Trident Series 65 grounding: theory and practice

leigh said:

Tonight, however, I have been taking a look at impedance balancing some of the 1/4" outputs that are currently simply unbalanced. I understand in principle how this is supposed to work, but in my testing, adding a terminating resistor between pin 3 (well, technically, the "ring" of the TRS output jack) and ground has only hurt the noise floor. ..... In testing, however, adding any resistance only increases the level of background hum. It's only by a couple dB (between 0Ω and 100Ω termination), but it's definitely not helping things. And so this practical example is not conforming to the theory of what I thought "impedance balancing" was.
................
2. When comparing 0Ω and 100Ω pin 3 termination, the same jumper wires are in place, running to the same trimpot. In other words, I am comparing different settings of a trimpot, not comparing the presence vs the absence of extra jumper wires (which could be picking up extra noise).

PS: Attached is a schematic of an impedance-balanced output, courtesy this Douglas Self article.

Click to expand...

- What is this stage feeding?  If its not feeding a properly designed Balanced i/p stage, this is what I'd expect.
- The 'balancing' resistor MUST be connected to the Clean earth used by the final OPA
- The physical area between the T & R connections & tracks, the 2 x 100R resistors, the final OPA, its Clean earth MUST be as small as possible.  It is a loop aerial and just right for picking up hum.  Using a trim pot will increase this area.
- Is the hum clean 60Hz or buzzy stuff?

Adding the impedance balanced resistance will only reduce noise picked up by the cable plugged into that output transferring signal to a balanced receiver elsewhere.

ASSuming the resistor is just added in series with the same exact ground/0V node that the jack (-) was attached to before, the only conclusion is that the added resistance is interfering with the receiving differential's ability to cancel ground differences.

I would first suggest testing the result into a few different balanced inputs to help isolate  where the problem is.  If the following input is unbalanced the extra signal - source impedance may not be helpful.

JR

ricardo said:

What is this stage feeding?  If its not feeding a properly designed Balanced i/p stage, this is what I'd expect.
- The 'balancing' resistor MUST be connected to the Clean earth used by the final OPA
- The physical area between the T & R connections & tracks, the 2 x 100R resistors, the final OPA, its Clean earth MUST be as small as possible.  It is a loop aerial and just right for picking up hum.  Using a trim pot will increase this area.
- Is the hum clean 60Hz or buzzy stuff?

Click to expand...

Good ideas of things to check, thanks. I had the insert send feeding an input of the AD converter (Lynx Aurora), which is definitely a balanced input stage. But thinking through these helped me find the fault in my testing method.

The fault was that on my patchbay, the Group inserts are half-normalled. So, although I had patched the insert send to the Aurora's input, the connection was still simultaneously normalled back to the Group insert's (unbalanced) return jack. The solution was of course to simply plug a dummy patchcord into the bottom row of the patchbay pair, breaking the normal.

With that loopback connection to the insert return broken, I finally saw the terminating resistor behaving as it should. The difference between the trimpot set for 0Ω vs. 100Ω is an ~8dB drop in the 60Hz peak, and about a ~4dB drop in the 120Hz peak. The 180Hz peak seems unaffected (that seems to always be the stubborn peak, in my testing this console).

- The 'balancing' resistor MUST be connected to the Clean earth used by the final OPA

Click to expand...

This was an interesting experiment. Previously I had only attached the jumper wire to the motherboard's ground in testing – so, at a point right next to the Group module being tested, but still this point is a "mixed" ground point, with signal and power connections both being made through it.

So, instead of referencing ground from the motherboard, I tried attaching the jumper wire directly to IC13's pin 3, which is the ground reference (+ input) for the final opamp before the 100Ω buildout resistor. By doing so, I was "jumping over" about 7" of PCB track, 1/8" wide, which seemed like a short and robust enough connection that it wouldn't be having significant effects.

But apparently it was having some effect - by referencing ground directly to IC13 pin 3, the 60 and 120 peaks, already quite low, disappeared into the general noise floor. And the stubborn 180Hz peak dropped 4dB (still is about 12 dB above general noise floor). Unfortunately, I don't know if I'll be able to figure out a sensible way to run that ground reference for the insert send separately, and still keep this console as a modular system without awkward hanging wires.

JohnRoberts said:

Adding the impedance balanced resistance will only reduce noise picked up by the cable plugged into that output transferring signal to a balanced receiver elsewhere.

Click to expand...

Right - I should have mentioned that the cable between the insert send and the converter's input is part of a 30' snake. This snake runs around behind the computer box and in back of a rack of outboard gear, so it has ample opportunity to pick up noise along the way.

Be aware that tying that output to your op amp pin 3 opens up the risk of injecting noise into that clean 0V. If the buffer op amp is inverting as it sounds like, connecting the output ground to chassis and then adding two resistors to make the buffer differential will reference the signal to that low impedance output ground while making that output low impedance. 

JR

PS: did i mention I like differentials?

JohnRoberts said:

Be aware that tying that output to your op amp pin 3 opens up the risk of injecting noise into that clean 0V. If the buffer op amp is inverting as it sounds like, connecting the output ground to chassis and then adding two resistors to make the buffer differential will reference the signal to that low impedance output ground while making that output low impedance.

Click to expand...

You lost me in the 2nd sentence there.

"Connecting output ground to chassis" – done, the new TRS jacks I installed have pin 1 to chassis ground.

"adding two resistors to make the buffer differential" – are you talking about adding two more resistors, or are you counting the 100Ω resistor I have already added between signal ground and the output TRS ring?

I'm attaching the schematic of the Group summing and insert send stage.

JohnRoberts said:

PS: did i mention I like differentials?

Click to expand...

Hmmm... maybe once or twice...  ;D

leigh said:

JohnRoberts said:

Be aware that tying that output to your op amp pin 3 opens up the risk of injecting noise into that clean 0V. If the buffer op amp is inverting as it sounds like, connecting the output ground to chassis and then adding two resistors to make the buffer differential will reference the signal to that low impedance output ground while making that output low impedance.

Click to expand...

You lost me in the 2nd sentence there.

"Connecting output ground to chassis" – done, the new TRS jacks I installed have pin 1 to chassis ground.

Click to expand...

OK my ASSumption was correct. IC13 is a simple inverting op amp. If you add two more resistors, a 4.7k between IC14 + and IC13+, then connect a 12k from IC13+ to bottom of 47k. Also connect that 47k to chassis ground instead of clean bus 0V. For your output audio - feed  connect the 100 ohm r to that same output chassis ground node.  What the differential op amp stage (ic13)  will do, is detect any voltage difference between the clean bus 0V and output chassis ground where you connect the 12k and then add or subtract that error voltage to the audio output so the audio is referenced relative to that low impedance output ground.  This way any noise coming back in is dumped to a low Z sewer line ground, and audio output differential compensates for any voltage caused by that noise current flowing into the ground.

Not bad for adding two resistors to the BOM, but good design does not have to be expensive... 8)

JR

Very tricky! So, like this? Or, would IC14's + input still connect to clean signal ground? Wait, it would need to, right?

Yes IC14+ connects to clean 0V.

JR

OK, so here is a revised schematic. IC14 references ground as before, but IC13 no longer does.

Seeing if this can feasibly be done as a PCB modification is another matter... I can tell already that cutting the existing IC13 trace to ground doesn't leave any space to hang resistor leads to the adjacent PCB trace, so it would mean hanging two resistor leads directly off the IC pin.

Your initial caveat was:

JohnRoberts said:

Be aware that tying that output to your op amp pin 3 opens up the risk of injecting noise into that clean 0V.

Click to expand...

And this is the whole reason to dump pin 3 to chassis earth with this scheme (if I understand you right). Is the concern mostly with the noise that gets picked up along the cables, or noise directly from the pin 3 of whatever outboard gear this insert send gets plugged into?

leigh said:

OK, so here is a revised schematic. IC14 references ground as before, but IC13 no longer does.

Seeing if this can feasibly be done as a PCB modification is another matter... I can tell already that cutting the existing IC13 trace to ground doesn't leave any space to hang resistor leads to the adjacent PCB trace, so it would mean hanging two resistor leads directly off the IC pin.

Your initial caveat was:

JohnRoberts said:

Be aware that tying that output to your op amp pin 3 opens up the risk of injecting noise into that clean 0V.

Click to expand...

And this is the whole reason to dump pin 3 to chassis earth with this scheme (if I understand you right). Is the concern mostly with the noise that gets picked up along the cables, or noise directly from the pin 3 of whatever outboard gear this insert send gets plugged into?

Click to expand...

Chassis ground is a low impedance sewer line so not very sensitive about noise current.  On the other hand any noise induced into the 0V connected to the virtual earth sum amp + input will get amplified by N+1 times, where N is the number of stems assigned to the sum amp. This is why we are generally careful about connecting sensitive local internal voltage nodes to the outside world. 

Using the differential amp allows the actual output signal + and - to be robust low impedance and clean while not exposing the inner circuitry to external noise.. 

Sorry it isn't easy but that is life when you try to modify old soldiers. If you know what you are connecting to will be well behaved you might just leave it alone.

JR

leigh said:

OK, so here is a revised schematic. IC14 references ground as before, but IC13 no longer does.

Click to expand...

Not sure why you've removed R91

I grovel at Guru JR's feet.

ricardo said:

Not sure why you've removed R91

Click to expand...

That is stock. The schematic shows a resistor symbol there, but in the parts list it says "R91 – link", and it's just a wire on the pcb.

JohnRoberts said:

Using the differential amp allows the actual output signal + and - to be robust low impedance and clean while not exposing the inner circuitry to external noise..

Click to expand...

Looking back through my notes about other line output circuits I had considered, your idea is quite similar to what Douglas Self calls a "ground-cancelling output" – see Fig 5a here (also attached): http://www.douglas-self.com/ampins/balanced/balanced.htm#4


Two key differences that I see are that:

1. In Self's circuit, the pin 3 terminating R (100Ω) is shunting only to ground, rather than being in series with the extra two resistors as it is in yours.

2. Self does not differentiate between using chassis ground vs signal ground – his primary stated goal with this circuit is to clean up the signal arriving at the other end, rather than to keep noise off what should be a clean internal 0v node.

Self also does not show the R87 (the 47k resistor), but my understanding is that's just there as a drain for the series output cap (C45), making that a trivial difference.

Are there other key differences that I'm overlooking? Or are these two variations on a theme, with different initial goals but very similar final functioning?

Work continues here on the Trident S65 grounding scheme.

1. I recently made a lengthy post on my tech blog, detailing how I rewired the Aux out XLRs to address the "pin 1 problem". This was a bit trickier than just lifting pin 1 from signal ground and shunting it to chassis ground, due to the presence of the 12k terminating resistors that need to run from the hot and cold signal legs to signal ground. Full details here.


2. After some positive results when testing with a 1/2" copper bus bar (laying it inside the board, and jumpering it to channel 1 ground and master ground), I decided to go ahead with a full install. See the attached photo for the results of that so far. I am finishing a couple other mods while I have all the channel modules pulled, so I haven't gotten to do a full test of the bus bar yet. I will post results when I do.

As you can see in the photo, I also added a barrier terminal block for power and ground routing (the big block on the left). Following an idea in Ciletti's article, I am separate running power wires (18ga) to the smaller motherboards, not visible in the photo. This is in lieu of the stock scheme of jumper wires connecting power "momma to momma," as Ciletti says – and those motherboards run power down a 1/16" PCB track. By my calculation, that means that (stock) the motherboard gives about 0.4 ohm of track resistance for the power lines by the time they get down to channel 1. The 18ga wire will improve that by roughly an order of magnitude, for those distant channels. I will be hooking those new power wires up *after* the initial round of testing the new bus bar, so I can see what additional effect, if any, can be measured from the improved power distribution.

There's one other mod going on here that I'm proud of, because it's my favorite kind of hackery: finding some unused resource, already in plain sight, and re-purposing it. On the right side of the photo you can see two other, 4-position, terminal blocks. Each of those 8 lines connects to motherboard pin #7 for groups 1-8. Stock, pin #7 is used for the master module to trigger the "Auto Mute" function to all channels that have been armed for auto mute. However, the group modules don't have auto mute, so pin 7 on the groups doesn't connect to anything. So, while I had the motherboard pulled out, I tacked a separate wire to each group pin 7, and cut the motherboard traces that were connecting them. Now I have a new, unused pin on each of the groups. My current plan for these pins is to provide a separate ground,  optimized for the inserts (as detailed above in this thread). However, if that doesn't prove to be an effective mod in situ, I could add some other group functionality on those new "free" pins.

leigh said:

JohnRoberts said:

Using the differential amp allows the actual output signal + and - to be robust low impedance and clean while not exposing the inner circuitry to external noise..

Click to expand...

Looking back through my notes about other line output circuits I had considered, your idea is quite similar to what Douglas Self calls a "ground-cancelling output" – see Fig 5a here (also attached): http://www.douglas-self.com/ampins/balanced/balanced.htm#4


Two key differences that I see are that:

1. In Self's circuit, the pin 3 terminating R (100Ω) is shunting only to ground, rather than being in series with the extra two resistors as it is in yours.

Click to expand...

Since that was a few days ago I don't recall the output being in series with 3 resistors, I think there was a chassis ground in there. 

2. Self does not differentiate between using chassis ground vs signal ground – his primary stated goal with this circuit is to clean up the signal arriving at the other end, rather than to keep noise off what should be a clean internal 0v node.

Click to expand...

I am apprehensive about critiquing Self's schematic (he is very highly regarded). This is difficult.  OK, what his circuit is actually doing is using a simple differential circuit to reference the hot output to the actual audio cold output. 

I see a few potential balance issues with this. The two Rs resistors will need to be different actual resistance values to present the same impedance to the + and - outputs since the low side also has the + input network in parallel. The second possible balance issue is that the + output feedback is taken from before the Rs, while the cold side feedback is grabbed after the Rs. Lastly the use of just one ground symbol is not reality. The ground on the cable shield is presumable connected to chassis at both ends (or perhaps not). But suggesting that this shield ground is also the same ground node as audio 0V may work in theory but not so well in the real world. 

I repeat I am uncomfortable criticizing Self.. He is probably just demonstrating a basic concept.

Self also does not show the R87 (the 47k resistor), but my understanding is that's just there as a drain for the series output cap (C45), making that a trivial difference.

Click to expand...

Yes. in fact for perfect balance you would need to add a cap in series with low side output, but that balance would only matter for very LF noise.

Are there other key differences that I'm overlooking? Or are these two variations on a theme, with different initial goals but very similar final functioning?

Click to expand...

I don't think Self's circuit is intended to be a finished cut and paste output, but he demonstrates the power of the differential, if we ignore differences between sundry ground nodes. . 

JR

At long last... a copper bus bar before and after photo!

Tonight I got all of the Trident's modules re-installed, after having had them all pulled out for a couple weeks. During that time, I installed a 1/2" copper bus bar along the top of the motherboard, and ran individual 16ga jumpers from the bar to each module's ground pin.

The "noise meter" plug-in shows an overall 9dB drop in noise. But the spectrum analyzer shows a much more detailed picture.

As in previous tests, the signal has been captured with the AD converter set to -10 dBV sensitivity (giving about 12dB hotter signal than a +4dBu input setting would), and then in the DAW it has been jacked up another 60dB. So we're looking far down into the noise floor here.

In both the "before" and "after" mod snapshots, all 28 input channels and 8 groups are assigned to the master bus, with the master bus fader at maximum. All channels and groups are muted.


PS: The white peaks are the left channel, the blue are the right channel. For the record, not sure what's going on there with the increased 120Hz in the left channel.

I think you will find your noise level is the reading shown plus 12 dB. Your setup is not adding another 60 dB of gain. So somewhere around -66dB relative to +4. Or -86 dB relative to +24.
Buy a used HP 334 on Evilbay and make some meaningfull measurements. Cheap quality measurements.

radardoug said:

I think you will find your noise level is the reading shown plus 12 dB. Your setup is not adding another 60 dB of gain.

Click to expand...

With all due respect, on what basis are you making that claim about my setup?

Once I have recorded the test signals into the DAW, they run through multiple Trim plug-ins that boost their gain by a total of 60 dB before the signal hits the spectrum analyzer plug-in.

(I'm just using Pro Tools' stock Trim plug-in, which is capable of a maximum of +12dB gain – so it runs through 5 instances in a row of the Trim plug, each one set for +12dB.)

Dig?


PS: If you're trying to have those numbers make sense, I should re-state that the dB markings are relative to 0dBFS.

So you are saying your noise level is -126 relative to 0 dBFS?
The Trident guys would be really impressed!
I think you would find if you did a proper noise measurement that you would be closer to the figure I suggested.