Lions and Tigers and Grounding Problems, Oh My! DoZ Grounding...

[EchoWars]

See if you can't help out a Kansas boy...

I built Rod's DoZ a couple of years ago, and was fairly pleased with the results after working out some bugs with the help of ilimzn. But it has always hummed...not loudly, and it can only barely be heard with most speakers. I've measured the hum as being something like -70db under a 1W level. This is all but inaudible with most speakers, but I currently have the amp connected to a high-efficiency set of speakers (101db), and the hum is quite audible.

Here's the schematic, and my colored notations on the grounding, which I know is screwy (the grounding, that is...the colored lines may be debatable).



The red lines represent the common ground plane on the PC board, which is returned to the star ground on the chassis via the violet lines. The one green line represents the signal return, which comes in through the input RCA jack, runs to the ground lug of the 100K volume pot, and then is returned to the chassis star ground (violet line). The signal return does not run to the PC board at all, except through the star ground.

I have a vague idea that I need to separate the signal return from the ground plane on the PC board via a 100 ohm resistor, but I also suspect that it will not be as simple as adding the resistor to the star ground line running from the 100K volume pot to the chassis ground. But then, maybe it is... Before I tear into it, I'd eagerly welcome any input that the people here might have...people who have encountered and solved these issues many times.

Thanks for your thoughts...

[Bonsai]

Echowars, I think if you connect your input pot as shown with your coloured lines you will get a ground loop.

I would return the ground side of your pot to the input ground on the connector and then connect this via a 10Ohm resistor to the input ground connection (not back to central ground).

Also, make sure your input connector ground is NOT in contact with your chassis - it must be insulated from and case metalwork.

[EchoWars]

Originally posted by Bonsai
Echowars, I think if you connect your input pot as shown with your coloured lines you will get a ground loop.

It IS connected as shown...and the result is a very low-level hum.

I would return the ground side of your pot to the input ground on the connector

It is...the green line shows that the ground of the volume pot is directly connected to the 'ground' (signal return) of the rear RCA jack.

and then connect this via a 10Ohm resistor to the input ground connection (not back to central ground).

OK...here's where you lost me... The 'input ground' is the signal return side of the input RCA jack, or at least that's what I always called it. The 'ground' runs from there to the ground side of the 100K volume pot, and then to the star ground (common ground). I'm not sure where this 10 ohm resistor is supposed to go...?

Also, make sure your input connector ground is NOT in contact with your chassis - it must be insulated from and case metalwork.

Roger that...input RCA connector is insulated from the case.

[AndrewT]

Oh Dear!!

separate the signal and NFB and very low level return circuits from the higher current circuit returns and separate both of these from the dirty returns including the decoupling and Zobel.

Starting at the left.
The chassis cap may work?
The next 7 returns should all connect to the left most purple wire.
The next two returns R5 & Q5 go to the middle purple wire.
The last two returns C7 & C5 goto the right most purple wire.
The final return R704 goes to a fourth purple return wire.
That leaves just two, C702 could probably be combined with the dirty return.
I am unsure of the best place to return the Zener. My feeling is that the return could be combined with the clean (left most purple) or a separate trace to star ground.
Any comments?

[EchoWars]

Wow...I'm not sure if I can hack up the PC board to accommodate that many separate returns... If it were flawed that badly, I'd have assumed that a lot less people would have bought Rod's board and built the amp.

Your first suggestion though -

separate the signal and NFB and very low level return circuits from the higher current circuit returns

- was closer to my own line of thinking. Join the returns for C2, R3, C6, VR1, R702, and VR701 to a common return path back to the star ground, perhaps connecting it with a 10 ~ 100 ohm resistor. It's just that I don't have a whole lot of experiance with grounding schemes, and it will be a lot of work to get it done, and really difficult to redo if it doesn't do the job.

Any other thoughts as to whether this is the way to go...joining the returns for C2, R3, C6, VR1, R702, and VR701 to a common return path back to the star ground?

[Bonsai]

Try this first:-

Remove purple wire going to VR701

Join input connectr GND to VR701 GND and R702 GND

Connect this group via a 10 Ohm resistor to VR1 and C6 GND

This should cure most of your problem.

After this, if you feel it neceassry, you can tackle the remaining ground issues.

Its important that your small signal ground connections are cascaded correctly to avoid ground loops.

[EchoWars]

What you described is what I almost did before I started seeing input grounding schemes that included pretty much the whole front end of the amp...much as AndrewT suggested.

Whichever way I go, it will be a difficult modification. What might be the pluses and minuses of going one way of the other (besides the fact that you're last suggestion will be by far the easiest)?

[ilimzn]

It would be helpful if you could describe the way the power supply is built (single for both channels, dual...). Also, if you have separate ground lines for the left and right channel, before they connect on the (common?) central ground point.
Finally, does it hum when the inputs are not conencted to anything?

[EchoWars]

Hums regardless of whether anything is connected at the input or not. I used a shorting plug at the input RCA for testing. Anyway, very low-level hum from the woofer (the high-efficiency speakers are off for re-foaming, and all I have for test purposes are some Dynaco A-25's).

Here's the power supply:


And here's the view of the underside of the amp (a mess, I know):


The big chunk of copper is the return point for all the grounds.

[ilimzn]

OK, there are a couple of things I can see that could cause the problem. Having a look at the hum on a scope should also help the diagnosis some, as you can check if it is a 60Hz sine, a 60Hz pulse, or a 120hz pulse.
Each suggests different mechanisms of hum injection. 60Hz sine would point at earthing issues or ground-earth loops (usually therough the mains power lead and interconencts), or inductive/capacitive feedthrough from the mains.
60Hz pulses mean either stray field (possibly due to transformer core saturation) induction, or induction via current loops formed by non-braided wires on either the primary or secondary of the trransformer.
120Hz pulses (all going 'one way' whereas with 60Hz you get one positive and one negative going pulse each 60Hz cycle), suggest filter cap charging current being shared with some part of the power/signal grond of the amp circuit, or the same current passing through unbraided wire that forms a loop sufficiently large to induce the same waveform in other circuits adjecent to it.

From what i can see in the picture of the underside, you have two problems:

1) Your rectifier lines to the main filter cap are not braided topgether, but instead form a rather large loop. Keep in mind that there are narrow pulse currents of many amperes here, so the inductance of this loop actually is quite signifficant, and due to the pulse edge rate, you get a rather unexpectedly high magnetic induction effect. ALWAYS look at these problems by drawing out the actual loop the current has to traverse. It is your goal to minimise the area of this loop as much as you can so that it does not end up being an AC magnetic field source. The wires from the rectifier to the main filter cap MUSt be braided or at least tied together in parallel. Conbtrary to what one would expect, it is not always prudent to keep the length of this wire very short or make it very thick - in fact some series resistance helps dampen the ripple current pulses, a little goes a long way. You get less noise and less strain on the transformer, rectifier and cap.

2) Never wire your ground so that the fil;ter cap charge current and any signal related current share a common path. Even fractions of a milliohm of resistance of a common path is enough to get a mV or so generated by the cap charging currents, intot he signal circuitry, where it risks being amplified. When you refer the result as dB vs 1W output, you'll be surprised how easy it is to lose 20-30dB of signal to noise ratio just on account of this.
If you look at your grounding bar, the rectifier ground attaches to it as well as the filter cap ground, but there is another ground attachment between it, as well as more grounds on the laft and righmost end of the bar. Hence, some part of a ground path is shared between signal and power supply.
The right way to do this is to keep the rectifier ripple currnt loop as much within the filter cap as possible - in other worde, the rectifier has to conenct directly to the cap pins. Since the cap has single pins, the only remaining shared path are the cap contact studs and internals, which canot be avoided anyway in this case.
The ground wire from the main cap then has to go to the two extra filters, and it has to be connected as a sort of T-formed wire link, similar to what you did on the positive side. The horisontal line of the 'T' should connect the two ground pins of the left and right final filter cap. Re-orient them so that the actual length of this wire is minimized (again, trying to minimize common paths of charge and audio currents). The vertical part of the T should then connect the mid point of the left/right final filter cap - pins to he main cap - pin. The central ground point will then be where the vertical and horizontal parts opf the T conenct. Ideally, all other grounds should connect to this point. However, connecting this point to the ground bar and them connecting the other grounds to the ground bar is also acceptable. In this case, the wire run from the central ground point to the ground bar is a shared signal current path for the left and right channel, which means that any resistance it adds reduces the left and right channel separation - but this is FAR less critical than injecting ripple current into the signal!

3) Tis I am not certain of as i cannot see it well enough, but it is a common problem with all amplifiers where the left and right channels are dislocated to one side of the enclosure, but the power supply is common. This invariably implies that at some point the ground connections for the two channels separate from the common star or bus ground. The problem that occurs here is that normally the grounding scheme is such that signal ground is derived from power ground. Typically, a power ground line goes from the central point to the output of the amp (point of maximum curent swing) and then from that point downstream towards the input stages, 'becoming' the signal ground. Because you have two channels phisically distant, you actually in the end get two signal grounds. When a signal ground wire is then routed from each channel to the input connectors, if you drw out the actual ground paths from one input ground to the other, you get a sort of a C shape, enclosing a whole lot of bits and pieces generating various nasty magnetic fields - very often the most obvious enclosed thing is the power transformer. Now immagine what happens when the C is closed into a full loop (O) by conencting the two grounds on the inputs together, which happens when you conenct an interconenct cable, which tends to have the grounds conencted on the 'other end', at the source device. This loop becomes a 'secondary' for the transformer stray field, and indeed for any current loop formed by unbraided wires carrying signifficant current.
There are two general solutions to this problem (excepting the obvious one, having separate isolated power supplies for each channel), one is to make the ground look like a sort of 'fishbone' structure, where one end is the attachment to the central ground (power ground), the other is conencted to BOTH input conenctor grounds. The end of the 'ribs' connect to various ground points in the amp circuits. The other possibility is to run the grounds in parallel to each other even though the lengths end up different, so that there is no chance of a loop being formed (or ends of a 'C' formed ground wire bundle becoming connected and forming a loop).