[Ian Bell[_2_]]

Patrick Turner wrote:

Ian Bell wrote:
Patrick Turner wrote:
Ian Bell wrote:
On several preamp PSUs I have built in that last couple of years I have
noticed the same thing. A very low frequency low level variation in the
output voltage. My latest PSU uses a series of five 100uF caps with 1K
resistors in between and it still exhibits this problem. The 'noise'
seems to be below 1Hz in frequency, fairly random and peaks typically
between +- 10mV with occasional excursions to +20 to 30mV.

Because the noise is so small, about the only place you can see it is on
the smoothed output with a scope set to ac input. I suspect this noise
is mains borne but I don't know how to see such small low frequency
signals on the mains itself. Any ideas what it is, how to look at it on
the mains and how to get rid of it?

Cheers

Ian
Because there are thousands of people hooked to the mains and they all
turn gear on and off randomly, the mains voltage level changes.
The changes are random, and when you rectify the mains after your power
tranny, you are creating a dc voltage which tracks the levels of applied
ac voltage. Its the very same effect that is used in detecting amplitude
changes of an AM RF signal.

Because power supplies for amp have caps between B+ and OV and after the
diode/magnetic circuit of mains and PT, there are two ways of getting
rid of LF signal if its in an output signal and from the B+ rails. One
option is make the circuit you have less responsive to LF, with every
stage R&C coupling with a cut off at 5Hz, Especially in Phono amps.
Another is to regulate the B+ rail after the first C stage, and perhaps
at sensitve stages like a phono amp.

But I have used several stages of R&C with C = 470uF and R maybe 220
ohms in a preamp and chosen poles for RC couplings that cut out very LF.

I am using five stages of 1K and 100uF which is pretty much the same as
220R and 470uF and I still see the LF noise.

Indeed, but if all you did was to replace the 100uF in your amp with
470uF, and leave the 1k R you have,
then you will see a big reduction of LF noise.

I agree. I just might try that.

But if you insist on response to 2Hz, you may need to regulate each
stage.
To maintain a response flat down to 20Hz over several stages implies
most will need to be flat to around 5Hz.

Yes, but if the response is -3dB at 5Hz at say 3 consecutive stages then
at 1Hz the response is becoming 18dB/Octave.


Yes but at present I am working with just one stage and this LF noise
makes it impossible to measure distortion of the order of 0.1% or less
at 100Hz .


Cheers

ian

There is no point in "lookong at it" on the mains. Assume the mains is a
sewer of noise. Simple filters and good design or regs will remove all
your noise.

I only suggested that as a means of confirming its source, but all the
replies I have had so far seem to confirm it is mains borne so I don't
now see a need to do that.

Remember that the mains 50Hz or 60Hz signal is a 'carrier' of LF noise
information just as audio information is carried by an RF carrier in a
radio set, Sometimes slow moving signals in mains are like dc variations
as well. Most of this is filtered out by the transformer.


Patrick Turner.
If you have LC filters, make sure their resonant F is below 7Hz or
lower. And maybe add some series R to damp the LC resonance.
With resonance, noise content around the LC Fo will appear in the B+
larger than at the resevoir cap unless you have applied critical
damping. This may mean the added series R = 1.4 x XC or XL at Fo; the
response won't be peaked at the pole F, ie, Fo.

Z100uF = 1,600 ohms at 1 Hz, so with R = 1k, there is little filtering
at 1Hz. If you had C = 470uF, ZC = 340 ohms and 4 RC filter sections
would give attenuation = approx 1/2 x 1/2 x 1/2 x 1/2 = 1/16. Its still
not a huge amount of filtering. But say your amp output
has 1mV of LF noise with 1k + 100uF, then with 1k + 470uF the noise
should be maybe 0.1 mV, and maybe OK for you.

A regulator will do better because they work right down to DC.

Patrick Turner.
Cheers

Ian