Trident Series 65 grounding: theory and practice

[leigh]

I am somewhere between knee and neck deep in overhauling my Trident Series 65's master section and grounding. Over the past few weeks I've done dozens of chip swaps, bypass cap experiments, and subtle grounding changes, focusing on the summing amp section so far and lowering the noise floor. After each change I record a noise sample into Pro Tools, so I can track exactly whether or not the changes are helping. (I have not yet done the "traditional" copper bus bar mod to improve the grounding between all the input modules and the master section, but that is likely next.)

I've also been reading a ton of articles and posts about grounding practices, and while some of this has helped, it has also made my head spin a bit, and there are still some conflicting principles I'm trying to resolve. So I'm hoping I could get a couple questions answered here...

Here are the specs:
- Trident Series 65 with Acopian power supply
- 28 input channels, 8 bus
- virtual earth summing through 12k channel output resistors
- summing amp ICs originally were TL071s, now subbed for LME49710s
- the schematic of the summing section is attached for part # reference

Currently there are three "grounds" onboard the Trident: the chassis/shielding ground, the 5V LED ground, and one for everything else, which we'll call the "main ground" (for both signal and +/-18 V power). There are separate PCB traces on the motherboard for the LED ground and the main ground.

Now, in Forssell's summing amp paper, he states:

It is very important to return the reference input of the summing amp to ground point that has the least amount of current flowing through it as you possibly can. An example of this could be a separate ground wire return to the central ground point in the console.

However, the Trident's summing amps don't have "isolated" ground references at all. In fact, pins 3 (the + input) of IC20 and IC22 are each within a 1/4" of PCB trace from the grounded end of C22 and C69, respectively. Those caps (not shown in the schematic attached) are the big local power bypass caps (originally spec'd as 100µF, now bumped to 470µF).

Anything other than a perfect ground on pin 3 of the summing amps gets amplified by something like 30dB. So, question #1, was placing the grounded end of power bypass caps right by those + inputs a poor/compromised decision, or is there some deeper wisdom to it?


Now, following Forssell's advice, it would seem the ideal grounding scheme for the summing amp would be to run separate wires from the + input pin of each summing IC to his "central ground point". Physically, that wouldn't be too crazy, running another small Molex connector to a central ground point, in order to provide the summing amps with the cleanest ground available. This would also require I cut any PCB traces connecting to the + inputs, and run local jumper wires on the board to re-establish regular ground to whatever sections of ground traces were interrupted by this change.

However, then I have to determine what that central ground point should be. And, reading his advice carefully, to use a "ground point that has the least amount of current flowing through it as you possibly can", I get confused. I would think that a "central ground point" would actually have the greatest amount of current flowing through it, since all the current leaving the console via the various ground paths has to pass through that single point. Is that correct?

I'll pause here... I feel I'm close to understanding grounding in a comprehensive way, but every article uses slightly different terms, or muddles one concept or another, and it just hasn't gelled 100% for me yet. Thanks for the help.

cheers,
Leigh

 

[gyraf]

Didn't Eddie Ciletti write an article on this topic?

 

[leigh]

Good memory! Ciletti did write about some ground mods for the Trident S65. Here is his article:

http://www.mixonline.com/mag/audio_analog_maintenance_upgrade/ (the article's photos unfortunately aren't showing there, but you can see them here)

In the section "Bright Lights, Big Torture", he describes a listening test where first you mute all channels, and then you de-assign them from the mix buss as well. That's actually what got me started on this current de-noising quest!

He talks about running a bus bar to improve the ground between channels. He also talks about running a beefier "local" ground within just the master section strip - "a new solid ground wire is paralleled with the original ground trace". I also took that idea.

However, he doesn't go into detail beyond that, hinting only that "three black wires branch off from the connector to reinforce the ground for the more 'distant' areas in need. You really can't have too much ground."

The more I've read about ground, and experimented with it, I've found that to not exactly be true. Running ground wires just anywhere isn't going to necessarily help you.

So, yes, good call, Ciletti's article is definitely relevant. My recent work has been to understand these issues in finer-grain detail.

 

[JohnRoberts]

Maybe you need to ask the guy who designed the Trident.  ;D

I have written a bit on the subject but there are not simple answers for re-engineering old finished consoles.

First I do not like calling signals, "signal ground". Ground is a concept not a voltage.

It sounds like the Trident is using what I call brute force grounding, so in theory if 0V signal reference at the channel sends and 0V signal reference in the master section are close to the same voltage it's all good. The problem with brute force is that it is pretty much impossible to do that.

Eddie Ciletti is probably a good source of minimally invasive mods to reduce the ground potential errors.

I personally think decades old consoles (including mine) should be turned into planters, but opinions vary.  8)

JR

 

[leigh]

Hi John, thanks for replying here. I have probably read more posts from you about this subject than anyone else!

I agree, re-engineering old consoles isn't an easy task, but I'm not quite ready to turn mine into a planter . I have read you advocating for the differential summing approach over the "brute force" approach, and wondered about implementing that retroactively in the Trident.

Attached is a schematic of the last active stage in each Trident input channel, before the signal runs to the panpot and assignment switches. At what point there would you take your 0V "signal ground" for a differential scheme?

Fader is 10k, R117 = 47k, R69 = 47k, R70 = 36k (opamp configured for 5dB gain)

It's easy to make changes in one spot, i.e. the master section, but the effort of adding that extra resistor to each input channel gets multiplied, x28 in this case... so, wondering where exactly it has to be snuck into the existing layout.

 

[leigh]

And, to put that potential effort/payoff in perspective, if I just "brute force" it via big copper bus bar, the greatest physical distance that the ground has to travel, between input channel #1 and the master section, is 4 feet.

Someone better with the maths could probably tell me what the noise benefit would be of:

- a proper 0V differential summing bus connected to each channel via 100ohm 1% resistors

versus

- making direct ground connections from each channel to a 1/2" x 1/8" x 4' grounding bar (which gives at most 0.0005 ohms of resistance)

 

[audiomixer]

I don't know whether the brute force approach is any good:

your 0V reference (or ground or whatever) will not carry any relevant current, so no current = no voltage drop across your ground path resistance. don't over do it. on the other hand you might consider the potential issues of creating a ground loop or stability issues... I do have a focusrite channel strip (19", 1HE) that does have ground distributed around the PCB to the different circuit areas via additional wires from the main ground point.

while I love the idea of making stuff as good as possible just for the fun of it, what disease are you trying to cure? is your desk noisy? any numbers on that?

cheers,

Michael

 

[radardoug]

You need to realise that with any particular design there is an end point to the amount of change you can make while getting meaningfull differences. What is your current noise floor? Is it below your practical recording noise floor? If so, you might as well stop making changes.
Thousands of good recordings were made on Trident 65's with none of your mods.

Yes, you could rewire it to a quasi balanced summing bus. But how much noise reduction would you get? Is it worth the hours of work it would take?

The quickest simplest method of getting a gain is to do the bus bar option with the biggest bus bar you can afford, but the difference might be small. Or buy another console with better specs if that is what floats your boat.

 

[JohnRoberts]

Hi John, thanks for replying here. I have probably read more posts from you about this subject than anyone else!

I agree, re-engineering old consoles isn't an easy task, but I'm not quite ready to turn mine into a planter . I have read you advocating for the differential summing approach over the "brute force" approach, and wondered about implementing that retroactively in the Trident.

Attached is a schematic of the last active stage in each Trident input channel, before the signal runs to the panpot and assignment switches. At what point there would you take your 0V "signal ground" for a differential scheme?

Fader is 10k, R117 = 47k, R69 = 47k, R70 = 36k (opamp configured for 5dB gain)

It's easy to make changes in one spot, i.e. the master section, but the effort of adding that extra resistor to each input channel gets multiplied, x28 in this case... so, wondering where exactly it has to be snuck into the existing layout.

I repeat that I have not abandoned my suggestion to make this a planter.

=====

From the screen shot you provided the where is obvious, where the fader and two resistors are connected together is a local channel 0V node. Of course the pan pot is between there and the actual sends, but i suspect it is still referenced to that same node.

As others have asked, what is the "problem" you are trying to fix... Crosstalk? hum in the bus noise floor? The differential bus topology is more accurate and might reduce ground (PS) related hum, but is not generally lower wide band noise

You may have already received a step improvement from using newer low noise opamps in the sum amp. Bus amp noise floor is generally lower than the noise floor from an open mic in a typical room. Making it a lot lower is not a good use of time (IMO).

JR

PS: Not to veer but noise is not the only way that sum amps can be not ideal. High closed loop noise gain can lead to increased phase shift and distortion, but the new opamps can help that (some) too.

 

[leigh]

To answer the question asked by the last three posts here, the problem I'm trying to tackle is the noise floor of the summing buss. The wideband noise is fairly low, but 60Hz and its harmonics are still showing up.

As I mentioned, I have recorded noise samples into Pro Tools, so I can tell you something about the numbers. With all the Trident's input channels de-assigned from the master bus, the overall noise on the mix outputs shows a *peak* (not RMS) meter value of around -96 dBFS. So RMS for the noise would be about 3 dB under that. I'm not certain how to convert that into a normalized SNR, but these converters are set up with 16 dB of headroom, so if you were to call a -16 dBFS RMS sine your "signal" level, SNR would be -99dB - -16dB or -83dB. ??

The rest of my mixing setup is Aurora converters with -117dB dynamic range, so as far as "how far should I take the Trident", I can always try to get closer to that.

I also admit I'm doing this for educational value as well, learning to "make stuff as good as possible just for the fun of it".

 

[leigh]

I don't know whether the brute force approach is any good:

your 0V reference (or ground or whatever) will not carry any relevant current, so no current = no voltage drop across your ground path resistance. don't over do it.

No "relevant current"? I'm not so sure about that.

What I'm finding is that there is enough current on the 0V "signal ground" (which I will henceforth refer to as "signal common") to be audible. Since the signal common is shared with the "power common", things like bypass cap ripple show up on the signal common line.

When it comes to installing a big, console-wide bus bar that will act as signal common (and, unless I go the differential summing amp route, will also act as a shared "power common"), I do anticipate an improvement in the noise contributed by the input channels. Currently, with the stock (motherboard) grounding, I have compared the noise when only channels 25-28 are assigned, versus the noise when only channels 1-4 are assigned (in all cases, faders are down and channels are muted). The higher set of channels is physically only a foot away from the summing amps; the lower set of channels is four feet away from the summing amps. The ones further away have about 3dB more overall noise on them, in the form of 60 cycle hum and its harmonics.

 

[leigh]

From the screen shot you provided the where is obvious, where the fader and two resistors are connected together is a local channel 0V node. Of course the pan pot is between there and the actual sends, but i suspect it is still referenced to that same node.

Right, that would be the general idea, but...

Would you disconnect that segment of "local channel 0V node" from the rest of the local channel, to isolate it from stray currents and power supply noise? So, cut some traces, and have only the cluster of the bottom of R117/R69/fader attach to a new 0v bus via the aforementioned 100ohm resistor? The new 0v bus would do this with every input channel, and it would attach to the + input of the virtual earth summing amp, and it would eventually attach to the board's "signal common" too, but only in one "central" place.

I'm attaching an expanded schematic of the input module's output section. In red I've circled the cluster of components I'm talking about above. (If the red circle isn't showing when you expand the image it means you'll have to scroll the image to the right.) The panpot, aux sends, and bus sends that follow the output amp are all shown now too.

 

[JohnRoberts]

From the screen shot you provided the where is obvious, where the fader and two resistors are connected together is a local channel 0V node. Of course the pan pot is between there and the actual sends, but i suspect it is still referenced to that same node.

Right, that would be the general idea, but...

Would you disconnect that segment of "local channel 0V node" from the rest of the local channel, to isolate it from stray currents and power supply noise?

No... Presumably the layout is decent and and the signal is clean at the channel output wrt to 0V at that point. If it isn't don't waste your time on the master sum bus. Garbage in would equal garbage out when perfectly summed. The assumption is that it is clean at the send from the channel strip. (If uncertain you might want to listen to the signal wrt that local ground but I suspect the signal is pretty quiet when the fader is pulled full down compared to where the fader is attached.)

The 0V bus is basically to sampled sum of all those 0V channel nodes to create a sum of all 0V, just like the sum of all channel signals in the normal bus. This sum of all 0V is differentially subtracted from the sum of all signals.

This cancellation is only perfect when all channels are assigned, but that also happens to be the worst case for noise gain, so less perfect cancellation when there are less channels assigned and less to cancel out generally works OK.

So, cut some traces, and have only the cluster of the bottom of R117/R69/fader attach to a new 0v bus via the aforementioned 100ohm resistor? The new 0v bus would do this with every input channel, and it would attach to the + input of the virtual earth summing amp, and it would eventually attach to the board's "signal common" too, but only in one "central" place.

no..  The sum of all 0V bus connects to the master section local 0V through exactly one same value resistor (100 ohm?). Otherwise the differential math is not correct. Perhaps not obvious but this sum of all 0V is just another signal bus, it just happens to be carrying "ground" noise to subtract out. 

I'm attaching an expanded schematic of the input module's output section. In red I've circled the cluster of components I'm talking about above. (If the red circle isn't showing when you expand the image it means you'll have to scroll the image to the right.) The panpot, aux sends, and bus sends that follow the output amp are all shown now too.

I wonder if you could grow petunias in there?  Just kidding some people like the EQ voicing of the old Tridents.  I never spent any time near one so am expressing general opinions about old consoles. Any old console. 

JR

 

[leigh]

Perhaps not obvious but this sum of all 0V is just another signal bus, it just happens to be carrying "ground" noise to subtract out.

I get that, and the whole scheme started making sense when I thought of it like that. So for this 0V "mix common" line, you WANT the noise on there. You don't want to grab a (possibly cleaner) reference point from a central ground, you want to grab it from right where the signal lives, and you want it to travel the exact same physical distance, over the same kind of conductor (ribbon cable, motherboard trace, whatever). You want the right kind of noise on there, namely, whatever noise is also on the signal line, so the noises can be cancelled out at the summing amp.

All of which makes it quite difficult to pull off as a retrofit. But, I'm glad to understand the concept.

 

[leigh]

So, taking a differential summing bus off the table for this Trident S65, I'm back to the brute force approach of copper bus bar for a cleaner (i.e. lower resistance) ground.

And, when it comes to sensitive points in the circuit, like the summing amps' 0V reference (+ input), I would also give those points direct connections to the bus bar. No more sharing PCB traces with power bypass caps.

The Analog Devices application note AN-202, written by Paul Brokaw, has been helpful in clarifying my thinking about better ground topologies. Anybody else read this?

Attached is a diagram from this app note, which separates the concept of "ground" into signal common and power common. (This is the terminology I borrowed earlier in this thread.) If you're going to go for "brute force" grounding, getting the topology right seems absolutely crucial.

I'll close for now with a quote from AN-202 that I think John will appreciate:

The term “ground” too often seems to be associated with a sort of
cure-all concept, like snake oil, money, or motherhood. If
you are one of those who regards ground with the same
sort of irrational reverence that you hold for your mother,
remember that while you can always trust your mother,
you should never trust your “ground.”

 

[joaquins]

Well, hard topic here, I'm building my own mixer, long term project, practically abandoned at the time. For my first build on that I will go with a pretty bad idea, basically brute force, starting with a thick PCB trace and once I tried and measure it I will add a cooper bar to it, but as it is completely modular I will only have to change one small board where pan and assign is being made, in each channel, which have 4 switches, one pot and a bunch of resistors and the bus board which are just straight traces picking up everything there. I plan to go for a ground sensing topology, the sum amp has a separate input for ground, so I won't need to change that, I need bigger switches and connect one resistor to the reference track for that ground input at the sum amp, one resistor for each channel to each bus, I think that will be enough, then probably connecting all with crimped terminals and a ribbon cable. As my opamps are discrete then I can play around with opamps there, even build a single transistor input and IC opamp on a board  that fits the 2520 footprint, or any other option, the idea is to have different summing amps in my different bus. All of this playing around is because it's my design and I have plenty of room to do that in my desk.

I think the think because is hard to understand this part of our mixers is because is not easy to think that with concentrated parameters, maybe applying some transmission lines methods may be helpful, I don't know. As JR says, brute force isn't a good idea, if you want to know if there will be any difference you first need to measure your noise, then tinning the ground trace with some bulky solder can bring the resistance on your trace by half or even more, if it's not already done. Then if you notice a difference on your measurements but still not enough for you you may go for the big cooper bar and try that out.

JS

 

[JohnRoberts]

For brute force to work well, the previous design still needs to be optimally designed with all the local circuitry referenced coming and going to that brute force ground node.

Someone mentioned that Eddie Cilleti had written up mods for that model. That is probably a good starting point for understanding what worked for him, and /or didn't.

Trying to clean up an old console with brute force, is more like that lump of dirt under a rug... you can stomp on the dirt to shift it around but rarely make it go completely away.

These kind of projects is what makes people think that grounds are magical and mostly guesswork. 

JR

 

[leigh]

For brute force to work well, the previous design still needs to be optimally designed with all the local circuitry referenced coming and going to that brute force ground node.

Agreed. Which is why the master section of my Trident now looks like this (or see attached photo). A heavy "local bus bar" (12 ga wire) is installed, that is soldered directly to the ground pins from the motherboard. From there, smaller runner wires connect to various points on the PCB. This accomplishes two things:

1. Drops the resistance of those connections by about an order of magnitude (roughly speaking, 10ths of an ohm per foot for PCB traces, vs 100ths of an ohm per foot for 22ga wire).

2. Allows separation of "signal common" and "power common" returns to ground (as per the diagram previously posted from AN-202).

Someone mentioned that Eddie Cilleti had written up mods for that model. That is probably a good starting point for understanding what worked for him, and /or didn't.

Yes, that was Gyraf and myself, in posts #2 and #3 above. And, as I said, Ciletti's article did in fact make for a good starting point. However, it did not go into detail about separate ground runs within the master section, so I've been doing my own work to figure out a better scheme.

These kind of projects is what makes people think that grounds are magical and mostly guesswork.

And how do we fight magic? With science!

I have made the effort in this modding process to quantify my results after each step. I didn't want to just shotgun a bunch of rumored "improvements" at the thing, and hope for the best (or, quite easily, make matters worse due to some overlooked or counter-intuitive process at work).

 

[leigh]

Here's one example of the measurements I've been doing. The attached graph shows two noise measurements made when I was "listening" to various ground nodes. I recorded the noise directly into a high quality A/D converter's differential inputs, attaching a "true" ground (master module pin 3) to the - input, and then a different ground node to the + input. So, what we see is of course the difference between these two points, the deviation of a particular ground node from the "perfect" central ground.

On the left is the ground node at the far end of my local 12 ga wire bus "bar". On the right is the ground note at IC20 pin 3, which is the ground reference input of the left channel summing amp.

As you can see, there is about 8dB more of 60Hz on the summing amp's ground reference input, along with another healthy peak at 180Hz. (The "absolute values" of the dB measurements aren't relevant here, they have booth been boosted quite a bit in the DAW to make them show up better on the graph.)

Yes, that 60 cycle noise is waaay down there. But it's measurable. And, being on the summing amp's reference input, it's going to get boosted up by about 30dB by the summing amp alone, not to mention the overall noise floor boost that will happen when 2-bus compression (and later, mastering limiting) is applied to the mix.

 

[JohnRoberts]

Be careful when sniffing around for noise, since it is always wrt some arbitrary 0V reference. I will probably sound like a broken record here but console design is about passing signal cleanly between sundry local 0V references. For brute force you are trying to force multiple local 0V nodes to be the same voltage.

As you have gathered things like PCB layout can be very important, while it isn't just a current thing, one common current mistake made by inexperienced console designers is to not account for the headphone jack ground current, that often corrupts the master section by leaking into everything.  ;D ;D

I like your methodical, empirical approach. Be consistent with what you use for your measurement reference 0V.

JR