naturerecordists
[Top] [All Lists]

Rolls PB224 noise floor (estimate)

Subject: Rolls PB224 noise floor (estimate)
From: "jpbeale" jpbeale
Date: Fri Sep 15, 2006 12:10 pm (PDT)
--- In  Rob Danielson <> wrote:
> Eric[...] He measured the noise floor with two MK 2/CMC6's [11
dB(A), > 15 mV/Pa] and found it to be -120 dBV.

For what it's worth, I estimate the Rolls contribution to be -134 dBV
at the switching frequency (well beyond audible range), assuming two
150 ohm mics drawing 10 mA each. Lower impedance mics and lower
current draw would mean less noise.

I don't have a Rolls and I haven't measured one. This is based on
a first-principles analysis of the schematic online at
http://www.rolls.com/data/pb224man.pdf  as outlined below.
---------------------------------------

Unknown: Q1 switching frequency Sf.
Noise is inversely proportional to Sf.
Sf assumed to be 50 kHz. Would have less noise if it is higher.

Unknown: mic supply current Is.
Noise is directly proportional to Is, assumed to be 20 mA
(worst case for phantom power: 10 mA x 2)

Unknown: mic impedance Z.
Noise is inversely proportional to mic impedance (for Z<<6.8K)
Z assumed to be 150 ohms at frequency of interest.
---------------------------------------

peak ripple Voltage at D2-C11 node:
I =3D C dV/dt, so  (I/C)*dt =3D dV
(20 mA / 47 uF) * (20 us) =3D 8.5 mV (at 50 kHz)

R7 & C6 form a one-pole lowpass filter at frequency
1/2*pi*R*C =3D 1/(2 * 3.14 * 100 ohms * 10 uF) =3D 159 Hz

A single-pole filter is good for 20 dB per decade, so
a 50 kHz signal is attenuated by 50 dB at the R7-C6 node

The phantom power supply has Rs =3D 6.8K in series with the mic.
if Z is 150 ohms, the voltage divider attenuates by 150/(150+6800)
which is 1/46.33, or a 33 dB attenuation.

So our 8.5 mV ripple is attenuated by (50db + 33dB) leaving
us 0.58 uVpp. If we ignore the higher ultrasonic and RF harmonics and
call our 50 kHz sawtooth ripple a simple sinewave, that would be
Message: 0.
Subject: 20 uVrms, which is -134 dBV.  That seems pretty quiet to me.

This is a simple analysis but probably pessimistic. Most mics don't
draw the full 10 mA, and the noise you actually care about is only the
fraction that gets by your anti-aliasing filter and is aliased down
into the audio band.

I only considered the main switching frequency and not any
subharmonics or other noise. I see the Rolls design has a feedback
mechanism through Q2 to regulate the power by pushing the U1F input
out of the active switching range when the zener current nears 0.7 mA.
This might be effectively a PWM control depending how the 4069 input
stage works in the linear region, but I don't know what impact if any
that has in the audio range. I'd guess negligible.  Zeners themselves
generate broadband noise but that would be much less than the ripple
voltage.

-----------
John Beale
www.bealecorner.com








<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU