Reference ABX hardware design and implementation

[mcdn]

Nice.
I see two diodes for each mosfet pair. Is it zenerers or TVR's? Have you done any calculations on the spike energy from the speakers?

Hey @Armand - One is a 15V gate-protection Zener, and the other is in the gate-drive circuit to open a lower resistance path to speed up discharge.

The idea is to switch while muted (even if the muting is only brief), but yes I have also done calculations, as we can't assume this is always the case! To summarise, the MOSFETs have a ~1.5J pulse avalanche rating, so whether switching under full load is safe or not depends on the equivalent inductance of the speaker+crossover, current at switching, etc. In the case of switching under load, 10A into 5mH of crossover inductor, if dumped back into the MOSFETs, could be 0.25J, and we have to account for the amp drive as well, so maybe 0.5J, which is a lot but you'd have to be really going some into big 3-way speakers.

TVSs were considered, but they have their own tradeoffs. Since switching is all under the control of the overall ABX device, which must be managing the line level outputs to the amps for level matching, I've gone without.

 

[mcdn]

Hi everyone. The boards have landed in Australia and are in customs so will be with me when I get back from a short holiday next week. I am fairly certain I'll need to do a respin of the line level board but this revision will be fine for functional testing. Hopefully the high level, digital and power/control boards will work out of the box.

In the meantime here's a walkthrough of the main user interface as seen on a phone. It's certainly not refined, and not all functionality is there, but there's enough that I'm seeking early feedback. There's a PDF document and the same as individual images of each page.

Note that there is also an IR remote, so ABX testing and many other things can be done without needing a phone/laptop/tablet, but for now I'm just showing the phone screens as I expect that will be the primary way of interacting.

Tagging @KSTR @hardisj @pma @Armand @amirm @Cascode

 

[KSTR]

Note that there is also an IR remote, so ABX testing and many other things can be done without needing a phone/laptop/tablet

I like this. I find it more practical to operate the A, B, and X switching with tactile buttons with eyes closed for less distraction, and with a remote this is feasible. For example, using the standard red/green/yellow/blue row of buttons.

 

[mcdn]

I like this. I find it more practical to operate the A, B, and X switching with tactile buttons with eyes closed for less distraction, and with a remote this is feasible. For example, using the standard red/green/yellow/blue row of buttons.

It will probably be a "topping style" remote, like this:

So I'm definitely interested in your preferences for which buttons should do what actions during an ABX session. It can all be adjusted in settings later of course, but it's best to start with good defaults.

 

[pma]

In the meantime here's a walkthrough of the main user interface as seen on a phone. It's certainly not refined, and not all functionality is there, but there's enough that I'm seeking early feedback. There's a PDF document and the same as individual images of each page.

I gave it several minutes and made notes. I like it, but I feel it might be complicated process to setup and perform the test. It might need some learning curve, compared to foobar abx. Regarding source selection and settings, it is a bit similar to iConnectAudio4+ that I own and its discontinued configuration SW. Unfortunately real parameters of this interface were not good enough.

 

[KSTR]

This is going to be professional test gear, and like any other professional test gear, say, an AP or Klippel test rig, it requires a learning phase and training for the operator, starting from a thorough understanding of the concepts, details and pitfalls of hookup, etc etc.
While Matt is aiming to incorporate as much automation and user guidance as possible some serious engagement and commitment from the user will always be required.

 

[KSTR]

It will probably be a "topping style" remote, like this:

View attachment 562945

So I'm definitely interested in your preferences for which buttons should do what actions during an ABX session. It can all be adjusted in settings later of course, but it's best to start with good defaults.

"Menu" --> Play A, "Back" --> Play B, "Home" --> Play X would look like an intuitive assignment to me once one has entered some specific mode (and exit with "Ok" from there)

 

[KSTR]

And then even "Vol-" --> X is A, and "Vol+" --> X is B, to cast a vote directly and proceed with the test.

As for procedure, it is open to debate whether the current voting result shall be optionally displayed directly after casting a vote, to give the user some feedback. Obviously during a training phase, but also when doing a real run. I personally prefer the instant direct feedback so I can abort the whole test quickly when I'm getting too many unexpected wrongs after a successful training phase, usually meaning too much "test stress" still present.
For me, this situation feels very similar to doing a recording session in a studio as a musician where I quickly want to decide when it is better to stop now and continue later without being forced to do all of the scheduled five takes.

 

[mcdn]

"Menu" --> Play A, "Back" --> Play B, "Home" --> Play X would look like an intuitive assignment to me once one has entered some specific mode (and exit with "Ok" from there)

You still to select what you think X is

 

[mcdn]

You still to select what you think X is

You beat me to it

 

[pma]

This is going to be professional test gear, and like any other professional test gear, say, an AP or Klippel test rig

Oh I see, I have not realized you are aiming so high.

 

[mcdn]

Oh I see, I have not realized you are aiming so high.

I love @KSTR's ambition and enthusiasm, but I think it's a bit early to be comparing it to a Klippel or AP! He is designing the balanced module for it (first version will be SE only), and I am sure that will have outstanding performance. It will be a drop-in replacement for the SE line board, with a different rear panel to go with it.

But either version will need some onboarding. It's not like a DAC that you just plug and play.

One of the features that's not in the UI yet is core profiles like "compare 2 DACs", which should narrow down the UI complexity for popular use cases. The user still has to follow a protocol to get accurate level matching and ensure the DUTs don't have obvious differences before starting a trial. Or if they choose not to then that is recorded in the trial report as well.

 

[pma]

iConnect config software uses a routing scheme method. You just connect rows and columns to get the I/O configuration. I find it quick and user friendly.

 

[hardisj]

Hi everyone. The boards have landed in Australia and are in customs so will be with me when I get back from a short holiday next week. I am fairly certain I'll need to do a respin of the line level board but this revision will be fine for functional testing. Hopefully the high level, digital and power/control boards will work out of the box.

In the meantime here's a walkthrough of the main user interface as seen on a phone. It's certainly not refined, and not all functionality is there, but there's enough that I'm seeking early feedback. There's a PDF document and the same as individual images of each page.

Note that there is also an IR remote, so ABX testing and many other things can be done without needing a phone/laptop/tablet, but for now I'm just showing the phone screens as I expect that will be the primary way of interacting.

Tagging @KSTR @hardisj @pma @Armand @amirm @Cascode

Hey, Matt. Excellent progress! Here are my notes.

Between Rounds:
a) Slide 2: Yes, the difference between End and Discard are clear.


The Results:
a) Slide 3: I don't see the need for any additional information unless you want to explain the p-score.
b) Slide 3: Might as well throw a % correct in there because people loooove percentages.

WiFi:
a) Slides 4-5: Looks good to me.


System Calibration (Slides 7-9):
a) In documentation, explain why this is needed. I understand, but others might need a short explanation for it to better help understand the difference between system and DUT calibration.
b) In documentation can you give a diagram/photos of this y-connector setup?
c) If I had the unit in hand today then I would try to find a 400Hz tone somewhere in Apple Music (my streaming source). If not found there then I'd search YouTube. Or I would simply create my own tone in Audacity and upload it to my YouTube and then stream off of that to whatever DUT I am using. Assuming this is a proper method for calibration.


ABX Trial:
a) Slide 11: I suggest placing the A/B next to each other and the X below; an upside down triangle. This is a more intuitive thumb motion for quick switching between the options when eyes are closed.


A & B Ports:
a) Slide 13: Personally, I think it would be more intuitive to have a dedicated section for the inputs and outputs rather than mixing them in the different analog/digital form you've created. I'd arrange it so the first screen is for the user to set up the inputs. Then a second screen for configuring the outputs.
b) Slide 14: USB Channels customization seems fine to me.


Transition Settings:
a) Slide 15: Maybe a dialog box explaining what the ramifications of changing the default are. Or a baseline recommendation. (is there any research or evidence of a faster/slower switching time?)


Levels and Distortion:
a) Slide 16: Okay, so this is not the same as the initial system calibration... I think I understand a bit more now. Is the System Calibration just a baseline reference to make sure everything is functioning properly?
b) As I understand it, the ABX device handles its own calibration and this step is essentially a sanity check so that the user knows levels are set up correctly before they begin the test?


Listening Modes:
a) Slide 18: This is very intuitive.
b) Slide 19: my comment about ABX Trial above applies here as well. Having the A/B options next to each other rather than above/below is more intuitive. It's easier to move your thumb left<-->right rather than up/down without pressing the wrong button *when eyes are closed*. This is because it's easier to detect the edge of the phone from side-to-side than it is going up/down. I speak from literal decades of experience using laptop-based DSP software and making eyes-closed changes.
c) Slide 20: This is very intuitive.

 

[hardisj]

iConnect config software uses a routing scheme method. You just connect rows and columns to get the I/O configuration. I find it quick and user friendly.

View attachment 562951 View attachment 562952

Having designed a few setups in Sigma as well as Rane's DRAGNET, and worked with these kind of diagrams in previous jobs I have to say that I absolutely loathe these things. While they may be intuitive to those who have experience with the system diagram type I/O panels, the average user will not find it intuitive. This is also based on having used DSPs with "mapping" type GUIs and watching users fumble through them.

I *much* prefer the checkbox setup Matt currently has in place. My only suggestion is to put the I/O in dedicated sections.

Edit: Not to mention there's not enough screen space for such a mapping diagram.

 

[mcdn]

Having designed a few setups in Sigma as well as Rane's DRAGNET, and worked with these kind of diagrams in previous jobs I have to say that I absolutely loathe these things. While they may be intuitive to those who have experience with the system diagram type I/O panels, the average user will not find it intuitive. This is also based on having used DSPs with "mapping" type GUIs and watching users fumble through them.

I *much* prefer the checkbox setup Matt currently has in place. My only suggestion is to put the I/O in dedicated sections.

Edit: Not to mention there's not enough screen space for such a mapping diagram.

It's an unfortunate fact of life that once you get to a certain number of I/O combinations the matrix is the only way to visualise it. I hate them too.

Luckily we are constrained enough we can offer a more specific interface. I'll post an updated walkthrough soon.

 

[mcdn]

ABX Trial:
a) Slide 11: I suggest placing the A/B next to each other and the X below; an upside down triangle. This is a more intuitive thumb motion for quick switching between the options when eyes are closed.

Would you be keen on a "long press" for selecting A or B as X in the phone or IR remote interface?

b) As I understand it, the ABX device handles its own calibration and this step is essentially a sanity check so that the user knows levels are set up correctly before they begin the test?

Yes, it's included here for completeness but all of the system calibration will be done "at the factory" before shipping. It can be done again by the user, but it's only required for absolute value calibration, it doesn't affect relative levels.

c) If I had the unit in hand today then I would try to find a 400Hz tone somewhere in Apple Music (my streaming source). If not found there then I'd search YouTube. Or I would simply create my own tone in Audacity and upload it to my YouTube and then stream off of that to whatever DUT I am using. Assuming this is a proper method for calibration.

Reference files will be provided and linked. I can potentially upload them to streaming services. This could be a useful community resource - a GitHub repository of known-good signals, which are also uploaded under well known names to streaming services.

Transition Settings:
a) Slide 15: Maybe a dialog box explaining what the ramifications of changing the default are. Or a baseline recommendation. (is there any research or evidence of a faster/slower switching time?)

I couldn't find any science on this. Some people prefer an instant cutover, some a fade in/out, some a fade with noise in the middle. In principle we're trying to allow users to find a difference between inputs without the ABX device giving any "tells".

The main reason we default to a short fade out/in is to eliminate the effect of DC offset differences between DUTs, which can otherwise generate a recognisable "tick". In my testing it's surprisingly hard to even notice a 100ms fade out/in, while a DC offset tick is immediately obvious



Thanks again for the thoughtful feedback!

 

[mcdn]

Boards are here! No time to test them until next week. Grain of rice at the top for scale - the main board is ~105x110mm

 

[mcdn]

What do you fine folks think of these wiring diagrams? Each one is a scenario the user selects, and then there's a diagram and a walk-through of the setup process. New scenarios are easy to add, this was just the four I thought were most obviously useful. @hardisj particularly keen for your input.