help with ground scheme please

[gary h]

Hi all,

I've been assembling, disassembling, and reassembling my system for the better part of a year. I have rebuilt, replaced, and redesigned every stage at least once and made many modifications but have yet to be able to fix the -60dbv hum at the speaker terminals.

I put 8 of Peter Daniel's lm3775 amps into an aluminum Mac G5 tower chassis. Each powers a driver of my MTWW loudspeakers. The source is actively EQed by a LR 24dB/oct 3-way crossover, (Rod Elliott's boards and PS design.)

The 6 amp transformers are on the bottom and the XO +/-15V PS, each woofer amp has a 120VA CT toroid and the T/M of each channel shares a 160 VA dual secondary toroid. The rectifier boards are on this level. The amps sit on a middle shelf attached to a thick aluminum heat sink. On the top shelf is the XO. The system also has a powered remote volume and a soft start circuit, both of which have been removed for simplicity.

Attached is a schematic of my star ground scheme. Nearly all ground wires have be replaced by 14 gauge (from romex.) They are all individually bussed to a copper plate inches from the earth ground. (I had previously implemented a ground loop breaker circuit without success.) I have read both the Davenport and Whitlock papers that are recommended and have endeavored to problem solve by taking measurements with both an analog oscilloscope and a soft FFT analyzer on my computer. The system only is connected to a dedicated GFI (Signalscope running on my mac is not connected to this circuit but the lead grounds are.)

All individual stages measure decently well when separated; the chipamps are quiet when fed by an ipod, and the XO measures under 20uV with no source. All this goes out the window when installed inside the chassis and connected to the system star ground. I have gotten rid of some high frequency noise with small capacitors and by moving wiring around but there is a persistent 60 Hz lump which manifests harmonics, (180 and 300 Hz being the biggest at over 100mV.)

I can post pics of the amp and scope measurements if necessary. I am hoping that someone has gone through this nightmare before and can catch the thing I have missed. At this point I must either concede failure or wait for a miracle. Any advice, however trivial or patronizing, is welcome. I have also read through dozens of DIYA posts regarding this topic and tried dozens of potential fixes.

Thanks,

gary

[jackinnj]

First order of biz, read Dave Davenport's DIY Article: Audio Component Grounding and Interconnection

Ground from the input should go directly to the preamplifier/xo. Another connection should be made from the XO to the star ground. It would help if the RCA jack is of an insulated variety, rather than connected to the chassis.

Also, can you rotate the transformer 90 degrees on its axis?

[gary h]

Hi Jack,

Thanks for your reply, I remember you helping me out in the past.

I have had trouble negotiating some of the info I have come across. For instance, the Whitlock paper shows a diagram of the input grounds going directly to the star ground. The differences are clearly due to my lack of understanding. I have tried it both ways without success. I have also removed the pot and selector and run straight to the XO. I'll reconfigure it the way you suggest. Help me understand this, I am under the impression that we want to reference the signal at every new stage to the same star ground point, without creating a loop at any point. The crossover then is the first stage, not the input jacks and pot?

I mounted all the inputs on a 3/16" ABS plastic board to insulate them from the chassis.

The XO power supply (shown) has moved around the chassis. You can see it is mounted on a protoboard with the rest of the PS components. It has been next to the XO, on the other side of the chassis next to the power in, out of the chassis on the desktop, and on top of the chassis. None of these positions seems to have made a difference. Currently it is isolated from the chassis, mounted on a plastic tray (a former divider inside the computer) and has it's ground soldered at the point to point wiring where the filter cap grounds meet.

What's next?

gary

[gary h]

btw,

my wife advises:

God ook zegene de vliegende Nederlandse vrouw

whatever that means.

[Redshift187]

In my design, the diode board is connected to the star ground as well as the center-tap of the transformer, and the amplifier board only connects to the ground from the diode board. The center-tap of the transformer does not connect directly to the diode board.

[gary h]

Hi Tyler,

I used the arrangement you mention at one point, (transformer CT, rectifier boards, and amp boards connected individually to the star ground.) I didn't notice a significant difference. Although this was on the bench.

I connected the transformer-diode board-amp board grounds in series after reading the Whitlock paper on grounding in which (as I understand it) he recommends a star of star formation or, at least, keeping the stage and its power supply grounds connected before going to the star ground. There's a section on the merits of a ground buss that illustrates this.

Here's the paper:

http://www.jensen-transformers.com/a...%20seminar.pdf

Please set me straight if I am drawing incorrect conclusions.

gary

[AndrewT]

Hi,
I have not read the other Members advice, but I would like to take you back to your post1 layout.

There is an input arriving at the RCA socket. This is a two wire circuit. It must have a flow from the source and a return back to the source.
You have a red connection from RCA to Selector and a red connection from Selector to Attenuator. These red "flow" connections each need a return.
A pair of flow and return wires must carry the signal through the various routes between stages and in stages.
The attenuator already has a two wire circuit to the XO PCB.
The three connections, from RCA to Selector to Attenuator to XO, should all be twisted pairs.

Now we have the input side with all the correct flow and returns.
Look at your two input grounds.
RCA to Star and XO to Star. These are returns/voltage references, where are the complementary flow? Ahh, maybe better for the time being to think of one of these as a voltage reference, later when you understand all the flows and returns, you might be able to identify the flow that complements this return.
Delete one of these grounding links. It will need you to experiment to decide which is better for your inputs. Either the RCA to Star or XO to Star will perform, but one is likely to be better than the other.

[gary h]

Hi Andrew,

thanks for your reply.

In reference to your first instructions: You are suggesting that I need a 4-pole selector switch so L in, R in, L shield/ground, and R shield/ground (from one of the three input pairs) all are switched at one time? I am using a selector that has no dedicated ground lug and is a 2P5T type. Or does this simply mean that the input ground/shield should "accompany" the signal to the selector but not actually land on it?

In regards to the second piece, I will reconfigure the drawing with the 2 possibilities you suggest to make sure I have it right. I'll do that tonight.

Thanks,

gary

[AndrewT]

effectively you are asking if the selector must switch the ground.

I would say no.

You must include the route of the returns with the flows for each input circuit.
They will all end up having a common star either at the selector switch or at the Group of input RCA's. That's introduced a third option for where the input reference ties in to the star ground.

[jcx]

can I be the 3rd to recommend Not connecting signal input gnd to chassis "star" directly?

all input paths (sig and sig_gnd) should 1st go to the XO board

your diagram doesn't show a XO PS gnd to XO gnd wire? - it is needed, twist all 3 XO PS wires

a XO board gnd then goes to the chassis star


I would connect each amp board to each diode board, xfmr ct with equal length twisted pairs (V+, PG+) and (V-, PG-) on Peter’s rev B board - possibly entering the amp PCB from the opposite side of the board from the signal wiring

twist each speaker output and load gnd return - I hope all output terminals are isolated from the chassis? - your illustrated shield on the output likely does nothing useful

for safety you need a gnd path from each xfmr secondary (when not using double insulated xmfr) to the chassis/safety gnd, taking the safety gnd connections from (through) each amp board could be OK if all of the connections are secure, heavy enough


the "Nuclear Option" for internal signal connections is to make each amp a differential receiver - not hard - only a few added components per amp board

for stability I would use noise gain compensation since the chip amps aren't unity gain stable - so in addition to the usual 4 resistor differential amp with gain there would be a series R1,C1 between the chip amp's +,- inputs and a cap C2 to local amp gnd to AC gnd the + amp input

lifting Ri and adding R2 gives a differential input – but not balanced input impedance – which could be OK as long as we only need to reject V in the dirty gnds – which is the case if stray magnetic field is reduced by twisting high current wires and by twisting the XO out signals and gnds

the 100 Ohm output series R on the XO buffers should be shorted or a matching 100 Ohms added in each XO output signal pair's gnd

the low frequency common mode rejection is limited by the ratio match of the signal input and feedback dividers