Input Transformer Advice for Common Impedance Coupling (Ground Loop)

[dandeliongold]

Set Up:
64 inputs, 64 outputs (28 balanced 'audio source' devices, 35 unbalanced 'FX devices')

I've been testing my recording studio interface for a while, and have concluded that I have ground loops when using unbalanced external devices in effect send/return configuration.
(For example, using a delay unit with it's output connected to the input of my RME AD converter and the delay unit's input fed from the RME DA converter, using TS connections.)

Identifying the issue:
Using the UNDERSTANDING, FINDING, & ELIMINATING AV GROUND LOOPS pdf, I determined that my ground loop buzzing is caused by 'Common Impedance Coupling'.
(As a side note, the issue also seems to be resolved when monitoring through the RME equipment without the computer powered up, suggesting that there may be a way to break the ground loop via the computer electronics. I'm attempting to pursue this as solution as well, but may be outside the scope of this forum)

Input Transformer / Isolator Options?
The ISO-MAX CI-2RR is recommended for breaking the ground loop. I am curious about other transformer options in regards to price and performance.

The issue is that I need to address this issue with 36 channels of audio. I also want to ensure that the potential coloration from the transformers is at a minimum, as these channels are likely cross patched and fed into each other (an equalized signal fed to a delay signal, both recorded and mixed in parallel). That means there is likely to be situations where audio is passed through these transformers multiple times, within reason.

How much extra work would be involved in ordering transformers and wiring them myself? Maybe I could test a few different units and then bulk purchase the unit I prefer?

What input transformers would you recommend?

Impedance Consideration:
How much consideration should I give to ensuring proper impedance matching to and from the transformer?
Would I need to get different transformer variation for each device to ensure best performance?
Example:
Strymon Timeline output impedance = 100ohms
Strymon Magneto output impedance = 1kohms
(I may not be able to confirm all devices output impedances.)

 

[JayGilb]

More details on your audio interface and pc as well as how it's connected to your pc would be helpful.
Fix the ground loop relating to your pc and there should be no need to add any isolation transformers to your system.

 

[dandeliongold]

This post's scope is mostly focused on advice in a given situation where isolation has been determined to be the solution.

Per this document ( https://www.jensen-transformers.com/wp-content/uploads/2014/08/generic-seminar.pdf ) my conclusion is that an input transformer isolator would be one potential solution. I would like to explore this solution while also attempting to address the other potential solution (computer electronics).

However, if you have any ideas that require specs on my system:
PC with a Asus mobo, i7 3.2gz CPU, 32gb of ram, Windows 10 pro 64bit
-Noise present with various video cards.
-Tested clean when using UFX+ without computer powered on.
-Computer on same or different circuit (as music equipment) has little-to-no affect.
-Interference introduced whenever computer is connected to audio interfaces (via USB2, USB3 or PCIe).
-Interference appears to be worse when running through MADI optical lines (But this is likely due to the fact that the computer has to be connected to monitor this configuration)

RME MADI FX card (master, all other devices fed via MADI), RME M32 pro AD, RME M32 DA, RME UFX+ (x2)
Also tested: RME Fireface UCXII

 

[antcollinet]

Do you need to input to the PC via USB as well as output from it?

If so, look at a USB isolator device such as the one just announced by topping.


If not, just get a USB to optical converter and use that to output audio.

 

[dandeliongold]

Do you need to input to the PC via USB as well as output from it?

If so, look at a USB isolator device such as the one just announced by topping.


If not, just get a USB to optical converter and use that to output audio.

The end master device will be a PCIe card (RME MADI FX). USB won't be plugged in with the final set up, but has been used for testing and was part of the old set up.

 

[JayGilb]

Try cleaning the metal around where your power supply screws into your pc chassis. I have found this to help with noisy pcie cards.
Make sure the screws that hold your mobo to the pc chassis are tight. Clean the contacts on your mobo pcie card slots and pcie connectors with a plastic safe contact cleaner
and re-seat the cards. Re-seat all the power connectors inside your pc.
Physically isolate all of your external RME boxes from each other to see if that helps. Sometimes metal rackmount devices can get ground loops when in electrical chassis contact.

 

[Speedskater]

a] First power all your components from the same AC wall outlet.
Computer SMPS's tend to dump noise onto the computer chassis.
b] so with a heavy wire (12/14AWG) connect the computer's metal chassis to the console's metal chassis.
Heck, connect all your components with an exposed metal chassis to the console.

 

[dandeliongold]

Try cleaning the metal around where your power supply screws into your pc chassis. I have found this to help with noisy pcie cards.
Make sure the screws that hold your mobo to the pc chassis are tight. Clean the contacts on your mobo pcie card slots and pcie connectors with a plastic safe contact cleaner
and re-seat the cards. Re-seat all the power connectors inside your pc.
Physically isolate all of your external RME boxes from each other to see if that helps. Sometimes metal rackmount devices can get ground loops when in electrical chassis conThat

I'll double check all of this, but we did recently do a cleaning on the hardware of the PC. I would be shocked if the RME boxes needed isolating but they already are.

The PCIe card is not noisy, at least generally speaking. It's channels with send return configuration where the noise is present. Devices that are not fed audio and then return that audio do not exhibit the noise. These signals are also always routed through other units (M32proAD and M32DA) and then fed to the card via MADI.

 

[dandeliongold]

a] First power all your components from the same AC wall outlet.
Computer SMPS's tend to dump noise onto the computer chassis.
b] so with a heavy wire (12/14AWG) connect the computer's metal chassis to the console's metal chassis.
Heck, connect all your components with an exposed metal chassis to the console.

All devices on are the same circuit, but the music equipment uses a Furman "Elite" Linear Power Conditioner and appropriate power strips to deliver power. PC is plugged direct into the outlet that the Furman is plugged into.

Seems like maybe you're suggesting the opposite approach (connecting the PC chassis to the "console") as @JayGilb as my "console" in this case would be the digital mixing aspect of the RME hardware. Instead of isolating, tying these chassis together.

 

[JayGilb]

You ground loop is most likely due to

All devices on are the same circuit, but the music equipment uses a Furman "Elite" Linear Power Conditioner and appropriate power strips to deliver power. PC is plugged direct into the outlet that the Furman is plugged into.

Seems like maybe you're suggesting the opposite approach (connecting the PC chassis to the "console") as @JayGilb as my "console" in this case would be the digital mixing aspect of the RME hardware. Instead of isolating, tying these chassis together.

I had both approaches work in reducing ground loop noise.
What happens if the pc is also plugged into the Furman ?
Are the daisy chained power strips of decent quality ? Some just have 3 strips of poor quality metal running in parallel and can easily be bent resulting in poor contact.

 

[dandeliongold]

You ground loop is most likely due to

I had both approaches work in reducing ground loop noise.
What happens if the pc is also plugged into the Furman ?
Are the daisy chained power strips of decent quality ? Some just have 3 strips of poor quality metal running in parallel and can easily be bent resulting in poor contact.

I don't THINK any of the power strips are actually "daisy chained", but rather spreading out from the Furman in parallel, though there may be one or two exceptions. Certainly no daisy chains in the equipment I'm using for my tests.

I did test the computer on the Furman, and results were either the same (or worse) but I didn't write them down. It may be worth re-checking.

I'll see about comparing the currently isolated chassis to a connected chassis scheme, but I'm skeptical. I'd love to be surprised!

 

[dandeliongold]

You ground loop is most likely due to

I had both approaches work in reducing ground loop noise.
What happens if the pc is also plugged into the Furman ?
Are the daisy chained power strips of decent quality ? Some just have 3 strips of poor quality metal running in parallel and can easily be bent resulting in poor contact.

Some interesting results so far.

First, I checked continuity with all ground pins and each devices chassis. All were fine.
Then I checked continuity between chassis of the devices, all fine, and ground pins of all devices (powered off of course) and all were fine. The resistance measured in continuity mode was less than 10ohms any direction (most was -9ohms).

When tying all chassis together with 12 gauge wire, I found that the noise got much worse! I was mistaken about the chassis of each RME unit being isolated. I had forgotten to take into account that they are racked in the same rack, and thus their metal rack ears were grounding the units together. However, when I tried isolating one of the units, the noise ALSO got much worse!

This lead me to loosening and tightening one screw rack screw holding the unit sending/receiving audio. The screw basically acted as a potentiometer to add or remove digital hash. Too tight, hash, too loose, hash. It happened to already be in a very good position, leaving only a background hum, which is what I was attempting to address in the first place.

Given how sensitive it is, I will attempt to match IEC power cable thicknesses, and put them all on the same power strip, etc.

However, the fact that the hum isn't really affected (only covered up by the introduction of more noise) I still suspect there is something I can do within the computer to clean the hum being fed into the system. I just don't know what that is.

 

[AnalogSteph]

The fact of the matter is that you're going from one solidly earthed device (DAC) to another (ADC) via unbalanced connections in a nice big loop (all x36 or at least x18), and as long as you've got part of your ground currents circulating through the unbalanced connections you'll only be able to change the problem but not eliminate it.

The least expensive off-the-shelf transformer-based option I can see would be a total of 18 Behringer HD400s, as janky as that sounds. (US pricing - $522 plus extra cables.) If you have some more budget, ART makes an 8-channel rackmount isolator called the T8 ($180 ea.). 4 of those plus 4 HD400s = $836, or 5 of them = $900.

I would run each RME output into the isolator before it goes on into the FX business.

If the price tag still sounds a bit steep, you could also tackle the problem with a bunch of custom cabling on the input (ADC) side, exploiting the presence of balanced inputs. Those would be like your regular old XLR cables, except one side would get a 1/4" TS instead of a female XLR, with the pinout T --> hot, S --> cold, and shield n/c (#13 in RaneNote 110).
If you feel fancy, include a series resistor matching the output impedance on the corresponding FX device between sleeve and cold for better CMRR, marking cables accordingly. Certainly not a bad idea in the 1 kOhm vicinity, which would result in a substantial imbalance otherwise.

(All of that is assuming that the FX units are all floating with their own individual power supplies or at least all galvanically isolated from each other.)

 

[dandeliongold]

The fact of the matter is that you're going from one solidly earthed device (DAC) to another (ADC) via unbalanced connections in a nice big loop (all x36 or at least x18), and as long as you've got part of your ground currents circulating through the unbalanced connections you'll only be able to change the problem but not eliminate it.

The least expensive off-the-shelf transformer-based option I can see would be a total of 18 Behringer HD400s, as janky as that sounds. (US pricing - $522 plus extra cables.) If you have some more budget, ART makes an 8-channel rackmount isolator called the T8 ($180 ea.). 4 of those plus 4 HD400s = $836, or 5 of them = $900.

I would run each RME output into the isolator before it goes on into the FX business.

If the price tag still sounds a bit steep, you could also tackle the problem with a bunch of custom cabling on the input (ADC) side, exploiting the presence of balanced inputs. Those would be like your regular old XLR cables, except one side would get a 1/4" TS instead of a female XLR, with the pinout T --> hot, S --> cold, and shield n/c (#13 in RaneNote 110).
If you feel fancy, include a series resistor matching the output impedance on the corresponding FX device between sleeve and cold for better CMRR, marking cables accordingly. Certainly not a bad idea in the 1 kOhm vicinity, which would result in a substantial imbalance otherwise.

(All of that is assuming that the FX units are all floating with their own individual power supplies or at least all galvanically isolated from each other.)

Great response!

When doing tests, it seemed as though isolating from the output of each device to the ADC would be the best option, but I can experiment with both of course.

That price tag isn't too bad at all. My main concern with the transformers is a coloration of phase or frequency response adding up when the devices are mixed in parallel, or in the occasional situation where one of these channels would be run into another, and both would be mixed in parallel (for example, running a delay unit into a compressor, then mixing the dry delay and compressor delay (which has now gone through two transformers).

I'll review the Rane link closely for any further tests.

I'm especially curious about your suggestion for the cabling into the ADC. I've tried a few configurations of lifting and merging various pins going to my ADC from the devices with no success.

Would your suggestion of a resistor (matching the output impedance of the device's output) have any audible affect on the signal?
Would this be a more transparent solution over transformers?

 

[KSTR]

@dandeliongold ,
Your RME M-32 Pro AD has balanced inputs and the M-32 (non-pro!) DA as well as the UFX+ have servo-balanced TRS outputs (the M-32's DB25 outs are not servo-balanced, also applies to the M-32 DA Pro, see note below). This means you can (and definitely should) run balanced cabling throughout, even from and to the unbalanced devices, which most likely completely solves your problems (if not, further tweaking involving chassis GNDs, power connection etc might be needed. Same applies to USB isolators and the use of MADI optical to isolate the digital side, as well as isolation mains transformers for unbalanced equipement that has a direct mains earth connection. Audio isolation transformers are really the means of last resort if everything else fails... good quality stuff is costly).

First things first, using balanced audio connections for everything is goal #1:

• Balanced out to balanced in is trivial.
• Unbalanced out to balanced in is also trivial, at the sending device (important!) just connect the cold line (Pin 3 of XLR, or R of TRS) to unbalanced shell and keep shield (Pin 1) connected to shell (only in very rare cases it is better to leave the shield open as per Rane app note suggestion).
• Servo-balanced out to unbalanced in uses the same idea, make the sensing (servo) connection at the proper end, the output end (at the receiving device) in this case, of the balanced cable, that is connect the cold line to shield there, not at the sending device. This only possible because of servo-balanced outputs (or transformer-balanced outs), with standard balanced outs there is no servo action and you're just shorting the cold line, not recommended. With the RME DA, single-ended mode of the servo outs restricts their max level to +21dBu which shouldn't be an issue I'd guess.

Summing it up, you need a set of special cables (adapters are not recommended but may be used for evaluation):

• Servo-Balanced out to unbalanced in: XLR-female (or TRS) --> balanced cable --> TS, with cold (Pin 3 from XLR or R of TRS) connected to S, and shield of balanced cable also connected to S. Mark those cables as they are only suitable to transformer- and serve-balanced outputs.
• Unbalanced out to balanced in: TS --> balanced cable (with cold and shield connected together to S at sending end) --> XLR-male or TRS. Impedance balancing and open shield usually is not required but should be tried if some hum/buzz remains.

Keep your cables as short as possible and uses ones with high-quality low-impedance shield, eg. good sturdy microphone cables with braided shields, not the simpler spriralled shields as used in cheaper, more flexible cables.

To repeat, it is very important that the cable run is balanced (2-wire + shield) over its whole length and the conversion to/from unbalanced happens at the proper end, therefore most of ready-made solutions one can buy are not adequate as they often use unbalanced cable, bascially cheating. With adapters you have to check the pinning (with a multimeter) and they tend to introduce additional contact resistance after the conversion from/to unbalanced, which is not optimal for best hum/buzz rejection.

If you can't make these cables yourself you need to find a competent technician.

And in case you chain several unbalanced devices, use the shortest possible and best shield quality unbalanced interconnects. Lowest shield impedance is paramount here.

 

[dandeliongold]

@dandeliongold ,
Your RME M-32 Pro AD has balanced inputs and the M-32 (non-pro!) DA as well as the UFX+ have servo-balanced TRS outputs (the M-32's DB25 outs are not servo-balanced, also applies to the M-32 DA Pro, see note below). This means you can (and definitely should) run balanced cabling throughout, even from and to the unbalanced devices, which most likely completely solves your problems (if not, further tweaking involving chassis GNDs, power connection etc might be needed. Same applies to USB isolators and the use of MADI optical to isolate the digital side, as well as isolation mains transformers for unbalanced equipement that has a direct mains earth connection. Audio isolation transformers are really the means of last resort if everything else fails... good quality stuff is costly).

First things first, using balanced audio connections for everything is goal #1:

• Balanced out to balanced in is trivial.
• Unbalanced out to balanced in is also trivial, at the sending device (important!) just connect the cold line (Pin 3 of XLR, or R of TRS) to unbalanced shell and keep shield (Pin 1) connected to shell (only in very rare cases it is better to leave the shield open as per Rane app note suggestion).
• Servo-balanced out to unbalanced in uses the same idea, make the sensing (servo) connection at the proper end, the output end (at the receiving device) in this case, of the balanced cable, that is connect the cold line to shield there, not at the sending device. This only possible because of servo-balanced outputs (or transformer-balanced outs), with standard balanced outs there is no servo action and you're just shorting the cold line, not recommended. With the RME DA, single-ended mode of the servo outs restricts their max level to +21dBu which shouldn't be an issue I'd guess.

Summing it up, you need a set of special cables (adapters are not recommended but may be used for evaluation):

• Servo-Balanced out to unbalanced in: XLR-female (or TRS) --> balanced cable --> TS, with cold (Pin 3 from XLR or R of TRS) connected to S, and shield of balanced cable also connected to S. Mark those cables as they are only suitable to transformer- and serve-balanced outputs.
• Unbalanced out to balanced in: TS --> balanced cable (with cold and shield connected together to S at sending end) --> XLR-male or TRS. Impedance balancing and open shield usually is not required but should be tried if some hum/buzz remains.

Keep your cables as short as possible and uses ones with high-quality low-impedance shield, eg. good sturdy microphone cables with braided shields, not the simpler spriralled shields as used in cheaper, more flexible cables.

To repeat, it is very important that the cable run is balanced (2-wire + shield) over its whole length and the conversion to/from unbalanced happens at the proper end, therefore most of ready-made solutions one can buy are not adequate as they often use unbalanced cable, bascially cheating. With adapters you have to check the pinning (with a multimeter) and they tend to introduce additional contact resistance after the conversion from/to unbalanced, which is not optimal for best hum/buzz rejection.

If you can't make these cables yourself you need to find a competent technician.

And in case you chain several unbalanced devices, use the shortest possible and best shield quality unbalanced interconnects. Lowest shield impedance is paramount here.

This response is gold! Thank you!

Could you elaborate on the use of USB isolators. I tried a couple things regarding some of my audio devices (synthesizers) using some USB isolators, and also attempting to lift the ground with a strip of tape. Perhaps there is a recommended product that I would give me the best chance of a good result.

Your suggestions mirrors some of what I've read in the above linked document, but also helps to clarify the importance of which side of the cables to modify and why.

Do you have any thoughts on the idea of adding resistors to ends of cables as suggested by @AnalogSteph? "If you feel fancy, include a series resistor matching the output impedance on the corresponding FX device between sleeve and cold for better CMRR, marking cables accordingly. Certainly not a bad idea in the 1 kOhm vicinity, which would result in a substantial imbalance otherwise."

I'm certainly curious about the use of resistors here, but will otherwise proceed as you have suggested.

 

[KSTR]

Could you elaborate on the use of USB isolators. I tried a couple things regarding some of my audio devices (synthesizers) using some USB isolators, and also attempting to lift the ground with a strip of tape. Perhaps there is a recommended product that I would give me the best chance of a good result.

Personally, I'm using USB 2.0 isolators by Intona successfully for years now, never had any issue with them with a myriad of USB devices.
For some time now, I also have the much cheaper Topping HS01 USB 2.0 isolators which also work reasonably well so far... though very occasionally I have connection problems with these, and they need an extra supply when connecting to USB bus-powered devices.

Isolating the digital side is always a good thing, that's for sure. This often reduces "ground loop" noise problems even without those speciality cables to low enough levels and then using the proper cabling will result in a totally silent system (unless you have some devices that have an intrinsic hum/buzz problem, not much you can do about).

Do you have any thoughts on the idea of adding resistors to ends of cables as suggested by @AnalogSteph? "If you feel fancy, include a series resistor matching the output impedance on the corresponding FX device between sleeve and cold for better CMRR, marking cables accordingly. Certainly not a bad idea in the 1 kOhm vicinity, which would result in a substantial imbalance otherwise."

The point is that balanced inputs are reduced in their ability to reject common mode-noise when the source impedance on cold and hot lines is not the same. As a secondary effect especially for very long cables, any noise ingress into the cable will not distribute evenly on both lines with an impedance imbalance. Therefore, matching these impedances restores the common-mode rejection ratio to the advertised amount. You need to know the output impedance of an unbalanced device and then add the corresponding impedance in the cold leg (ideally including any coupling capacitors).

In practice, with short cables and when the common mode voltage is low to begin with, the reduced rejection is not that much of a problem. And by keeping the GND/shield connection between gear not that much common-mode noise can develop, a few millivolts at most. OTOH, with an open shield/GND connection much larger common-mode voltages can be present and then impedance balancing becomes more important... but it's not always possible to find output impedance without inspecting the circuit and/or proper means to measure it.

 

[dandeliongold]

I've done the following tests:

1.) Device Out > TS > RME In. RME Out > TS > Device In (original configuration)
2.) Device Out > TRS > RME In. RME Out > TRS > Delay Unit In (cold shorted to sleeve/ground at the I/O of the Device) @KSTR
3.) Device Out > TRS > RME In. RME Out > TRS > Delay Unit In (ground opened at the I/O of the Device)
4.) Device Out > TRS > RME In. RME Out > TRS > Delay Unit In (100 ohm resistor shorting cold to ground of cable, sleeve of jack open) @AnalogSteph
(NOTE: 100ohms is the output impedance of the device according to specs)
4.) Device Out > TRS > RME In. RME Out > TRS > Delay Unit In (100 ohm resistor shorting cold to ground of cable, sleeve of jack connected to ground of cable and leg of resistor)

The only improvement was a 4db drop in hum when using TRS with the cold shorted to ground. Beyond that point there was no further improvement.
Putting the computer on a different circuit has little or no affect on the noise.

I can't help but wonder about splitting the hot of the output of the device and inverting one side to create a balanced output. However, since I can find nothing confirming this would work, or how to do it (without a transformer) I have to assume this is not actually practical/possible.

Did I configure these tests correctly?
Is "star grounding" something I could attempt with various unbalanced devices like this? If so, how would the logistics of that work?
Should I attempt to lift the power ground on any of the devices, maybe the computer, with a 3-to-2 prong adapter (screwing it into the outlet screw)?

 

[thecheapseats]

doing a quick read of this thread as you search for solutions - a few things occured to me...

(1) a bunch of balanced i/o and a lot of unbalanced ?fx? well ok - why so much un-bal? (that's rhetorical - no answer necessary)...

(2) bill whitlock's pdf (linked above) as well as his aes papers are 'THE" resources - that you have them is a great start... dean jensen was kind enough to help me personally when I was starting out fifty years ago with application suggestions - convenient we were both in L.A. at the time...

(3) I've bought a lot of jensens xfmrs over the years for various applications... warning - they aren't cheap if you want the 'good' stuff... for that many fx sends, buy them and wire them up yourself to save money (if you have the assy skills)... a pair of sends or receives is one thing - but over 30?... damn...

(3) to my reading @KSTR gave great examples and comments and skillfully outlined the practical problems and solutions most have had to face over the years when dealing with unbalanced stuff in a balanced world... never fun but sometimes you gotta' get it done/interfaced for some reason(s)... but you have a ton of un-bal fx? - I don't get that, but no matter...

if applying a transformer solution with minimal phase error and preservation of frequency response is #1 - well you know what you have to do (I suppose)...

(4) galvanic isolation is/can be your friend (re: usb iso devices - that's a different discussion)... but consumer pc noise if that's part of your issues IS evil around high-end audio...

(5) generally, implementing a comprehensive a/c grounding system for pro-audio (star, capacitvely-decoupled, balanced a/c, etc.) is always difficult after the fact/installation - but it can be done... hang in there...

(6) an afterthought - bal to un-bal is one thing - but have you also considered operating level issues (+4 to -10, and -10 to +4) and inevitable gain matching/staging...

 

[dandeliongold]

doing a quick read of this thread as you search for solutions - a few things occured to me...

(1) a bunch of balanced i/o and a lot of unbalanced ?fx? well ok - why so much un-bal? (that's rhetorical - no answer necessary)...

(2) bill whitlock's pdf (linked above) as well as his aes papers are 'THE" resources - that you have them is a great start... dean jensen was kind enough to help me personally when I was starting out fifty years ago with application suggestions - convenient we were both in L.A. at the time...

(3) I've bought a lot of jensens xfmrs over the years for various applications... warning - they aren't cheap if you want the 'good' stuff... for that many fx sends, buy them and wire them up yourself to save money (if you have the assy skills)... a pair of sends or receives is one thing - but over 30?... damn...

(3) to my reading @KSTR gave great examples and comments and skillfully outlined the practical problems and solutions most have had to face over the years when dealing with unbalanced stuff in a balanced world... never fun but sometimes you gotta' get it done/interfaced for some reason(s)... but you have a ton of un-bal fx? - I don't get that, but no matter...

if applying a transformer solution with minimal phase error and preservation of frequency response is #1 - well you know what you have to do (I suppose)...

(4) galvanic isolation is your friend, re: usb iso devices... consumer pc noise is evil around high-end audio...

(5) generally, implementing a comprehensive a/c grounding system for pro-audio (star, capacitvely-decoupled, balanced a/c, etc.) is always difficult after the fact/installation - but it can be done... hang in there...

(6) an afterthought - bal to un-bal is one thing - but have you also considered operating level issues (+4 to -10, and -10 to +4) and inevitable gain matching/staging...

(1, 3) My set up is a little unique in the number of FX send/returns, but it's very functional for my needs. It just turns out that a lot of gear happens to be unbalanced.
Examples: Tascam Reel-to-Reel Tape machines, Strymon pedals, Moogerfooger pedals, synthesizer with audio inputs (to access to their analog filters and gain stages), and modular gear (filters, phase shifters, frequency shifters, delays and reverbs).

(6) The RME inputs are set to +4dBu, which has proven deal for my needs. RME outputs are set so that a loopback will return at very close to the original signals amplitude. This is essentially a line/synth level I/O, as all my synths, pedals, modular gear, as well as preamps, eqs and compressors work well at this level. (The one exception might be that the Tascam machines are technically -10... so technically the send to them is the wrong level, but I almost always use a Neve line amp between the RME and the Tascam when running signal in that direction) Regardless, I haven't noticed the ground noise affected by the input or output levels of the devices, though I haven't done any thorough comparison.

After this thread, I'm motivated to solve the ground issue without transformers if possible. However, I'm unsure about how I can modify my grounding scheme further.

The cold-shorted-to-ground at the unbalanced device definitely helps to a degreer, but it's not removing the noise entirely like powering down the PC does.

I'm about to do a fresh PC build for other reasons, but I haven't been able to find any information addressing ground noise at the PC (aside from USB isolators). The noise seems to be caused in the relationship between PCIe (or USB in tests with USB interfaces) and the rest of the AD/DAs. as it can be eliminated by monitoring audio with the PC off.

Do you have any suggestions on next steps?
Are there ways of addressing the ground noise at the computer end (specifically PCIe)? (perhaps modifying or improving the grounding scheme from the PSU or MOBO?)
Is there some other way I can modify my current grounding scheme, either at the unbalanced devices or at the computer?