Testing for DAC noise

[stew1234]

I am experimenting with various transformer sizes and layouts for a DAC. Other than simply listening to various tracks, does anyone recommend a good method for testing for noise (whether resulting from the transformers, i2s lines or otherwise)?

 

[duncan2]

Read-
2016-FP-Understanding-Transformer-Noise.pdf

 

[Markw4]

I am experimenting with various transformer sizes and layouts for a DAC. Other than simply listening to various tracks, does anyone recommend a good method for testing for noise (whether resulting from the transformers, i2s lines or otherwise)?

If you are talking about electrical line noise passed through power transformers, R-core transformers have very good isolation between primary and secondary windings. IME line noise incursion problems are far less an issue using R-cores and also common mode line filtering before the transformer primary.

Other than that type of noise, there can also be noise from transformer rectifier diode commutation which can produce high frequency ringing that then can then find its way into various circuity by conduction and radiated coupling. Snubbers and or soft recovery diodes can help a lot to prevent that problem before it causes symptoms.

The last type of transformer related noise (I can think of at the moment) is line frequency and its harmonics appearing on power rails, usually from insufficient power supply output filtering and or regulation.

Some of the noises can be seen on an FFT, other types can be harder to pin down. Using good practices such as R-cores, line filters, etc. can help. Sometimes you only know if you have fixed a problem when you hear an audible improvement.

It can be handy to get a used Monster HTPS 7000 Mk.II power conditioner (they work exceptionally well) to use for A/B testing and for some types of bench work. Also, a line filter in a box can be switched in and out to see if any audible or measured differences are detectable. If so, you may have a noise issue to track down and fix.

Regarding noise from HF/RF sources (say, clocks, I2S, etc.), sometimes connecting a scope ground clip to the scope probe tip can make a little 1-turn pickup coil that can be waved around circuitry to probe for RF emissions. Sometimes an RF spectrum analyzer can be helpful, other times a shortwave radio with an external antenna wire used as a probe can be used to find and trace RF noise that isn't easy to find by other means, etc. Don't know of any universal tool that always works for finding HF/RF noise problems.

Of course, a lot of noise problems can be prevented by good design in the first place. If building dacs it can help a lot to have a better dac on hand as a listening reference, an ultra-low distortion headphone amp, and some low distortion headphones. Unfortunately, those things can be expensive just some like other high quality pieces of test equipment can be expensive. However, listening tests are an important part of audio design IMHO, its hard to know if an audio design is really good or needs more work if you can't measure it and or if you can't critically listen to it (or have someone else you trust do that stuff for you).

 

[jackinnj]

Regarding noise from HF/RF sources (say, clocks, I2S, etc.), sometimes connecting a scope ground clip to the scope probe tip can make a little 1-turn pickup coil that can be waved around circuitry to probe for RF emissions. Sometimes an RF spectrum analyzer can be helpful, other times a shortwave radio with an external antenna wire used as a probe can be used to find and trace RF noise that isn't easy to find by other means, etc.

Jim Williams demonstrated that in a discussion of probing techniques!

IDon't know of any universal tool that always works for finding HF/RF noise problems.

Remember those crystal radios of your youth ? (well, my youth at least) -- germanium diode has junction capacitance of under 1pF and can be used as a detector into VHF, you can also wire two diodes as a doubler. Just remember that fluorescent lighting will modulate the junction at ~30kHz.

ARRL has devoted gallons of printing ink on the topic: Sounds of RFI

One of the best shielding products -- that bimetallic food can. I have used tuna tins for ultra-low noise amplifiers, altoid mints for RF attenuators. Not the most elegant solution, but it really does work!

 

[stew1234]

Thank you all for the responses. The issue that prompted my first post may be caused by any number of things but I have been working on the layout of a DAC and am experimenting with using 50VA or 100VA transformers for the i/v stage and potentially individual transformers for the clock and DAC. 100VA may be overkill and having dedicated transformers may radiate EMI/hum into the nearby signals negating their advantage.

When listening to headphones directly out of the DAC, things are generally silent, even with the DAC turned up using its digital volume control. When connected to my speaker amp via XLR, I hear an audible noise from the tweeters with my ear close to the tweeters. This could be the amp, not the DAC, or it could be some characteristic of my DAC layout.

The headphones are somewhat difficult to drive, so any noise from the DAC may not come through versus the sensitive tweeters.

 

[phase]

Sometimes line filtering is more effective with a slight load added as well, experimenting will tell.

 

[Pano]

If you disconnect the DAC from the amp, is the tweeter silent?
If you turn down the volume on the amp, is the tweeter silent?
What are you using to drive the headphones? What's between the DAC and the headphones?

 

[stew1234]

There may be some baseline noise in the AC lines. I am located in New York City. As soon as the DAC is disconnected there is audibly less noise, but it is still there. I've included two recordings of the speakers with the DAC plugged in and with the signal lines unplugged. The mic is about 6 inches from the tweeter and starts with one monitor off - I then turn it on a few seconds in.

Dropbox - DAC plugged in.m4a - Simplify your life

Dropbox - DAC unplugged.m4a - Simplify your life

I have tried in two set-ups, but in both situations the DAC is run directly to the amp. I am using self-powered monitors and speakers with an external amp but no pre-amp other than the DAC. The DAC volume doesn't impact the background noise. With the monitors, everything is using the same power strip as a desktop. In the other set-up, all audio components are plugged into a power conditioner separate from other electronics.

I have been running the headphones balanced out directly from the DAC. I think the higher impedance of the HD650s prevents the noise coming through the same way it does with the monitors and other speakers, which have a very sensitive tweeter.

I'm not expecting to get to silent, but I am trying to optimize the location, size and number of transformers. As you can see from the attached picture, the Antek transformers I am using have very long primary and secondary lines. I am trying dual mono and dedicated transformers for the DAC and clock, although this may be a case of diminishing returns if the long AC are emitting and/or picking up noise.

 

[Markw4]

As you can see from the attached picture, the Antek transformers I am using have very long primary and secondary lines.

Torroid transformers, particularly Antek (maybe the lowest cost option available), are not the best for minimizing AC-line noise incursion IME. Again, I would suggest R-cores.

Also, most power conditioners are not great. The one I mentioned previously is the one to get (used from ebay, as they are no longer made).

Another thing to consider is shielding. Dacs often sound cleaner if inside a metal box.

Cronus, if that's one of the PCB stacks in the pic, seemed noisy and jittery IME as compared to some other USB-to-I2S isolation and clock distribution options. At least though, some improvement was noticed with it placed inside a metal box. Still not my first choice.

Many voltage regulators operating as a single stage do not have sufficient PSRR to deeply suppress line frequency harmonics. Don't know about the ones you have there, but an FFT should show if line harmonics are an issue at all.

 

[duncan2]

That says a lot stew1234,

I did an investigation into earth generated interference in the UK and America it didn't take long to find out NY is one of the worst (in the USA ) for noise pollution .


In every electronics branch this was a continual complaint by those operating sensitive electronic equipment , I also found due to the enormous scatter of RF radiation it was also introduced into the NY mains earth via radiated induction .


Hospitals need special earthing principles for their equipment so much so they have industrial mains supplies which are isolated from direct power company earthing supplies provided for the general public like metallic armored supplies also for large high tech. computer companies .


Various power companies have their own special methods for business/industrial /etc organizations.


This is not suitable for the general public .


And don't even mention SMPS.s for LED lighting many horror stories including traffic lights (LED ) interfering in car radios .

 

[stew1234]

I plan to box it up eventually in a metal case. I have looked into r-cores, but they are tough to source in the US. Audiophonics has them for US voltage. After checking again, it looks like they have 30VA r cores with 15v and 9v secondaries. I'd like slightly higher VA, but if running dual mono and one transformer to each the DAC and one to clock that should be plenty of power.

The Anteks have an electrostatic shield and shield band around the outside, which I thought might be better than the "naked" Avel Lindberg transformers I was using. I have read about getting custom ones made with extra primary windings and may try that from SumR at some point. Canterbury Windings has a big delay due to Covid and Piltron is probably cost-prohibitive.

Torroid transformers, particularly Antek (maybe the lowest cost option available), are not the best for minimizing AC-line noise incursion IME. Again, I would suggest R-cores.

Also, most power conditioners are not great. The one I mentioned previously is the one to get (used from ebay, as they are no longer made).

Another thing to consider is shielding. Dacs often sound cleaner if inside a metal box.

Cronus, if that's one of the PCB stacks in the pic, seemed noisy and jittery IME as compared to some other USB-to-I2S isolation and clock distribution options. At least though, some improvement was noticed with it placed inside a metal box. Still not my first choice.

Many voltage regulators operating as a single stage do not have sufficient PSRR to deeply suppress line frequency harmonics. Don't know about the ones you have there, but an FFT should show if line harmonics are an issue at all.