Phono Preamp 44.1 vs 96

[Marc Hildebrant]

Group,

When I'm listening to a song through the Phono Preamp filter (flat to RIAA), it seems to me that the same song sounds different if its 96 kHz vs 44.1 kHz.

The setup...I take a song recorded at 44.1 and then create a version at 96 using the change sample rate filter. Then I play both back through the same Phone Preamp setting.

It seems that the 96 version is louder ?

Marc

[Craig Maier]

Hi,

You can run a test to see as follows:

Make a white (random) noise file at 44.1 kHz and save it

Make a white noise file at 96 kHz and save it.

Play each through the flat to RIAA curve and observe the signal output on the VU meter, time display, spectrum analyzer, etc. That will show if there is a difference.

Craig

[Marc Hildebrant]

Craig,

Experiment #1

Create white noise file at 44.1 kHz. Measure average spectrum value and VU meter peak.

Change file to 96 kHz via resample option. average spectrum value is 1 dB higher, and VU meter peak is 4 dB higher. Waveform has more noise on top.

Marc

[Craig Maier]

Marc,

Do you get the same result if you make two separate files directly via the Make Waves Generator rather than re-sampling?

Craig

[Marc Hildebrant]

Craig,

If I make separate files at 44.1 & 96 they have the same amplitude.

It seems to be the conversion from one sample rate to the other. By the way, if I start at 96 and convert down to 44.1, the peak value on the VU meter changes and the top of the waveform has a "rougher" look to it.

96 to 44.1 isn't an integer relationship...most likely its a minor problem with interpolation.

Marc

[Dan McDonald]

I've noticed slight differences over the years. I think at one point, there was a suggestion that 88.2 might be a better resolution to use if you're converting to 44.1 as the end result.

[Craig Maier]

Hi,

Just as a general mathematical concept, the rectified average (the VU meter) or the RMS value of a given frequency span of white noise will be different depending on the bandwidth of the signal source. Put another way, white noise signals are defined as having random amplitude and random frequency from "DC to daylight". But, that is not possible to have in real systems. All real systems are bandwidth limited. So, the more bandwidth limitation applied to random (white) noise, the lower the rectified average and/or RMS values of said noise will be.

Does that make sense? It is a bit confusing, but that is part of the math underpinnings of this sort of a test.

It would be interesting to see if there is a difference before and after file conversion using a 1 kHz pure sine wave. In that case, the signals should measure the same before and after file sample rate conversion. If they are not the same, then there is a bug in the sample rate converter.

Craig

[Craig Maier]

Hi Again,

I just ran the file converter on a pure 1 kHz sine wave (both directions). I see no amplitude deviations due to the file sample rate converter. Thus, I conclude the difference with white noise is the power spectral density issue that I mentioned earlier (RMS and Rectified average values of noise will vary with the bandwidth of the system through which it passes).

Craig

[Marc Hildebrant]

Craig,

A pure tone will not show the effect (if present). I would suggest that you try a music file.

Marc

[Craig Maier]

Hi,

Well, music, when averaged over long time intervals, is basically weighted noise in a mathematical sense. So, it is possible that the power spectral density effect of system bandwidth on rectified average (or RMS) measured signal could be present with musical content just as can be observed with white noise. Remember that musical recordings can contain ultrasonic signals which will be maintained in a 96 KHz sampled signal with greater spectral density than musical signals recorded at 44.1 kHz.

Craig