Reference ABX hardware design and implementation

[KSTR]

Do I see feature creep setting in? ;-)

The following is food for thought, not intended as a literal proposal.

I'd strongly advise a very modular approach when trying to pack more and more features and functionality into this. This reduces risks quite a bit, the penalty would be a bit higher production costs.

Back in my instrumentation days in the 80ies and 90ies we would have certainly used a standardized 19in 3HE module carrier with a DIN 41612 connector backplane and basic power supplies. And then add each basic and somewhat independent block of functionality on Eurocards (100x160mm), single-width or stacked modules for stuff that requires more space. This way of doing things is still standard practice for specialized lab equipment that is manufactured in small quantities.

Usually we opted for a complete nice-looking carrier with handle and feet, like the one above. Today you can have a lot of variants so you are not limited to full 19" width or 3HE height (for example, see https://www.bopla.de/fileadmin/product_data/00_the_red_book/internorm-stil_en.pdf)

That would allow for a future-proof whole ecosystem, assuming the whole project would go open-source and documented so that a community could add new stuff.

 

[KSTR]

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.

I can say that we're currently designing a complex audio system (musical instrument, actually), based on Raspberry Pi CM5. Thus we have a good onboard class-II mains --> 5V SMPS supply (with functional earthing in this case). The whole audio I/O is isolated (not medical level tough), using a 5V-->5V 10W isolated module, and from there +-7.5V and +-20V are generated with a CUK/SEPIC/MULT converter design of mine, feeding linear regulators. The 3.3V local digital and codec supplies are derived from 5V with another set of regulators.

There are many ways to do this, of course. The Murata DC/DC stuff usually is excellent. Sometimes with a bit low switching frequency (45kHz in the case of MEJ2).

 

[KSTR]

but I can't think of a desktop class USB DAC that doesn't also have a TOSlink input

Some have SPDIF (coax or opt.) outputs, not inputs. For example Topping D10 series.

 

[PristineSound]

Do I see feature creep setting in? ;-)

The following is food for thought, not intended as a literal proposal.

I'd strongly advise a very modular approach when trying to pack more and more features and functionality into this. This reduces risks quite a bit, the penalty would be a bit higher production costs.

Back in my instrumentation days in the 80ies and 90ies we would have certainly used a standardized 19in 3HE module carrier with a DIN 41612 connector backplane and basic power supplies. And then add each basic and somewhat independent block of functionality on Eurocards (100x160mm), single-width or stacked modules for stuff that requires more space. This way of doing things is still standard practice for specialized lab equipment that is manufactured in small quantities.
View attachment 552033
Usually we opted for a complete nice-looking carrier with handle and feet, like the one above. Today you can have a lot of variants so you are not limited to full 19" width or 3HE height (for example, see https://www.bopla.de/fileadmin/product_data/00_the_red_book/internorm-stil_en.pdf)

That would allow for a future-proof whole ecosystem, assuming the whole project would go open-source and documented so that a community could add new stuff.

Superb idea and I want it like this REALLY BAD

But the reality is, for a low demand device like this, these future proof extension cards are highly unlikely to (a) be developed (b) be sold (c) supported for long.

We be lucky if there is even 100 units of this thing being made within the year.

 

[KSTR]

We be lucky if there is even 100 units of this thing being made within the year.

100 units?.... I wouldn't dare dreaming of such amounts. More likely 10. But one can never know, all we might need is thorough promotion from some prolific influencer to get the stone rolling.

 

[PristineSound]

100 units?.... I wouldn't dare dreaming of such amounts. More likely 10. But one can never know, all we might need is thorough promotion from some prolific influencer to get the stone rolling.

I was hedging myself by saying 100 units, just in case miracles happens

 

[mcdn]

Do I see feature creep setting in? ;-)

Very good question! Adding the full suite of supply isolation (5V DC/DC converter, USB isolator, I2C and I2S isolators) is not cheap at all, but is probably justifiable for a device that needs to be "blameless" in use. About $20 in parts and $10 in additional assembly cost for prototype volumes.

I'm definitely drawing a line under the core hardware at this point. Digital and line I/O, speaker level I/O, ADC, DSP processing, AWG and control processing are all nailed down. Two enclosures, one for speaker level and one for everything else.

But thanks to the software defined nature of the device we can add functionality that doesn't need new hardware very easily, so I think that sort of feature creep is fine.

 

[mcdn]

100 units?.... I wouldn't dare dreaming of such amounts. More likely 10. But one can never know, all we might need is thorough promotion from some prolific influencer to get the stone rolling.

10 is as high as I'll let my ambition get for now! But the more I talk with everyone about it, the more use cases emerge, so you never know. There's an unfilled niche, the question will be how many people care about that niche.

 

[mcdn]

Specific call for help here on the power architecture. Power supply is the standout BOM cost right now. The main isolated +-15V medical grade supply is $50, then $10 for the 5V/5V isolated converter. Good power is of course the foundation of other performance, but I can't help feel there should be a slightly cheaper way to get there. I'm happy to fix the external DC supply at 12V to simplify things.

 

[KSTR]

If 2x 2W (130mA) output power is enough, you could use two MEJ2S0515SC which would save some cost. And 5V USB-C input hardwired for 5V/1.5A PD.
But, no matter what we'd choose, there is always the risk of stuff being out of stock with long lead times the day you want to start a production run. One of the reasons I'd go for as much modularity as possible, to allow for switching horses during the race.

 

[mcdn]

While I sweat my way through the details of PCB routing, here's some sample UI screens. Each image shows both the web interface and the front panel interface. Notionally these are analog DUTs like DACs. The "levels matched" screen would also show phase, delay and distortion issues before the trial starts.

There will be open listening modes for when its just being used as a smart level matched switcher rather than conducting an explicit ABX trial.

 

[mcdn]

Fantasy vs reality:

 

[PristineSound]

Out of curiosity, how much harder is it to add balances XLR.

 

[mcdn]

Out of curiosity, how much harder is it to add balances XLR.

It's not hard from a design POV, but it's double the components if you go balanced end to end. Otherwise it's a balanced to SE conversion on input, and reversing that on the output, which is how a lot of balanced devices operate anyway. Both those conversions add noise and distortion, and require very close part tolerances to be done transparently.

 

[mcdn]

Out of curiosity, how much harder is it to add balances XLR.

Oh and I should say, you can totally connect a balanced output to this, it just needs the right cable to convert XLR to RCA. There’s a drop in SNR and you lose the common mode noise rejection inherent to balanced connections, but RCA is perfectly capable of -140dB performance.

 

[PristineSound]

It's not hard from a design POV, but it's double the components if you go balanced end to end. Otherwise it's a balanced to SE conversion on input, and reversing that on the output, which is how a lot of balanced devices operate anyway. Both those conversions add noise and distortion, and require very close part tolerances to be done transparently.

Something to keep in mind. Some devices have different performance between balanced and single ended. And some device does not use single ended at all, like the benchmark amp.

This unit will be a low volume unit. This is a specialty, precision tool. I suspect the people wanting to buy this aren't the ones on a race to the bottom of the barrel and willing to spend money to make sure it performs and haves the functionality.

I personally and willing to drop the cash on a device with end to end balanced.

But then again, I am one single person making this comment.

 

[mcdn]

Something to keep in mind. Some devices have different performance between balanced and single ended. And some device does not use single ended at all, like the benchmark amp.

This unit will be a low volume unit. This is a specialty, precision tool. I suspect the people wanting to buy this aren't the ones on a race to the bottom of the barrel and willing to spend money to make sure it performs and haves the functionality.

I personally and willing to drop the cash on a device with end to end balanced.

But then again, I am one single person making this comment

Let’s get this version worked out first. I’m absolutely open to a balanced version later. A note on amps - the speaker level switching module works with any kind of amp output. But yes, it would need an RCA to XLR cable to drive the input of a benchmark balanced only amp. It will drive it at 4Vrms though, so we’re only giving up some common mode noise performance.

 

[somebodyelse]

Oh and I should say, you can totally connect a balanced output to this, it just needs the right cable to convert XLR to RCA

There isn't a universal way to connect a balanced output to an unbalanced input with just a cable, as has been discussed at length in several threads. If you don't know the details of the DUT's output stage then the only safe and reliable way is a balanced to unbalanced converter. That could be a transformer (distortion penalty, especially for more affordable transformers) or an active converter before the unbalanced input, which would perform worse than having the same converter as the input.

 

[somebodyelse]

But yes, it would need an RCA to XLR cable to drive the input of a benchmark balanced only amp. It will drive it at 4Vrms though, so we’re only giving up some common mode noise performance.

I'm slightly torn here. On one hand a single ended output is only a few passives and a connector away from an impedance balanced output. On the other an impedance balanced output won't work fully with a subset of amps with a 'balanced' input like the bridged Schiit Vidar or the Topping B100 as only one side of their bridge will be driven. Would it be viable to design for all three, leaving it up to the builder whether or not to populate for either of the balanced options depending on need?

 

[KSTR]

I'm with @mcdn on the balanced I/O thing. Universal I/O makes the design quite a bit more complicated, and with a modular approach an analog module that covers this can still be added later. For now, we just need to make sure we have spent enough thought on mechanical, form-factor and mounting details for such a modular concept so that a balanced module can be retro-fitted in existing devices.