? You may be able to set SoundBlaster in Windows Sound to use 44100 Hz Default Format (on Windows Vista and later).

Please post the online references you found as to why SoundBlaster adjusts sample values down when resampling or when using higher sample rates.


Gale

Thanks Gale,
Sorry about the typo. Yes I meant 44,100 and not the old telephone modem rate reference). For newbies, 44,100 hz (hertz or samples per second) is the same as 44.1 khz (kilo-hertz).
Originally I thought setting a capture rate of 44,100 and depth of 16 bits might be best since this is standard wave format of audio CD's and is the final destination of my music. At that rate I had a bad result of the red line clipping indicators didn't show. I have since read that the software in windows actually does math and transformations quicker and more accurately in the 32-bit floating point rate. Probably due to the onboard math coprocessor. So, my preferences are to set my projects at 48000 capture rate and a bit-depth of 32-bit rate. I'm guessing this costs in file size.

My other settings;
Audio host: Windows DirectSound
Output decic: SB Live!
Input device: SB Live! - Line In

On my computers with SoundBlaster cards I'm running Windows XP with all updates for windows as well as the latest updated drivers from Creative SoundBlaster for the card.
You suggested "You may be able to set SoundBlaster in Windows Sound to use 44100 Hz Default Format (on Windows Vista and later)." That isn't the problem because the sound card does allow it's output to be configured to deliver 44,100/16 bit output to Windows and the Audacity application. But, and I was unable to find any direct reference from Creative, the card actually samples at a higher rate and internally does the conversions with round down errors.

A suggestion to the Audacity development team is to allow the red-line clipping point to be configurable to a value to compensate for in-line hard and soft clipping points.

I will continue to look for a knowledgable reference to the SoundBlaster condition.

Thanks, Rusty

RustyL wrote:If Audacity is set to a project rate of 44100, I found out that my Sound Blaster card actually captures at 48k and internally converts to 44.1. In so doing, the clips have been converted to a value less than Audacity would display with red bars. Changing to 48000 project (capture) fixed my problem.

I've only skimmed this topic, but that sounds like the old EMU10K1 / EMU10K2 chips. They were very popular for mid-range Creative sound cards for many years (I've still got a couple of old AWE 32 cards somewhere). The same chips were used in more expensive EMU sound cards. Internally they run at 48 kHz and include on-board DSP (digital signal processing) which is best avoided if you want to do straight recording / playback (but very useful if your interested in sound synthesis).

The drivers for those cards were something of a let-down, but for anyone with > 80% geek rating there were alternative drivers available from the KxProject (http://www.kxproject.com/).

As you describe, one common quirk was that they clipped below 0 dB. It was possible to avoid that by carefully tweaking the settings (some of which are hard to find). If I remember correctly, the optimum settings on the AWE 32 was to set all levels at 72 %.

Another quirk of several in that range (including the AWE 32) was that the rear line output was higher fidelity than the front line outputs (the Kx drivers swapped the front and rear outputs so that the "main" outputs were from the rear output socket (so as to take advantage of the better sound quality).

There's a fair bit of technical information on the KxProject website (better to read it than rely on my memory - this is going a long way back in history)

RustyL wrote:my preferences are to set my projects at 48000 capture rate and a bit-depth of 32-bit rate. I'm guessing this costs in file size.

32-bit float is the default for Audacity and recommended. Yes 48000 Hz project rate takes about 9% more disk space than 44100 Hz.

RustyL wrote:I'm running Windows XP with all updates for windows

Please see http://forum.audacityteam.org/viewtopic ... 46&t=84123 for more information about Windows XP and Audacity.

RustyL wrote:A suggestion to the Audacity development team is to allow the red-line clipping point to be configurable to a value to compensate for in-line hard and soft clipping points.

OK we'll count that as a "vote" for that enhancement.


Gale

steve wrote:

RustyL wrote:If Audacity is set to a project rate of 44100, I found out that my Sound Blaster card actually captures at 48k and internally converts to 44.1. In so doing, the clips have been converted to a value less than Audacity would display with red bars. Changing to 48000 project (capture) fixed my problem.

I've only skimmed this topic, but that sounds like the old EMU10K1 / EMU10K2 chips. They were very popular for mid-range Creative sound cards for many years (I've still got a couple of old AWE 32 cards somewhere). The same chips were used in more expensive EMU sound cards. Internally they run at 48 kHz and include on-board DSP (digital signal processing) which is best avoided if you want to do straight recording / playback (but very useful if your interested in sound synthesis).

The drivers for those cards were something of a let-down, but for anyone with > 80% geek rating there were alternative drivers available from the KxProject (http://www.kxproject.com/).

As you describe, one common quirk was that they clipped below 0 dB. It was possible to avoid that by carefully tweaking the settings (some of which are hard to find). If I remember correctly, the optimum settings on the AWE 32 was to set all levels at 72 %.

Was that tweak possible without using the Kx drivers, do you recall?


Gale

Gale Andrews wrote:Was that tweak possible without using the Kx drivers, do you recall?

Setting levels for undistorted recordings producing a 0 dB waveform was possible with standard drivers, provided that the audio source did not overload the soundcard inputs. I believe that was the case even after I upgraded to Win XP (from Win 98).

Some followup of my previous post; the article at Wikipedia on the SoundBlaster Live! http://en.wikipedia.org/wiki/Sound_Blaster_Live! does a good job of explaining the shortfalls of the EMU10K1 chip used in this and other SB cards. It notes the DSP internal sampling rate is fixed at 48000 and that the card has audible problems in the rate-conversion step which can be overcome by converting as a seperate process (i.e. capture at 48000 and allow Audacity to export the results to other rates such as 44100 for burning to CD).
I have looked for the third-party drivers at Kxproject (http://www.kxproject.com/) but run into a "502 Bad Gateway" error when clicking on the filename. Anyone have any suggestions?
Many thanks for the feedback,
Rusty