USB plus Audio Power on distributed DC System - Input wanted!

[amplexus]

So here's the sitch.

We have a system installed in a studio where there are a number of movable 'podiums' which contain a headphone amp, and hold a tablet which wirelessly talks to an X32 for artist cue mixing. Audio and power are conveyed to the podium via a shielded CAT5e line. A 20A 15Vdc SMPS resides in a machine room, and DC is distributed around the room to various patch panels.

I am adding boards to each podium that will take our 15Vdc distributed power, and regulate down a 12V rail for the headphone amp with additional filtering, and a 5Vdc line for USB charging to keep the tablets charged during long sessions.

There are two main 'hurdles' I am attempting to overcome- First, that the 12VDC rail for the headphone amp is dead quiet and free of switching noise AND any induced noise that the tablet charging might throw out which is seeming to be an issue with our current 'dumb' system, and that switching noise in the main distribution line is quiet enough not to contaminate the audio lines while they travel together for 15ft in the CAT5 cable. Second, that we can reliably have the mainstream range of devices that may be used always go into fast-charge mode either 5W or 10W

I will add a hefty amount of CLC filtering at the SMPS to clean up the output for a start, and I'm using an SMPS that can function correctly at no-load. What I've got is the following circuit to go into the boxes to deal with the headphone amp and Vbus rails as well as the charge port controller chip and USB connections. What I am hoping is if someone can tell me if i've done enough/too much/wrong things to decouple the charging rail from the headphone rail, and what other considerations I might need to keep noise to minimum or zero.

Or if I'm just way off base and need to rethink the whole board

 

[cyrano]

I wouldn't worry too much about noise from the SMPS. I guess the cabling might very well pick up way more noise. Add to that the noise coming from the USB devices and you're perhaps barking up the wrong tree?

 

[amplexus]

I wouldn't worry too much about noise from the SMPS. I guess the cabling might very well pick up way more noise. Add to that the noise coming from the USB devices and you're perhaps barking up the wrong tree?

The whole space is wired on AES72 analog audio over CAT5/6 with something silly like 96 channels of lines out to the silly amount of panels for a relatively small space- there are no noise issues and there are no noise issues on the headphone signal lines until power is introduced. We've probably got four or five thousand feet of shielded twisted pair in the walls. The "format" and charging setup are requirements of the design hence the need to engineer a solution.

 

[cyrano]

I suppose without power, the tablets and HP amps don't function? The tablets have a battery, but aren't connected?

Sounds a bit like you have a gigantic ground problem.

Stupid question: "What's nextdoor?"

 

[amplexus]

I suppose without power, the tablets and HP amps don't function? The tablets have a battery, but aren't connected?

The HP amps need power yes, ideally 12VDC alongside the 5VDC Vbus power for USB. The tablets have battery obviously, but during sessions they are being used in 'kiosk' mode where the screen is up all the time and running the X32 monitor mix software so they chew through battery fairly quickly, and in a long session it is desired that they wont' have to stop for 2 hours to remove the tablets to a charging station as the whole cue system relies on using them for control/routing of the X32. Thus they need to be powered while in use.

Sounds a bit like you have a gigantic ground problem.

Stupid question: "What's nextdoor?"

There's no ground issue. This is a foundation-up build and both the power and data/signal runs were done by a competent electrician. We have zero noise issues in any other part of the system. If the headphone amps are fed their 12VDC rail from the current workaround, everything is fine and quiet, its only when the tablet power is connected and drawing current that we get noise- so it's either being radiated into the audio path, or it's demodulating its way in on the power rails. My current proposed solution is intended to simplify the distribution system while (hopefully) isolating the power rails from each other- but USB/digital power systems are not in my wheelhouse- hence the post.

As to what's next door.... lol absolutely nothing, and I do mean that. This studio is built on a private island with boat only access.

 

[cyrano]

I said it was a stupid question...

Must be pretty noisy tablets. I'm out of suggestions, sorry.

 

[Khron]

I'm more... concerned about the linear regulators for both sides. Maybe i could justify that for the headphone amps, but i see zero reason to use a linear regulator for the usb power outputs. Those end up feeding a boatload of other switching regulators inside the tablets, which really don't give that much of a toss about how "clean" that 5V is or isn't.

Add to that, those might want to draw at least 1A each, maybe up to 2A, so you're looking at burning off up to 20W(!!!) between R1 and U1; double that, if you're powering two "hungry" usb devices...

 

[amplexus]

Well one way or the other I have to knock 15VDC down to 5V for the USB devices. Dc-dc converter i suppose but i’m trying to limit noise inside these boxes

 

[Khron]

Shielding high frequencies is loads easier than 50/60Hz - worst case chuck them into some boxes made out of bare PCB material, or why not, some diecast metal guitar pedal enclosures. My bigger concern would be the grounding arrangement, but since you shouldn't need to connect "external" devices that may or may not be earthed, that's probably a non-issue.

 

[radardoug]

You can buy switchers with isolated outputs. Use them for your USB, then no ground currents.

 

[Voyager10]

So the tablets have analog audio out via a headphone jack which goes to the headphone amp?

If so there will be two paths between GND on the board and ground on the headphone amp (one direct, and one via the headphone jack) and this might cause trouble.

Options might be to transformer-isolate the tablet output jack from the HP amp input, or build a differential inout for the HP amp. Alternatively, use an isolated stepdown converter for the 5V.

 

[Voyager10]

Here's a quick sketch showing what *might* be going on...



I've just shown the ground paths, for clarity.

The digital gubbins in the tablet can cause a noise voltage to appear between USB ground and HP Jack ground. If there is a loop in the ground paths, a current will flow round the loop (i.e. noise voltage / loop resistance). This noise current will flow through the HP amp (between audio-in ground and power ground) and - depending on how it's built - the noise can get coupled in to the audio.

You can test this theory by making a cable which just connects the ground between the tablet HP jack and the HP amp audio input. If this makes audible noise appear, it's some form of ground-loop problem.

 

[thor.zmt]

We have a system installed in a studio where there are a number of movable 'podiums' which contain a headphone amp, and hold a tablet which wirelessly talks to an X32 for artist cue mixing. Audio and power are conveyed to the podium via a shielded CAT5e line. A 20A 15Vdc SMPS resides in a machine room, and DC is distributed around the room to various patch panels.

I am adding boards to each podium that will take our 15Vdc distributed power, and regulate down a 12V rail for the headphone amp with additional filtering, and a 5Vdc line for USB charging to keep the tablets charged during long sessions.

Sounds to me like a lot of potential for ground/earth loops.

Step one, use transformer input (balanced) on the headphone amplifier. You can probably find a reasonably inexpensive 600:600 transformer that will do the job. The HP Amp does get the signal via a CAT5 pair, right? Transformers could be on the outputs of whatever produces the cue mix with an isolated floating output, which eliminates ground loops.

Step two, use LC filtering on the power feed to the HP Amp, regulator to suit, could be a switcher - they can be very quiet and use very high switching frequencies. The Regulator should eliminate the noise from the 15V/20A SMPS (likely ~ 67kHz or variable), or have the LC filter deal with it.

Step three, use a modern step-down switcher (>1MHz) for the 5V charging power, You can even use a variable voltage circuit with USB-C PD if the tablets have USB-C, this can allow faster charging.

Thor

 

[amplexus]

Appreciate the response but this isn't actually what the system looks like. The tablet is strictly for control. No audio comes from the tablet and the audio system distribution is fed isolated direct from the X32 that handles monitor duties straight to each cue output. The cue boxes don't have any connected audio ground except to the chassis, and the HP jacks are isolated.

At any rate- I did ultimately solve the issue a few weeks ago. We were realistically just getting too much switching noise on the DC lines and needed to both clean up the distributed DC and also integrate a DCP controller for USB charging. I should have gone that way in the first place but I was still languishing under an old idea of how USB charging operated. I tried to keep things simple and ultimately made them more complicated. The solution I ended up with is actually pretty slick and we're going to create an even more refined version for future builds!

Here's a quick sketch showing what *might* be going on...



I've just shown the ground paths, for clarity.

The digital gubbins in the tablet can cause a noise voltage to appear between USB ground and HP Jack ground. If there is a loop in the ground paths, a current will flow round the loop (i.e. noise voltage / loop resistance). This noise current will flow through the HP amp (between audio-in ground and power ground) and - depending on how it's built - the noise can get coupled in to the audio.

You can test this theory by making a cable which just connects the ground between the tablet HP jack and the HP amp audio input. If this makes audible noise appear, it's some form of ground-loop problem.

 

[amplexus]

Will preface this by saying I actually built and tested a solution a few weeks ago that worked perfectly, and it was installed in the facility a couple of weeks ago and is working flawlessly as designed. I'm bad at coming back to update these things...

Sounds to me like a lot of potential for ground/earth loops.

There are no real issues with ground loops. There has been a misunderstanding of how the system is laid out I think, which may be that my description is not clear enough. The tablets do not handle audio at all, they are strictly for control. The headphone amp input is fed from a dedicated line directly from the X32- it does not end up sharing grounds anywhere, and the sheild only terminates to the chassis of the cue-box enclosure.

Step one, use transformer input (balanced) on the headphone amplifier. You can probably find a reasonably inexpensive 600:600 transformer that will do the job. The HP Amp does get the signal via a CAT5 pair, right? Transformers could be on the outputs of whatever produces the cue mix with an isolated floating output, which eliminates ground loops.

I had contemplated this, but it turned out to be unnecessary. The issue was not ground/signal isolation related.

Step two, use LC filtering on the power feed to the HP Amp, regulator to suit, could be a switcher - they can be very quiet and use very high switching frequencies. The Regulator should eliminate the noise from the 15V/20A SMPS (likely ~ 67kHz or variable), or have the LC filter deal with it.

This was a big part of my redesign's step one. I built a completely new power supply, distributing at a single rail higher voltage (16VDC), followed by a hefty CLCLC filterbank at output of the SMPS, and then......

Step three, use a modern step-down switcher (>1MHz) for the 5V charging power, You can even use a variable voltage circuit with USB-C PD if the tablets have USB-C, this can allow faster charging.

.....Inside each cue box now is a local PSU set that takes the heavily filtered 16VDC and splits off into two decoupled rails: one has a common mode choke and another set of LC filtering to isolate a 3-terminal DC converter to give me the 5V Vbus power for usb charging, and feeds a DCP controller (protected by TVSS diodes), and a common mode choke, all feeding the USB port, and the second feeds a basic RC filter into a bog standard 7812 regulator arrangement to feed the HP amp. The -ve side of each rail's output is brought back to a 'star' point at the DC input common and i made provision to be able to isolate them if needed (it wasn't however)

The big key was switching the whole topology to use a higher distributed single rail DC supply, and make it as clean as possible, then derive the needed rails inside each box with lots of isolation to keep any switching noise currents out of the audio rail. I also probably added wayyyy more filtering and common mode chokes than were strictly necessary- but the system is dead quiet even when all 8 tablets are pulling full fast charging current!

 

[thor.zmt]

Will preface this by saying I actually built and tested a solution a few weeks ago that worked perfectly, and it was installed in the facility a couple of weeks ago and is working flawlessly as designed. I'm bad at coming back to update these things...

Excellent. You illustrate the key to trouble shooting - identify and solve the root cause.

Thor