ACMP-81 hum sounds

[jdbakker]

There's no such thing as a "ground loop". That's a total misnomer.

That's a pretty unique perspective.

On a rack level, ground loops occur when interconnected devices have different ground potentials. This can happen when neither, some or all of the affected devices are connected to mains ground. For current to flow there must be a loop; if not through mains/safety ground this often happens through common-mode capacitive coupling in the PSU transformer.

A second mechanism is induction; wherever you have a loop and a varying magnetic field you'll get a current flowing. This current, combined with non-zero impedance, means that not all points of the loop are at the same potential. This is an extremely common mechanism of both transmitted and received EMI, and a common cause of hum in DIY and (proto versions of) professional equipment. For hum, you can think of it as an extra parasitic winding on the PSU transformer, this is also one reason why there can be more/less hum when a device is racked vs. on the bench. To eliminate this, either kill the loop (as with star grounding) or minimize the loop area.

Within the context of this device, the best thing that could be done to avoid ground hum is to star ground each of the boards and slice the grounds on every jumper board, but if you star ground it with heavy enough wire, slicing the existing ground traces (which do parallel the signal traces) shouldn't make things much better at all.

The individual boards have unbalanced ground-referenced signal I/O. If you star ground each board, the current consumption fluctuations of each board multiplied by the inductance of your star grounding get added to the signal. Best case this injects both PSU noise and signal-related components (possibly overtones); worst case it turns the system into one big oscillator.

Very thick/short star ground links 'only' minimize this. To make it go away altogether, you need either ground-follows-signal (current situation), or a whole bunch of interstage transformers.

BTW, alexc on Prodigy-Pro has done some more investigations into the whole hum issue. This is a good starting point for his discoveries (which mostly match what has been found before, but there are some surprises).

JDB.

 

[crazydoc]

I've been able to do some work on the 81 this weekend. I guess this thread is where we're posting this stuff now.

I grounded each board ala dgatwood, including board 1. I also added a 22uF filter cap to board 2 (which was what I had on hand.) Though all their ground traces are now within 0.1 mV of chassis potential, I hear no difference in the hash noise from the preamp board and eq boards, or hum from that damned lo freq inductor on board 5.

Here's a pic after grounding, with Mumetal shields in place on inductors and PS transformer.

 

[crazydoc]

So I removed the inductor from board 5 to make it remote, to see if that diminished the hum. There was so much noise from the 3 feet of unshielded wiring I couldn't tell.

 

[crazydoc]

I then thought to try to reorient the inductor with respect to the PSU transformer. This helped a little, but what made a major difference was being able to extend the shielding on the lower part of the inductor, so it is now enclosed in a long cylinder. (I'd tried capping the tops of all the shields last week - it made no audible difference.)

 

[crazydoc]

So now I've got the major noise under control - Zmix's suggestions for reorienting the power transformer and swapping the transistors on the eq boards (I only did boards 4 and 5, and used different transistors), shielding the inductors, and now being able to extend the shield on that pesky low frequency one.

I've socketed the 1Q1 and 1Q2 transistors in anticipation of receiving a bunch of low noise transistors this week, to see if replacing these, or some on the the eq boards, will help resolve the remaining hash noise at high gain.

But I think what I have now is very usable.

 

[antichef]

I've socketed the 1Q1 and 1Q2 transistors in anticipation of receiving a bunch of low noise transistors this week, to see if replacing these, or some on the the eq boards, will help resolve the remaining hash noise at high gain.

Please follow up with what happens -- assuming the preamp board is the same general circuit as the "original", does anyone know which transistors were "originally" specified for these two spots?

But I think what I have now is very usable.

Yeah, all 4 of my units are in great shape. I played DI electric guitar through a 73 for hours today -- it was like candy! Dead quiet, except for the hum of the P94 pickup (and when I put the switch in the middle position, which sets the two P94s into humbucking mode, it was dead quiet altogether), and I used some EQ - really, really good!

 

[dgatwood]

For current to flow there must be a loop; if not through mains/safety ground this often happens through common-mode capacitive coupling in the PSU transformer.

Fair enough, but I would argue that what you're talking about isn't a ground that is looped with another ground; it's the source of power on one side, ground on the other.

A second mechanism is induction; wherever you have a loop and a varying magnetic field you'll get a current flowing. This current, combined with non-zero impedance, means that not all points of the loop are at the same potential. This is an extremely common mechanism of both transmitted and received EMI, and a common cause of hum in DIY and (proto versions of) professional equipment. For hum, you can think of it as an extra parasitic winding on the PSU transformer, this is also one reason why there can be more/less hum when a device is racked vs. on the bench. To eliminate this, either kill the loop (as with star grounding) or minimize the loop area.

Bear in mind, however, that no loop is necessary in such a case. It is merely necessary to have a path to ground that flows through an audio cable. I've experienced this with devices that have no third prong at all. That said, the majority of the problems I've seen have involved three-prong devices in which the signal ground was floating and the signal was ground referenced to squat.

The individual boards have unbalanced ground-referenced signal I/O. If you star ground each board, the current consumption fluctuations of each board multiplied by the inductance of your star grounding get added to the signal.

The problem is, as it was originally, the ground on the decoupling caps between stages were not able to function effectively because the ground was just plain not good enough. As a result, all that noise was going straight into the audio signal by way of the V+ lines. I'm not star grounding the boards. I'm star grounding the decoupling caps to improve their ability to smooth the supply rails between boards. That may not be the ideal solution because as you note, I'm also messing with the signal ground a bit in the process, but it has made the hum much, much better and doesn't appear to have made any of the other problems any worse.

 

[dgatwood]

Wow, is that hum nasty looking. When I look at the output of board 5 with boards 4 and 6 out of circuit, I'm seeing overtones as high as 300kHz or so. I can't quite tell if I'm seeing parasitic oscillation on top of that given how dirty the signal is. Looking for parasitic oscillation in that crap is like looking for a diamond in a room full of clear marbles....

What's fascinating is that neither board 5 goes crazy only if board 3 is there, board 4 goes either way when it is at the end of the chain unless I disconnect the pot at which point it goes silent. So basically we're probably getting parasitic oscillation as a result of the feedback loop.

In diagnosing it, I've noticed that the schematics are subtly wrong. There is an extra resistor in the feedback path, 4R39, that is not present on the schematics at all. My ohmmeter reads it as being somewhere between 0 and 0.1 ohms. Looks like that's their way of avoiding a jumper. Okay, so I guess that doesn't count... but it does count as resistance, which could be useful....

Aha! I think I just solved the oscillation problem in board 4. Once I realized it was only oscillating when the feedback loop was in place, I realized that there needs to be a filter to limit the bandpass to the audio range. To do this, I put a 37pF cap between the leftmost pin on the pot and ground. If my math is right, that should have a cutoff at something like 1.6 GHz. Wait... it can't be oscillating that fast. So I guess this isn't a bandpass issue at all, but apparently a little extra capacitance in the feedback loop makes a difference. It seems to have no impact on the sound other than making the hum on that board drop to zero instantly as soon as I touch the contact... and no swishing, etc.

Could somebody take a look at the schematic and tell me if my math and reasoning seem right? I want to make sure I'm not steering people down a completely wrong path here... and before I order a bunch more 37pF silver mica caps.

 

[dgatwood]

One more thing. It looks like there's a little oscillation causing the hum on board 6, but that fix doesn't cure it. Instead, I had to put a 37pF cap between the return contact (top pin on the jumper boards) and ground. So it takes five 37pF caps to finish off the mod, and with that, I think I'm going to call this mod done unless it isn't working as well as I think it is.

 

[tarnationsauce2]

Dgatwood, do you have the stock Q4/Q5 transistors on each board swapped out with BC441/BC461 transistors yet?
I assume you already know that the stock transistors are under-spec'd and will oscillate.
You can see in Antichef's pics that he has swapped his transistors. More reading if you haven't already: http://www.prodigy-pro.com/diy/index.php?topic=27791.140

 

[crazydoc]

There is an extra resistor in the feedback path, 4R39, that is not present on the schematics at all. My ohmmeter reads it as being somewhere between 0 and 0.1 ohms. Looks like that's their way of avoiding a jumper. Okay, so I guess that doesn't count... but it does count as resistance, which could be useful....

Those blue "resistors" with one black stripe are jumpers. There's a bunch of them on most if not all of the boards. Maybe they're easier to handle than wire for the assemblers?

And as tarnationsauce2 asks, did you swap out the transistors? That's the cure for the oscillation on most of the boards, though if your solution works, go for it.

 

[mshilarious]

Those blue "resistors" with one black stripe are jumpers. There's a bunch of them on most if not all of the boards. Maybe they're easier to handle than wire for the assemblers?

I use lots of resistor-jumper thingies. They are oxymoronically called "zero ohm resistors" Anyway, they are much more convenient than resistor leads (which I use lots of too), because you can bend the leads with your fingers quickly, vs. having to use pliers with the smaller length (and lack of ceramic body) of a resistor lead. They only cost about a penny each . . .

 

[dgatwood]

And as tarnationsauce2 asks, did you swap out the transistors? That's the cure for the oscillation on most of the boards, though if your solution works, go for it.

No, this is with the stock transistors. The goal of my hacking was to try to find a way to prevent the oscillation without requiring folks to buy parts that can't easily be sourced locally.

 

[tarnationsauce2]

No, this is with the stock transistors. The goal of my hacking was to try to find a way to prevent the oscillation without requiring folks to buy parts that can't easily be sourced locally.

That could be quite a task. The problem is that there is too much current running through the stock transistors to handle.
Look at the data sheet for the stock BC557/BC547 transistors and compare to the BC461/BC441. Look check the collector current.
The BC461 is is capable of 10x the current and 20x the total power dissipation of the stock BC557.
Having said that, all that truly matters is how much actual measured current is running through the stock transistors as well as actual measured power dissipation... not the specs of the transistors. But I speculate that it is very near or over 100mA and/or 500mW (the max of the stock transistors). As per specification, anything over 100mA and/or 500mW the transistors will not and can not perform in a predictable linear fashion.

Datasheets:
BC461: http://pdf1.alldatasheet.com/datasheet-pdf/view/75200/MICRO-ELECTRONICS/BC461.html
BC557: http://pdf1.alldatasheet.com/datasheet-pdf/view/75207/MICRO-ELECTRONICS/BC557.html

You will probably have to change the bias of the transistors (diff resistors). But that may significantly change the way the circuit preforms and sounds. Some heat sinks may help a bit.

FYI the parts are readily available at newark.com
BC461: http://www.newark.com/magnatec/bc461/_/dp/25M7693
BC441: http://www.newark.com/magnatec/bc441/_/dp/96K6848?_requestid=62823

 

[mshilarious]

100mA is the normal bias current? I thought that was a result of the oscillation! Given that the incorrect transistors can't handle whatever the proper bias current is, but hopefully that's a much lower figure. 500mW is an obscene amount of power to waste on each EQ stage. The output stage bias is 2V/47 ohms = 43mA, no?

Does Al Gore know about these pres?

 

[tarnationsauce2]

100mA is the normal bias current? I thought that was a result of the oscillation! Given that the incorrect transistors can't handle whatever the proper bias current is, but hopefully that's a much lower figure. 500mW is an obscene amount of power to waste on each EQ stage. The output stage bias is 2V/47 ohms = 43mA, no?

Does Al Gore know about these pres?

Yeah I'm not sure what the bias current is supposed to be, but obviously lower than the collector current. Good point that the oscillation can cause the collector current to shoot up when it otherwise would not.
BTW the 100mW/500mWfigures I gave were for maximum collector current. I was just trying to illustrate how under-spec'd the stock transistors are.

 

[rodabod]

Either way, I'm not sure that band-limiting the circuit is maybe the best option for stopping the oscillation. The logical step in my opinion is to take the design closer to the original if this helps.

 

[dgatwood]

Either way, I'm not sure that band-limiting the circuit is maybe the best option for stopping the oscillation. The logical step in my opinion is to take the design closer to the original if this helps.

While I agree that it would be nice to bring the design closer to the original, that's probably a rather large project....

I'm not too worried about the band limiting. Unless I'm miscalculating, any band limiting with a 37pF cap, even if you include the two 7.5k resistors as part of the RC filter, is still band limiting at almost 300 kHz---well outside the desired operating range of the amplifier.

The bigger concern is that this is effectively adding a capacitive load on the output. I have no idea what this is doing to the phase response of the amplifier stage at high frequencies---it would probably be wise to go with the smallest value that works to avoid undue ringing, etc., though the effects might be outside the band we care about, in which case I guess it doesn't make any difference as long as it isn't oscillating.... Dunno. Maybe I can dig up some 1pF caps and see if they'll do the job. Hmm. Yeah, I think I ripped some out of a mic a while back. I'll dig around.

 

[crazydoc]

I've socketed the 1Q1 and 1Q2 transistors in anticipation of receiving a bunch of low noise transistors this week, to see if replacing these, or some on the the eq boards, will help resolve the remaining hash noise at high gain.

Well, I tried 3 different types of low noise transistors, several of each, substituting for 1Q1, 1Q2, and 1Q3. No appreciable difference. I swapped 1C1 for several different brands - no difference in noise.

I seem to remember this individual unit as being noisier that my other two - I'll have to get them out and recheck them. If no difference, I guess this noise is just what you get when you try to amplify by 80dB.

 

[dgatwood]

Might be worth trying the low-value capacitor hack in the boards' feedback loops and on the EQ return line, particularly if you still have any hum coming from the EQ boards. I doubt it will make much difference on the hiss, but if what you're hearing turns out to be a side effect of the oscillation, it might.

The nice thing about it is that it is trivial to try it in a temporary fashion to see if it works---stick one lead of the cap into the back of each connector and hooking a wire from the other lead to any screw.