Topic: sony pcm-m10 dynamics/quantisation question

[gmarinov]

hi guys.
first, thanks for a great forum, i've been lurking for a while.
i have what i believe is a more advanced question, and one that guysonic might be able to answer.

I recently did a 24/96/wav recording of a passing train using the pcm-m10, with external microphones placed very close to the tracks. The mics were Core Sound binaurals and had their own 9V power supply. The nature of the recording necessitated low sensitivity and level of 2 or so. It peaked to -6dB so that was fine.

When I opened the recording in a wave editor, I was interested to see what the quietest level of recorded sound was. Zooming in quite a bit I saw that there's significant quantisation taking place at -65dB (FS). This is roughly in line with the findings on Avisoft Bioacoustics on dynamic range but in a roundabout way?

I wonder not only what that means for bit-depth in recording, but also what the reason for this could be? The input level control is analog to my awareness?I know I shouldn't expect a machine of this caliber to handle -65dB in outstanding manner, but still.. any ideas?  The digital limiter was engaged on this. (that's another 12dB of protection, and I know it's 20dB on the D50). Could that be related?

feel free to throw technical / engineering terms at me.
thanks in advance.

[andromedanwarmachine]

sorry-what are you actually querying gmarinov?

Are you saying there was alot going on at -65dB and you're wondering why, or that because there was alot going on you don't think the machine will have the resolution to be able to accurately record further up the dynamic range?

And as mshilarious points out, the audio signature of quantisation would not really be enough to comment accurately on how much work was going on.

Is that what you mean?

[gmarinov]

hi guys

I mean that at -65dB the waveform looks quantised as if the bit depth was 4 bit not 24.
-65dB is some 11 bits taken off, this doesn't look like 13 bits to me:

[gmarinov]

(at 6dB/bit, 24 bits = 144dB theoretical dynamic range. at -65dB, a signal only uses the rightmost 13 bits. )

The software is not at fault. I normalised a 24/96 file to -60dB, saved it, and reopened it in the same app, with the waveform looking just fine.
The plot above is coming straight from the recorder. Notice the settings: low sensitivity, rec level at 2 or less. also, I suspect the sine wave smoothing is not the software..

I did some further tests. This looks like digital noise which falls outside of the 20-20k spectrum. A lowpass EQ at 20khz immediately knocked 10dB off the meters and restored the waveform to a much more normal-looking state:



this was done with external microphones w/ plug-in power switched off. I cannot remember the actual circuitry that powers them. I'm guessing I now need to find whether this is coming from inside the recorder or from the external mics. I've ruled out interference (tested w/ outdoor recordings away from any other devices or power lines).

Need.. more... tests.

[gmarinov]

That's a bit more interesting.  Run a FFT on the 96kHz file.

it's a straight line from 0 to 20khz, but above that:



edit:



LP @ 30kHz, Q 3-4 seems to hit a sweet spot..

[gmarinov]

Again, try FFT rather than the spectrogram, that will show oscillation vs. broadband noise more clearly.

It's likely there is switching power supply noise at ultrasonic frequencies.

I did try FFT, no specific frequency stands out, but the hiss gradually goes up in level from 30k and onwards. I don't think there are any distinct oscillations or pulses.
When pitch-shifted down it sounds like an old-school modem.
There was no switching power supply (or any other source of interference) in proximity in any of the recordings, and the microphones use Panasonic WM capsules, with (to my awareness) simple 9V circuit power supply box.. The LCD backlight was off and so were the LEDs.

[gmarinov]

Then it's just broad spectrum noise which probably isn't an issue, unless you were counting on ultrasonic response above 30kHz.

The switching power supply is inside the recorder; they will typically generate several different voltage rails from the battery voltage.

That's the thing. I do count on ultrasonic response above 30kHz more often than not..
Wondering if the line-in will have the same issue.

[andromedanwarmachine]

I can't lie to you- this is out of my depth!!

But I can say, relative to your last comment, that I ran a Sony TCD D3 once which very definitely introduced an HF component into analogue sourced material when the backlit screen was on...

[gmarinov]

if anyone else has done (or is doing) similar tests, especially with regards to high frequency response, i'd appreciate if you shared your findings. thanks.