DIY ABx Tester

[GreenKiwi]

(So I posted this in a separate forum, but didn't get much traction there. So I'm posting here too. I'm most interested in finding the right relays, since the rest of it should be fairly straight forward)
I'm thinking about creating my own ABX tester.

I'm going for something really simple, a small inexpensive microcontroller that controls the test and provides results and a set of relays that the controller will flip.

My thinking is that if the controller can control one or more sets of relay pairs, one could easily attach one headphone to two separate headphone amps, or two separate cables to one pair of headphones.

At the moment, I'm just in the planning stage. My top questions are:

What micro controller?

• arduino - www.arduino.cc/
• raspberry pi
• tessel - tessel.io/ (or other javascript microcontroller)
• something else?

(I don't care about the actual programming language, but good IDEs are important, as are something with good community support and I'm all for keeping it as simple as possible. I want this to be something that others can contribute to.)

What relays to use?

Here's the part that I am least sure about. Given that I want to test extremely high quality audio devices, this part needs to be as transparent as possible. At the same time, they can't be too expensive.

Level matching
In thinking about this, I think that I could use an onboard ADC to help level match multiple inputs. Play a test signal through both and then measure that test signal. Could just be playing some white noise. This wouldn't matter for comparing cables, but would be useful for DACs/Preamps/Amps/etc.

How to control/see results?

There seem to be two approaches here

• Web/Smart phone app
• Onboard display/buttons (this might be a bonus, but decided on web/app control)

BOM - $50-100

I'd love to hear thoughts from others.

[audiobill]

Why do you want to do this?

[Boomzilla]

Arduinos are common, inexpensive, and reliable. I'd go there, were I you.

Good, low resistance, low-noise relays are also commonly available (Parts Express?).

Shouldn't be too big a deal to do.

[wizardofoz]

this was made by a friend of mine ... maybe you could make it automated with an Arduino

sound.westhost.com/abx-tester.htm

[GreenKiwi]

audiobill - I want to be able to compare audio equipment and remove as many parts of the equation as possible. In the near future, I'm going to have a DAC that I want to compare with my DC-1. If I had this device, I could set it all up and then run a series of tests. Then look at the results and see what I think. I'd probably have a simple website too, where devices could report results. And I'd plan on making it all open source, so that people could enhance and/or make their own, for their own testing purposes.

Boomzilla - was definitely leaning toward Arduinos. Is there anything besides low resistance/low-noise to look for in the relays?


Also, I added one more feature that I think would be useful, the ability to perform level matching. Most of these boards either have add on boards that do ADC or have them built in. I figure that I could have the signal get measured by the ADC in a level check mode, and that would make it easier to get things matched accurately for testing.

[Boomzilla]

Doing level matching without introducing audible variables will be much harder than just doing ABX switching.

How the impedance interacts with the source may well cause more audible difference than the two source components under test.

To get level matching, I'd think that you'd need to build identical buffers for both inputs, then level match via whatever means, and then build identical output buffers for both outputs. This adds cost, complexity, and LOTS more variables to the project. A simple variable attenuator will be unacceptable because it would provide some impedance variation to both the source and the output.

[KeithL]

For a relay you want low resistance and reliable operation (assuming that the contacts close properly every time, and have low resistance, there is little opportunity for noise to occur). Depending on your source, and your topology, there is also the possibility of capacitance between contacts becoming an issue (if you use DPDT relays).

Assuming that we're talking about a "computer controlled ABx" test system, here's what I would want it to do.

First, I connect my devices under test and adjust the levels of each.
Then I tell the test set how many trials I want (let's say 10).
Now, for each of those ten trials, when I hit a button, the test set randomly picks a device for me, and "remembers" what it picked each time.
It would say "Trial 6; device chosen; please record observations" and remember which device it had selected (storing that in a file).
During the test I would record my observations for each Trial - by number.
Then, once the test was finished, I could ask to be presented with the list so I could compare my results to the devices chosen.

Once we've got a microprocessor controlling the relays, the additional coding required to randomly select devices, and to record the list, should be minimal.


However, in the specific case we're talking about, the analog inputs on the DC-1 are actually relatively transparent.
Therefore, if your other DAC has a level control, you can simply connect its output to the analog input of the DC-1 and match the levels.
By simply switching between the analog and digital inputs on the DC-1, you will be comparing the DAC in the DC-1 to the other one.
(Unfortunately, with any sort of DAC test, you're going to have to arrange for two identical signal sources;
if you use USB and a computer, you're going to have to either use two computers, or two instances of a player on one computer,
or there will be a noticeable lag while you stop the one player, change the output device, then restart the player. )

Feb 3, 2015 12:45:53 GMT -5 GreenKiwi said:

audiobill - I want to be able to compare audio equipment and remove as many parts of the equation as possible. In the near future, I'm going to have a DAC that I want to compare with my DC-1. If I had this device, I could set it all up and then run a series of tests. Then look at the results and see what I think. I'd probably have a simple website too, where devices could report results. And I'd plan on making it all open source, so that people could enhance and/or make their own, for their own testing purposes.

Boomzilla - was definitely leaning toward Arduinos. Is there anything besides low resistance/low-noise to look for in the relays?


Also, I added one more feature that I think would be useful, the ability to perform level matching. Most of these boards either have add on boards that do ADC or have them built in. I figure that I could have the signal get measured by the ADC in a level check mode, and that would make it easier to get things matched accurately for testing.

[GreenKiwi]

wizardofoz thanks for that pointer. Lots of good information there.

Boomzilla - I'm not thinking that the tester will actually perform the level matching. Just that it would have the ability to actually measure the signal going through so that you could perform the level matching yourself. e.g. with two DACs, assuming they have volume controls, you could adjust them accordingly. Or update their gain coming from a computer.

@keithl - thanks for all the feedback. Most of it aligns with my thinking.

First, I connect my devices under test and adjust the levels of each.

I'm thinking that using an ADC, I could help users set these

Then I tell the test set how many trials I want (let's say 10).
Now, for each of those ten trials, when I hit a button, the test set randomly picks a device for me, and "remembers" what it picked each time.
It would say "Trial 6; device chosen; please record observations" and remember which device it had selected (storing that in a file).
During the test I would record my observations for each Trial - by number.
Then, once the test was finished, I could ask to be presented with the list so I could compare my results to the devices chosen.

Once we've got a microprocessor controlling the relays, the additional coding required to randomly select devices, and to record the list, should be minimal.

This is exactly what I'm thinking it will do. Once it's running and able to switch, you'd setup your trial and for each iteration, I'm imagining that you would pick a high level data point:
1. I think that it is A or B or can't tell
2. I think that it sounds better or worse or the same
3. Additional observations. There will probably be a text field, but I could easily see having a few other controls for easily noting specific characteristics, for example, a simple sound stage slider where you could just set an arbitrary value

(Unfortunately, with any sort of DAC test, you're going to have to arrange for two identical signal sources;
if you use USB and a computer, you're going to have to either use two computers, or two instances of a player on one computer,
or there will be a noticeable lag while you stop the one player, change the output device, then restart the player. )

At least on the Mac, it's fairly easy to do with the Midi Settings. One can create a virtual device that will send the same data to both. Though one might have to be a bit careful with sampling rates. And with JRMC, one can create a linked zone that generally matches the signals quite well. And would also allow for level matching on DACs that didn't have volume adjustment.



As for using the DC-1, this is definitely one of my favorite features of the DC-1. I'm thinking about this also allowing for me to test a few different things, for example headphone cables, or interconnects, or speaker cables. I'm personally of the opinion that they don't make much difference, if they are appropriately sized, same lengths, etc But given that I inherited some big fancy MIT cables from a friend, I figured I'd have the opportunity to play. I'm planning on starting with line level switching, then doing headphone level switching, then speakers.


I'd also love to have a web site that would allow people to setup and perform these listening tests easily.

Ideally, I'd also have an iOS and/or android app... but for now, I think it can easily be done via the web.