Reference ABX hardware design and implementation

[PristineSound]

it is possible to make a recording using main system loudspeakers and taking signal from their binding posts. Then the recordings can be level matched and abx tested.

If the device is not ABXing the actual physical DUTs or is not getting the actual live signal from sources then that is not a product I would buy, in fact, a hard no for me. I suspect many potential buyers would feel this way.

 

[pma]

You, of course as anyone else, have a full right of your personal opinion. I am posting based on my tests, measurements and controlled listening tests performed.

 

[PristineSound]

You, of course as anyone else, have a full right of your personal opinion. I am posting based on my tests, measurements and controlled listening tests performed.

You will open the door for people to invalidate and write off the test results.

And does it comply with proper ABX protocol? I don't have an answer to that question.

But anyway, if other's would be interested in such product, I'm just one person.

 

[mcdn]

@pma your point is well made that recordings through an ADC are in many ways a better starting point for ABXing amps for example. Or DACs, or anything else for that matter. But this isn't the thread for that

 

[voodooless]

Maybe add an SD card holder to the pico board, just in case

 

[PristineSound]

Also don't forget ABXing of recordings from actual DUT already exists today. It's the Cosmos ADC and foobar ABX. Just throwing that out there.

 

[mcdn]

Maybe add an SD card holder to the pico board, just in case

Ha! Don’t forget the perfect headphone amp, ADC and DAC, and the balanced inputs and outputs, and the lab-grade reference signal generator!

 

[mcdn]

Also don't forget ABXing of recordings from actual DUT already exists today. It's the Cosmos ADC and foobar ABX. Just throwing that out there.

Yep, and there are other threads for that. Hardware solutions have their own qualities, and remember this hardware does a lot more than DAC switching.

 

[voodooless]

Ha! Don’t forget the perfect headphone amp, ADC and DAC, and the balanced inputs and outputs, and the lab-grade reference signal generator!

The last one is actually fairly easy with the Pico Costs nothings, hardware wise.

 

[Armand]

About the clock difference problem (out of sync). Let's measure it!
We still want to isolate the two DUT, but this can be achieved by doing it like this:

Since the only interesting thing here is timing there is no need for expensive parts.
* Two isolated power supplies. RFM-0505S (USD 4)
* Two PCM1808 (USD 1.5)
* One digital isolation chip. ISO7741 (USD 1.5)

There is a issue with the 75pF leakage in the isolated power supplies. With an extreme situation of 4Vrms common mode noise and a very high ground resistance of 1 Ohm this system can introduce audible noise at -112.6dB.
A more realistic scenario: ground resistance of 100mOhm and a CM signal of 1Vrms@15kHz gives us -141dB. At lower frequencies the problem becomes even less of an issue.

Another fun thing we could do is to measure the ground current. With both DUT connected we could switch in a AMC3301 (USD 3) and actually measure the ground current with accuracy up to about 400kHz.

I know, I know. Not what you asked for @mcdn but with these additions we could also measure things that are often discussed as problems with ABX setups.

 

[ppierre]

About the clock difference problem (out of sync).

Can't the problem be avoided by "resetting" the stream at each switch (or simulated switch) ?

 

[mcdn]

The last one is actually fairly easy with the Pico Costs nothings, hardware wise.

Ah well yes I have included an ASG but I don’t think most metrology labs would be OK with it! It will be better than 1% accurate for voltage, low noise, and bandwidth limited to about 8kHz. Similar approach to how DSPi does its “sub out”

 

[mcdn]

Yeah, I've gone back and forth on this a little bit. Would it be better to use the _very expensive medical grade_ +-15V rails to derive the 5V and 3V3 digital supplies, or put them behind a separate isolated converter, thus keeping their ground return separate? Either seems to have pros and cons.

Facepalm. I forgot about the tie between audio ground and digital ground. Now thinking of using a Murata MEJ2S0505SC, unless anyone has a stronger suggestion?

 

[mcdn]

And a USB isolator IC. The supply picture for the non-audio side (after the 5V buck) is proposed as:

The MEJ2 has only a 4pF isolation capacitance. Ripple is a bit crap but we only use it directly for things that don't care much.

 

[voodooless]

Ah well yes I have included an ASG but I don’t think most metrology labs would be OK with it! It will be better than 1% accurate for voltage, low noise, and bandwidth limited to about 8kHz. Similar approach to how DSPi does its “sub out”

You can have your DUT DAC give the signal, just use the Pico to generate to digital data, output via SPDIF.

 

[mcdn]

You can have your DUT DAC give the signal, just use the Pico to generate to digital data, output via SPDIF.

Well yes, but that's not built-in. I did consider putting a basic DAC in there, but along with the headphone output it got put to one side as something people can add very easily themselves with one of the spare outputs or a Y-cable. There will be an option in the firmware to tell it the 0dBFS voltage of a DUT, which together with the factory calibration of the ADC will give very accurate voltage readings. The exact voltage of the built-in AWG will be written to each device during factory calibration.

 

[mcdn]

About the clock difference problem (out of sync). Let's measure it!
We still want to isolate the two DUT, but this can be achieved by doing it like this:
Since the only interesting thing here is timing there is no need for expensive parts.
...

Thanks @Armand!

Clock drift between USB DACs is a thing, but I can't think of a desktop class USB DAC that doesn't also have a TOSlink input, so the preference would be to send the single source to the optical input, let us duplicate it, then send the duplicated optical outputs to the DACs under test. That does assume the user isn't interested in the possible performance of the USB section of the DAC, but we can't solve every problem in one box!

We can measure the drift over time using the ADC we already have though. A 20ppm difference in DAC clocks at 48kHz accumulates to ~70ms over an hour, which could credibly be audible if hard switching. Adding 250ms of fade/switch time (the likely default being 75ms fade out, 100ms muted switching time, 75ms fade in) _might_ mask it.

Either way it's very measurable because the ADC can tap the signals from both DUTs during the muted time and we can issue a warning if the sync mismatch goes over a set limit. The user should then restart playback on their USB host to resync the clocks. Being able to check parameters like this during trials is a big benefit of the device, as is the ability to set preferred protocols for different scenarios, avoiding users accidentally doing tests that turn to to contain unintentional cues.

 

[mcdn]

Another fun thing we could do is to measure the ground current. With both DUT connected we could switch in a AMC3301 (USD 3) and actually measure the ground current with accuracy up to about 400kHz.

I definitely hadn't thought of that. But since we never have both DUTs connected at the same time in ordinary use as they are relay-isolated, I'm not clear on the value.

 

[mcdn]

Maybe add an SD card holder to the pico board, just in case

I know I gave a cheeky reply before, but we can save a few seconds of audio from the PCM1808 onboard for later download - calibration evidence for example. More importantly we will be providing a full audit log with timestamps, so an externally made recording can be synced exactly after a listening session is done, with the ASG providing a sync chirp at the beginning of the session.

 

[patoulol]

Adding the timestamped audit log + ASG sync chirp is actually huge for scientific validity

Being able to cross-reference the physical ABX switching log with a continuuous capture from a Cosmos ADC after the fact bridges the gap between hardware switching and recorded file testing nicely