Ridiculous amounts of EMI at single core load with USB DACs

(Recording of EMI available upon request, it sounds like Tinnitus. I confirmed it wasn’t me by plugging my ears after shutting the system down.)

So, I unexpectedly got audible levels of EMI coming through on my audio signal on my Zoom U-22 DAC when used in 96khz/32bit. 48Khz/32bit doesn’t have the same issue, (48khz also has the issue when signle core loaded) but I expected to use my DAC in all frequencies without ridiculous EMI. I know 96Khz and higher uses more CPU, but I didn’t expect audible levels of EMI to result. Same thing happens when a single core load is constant. I even used a ground loop breaker but that amplified the EMI coming out of the USB port.

Could this be partly responsible for the USB 2.0 dropped packets on Ryzen 5000? That the EMI produced by the CPU was interfering with the entire USB system?

Using a Ryzen 5800X on a Gigabyte X570 I (ITX). F31K BIOS.

Edit: Also affects the Intel 10920X. Maybe newer platforms like X570 and X299 increase parasitic resonance frequencies.

1 Like

Your DAC may not have sufficient isolation.

Have you considered powering it off the back panel to see if there is a change?

A higher end USB power brick is good solution too.

If you can power it off your phone. That is a good low EMI power source. This will help others determine what your issue is. Right now you are flying blind :wink:

Dont use your USB bus power. Power it off a solid 5V in

ITS DEFINITELY not this. Thats a hum. You are having EMI or jitter. :slight_smile: best of luck hope something works

Yeah, I just confirmed it is most present when single core boost happens on Ryzen 5000 regardless of output sample rate, 48Khz and 96Khz both have the same issue when the CPU is single core loaded, so it’s definitely CPU related.

May have to get this then:

I use their iDefender as a separate 5V source to break ground loops and hum.

I’ve been using the back USB ports exclusively, but I tried all ports and got equal amounts of EMI.

Most PC power supplies run a major 12V rail of which they generate the 5V and 3.3V rail via DC-DC stepdown. That way you could totally feed VRM switching noise out the minor rails.

ifi products are very disappointing from a hardware to price standpoint. A powered USB hub will give you the same result.

For example, take the Jitterbug:
audioquest_jitterbug

That is less than 5€/$ in parts…

I have a powered Orico USB hub and it adds 60hz hum. So it’s barely filtering the 12V input. I already spent the money on the iDefender, so now I need a EMI ANC.

My cheapo USB-hub disconnects everything except USB-Data when you plug in 5V to it.
Gets rid of all EMI from the PC.

Throw some ferrite chokes on the usb cable maybe. Might not do anything but it should help with HF switching noise.

1 Like

Threw some ferrites on a USB extension before the iDefender and it helped with bass due to filtering out the super high frequencies, but not the tinnitus frequencies.

Those frequencies only actively engage and are audible when the CPU is under load. Tried with my 10920X and it’s the same thing. It’s not limited to Ryzen 5000, but it could be limited to the newer platforms.

My benchmark track is “Ginger Island” from the Stardew Valley Soundtrack. I run my U-22 at 96Khz to upsample the audio. The track has a lot of good bass.

The other hail mary is potentially the cable is draped over something that is inducing this noise and simply rerouting it might help reduce that.

Otherwise I think it’s time to find a new dac and/or get a more isolated powered usb hub.

It’s not the cable. It only engages the noise when connected to a PC source. Only the iDefender DC power is super quiet.

You may also be experiencing some form of aliasing with the noise itself at that sample rate which would make it seem worse. Moving to 192khz might be worth a shot.

This DAC’s limit is 96Khz 32bit. I don’t believe it’s that because the noise is also present during CPU load when no audio is playing.

Noise aliasing would occur regardless of if something is playing or not. It’s an issue with sample rate.

It’s weird that it’s louder when the CPU is under load though. And remember, I had the same noise at 48Khz with enough CPU load.

It’s likely switching frequency noise which is induced from current flow in general. It’s a hard problem to solve these days with ever increasing frequencies. Some boards include isolated USB ports for this very reason.

Does the dac have toslink? That would optically isolate it.

Speaking in terms of modern boards, only my X299 board’s ALC1220 has Toslink, but my 2001 JVC AV receiver is slow to decode from null audio. Have this problem constantly in Stardew Valley where some places just have no sound, and then the PCM decoder is late decoding when there is sound.

It seriously looks like I need an iSilencer+.

Honestly, I would start to look at the U-22 being the culprit here.

Can you run the U-22’s output to some speakears or line-in of a computer to plot the spectrum?
Audacity has a built in function, Friture does it in real-time.

What to look for:

Note: These are fabricated with a buzzer on an arduino and then recorded with my mic.

I had the same experience on audio out through my Rode NT-USB mini with the board Furry’s using, and still with the B550 version, so doubt it’s his U-22. It’s much more noticeable using the 3900x over the 3600.

Yeah, I need to do noise cancellation then use a different power source to remove ground loops. Good to know it’s very likely not the U-22.

Did you solve this in the end?