Cooling heatsink with a fan, noise gets into the power supply.

Started by jrc4558, June 03, 2007, 10:05:48 PM

Previous topic - Next topic

jrc4558

Hi everyone! I have finally mounted the LM3876 amplifier into the chassis and (in need to save space) decided to use a CPU heatsink from a 486. I mounted it o nthe outside of chassis, and the system runs nice and cool. The noise from the fan is virtually absent, so I got a small heatsink with temperatures not exceeding 35-40 C.

Howecver, another problem emerged. The amp runs at 41-0-41 VDC. I tapped the (+) supply and brought it down to 18VDC with a voltage regulator (LM7818). I used that to power the preamp, and after that I took the output from LM7818 and brought it down to 5VDC (LM7805). The fan works and the preamp does what its supposed to, BUT I get audible whining noise in the audio range. The noise changes with the frequency of the fan's rotation. I gather its because the fan's and the preamp's power comes from the same rail. The fan's supply, however is BEHIND the voltage regulator, and a huge cap (1Farad, 5V) is buffering it. I tried using separate voltage regulators (on 18 for preamp, and 18 to 5 for fan - because the regulator doesn't seen to be able to go from 40 to 5 VDC) but that didn't help.
I still want to try using a negative supply and running the fan from another rail, but it will have to wait a comple of days before I get to the studio again.
Before then - I ask you noble gentlement for any advice on how to eliminate this noise.
Has anyone had a similar problem?

Eager to hear from you! Thanks!

Rob Strand

It could be electrical noise being capacitively or inductively coupled into the preamp from fan's motor/electronics.  The fan might be too close to the audio signals.

I don't think it's the power supply.  You could verify this by powering the 5V regulator from a battery or other supply.  You should check with the second supply being isolated from the amplifier's supply and also when the grounds of the two supplies connected together.  You should do both tests because an isolated supply can reduce capacitively coupled noise - it's got to do with how noise currents flow through capacitive paths.

You might need to move the fan away from the audio.

I have my doubts a small 486 fan will provide enough cooling for a 50W amplifier running flat-chat.
Send:     . .- .-. - .... / - --- / --. --- .-. -

runmikeyrun

do you have a brushless fan?  I'm assuming it is.  But yes- currents produced by the fan's motor are being inducted into your circuit somewhere.  Move it further or try to shield the circuit.
Bassist for Foul Spirits
Head tinkerer at Torch Effects
Instagram: @torcheffects

Likes: old motorcycles, old music
Dislikes: old women

R.G.

1. Ground noise
2. Radiated from magnetic loop
3. (distant third) capacitive radiation
R.G.

In response to the questions in the forum - PCB Layout for Musical Effects is available from The Book Patch. Search "PCB Layout" and it ought to appear.

Paul Perry (Frostwave)

Maybe a common earth path, or partly common earth path, for the fan and some part of the effect circuit.
Imagine the current going to the fan, then back through the ground path.
This develops a few millivilts of noise along the ground path, and if part of this ground path is used by an IC or whatever, then the noise gets injected into the system.
As Rob says, try with a completely separate supply for the fan.

jrc4558

Thank you gentlemen!
I have several questions on your suggestions.
R.G. : You suggestion in number 2. Could you expand it a little perhaps? I get 1 and 3 - but scratch my head on 2.
Rob Strand : "capacitively or inductively coupled" - how does such coupling occur?

Thanks again for valuable advices!

Will definitely try a separate supply.

Sir H C

It is a DC fan, so I assume that it is not brushless.  These things are noisy.  They put out a lot of hash on the power and ground lines.  Best to run it off a completely separate supply, that is some high frequency noise, and big electrolytics will probably not filter much of it out (they are usually too "slow").

R.G.

QuoteR.G. : You suggestion in number 2. Could you expand it a little perhaps? I get 1 and 3 - but scratch my head on 2.
A DC motor will pull intermittent DC with little glitches and spikes in the current and voltage. If the wires which run to the motor form a big loop with a lot of area inside the loop, then the intermittent current causes an intermittent magnetic field to pass through the loop. IF there is another loop of conductor which shares the area of that first loop, it forms a transformer and a current/voltage is induced into the second loop. If that loop is anything to do with your circuit inputs, you hear it.
QuoteIt is a DC fan, so I assume that it is not brushless.  These things are noisy.  They put out a lot of hash on the power and ground lines.  Best to run it off a completely separate supply, that is some high frequency noise, and big electrolytics will probably not filter much of it out (they are usually too "slow").

If there is only ONE place where the current loop containing the motor is connected to the rest of the circuit, it is not possible for it to pull current into/out of the rest of the circuit and no conductive interference is possible. Inductive and magnetic are still possible. ONE connection to a common ground with no shared conductor path will not cause noise to be induced.
R.G.

In response to the questions in the forum - PCB Layout for Musical Effects is available from The Book Patch. Search "PCB Layout" and it ought to appear.

jrc4558

Quote from: R.G. on June 04, 2007, 01:16:32 PM
A DC motor will pull intermittent DC with little glitches and spikes in the current and voltage. If the wires which run to the motor form a big loop with a lot of area inside the loop, then the intermittent current causes an intermittent magnetic field to pass through the loop. IF there is another loop of conductor which shares the area of that first loop, it forms a transformer and a current/voltage is induced into the second loop. If that loop is anything to do with your circuit inputs, you hear it.

Thanks for reply. No, no loops. Star ground, fan to voltage reg, volt reg to star...

I found this: http://www.comairrotron.com/brushless_dc_fans.shtml
and this: http://www.comairrotron.com/emi_note.shtml
In the first one on the bottom - there's a suggestion for a pi filter before the fan. Will it have effect in the case of audio frequencies?

And also - would photos of internal wiring help?

jrc4558

The noise was the current ripple, RG is correct!
Being unable to fit another transformer in the chassis, I went with a huuuge heatsink and trashed the fan idea.

Thank you all gentlemen for your words of advice!

Once on the subject - how hard can one drive the LM3886-3876? I am using a fet preamp (single transistor), but have a hunch that I gould throw another 5-10 volts of p2p signal in there to get more volume. Anyone knows what's the safety margin? My rails are at 41VDC each. Running the fet at 18VDC, I get loud signal, but doesn't seem much louder than my Marshall 18 watt clone, really...

R.G.

Quotehow hard can one drive the LM3886-3876?
That's not really the way to think about this.

The LM3886 is an opamp that happens to be able to run from high voltage power supplies and provide large output currents. I don't know what schematic you used, but most power amps are set up for a voltage gain of about 20-30. The LM3886 is only stable for gains higher than 5, so it's sure to be more than that.

With a gain of 20 (which is what all the LM3886 ap notes use) and a 41V power supply, the signal peaks would only need to be 41/20 = about 2V peak, 4V peak to peak to drive the LM3886 to clipping.

QuoteRunning the fet at 18VDC, I get loud signal, but doesn't seem much louder than my Marshall 18 watt clone, really...
Once again, that's not really the way to think about this. If you're getting 56W into 8 ohms from  your LM3886 and eighteen actual watts from the clone, then the LM3886 is only putting out 56/18 = 3.11 times the power, or 4.9db more power.

You have two things involved: one, you're not using the same speaker. There can easily be 5-10db of efficiency difference in the speakers; if the clone has the efficient one, it could well sound louder. Two, the human has a logarithmic response to the power of a signal. A 3db difference in power is just a perceptible difference in loudness. A 10db difference in power (that is, 10x the power) sounds twice as loud. So you'd need a 180W amp with the same speakers to sound twice as loud as the clone.
R.G.

In response to the questions in the forum - PCB Layout for Musical Effects is available from The Book Patch. Search "PCB Layout" and it ought to appear.

jrc4558

Quote from: R.G. on June 06, 2007, 11:05:04 PM
Quotehow hard can one drive the LM3886-3876?
That's not really the way to think about this.

The LM3886 is an opamp that happens to be able to run from high voltage power supplies and provide large output currents. I don't know what schematic you used, but most power amps are set up for a voltage gain of about 20-30. The LM3886 is only stable for gains higher than 5, so it's sure to be more than that.

With a gain of 20 (which is what all the LM3886 ap notes use) and a 41V power supply, the signal peaks would only need to be 41/20 = about 2V peak, 4V peak to peak to drive the LM3886 to clipping.

QuoteRunning the fet at 18VDC, I get loud signal, but doesn't seem much louder than my Marshall 18 watt clone, really...
Once again, that's not really the way to think about this. If you're getting 56W into 8 ohms from  your LM3886 and eighteen actual watts from the clone, then the LM3886 is only putting out 56/18 = 3.11 times the power, or 4.9db more power.

You have two things involved: one, you're not using the same speaker. There can easily be 5-10db of efficiency difference in the speakers; if the clone has the efficient one, it could well sound louder. Two, the human has a logarithmic response to the power of a signal. A 3db difference in power is just a perceptible difference in loudness. A 10db difference in power (that is, 10x the power) sounds twice as loud. So you'd need a 180W amp with the same speakers to sound twice as loud as the clone.

Dang, completely forgot about our hearing. I studied that in perception class. :(
Okay, will try with same speaker cabinet.

I am currently running it through a 2-10 alnico speakers. If i remember correctly, alnico magnet speakers tend to have better efficiency to them. So, will try to set the gain a little higher.

Thanks for the advice!

petemoore

 Feasably functional yet funky..maybe OT...
  "Smokestack" air circulators...metal tubes.
  Heat up some metal tubes and they'd work just like a hot chimney,  heated air rising causing circulation, not as much as a fan.  So, instead of a fan, heating vertical metal tubes by placing them adjacent to and over the hot glass ones, their inlets could sucking air through the heatsink or across the items that require the cooling...even a small flow of cool air can do alot of heat dissipating.
  The untested, long way around...if you've got alot of heat to start with, right over the area you want cooled, it could be used to cause cooler air to flow right there.
Convention creates following, following creates convention.

jrc4558

Gentlemen,
It is my pleasure to inform you that after building an IC preamp around a LF351 (gain of 11), the amplifier is loud and clean, and working perfecly!
I thank you all wholeheartedly for all the help!
Pictures forthcoming.