-
Hi everyone,
I am working on a small microphone preamp for my PC which can be powered either from a battery
or from a USB port. When it is powered from the battery everything is ok and the main source of noise
if the electret itself. As you can imagine the USB power line is not the cleanest... so I am trying to
filter it. I don't know the best way to do it, but I've used a Pi filter, a common mode choke and finally
a notch filter for a specific frequency.
Below are the results from each filter in succession. (The 5V are AC coupled with a 100nF capacitor
and fed to the mic in of the PC. On the PC no mic gain and 50% levels .)
5V USB, no filtering
5V USB and Pi filter
5V USB, Pi filter & Common Mode Choke
5V USB, Pi filter, Common Mode Choke and 12.7kHz Notch filter
At this point the 5V rail is much better and you can barely hear anything
even if you stick your ear to the HiFi speaker.
Now when I use an LDO to lower and regulate the voltage to 3V,
its output adds approximately 15db of white noise! LDO is the LP29050ACZ
and it is bypassed with a 100nF capacitor. Here is the result :
You can hear it as a slight quiet hissing sound.
https://soundcloud.com/hgg2k/5v-usb-noise-pi-filter2-3v-ldo
Its not loud but this will get amplified through the mic preamp.
Is this a usual behaviour of an LDO?
Is there a better way to clean the 5V USB?
Thanks!
George.
-
The LP2950 may need bigger input and output capacitors.
Check chapter 9.2.1.2.1 of the data sheet. -
I had some great advice about using LDOs a few years ago from a sage in RF, and that was to be wary of noise from LDOs. For digital circuits it's not usually much of a problem, but for RF and audio it is.
I use the TPS79901 series extensively as a result, but there are plenty of others available these days. Look for "high PSRR", often this also indicates that the LDO's output noise will be pretty low too. They often have bypass pins on them for an external capacitor to bypass their internal voltage references, reducing noise. -
mij59 you are right. The 100nF cap was not enough. With a smaller value capacitor the LDO oscillated
pretty badly. With the 100nF the oscillation stopped but you could not see any hint of the white noise
on the oscilloscope! You had to use a spectrum analyser or listen to the audio signal...
Its amazing how the human ear can detect such subtle changes.
Below is an animated photo that shows the drop in the signal noise as I increase the capacitor size.
With 330uF the signal was back to normal. (At 820uF the oscillation started again.)
Howardlong thank you very much for your advice, which I will of course follow it.
Instead of solving a problem I prefer not to create it in the first place. I will search for a better LDO.
Thank you very much!
George.
(I had a look at Dave's eevblog audio stream and his noise floor is excellent. Either he is using
noise suppression at post processing or an excellent microphone and preamp.) -
Using a lower supply voltage can cause more problems with noise. Another thing to bear in mind is the electret microphone: are you sure it'll work reliably at 3V?
-
With the 100nF the oscillation stopped but you could not see any hint of the white noise on the oscilloscope!
Not surprising. From the pics, you're looking at the difference between ~-90dB & ~-110dB down; that's into the noise / below the resolution of what most scopes (analog or digital) have any chance of showing you.
It's a case of knowing your tools & what they're capable of... -
Hero999 As far as I know most electrets work fine with 3V and even lower 1.5V to 2.0V for most of
them. 3V for the Primo EM172. I was not going to power the microphone though from the 3V.
Instead I was thinking of using fixed regulated upper and middle voltages for the op-amp in place
of the resistor divider biasing, but after hearing the noise injected by the LDO, it doesn't
seem like a good idea...
Also the noise spectrum graphs were from the regulated and filtered 5V USB power alone.
AC coupled with a 100nF capacitor and fed to the PC's microphone input.
No microphone and no amplification.
Tac Eht Xilef so it looks like that the best instrument for testing the 20hz-20khz band is our ear.
-
What does your circuit look like? Is the noise itself (on the 5V supply) consistent or does it vary e.g. with hard drive activity and everything?
The common mode filter seemingly produced a way stronger effect in the >10kHz range than the pi filter (which doesn't appear to do anything at all?), but I have no idea if that's due to the source (statistically insignificant) or not. Also, common mode would suggest that you have bigger problems with layout or grounding; which may not even be "bad", it's just that -100dB is a very sensitive baseline to measure against, and every little bit counts!
Tim -
T3sl4co1l this is the actual circuit of the preamp:
and this is the filter that gets power from USB and gives power to the preamp:
The common mode choke has a value of 2x13mH and does a good job especially on the high
frequency noise.
When I use the USB from the LCD screen hub, power is more stable and does not change
with HDD activity, but when I use the PC's USB directly, its worse and has some variations as well
depending on activity.
I think that there is a ground loop problem. I've also ordered an isolated DC-DC converter block
that might help. 5V USB --> Common Mode Choke --> DC-DC converter --> Filter --> Regulator. -
Well, if the PSU schematic is correct, then C8 is your problem. It passes only AC. Add an LC filter instead.
-
The reason is that I am testing the noise of 5V USB power line, before and after the filter.
I am only interested on the ripple on the 5V rail. When I find a filtering solution that works
reasonably well I will remove C8 of course. (I cannot connect the 5V rail directly to the
PC's microphone input. I connect the output from C8 to mic in, record 30 seconds and look
at the result with a software spectrum analyser.) -
Using components outside their specifications doesn't help much:
The LM2940T-5 is only specified down to 6V and won't so much when run off 5V, other than just draw a high quiescent current to saturating the output transistor, dropping between 100mV to 1V, depending on the temperature and current.
The TL072 is only specified down to a total supply voltage of 7V (+/-3V)
The N5532 is not specified below 6V (+/-3V)
Eliminate the LM2940T-5 and use an op-amp which is properly specified down to 5V such as the TLC2202 or MCP602.
-
Sorry. The regulator is the LP2950 with a 3V output.
I could not find something similar when making the schematic...) -
What does the connection to the USB connector (including shield) look like?
You should also have a pull-down on the regulator output to perform the noise test, otherwise you are measuring a potentially floating output.
3V isn't nearly enough to run a TL072, what are you using? If you also aren't using exactly what's on the schematic, please be meticulous and mention every difference!
One way to break the ground loop would be, to add a small resistor (perhaps 10-100 ohms) in series with the output connector ground. This will drop some voltage when a load is connected (for example, if this circuit were used to drive headphones instead of driving another audio input), but you can compensate for this, and for the difference in ground loop voltage, by ground-referencing the mic input and op-amp feedback nodes to this ground, rather than power ground.
Tim -
Sorry for the confusion. In the schematic of the preamp I am using the LM4562 op-amp but I am
trying to filter the noise of the USB first. The preamp is not even connected and I will not power it
of course from 3V.
I get the 5V USB power line from a USB to 3.5" DC cable (no shield), apply the filter on that and then
after I AC couple the output signal with a 100nF capacitor (C8) I pass it to the microphone input of my
soundcard. So actually I am recording only the ripple and noise of the 5V USB power line.
I was thinking of using the 3V from the regulator as the middle biasing point on the opamp instead
of the usual resistor divider, but after I used it, I realised that LDOs are not the quietest of components...
It introduced white noise in the whole audio frequency range. The two 330uF capacitors corrected that,
but not completely.
-
Sorry for the confusion. In the schematic of the preamp I am using the LM4562 op-amp but I am
That's better, the LM4562 is specified down to a total supply voltage of 5V (+/-2.5V)
trying to filter the noise of the USB first. The preamp is not even connected and I will not power it
of course from 3V.
I get the 5V USB power line from a USB to 3.5" DC cable (no shield), apply the filter on that and then
after I AC couple the output signal with a 100nF capacitor (C8) I pass it to the microphone input of my
soundcard. So actually I am recording only the ripple and noise of the 5V USB power line.
I was thinking of using the 3V from the regulator as the middle biasing point on the opamp instead
of the usual resistor divider, but after I used it, I realised that LDOs are not the quietest of components...
It introduced white noise in the whole audio frequency range. The two 330uF capacitors corrected that,
but not completely.
A voltage regulator will be more noisy than a potential divider. If you're worried about the thermal noise from the resistors in the divider, then bypass one of them to the supply, rail via a capacitor which is large enough to make the cut-off well below the audio band.
For example if the resistors in the potential divider are 1k, the output impedance will be 500R, so for a cut-off of 10Hz:
C = 1/(2pi*Fc*R) = 1/(2pi*10*500) = 31.8*10-6F
In practise a 47µF capacitor would do. -
I will try that as well, although the main source of noise is the USB 5V rail which I am trying
to filter.
Do you think that using an isolated DC-DC converter to pass the 5V through it will make it
more quite or that the switching inside the DC-DC converter will produce more noise again?
-
I think the switching inside the DC-DC converter will make it more noisy.
-
...it seems that cleaning the USB power line is not so easy after all...
-
Why not use an R/C filter (resistor in series and cap to ground) to derive the mic supply? Maybe try 1K and 100uF to start with.
-
I've just tried it. It had no effect.
Is there any standard guide or reference on filtering such kind of noisy lines as the 5V USB?
-
Then the noise isn't from +5V, it's from USB_GND.
Tim -
Then the noise isn't from +5V, it's from USB_GND.
Tim
Can you please explain the reasoning behind that conclusion?
(The USB ground is the same as the mains ground and I can measure 33.5mVDC potential
from mains earth wall socket nearest to the PC)
Thank you. -
Sounds like the layout needs improving then?
-
Try filtering both the positive and negative rails to the microphone and use a differential amplifier?