DIY amp testing (with AES17 filter)

[sarieri]

I think, you don't need the coilcraft inductor at all but a coreless inductor, which has the best THD.

Yah. Those are way bigger though. I'm still confused about rising H3 when measured with APU-Scalar-ADC comparing to Scalar-ADC.
Also, the AUX0025 is clearly having higher rolloff at 20kHz. Not sure which is the cause.

 

[IVX]

ScalEr isn't scalAr, because it's from scale but not about scalar/vector.

 

[mcdn]

The new v2 board has the inductors way more spaced out, amongst a few other layout and component improvements. It'll be interesting to see how it works out and if it solves the excess HD3 and the higher 20Kz rolloff. Bear in mind we are talking about stuff below -120dB, so it's not really a concern for amp measurement, but better is better!

Repository is at https://github.com/mattclarkdotnet/aes17_filter_25/tree/v2

 

[Solveit]

Nice, I ordered some boards.
You dont have any ready made BOM for Mouser?

 

[mcdn]

Nice, I ordered some boards.
You dont have any ready made BOM for Mouser?

There’s a list with links in the readme, but no uploadable BOM.

 

[mcdn]

Holly molly!! What kind of amplifiers you guys test? that signal would be an 5000W @ 4Ω amp

My 15A breaker will trip at 1800W and my most powerfull amp is 300W

For my application I'm looking more for crossover distortion and maybe THD+N @ 1w

It is a bit silly, but those are the specs of the AP device, so here we are! In countries with 240V mains it’s normal to have a 15A circuit, so 3.6kW.

 

[mcdn]

I thought for Single End AMPs like Hypex offerings, you are basically using only half of the LPF. So actually only 1200W.

Hypex uses balanced power supplies (e.g. +/- 48V referenced to neutral AC input) which naturally gives a 0 DC offset at the output - you should connect it across the whole LPF. Even TPA3255 which uses a single ended power supply has ~20mV DC offset at the speaker terminals at idle. This is achieved through clever modulation schemes. Simplistically put: at idle both sides of the output switch high and low at the same time, giving zero offset overall, plus the switching noise. Then as the signal level rises the outputs are offset more and more, until at full level one is on while the other is off. In either case (Balanced or SE supply) you should connect your amp across both branches of the LPF, and reserve the GND connection for solving any noise/hum issues.

 

[sarieri]

Just tested with new inductance. Unfortunately, still -0.7dB at 20kHz. And -127.4dB H3. No actual improvement.

 

[mcdn]

Just tested with new inductance. Unfortunately, still -0.7dB at 20kHz. And -127.4dB H3. No actual improvement.

I'm getting -133dB H3 with D10 Balanced -> v2 filter -> ADCiso. The new board layout does seem to help, but I need more time to do a full set of measurements before I'm confident. Also got an APU ordered so we can compare then.

 

[mcdn]

Good news @sarieri! v2 seems to solve the HD3 issue, although the rolloff at 20kHz is still higher then expected at 0.5dB despite the tighter component tolerances. Here are my measurements of the new filter version. The chain is D10 balanced -> DIY AUX-0025 v2 -> Cosmos Scaler @ 0dB -> Cosmos ADCiso, first without and then with the filter:

 

[sarieri]

Good news @sarieri! v2 seems to solve the HD3 issue, although the rolloff at 20kHz is still higher then expected at 0.5dB despite the tighter component tolerances. Here are my measurements of the new filter version. The chain is D10 balanced -> DIY AUX-0025 v2 -> Cosmos Scaler @ 0dB -> Cosmos ADCiso, first without and then with the filter:View attachment 354451View attachment 354452View attachment 354453

Looks good

 

[mcdn]

I figured out the rolloff issue. It's because of the output impedance of the D10 balanced (88 ohms). Which makes sense as the filter is designed primarily for amp testing. The rolloff is actually extremely sensitive to the exact combined impedance, but since any non-broken amp will have an output impedance <1ohm it's not usually a big deal. For DAC testing I suppose one could place a trimpot in parallel with the each input resistor. One with a 5k range would work nicely.

 

[sarieri]

I figured out the rolloff issue. It's because of the output impedance of the D10 balanced (88 ohms). Which makes sense as the filter is designed primarily for amp testing. The rolloff is actually extremely sensitive to the exact combined impedance, but since any non-broken amp will have an output impedance <1ohm it's not usually a big deal. For DAC testing I suppose one could place a trimpot in parallel with the each input resistor. One with a 5k range would work nicely.

You are right. I did another measurement: D6s--Scalar L in--Scalar L out--AUX0025--Scalar R in--Scalar R out--ADCiso. And here is the FR I got after calibrate D6s's FR in REW.
+- 0.05dB 20Hz-20kHz. I guess we can forget about the FR issue.

.

 

[mcdn]

You are right. I did another measurement: D6s--Scalar L in--Scalar L out--AUX0025--Scalar R in--Scalar R out--ADCiso. And here is the FR I got after calibrate D6s's FR in REW.
+- 0.05dB 20Hz-20kHz. I guess we can forget about the FR issue.

I will update the documentation to be clear about the potential issue so users aren't surprised when they measure DACs and get a bit of HF rolloff they weren't expecting.

Thank you so much for your feedback and collaboration, it's made the end result much better than the original!

 

[sarieri]

I will update the documentation to be clear about the potential issue so users aren't surprised when they measure DACs and get a bit of HF rolloff they weren't expecting.

Thank you so much for your feedback and collaboration, it's made the end result much better than the original!

As long as they use something like scalar, I think impedance is just fine to test DAC. More worried about H3 but here we are talking about -140dB level. Also, the reason we need a LPF when testing DAC is because Cosmos APU’s preamp is very sensitive to HF noises many DACs have. We don’t need LPF anywhere else.

For amp use, I don’t know what is everybody using to attenuate the signal. As far as I see it, as long as you use a passive voltage divider, I don’t see anyway to get away with the FR noise issue if attenuate before LPF and noise issue if attenuate after LPF.

 

[mcdn]

For amp use, I don’t know what is everybody using to attenuate the signal. As far as I see it, as long as you use a passive voltage divider, I don’t see anyway to get away with the FR noise issue if attenuate before LPF and noise issue if attenuate after LPF.

The self-noise of a 10k divider is -115dBV (waves hands about how much of that leaks to the ADC, but presumably not all of it). Assuming use of ADCiso ADC with 10V input range, that will be -135dBr for a 100V amp output. The state of the art for amps is 137dB of dynamic range I think (Hypex NCx500: https://www.audiosciencereview.com/.../hypex-ncx500-class-d-amplifier-review.41007/). So I think it's OK. Certainly better than the AP516, but not as good at the AP555 with its active auto ranging input buffer. What we need is a Cosmos Scaler with -20dB of gain...