Hunting down noise sources in my current design...
I was disappointed with a noise output figure of about .6mV on my most recent project and went looking for the source. I'd worried about this one because I'd really crammed the componentry close together-- I was especially worried about the input transformer picking up stray noise from the power toroid, or the fact that I was also running the V+/G/V- as a twisted bundle right under the input transformer...
Nope and nope. The source of my noise-- the bloody input coupling cap. Apparently when the film/foil cap is half the size of your fist, it becomes a dish-freakin'-antenna for stray signals. Simply putting it inside a 1/4" thick metal case is not enough. Apparently one also needs to shield it from everything inside the case too (the things I do to completely kill DC offset).
The test that identified the input coupling Orange Drop as the interference source was to wrap it in aluminum tape and ground the tape to the chassis. That alone killed off all last traces of hum/hash (measured by sticking my ear right up onto the cone, in the dead quiet basement workshop with everything else electric turned of at 3am) although there's still stray measurable white and RF noise on the scope...
Now I'm wondering-- do I add a sheet-metal 'clamshell' shield over and under the entire input stage PCB? Do I ground that to chassis or signal ground? Maybe it's also time to go to shielded cable for internal connections...
Something 'new' every day
I was disappointed with a noise output figure of about .6mV on my most recent project and went looking for the source. I'd worried about this one because I'd really crammed the componentry close together-- I was especially worried about the input transformer picking up stray noise from the power toroid, or the fact that I was also running the V+/G/V- as a twisted bundle right under the input transformer...
Nope and nope. The source of my noise-- the bloody input coupling cap. Apparently when the film/foil cap is half the size of your fist, it becomes a dish-freakin'-antenna for stray signals. Simply putting it inside a 1/4" thick metal case is not enough. Apparently one also needs to shield it from everything inside the case too (the things I do to completely kill DC offset).
The test that identified the input coupling Orange Drop as the interference source was to wrap it in aluminum tape and ground the tape to the chassis. That alone killed off all last traces of hum/hash (measured by sticking my ear right up onto the cone, in the dead quiet basement workshop with everything else electric turned of at 3am) although there's still stray measurable white and RF noise on the scope...
Now I'm wondering-- do I add a sheet-metal 'clamshell' shield over and under the entire input stage PCB? Do I ground that to chassis or signal ground? Maybe it's also time to go to shielded cable for internal connections...
Something 'new' every day
AndrewT said:
Hi AndrewT. It's kinda disturbing how you see and respond to practically everything that gets posted to the forum...
Having read more from some old Bell Labs treatiese on proper shielding, my 'signal or chassis' question is answered very simply: If you're shielding from RFI/EMI, you always ground to chassis. Signal ground is for actual signal path, period. No exceptions. So sayeth Bell Labs, so it was written, Amen.
The cap is an Orange Drop, so no shielding foil. It's a naked, dipped polyprop/foil cap. Reversing it makes no difference.
The noise is present identically whether it's open circuit, shorted input, resistored input or actually hooked up to a source. When grounded (and I'm using a three-conductor mains input, so it is always grounded) neither handling the case nor the input cables affects the noise [bonking the case causes some noise, but you expect a foil cap to be microphonic]. So I'm reasonably certain my minor noise issue is purely of an internal nature. Of course, the output noise level explodes to multi-tens of mV when I open up the case and the Big Cap and other innards are suddenly exposed to the ambient RF. Again, exactly as you'd expect. Also, eliminating the cap, coupling directly, and accepting the DC offset also eliminates the majority of the noise problem. However, I have need to couple in the larger scheme unless I want to live with 100mV or so offset [that will crackle more and more as the attenuator ages].
Granted the noise 'problem' is only noticable under absolutely ideal circumstances. .6 mV or so with the selected matched full range driver (85dB/W) yields an absolute noise SPL of 20 dB; I don't think I've ever purchased a mass-produced consumer system that only put out 20dBSPL noise idle at full volume. To hear it, I have to get close enough to the speaker that my beard brushes the cone, and that's *much* louder ;-)
But this particular project is a project that is all about design excess and showing off... else I'd just go [baaaa] buy another Bose system [baaaa] and think that was good enough.
I have already spent ten months just on hand machining the cases from raw aluminum... So I want to hit -80dBV noise or ~10dBSPL [one more order of magnitude]. I'm just trying to decide if clam-shelling the input stage in metal or just wrapping the big cap is enough to hit that target.
...or am I actually getting a little too insane? Is everyone else happy with a full-range idle output noise of 20 dBSPL? Most of the high end amps I see advertise an output noise level of ~-80dBV, and I'm only at about -65 dBV. It still seems reasonable to try to get more....
xiphmont said:
With any larger poly cap, I tend to put the signal ground wire (insulated) one turn around the cap body. I had no idea there was any good reason to do so. Its just that I believe that all "hot" lines need to be within 1/4" or less (preferably less) of a ground, and those caps were too big to be protected by the board's ground plane sheild.
Sometimes I'll even make a ground line run along with the rest and then it stops where its usefulness has ended (instead of a ground loop). This appears to connect to nothing, but I call it a sheild. Oddly enough, that's just like the foil (stove pipe tape?) on your cap. Honestly, I had no idea that it was a good practice. In my case, I did it out of imagination, which has been remarkably more reliable than my calculator.
Here's a couple of questions:
Do you have RF filtering on input, output, and power supply?
Do you have a load at both sides of the input filter cap?
Is your input circuit in "load, pass, load, pass, load" layout?
danielwritesbac said:
There isn't any good reason to do what you're describing. Winding a ground wire one time around a large cap will not provide any effective shielding. Google "Faraday cage".
danielwritesbac said:
Your calculator must not be a nice scientific one. Those are usually quite good
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.