Review and Measurements of Orchard Audio Gala DAC

[rajapruk]

As I understood it, there were some issue in the noise-floor of the jitter measurement? Does this fix fix that?

What is the introduction-price for bare pcb without box? I am possibly interested in 4 of them.

 

[orchardaudio]

What is the introduction-price for bare pcb without box? I am possibly interested in 4 of them.

Bare boards will be $275

 

[andreasmaaan]

Hello,

Dynamic Range and SNR is not the same thing as many of you point out, also Amir does not use A-Weighing which would further improve the result. So my 130dB SNR (0 dBFS, A-Weighted, 10Hz - 22kHz bandwidth) spec is accurate, and I stand by it.

Another thing to keep in mind is that the APx555 that Amir uses has it's own residual noise of 1.1uV. This noise is present in all of the unit's measurements. This becomes more and more critical as a DACs noise gets lower and lower.

For example if you have a DAC with a noise of 1.1uV the unit will actually measure a noise of sqrt(1.1^2 + 1.1^2) = 1.56uV that is 1.41 times higher than the actual noise of the DAC.

But if you have a DAC with a noise of 10uV the unit will actually measure a noise of sqrt(10^2 + 1.1^2) = 10.06uV, almost no difference.

As for the SINAD measurement, I went back and double checked what the manufacturer did on the unit that I shipped to AMIR, and they used an old bill of material, which means the power supply for the negative rail of the op amps is not sufficient and will cause them to slightly clip or saturate as AMIR put it. That is why the SINAD measurement is lower, when I measured a different unit with the correct power supply I got -110dB for THD+N. Here is the plot. As you can see THD+N at 0 dBFS is essentially 0.0003% which would equate to -110dB.
View attachment 16438

The final unit will have the LED's significantly dimmed as well. Actually I was planning on removing 2 of the 3 and dimming the remaining one, just so that there is some indication that the unit is running. (I would like some more feedback on this as well)

In summary the specs that I have for the unit will be correct for the final product.

Another thing is that if sign up for the mailing list here, you will be able to purchase one at $349:
https://www.indiegogo.com/projects/gala-ultra-high-performance-stereo-dac/coming_soon/x/19247162

Would you be able to post an SMPTE IMD % vs FFS/dBfs graph for comparison to Amir's, please?

 

[orchardaudio]

Would you be able to post an SMPTE IMD % vs FFS/dBfs graph for comparison to Amir's, please?

Unfortunately I cannot, I currently do not have access to and audio analyzer.

What I can do is fix the power supply inside the unit that Amir has and send it back to him.

 

[orchardaudio]

For those who are interested here is some more info:

GALA - What's Inside?

Full GALA Specs

• SNR: 130dB (A-Weighted, 10Hz - 22kHz bandwidth)
• THD+N: <-108dB (0dbFS)
• Output Level: 5 Vrms (XLR)
• Output Impedance: 100 ohm
• Frequency Response: DC - 22kHz
• Input: Gold Plated Coaxial S/Pdif
• Sample Rates: 44.1, 48, 88.2, 96, 176.4 and 192kHz
• Resolution: 16 and 24 bits
• Size: 5.5'' x 4.25'' x 2.375''
• Volume Control: 0 to -80dB, and mute (knob turned fully counter-clockwise)
• Manufactured in USA

 

[pos]

Will these modifications solve the high noise floor that we see in Amir's jitter test, as well as in the lower part of the IMD vs level test?

 

[orchardaudio]

Will these modifications solve the high noise floor that we see in Amir's jitter test, as well as in the lower part of the IMD vs level test?

The IMD issue will definitely be solved, as that is a clear result of saturation.

For the jitter, I am a bit shocked, considering that the dynamic range already measured much better than on ApplePi DAC. So how can the noise floor be worse? I would like to think this is also attributed to the op-amps clipping / saturation.

GALA is better than ApplePi DAC with respect to Jitter and SNR. This is because it has lower noise linear supplies, and the clocking scheme is much better than on the Raspberry Pi.

 

[pos]

The IMD issue will definitely be solved, as that is a clear result of saturation.

I am referring to the lower level part, on the left-end side of the graph.

 

[orchardaudio]

I am referring to the lower level part, on the left side of the graph.

After a closer review of the data, I have a very strong suspicion that the manufacturer also put some wrong value filtering capacitors on the board. This is because the noise seems way too high.

That should not be there, at low levels the IMD should be better than -60dB not anywhere near -40db where it is now.

I really should have measured the device myself before sending to Amir, but renting an audio analyzer is way too expensive for a small manufacturer like myself.

The noise floor of GALA is better that the ApplePi DAC (I have verified this on multiple occations), so the unit Amir has must be defective.

I can guarantee that every production unit will be fully tested and meet the specs.

 

[gvl]

Amir, if you have some spare cycles, can you please do a quick test of a single leg of the XLR output? Curious if there's common mode noise is present on the outputs.

 

[gvl]

After a closer review of the data, I have a very strong suspicion that the manufacturer also put some wrong value filtering capacitors on the board. This is because the noise seems way too high.

That should not be there, at low levels the IMD should be better than -60dB not anywhere near -40db where it is now.

I really should have measured the device myself before sending to Amir, but renting an audio analyzer is way too expensive for a small manufacturer like myself.

I can guarantee that every production unit will be fully tested and meet the specs.

This is somewhat disconcerting, how can one be sure they get a unit that performs up to the spec if the manufacturer can change the parts without notice? At $500 a pop I'd expect more attention to the quality control.

 

[orchardaudio]

This is somewhat disconcerting, how can one be sure they get a unit that performs up to the spec if the manufacturer can change the parts without notice?

I dont think they did it intenianally, but mistakingly. These boards were populated by hand not by pick and place, so human error is very possible.

Also all production units will be fully tested before shipping.

 

[rajapruk]

Maybe you can provide a test protocoll for each device when it is sold, for assurance?
Just put a paper with it within each shipping package.

 

[orchardaudio]

There is still one thing that bugs me, the dynamic range measured much better than ApplePI DAC, yet noise floor is much worse, those two are contradictory.

 

[orchardaudio]

Maybe you can provide a test protocoll for each device when it is sold, for assurance?
Just put a paper with it within each shipping package.

I like this idea, something for me to brood on.

 

[amirm]

Another thing to keep in mind is that the APx555 that Amir uses has it's own residual noise of 1.1uV.

My actual performance using the same settings as for DAC tests in dashboard, has less noise:

Or about 0.88 uV.

If I go up to 45 kHz bandwidth, it will then get up to 1.2.

 

[orchardaudio]

My actual performance using the same settings as for DAC tests in dashboard, has less noise:

View attachment 16464

Or about 0.88 uV.

If I go up to 45 kHz bandwidth, it will then get up to 1.2.

Ok, but even the 0.88uV will still make a difference at lower noise levels.

What is the noise level measured with GALA connected and receiving 0FS signal.

Also as per my other post can you shed some light on why the dynamic range measures much better than ApplePi, yet the noise floor is much worse. To me those two things are contradictory.

 

[amirm]

What is the noise level measured with GALA connected and receiving 0FS signal.

Here it is with S/PDIF off (relay disconnected in the analyzer):

Here it is with S/PDIF on but generator off:

Results are the same for a -150 dB test tone at 1 kHz.

 

[orchardaudio]

Here it is with S/PDIF off (relay disconnected in the analyzer):

View attachment 16465

Here it is with S/PDIF on but generator off:

View attachment 16466

Results are the same for a -150 dB test tone at 1 kHz.

I assume this is not A-Weighted

 

[amirm]

I assume this is not A-Weighted

It is not.