ACMP-81 hum sounds

[crazydoc]

So it seems to me that the key to reducing the hum may well be getting that power cable as far away from the transformer as possible. Since I don't think I can realistically keep the power cable routed so far away from everything, I'm considering soldering a right angle Molex connector onto one of the jumper cards near the left end and feeding power in from the opposite end of the device. (The power rails run straight across without resistors, diodes, or voltage regulators, and there are decoupling caps between each stage, so it shouldn't really matter which end of the bus provides power.) Thoughts?

I have a piece of mumetal braided shielding. This week sometime I'll try routing the DC cabling through it on my 81 that has the toroid and inductor shielding (that is now essentially hum-free) and remove the their shields to see what happens.

 

[dgatwood]

So a few minutes with the Dremel and a few minutes of painful bridge soldering later, I got the ACMP-81 hooked back up with the DC power cable plugged into the second board from the left (because there was plenty of space to drill holes on that board). Apart from a bzzzert sound for a fraction of a second at startup, the hum level is zero with the output turned all the way up and a medium input gain setting. The EQ hum that plagued me before is just plain gone.

Something occurred to me today as I was walking from lunch to a meeting in another building. All this effort to redesign the toroid is massive overkill. All we really need to do this right is a single large +24V secondary with no center tap.

1. Use a simple charge pump (voltage doubler) to provide 48VDC off the 24VDC for phantom. This is a trivial circuit.
2. Use a simple voltage regulator to provide +12VDC.
3. Don't bother with the -12VDC.

#3 may seem shocking, but I'm pretty sure it's a valid thing to do. Here's why: we're only using one side of the output of the op amps to drive the LEDs. Therefore, we should only need one rail. We really don't need the -12VDC rail at all. It's just wasting current and unnecessarily complicating the power supply... a lot.... Just ground the -12V rail and you can get a replacement power transformer at Radio Shack. Well, for proper 125/250V operation, you'd need one with a 220V center-tapped primary, but either way, we're talking about a $10-15 power transformer here.

I'm utterly horrified to realize just how absurdly over-complex this power supply design is. Terrible design. Horrible.

 

[dgatwood]

Here's a reference design for a 48V supply that could draw off the 12V regulated output, though this is way more complicated than is really necessary given the minimal current requirements:

http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1003,C1042,C1031,C1115,D26886

You could also do a simple charge pump like this one, filter the output with a few caps, and use a Texas Instruments TL783C to cleanly regulate the resulting output at 48VDC.

http://pdf1.alldatasheet.com/datasheet-pdf/view/28837/TI/TL783CKC/datasheet.pdf

Probably a much simpler solution given that we're talking about one chip, one voltage regulator, and ten random parts---resistors/caps/diodes---not counting any added filter caps.

P.S. That charge pump circuit needs one small change to be usable for these purposes. Change C1 to be .001uF instead of .01uF to raise the clock rate from 4.4 kHz to 44 kHz. Or even go as low as 500pF. The last thing you want to do is introduce another source of audio-frequency noise in these things.

 

[dgatwood]

Put another way, the power supply I would have designed for this would cost:

48VDC regulator:
$1.60 x 1

diodes:
~30 cents x 2

Resistors:
~5 cents x 4

Caps:
5 cents x 1
~60 cents x 3

Transformer: $10.49

555:
48 cents x 1

Full wave bridge rectifier:
$1.50 x 1

24VDC regulator:
93 cents x 1

12VDC regulator:
42 cents x 1

Grand total:
$18.07

This is in single-digit quantities from Digikey and/or Radio Shack. In bulk, it would probably be closer to $10-12. Now could somebody tell me what they could possibly have been thinking designing this power supply with a custom $100+ transformer containing four secondaries!?!?!?!?!

 

[dgatwood]

I think I might even be able to do one better, assuming I'm not missing some subtle use of the +12VDC supply....

Take out the DC blocking caps like 3C31. That way, the signal is biased to fall between 0 and 24VDC like the rest of the circuit. Then, feed +24VDC to the positive rail on the op amp instead of +12VDC. Put the DC blocking cap at the output of the second op amp.

Next, change the current limiting resistors on the switched LEDs so that you can use the +24V to drive them, too.

With those changes you should be able do away with the 12VDC regulator entirely. There went another half a buck. You're now down to a simple 24VDC single-ended supply and a charge pump to get 48V phantom. The power supply just went from being a complicated monster from hell to being something your average junior high electronics geek could put together with parts from Fry's and Radio Shack....

Am I missing something fundamental here, or is this the most embarrassingly overbuilt power supply in the history of the human race?

 

[mshilarious]

Something occurred to me today as I was walking from lunch to a meeting in another building. All this effort to redesign the toroid is massive overkill. All we really need to do this right is a single large +24V secondary with no center tap.

I had tried to point that out, to no avail. Steve Hogan obviously can design and build any transformer he wants, why cost is not a consideration I don't know. Your point is well taken; an off-the-shelf quality toroid should work provided you don't care about the silly LEDs, and it will be about 1/3 the cost of Hogan's custom unit.

1. Use a simple charge pump (voltage doubler) to provide 48VDC off the 24VDC for phantom. This is a trivial circuit.

A charge pump and a voltage doubler are really different circuits. Well, at least the "pump" part of the circuit is only necessary when the only supply you have is DC. Since we have 24VAC, you only need the voltage doubler, not the 555 chip. Therefore, you aren't adding a very significant noise source, and you only need a few caps and diodes. This is a near universal solution for phantom power. Oh, and the current available is much higher than the 555 circuit.

If you go the route of a DC converter, a 555 chip is a cheap but relatively terrible solution--lots of off-the-shelf converters that work much better.

#3 may seem shocking, but I'm pretty sure it's a valid thing to do. Here's why: we're only using one side of the output of the op amps to drive the LEDs. Therefore, we should only need one rail.

Yes, but you'd need to bias the inputs or the output will just latch to one rail. That adds a quite few parts, although it should have been done in the original design. I would just dump the stupid LEDs myself . . .

Edit: yes, your second solution for the LEDs should work, although you need to watch the DC offset at the opamp's input. The bias of the preamp circuit at various points isn't necessarily exactly 12V. I haven't looked at the LED circuits, so I don't know if there is a problem there or not.

I'm utterly horrified to realize just how absurdly over-complex this power supply design is. Terrible design. Horrible.

Yes.

 

[dgatwood]

Yes, but you'd need to bias the inputs or the output will just latch to one rail. That adds a quite few parts, although it should have been done in the original design. I would just dump the stupid LEDs myself . . .

Thanks. I wasn't aware of that phenomenon. I can't say I've ever tried a single-ended op amp circuit. On further looking, the TL072 that they used is described as providing "latch-up-free operation", so I wouldn't expect that to be a problem (though it may still require a diode on its input to avoid negative input signals---not sure).

 

[mshilarious]

Thanks. I wasn't aware of that phenomenon. I can't say I've ever tried a single-ended op amp circuit. On further looking, the TL072 that they used is described as providing "latch-up-free operation", so I wouldn't expect that to be a problem (though it may still require a diode on its input to avoid negative input signals---not sure).

Single-ended is all I do!

Latch-free probably means it doesn't get stuck to a rail after the inputs move off the rail. I didn't pick my terms too well; I think the danger is that the input would simply effectively stay on one rail all the time. It depends on what the meter circuit is trying to do. To operate a meter properly, you need to full-wave rectify the incoming signal. They may have skipped that nicety, I don't know. Then you need a comparator circuit. Either stage could be very problematic with improper bias, especially since there is probably an inverting stage somewhere (grounded noninverting input). DC coupling of a single-ended multiple opamp circuit is rather tricky.

Regarding using a voltage doubler, the only ways I've seen to do this involve changing the ground reference. I don't think you can use a true voltage doubler and a FWB for a voltage regulator on the same secondary

Sure you can, but you would need a separate rectifier for the other supply rail:

http://www.play-hookey.com/ac_theory/ps_v_multipliers.html

I couldn't find any that could go over about 10-12VDC on the output. Any that you would suggest?

I don't think there are too many off-the-shelf charge pump ICs that go over 12V. I haven't looked in a while. But the high-voltage high-current solutions are boost converters, using inductors.

Funny you should ask, because I've spent the last two months slaving over a high-efficiency low-noise DC converter design to generate +48V phantom from a 9V battery. I am using MC33063A together with a MOSFET to buffer the IC's internal switch from the +48V (really more like +58V) which exceeds the chip's rating. Another chip that looked more promising on paper but didn't work out as well for me in real life is MAX1771.

I can share the schematic if you would like, but again if you already have an AC source, it's totally the wrong solution to use a DC converter.

 

[dgatwood]

A charge pump and a voltage doubler are really different circuits. Well, at least the "pump" part of the circuit is only necessary when the only supply you have is DC. Since we have 24VAC, you only need the voltage doubler, not the 555 chip.

Yes, I meant that we'd be using the charge pump to double the voltage, not that it was a true voltage doubler circuit. Sorry, could have worded that better. Regarding using a voltage doubler, the only ways I've seen to do this involve changing the ground reference. I don't think you can use a true voltage doubler and a FWB for a voltage regulator on the same secondary, and as soon as you add a second secondary, the transformer cost takes a giant leap. Am I interpreting that wrong?

If you go the route of a DC converter, a 555 chip is a cheap but relatively terrible solution--lots of off-the-shelf converters that work much better.

I couldn't find any that could go over about 10-12VDC on the output. Any that you would suggest?

 

[dgatwood]

This is weird. They must have been messing with the clock on the server. That last post by me was actually posted before the other post by me, and several hours before the response by MsHilarious. Weird.

 

[mshilarious]

This is weird. They must have been messing with the clock on the server. That last post by me was actually posted before the other post by me, and several hours before the response by MsHilarious. Weird.

The whole board is pretty much hosed. I enjoy the surreality it creates

 

[thajeremy]

I have 5 of these things that I havnt even fired up yet...just because I havnt needed them....I have 3 73's and 2 84's....From everything Im reading, Im not going to have as many problems with these as you are with the 81's, correct?

and if I do have the same problems...is anyone taking on orders to fix them???

 

[dgatwood]

I have 5 of these things that I havnt even fired up yet...just because I havnt needed them....I have 3 73's and 2 84's....From everything Im reading, Im not going to have as many problems with these as you are with the 81's, correct?

and if I do have the same problems...is anyone taking on orders to fix them???

AFAIK, the 81s are the real problem children. You might have a one-off problem with the others, but probably not. Both my 73s are fine, requiring no mods at all.

 

[thajeremy]

AFAIK, the 81s are the real problem children. You might have a one-off problem with the others, but probably not. Both my 73s are fine, requiring no mods at all.

koo....

maybe Ill finally get around to firing them up over the weekend....

 

[crazydoc]

AFAIK, the 81s are the real problem children. You might have a one-off problem with the others, but probably not. Both my 73s are fine, requiring no mods at all.

Except, of course, the gain switch oscillation blast.

And my 73 had inductor hum requiring toroid and inductor shielding.

 

[iownrocknroll]

Can someone explain what is ment by rotating the transformer.

I'm also curious, did you guys change the transistors on the first EQ board, or board number 2. where the HPF and LPF are located.

Thanks!

 

[dgatwood]

Except, of course, the gain switch oscillation blast.

Which is relatively minor

And my 73 had inductor hum requiring toroid and inductor shielding.

Interesting. This at least suggests that all of these hum problems are caused by variations in the power transformer. All the more reason to change the circuitry to with with a single-secondary transformer that can be purchased off the shelf instead of the current (junk) custom power transformers.

 

[crazydoc]

Can someone explain what is ment by rotating the transformer.

Loosen the bolt holding the transformer down. Then rotate it clockwise or counterclockwise while listening, to find the place where noise is minimized. You have to do the transistor swap first on the ACMP81 - otherwise the oscillation noise overpowers the inductor hum.

I'm also curious, did you guys change the transistors on the first EQ board, or board number 2. where the HPF and LPF are located.

Thanks!

The transistor swap is for the Q4 and Q5 transistors on boards 3, 4, 5, and 6 on the ACMP81 only. The transistors are numbered 3Q4, 3Q5, 4Q4, 4Q5, etc. The 84 and the 73 don't need transistors changed.

Schematics here if you don't have them (but not needed to do the fixes):
http://recordinghacks.com/tnc-audio-preamps/

 

[dgatwood]

Can someone explain what is ment by rotating the transformer.

The transformer is round. It is held down by a bolt in the center. Loosen the bolt and twist. That said, this is not likely to fix these problems. As soon as you put the case back on, the case becomes a wave guide. In order to do any good, you have to test hundreds of times, putting the case back on for each test....

I'm also curious, did you guys change the transistors on the first EQ board, or board number 2. where the HPF and LPF are located.

The official transistor mod plan only calls for substituting parts on boards 3-6, not board 2.

 

[iownrocknroll]

thanks crazy doc, I noticed there is a Q4 and Q5 on board number 2, so 2Q4 and 2Q5 wasn't sure if those needed changed to

If I engage the EQ and turn the Output Gain "far right" the whole way up I get some hiss/hum, If I turn it mid-way up, I can hear a slight bit of hum/hiss however nothing registers on the meters in pro-tools.

Do you think rotating the transformer/shielding will take care of the last little bit?