DIY amp testing (with AES17 filter)

[Hipocrates]

Something like a U-Pad attenuator

Got it.

I think Ivan’s LPF is good enough.

I don't measure DACs and mostly of my work is around vintage amps (lately some class D amps too), so you would imagine what kind of tolerances and THD-N I'm use to deal with , what these devices and filters can achieve is from my point of view, unreal... But I get what are you saying, I'll see what can I do.

For the AUX0025, I’m more worried about its EMI sensitivity, I need to change its position every time to capture a FFT without much lower frequency noise, I’m suspecting that the inductance used is causing the issue.

I will pay specially attention to that issue, we're talking about LF EMC (2 ~150 kHz) right?

Thermal noise on MELF resistors doesn't seem to be too much of an issue, but I don't know how hot those inductors will get. That said I will adjust the layout.

 

[sarieri]

Thank you and @IVX for sharing. Great work! It would be great if you could share the BOM for the actual layout because of the best part fit and reproducing the test results.

Here is the BOM for Ivan's LPF

 

[sarieri]

I will pay specially attention to that issue, we're talking about LF EMC (2 ~150 kHz) right?

Thermal noise on MELF resistors doesn't seem to be too much of an issue, but I don't know how hot those inductors will get. That said I will adjust the layout.

That's right, LF EMC. I did some simulation here: https://www.circuitlab.com/circuit/sw84ff24hzkr/test-rack-class-d/

 

[Hipocrates]

That's right, LF EMC. I did some simulation here: https://www.circuitlab.com/circuit/sw84ff24hzkr/test-rack-class-d/

Do you want it THT or SMD?

 

[sarieri]

Do you want it THT or SMD?

SMD should be fine. I have a DIY heating station that can do overflow solder. But honestly, I would just have LCEDA to do the job for us. I had them made 10 of Ivan's LPF a month ago and they cost around $5 each pieces. And that's including everything including the parts and PCB.

 

[Hipocrates]

SMD should be fine. I have a DIY heating station that can do overflow solder. But honestly, I would just have LCEDA to do the job for us. I had them made 10 of Ivan's LPF a month ago and they cost around $5 each pieces. And that's including everything including the parts and PCB.

Cool, I like it... I'll start tonight with your simulation. Let me know anything that comes to your mind regarding terminals and hardware.

What is LCEDA? do you mean LJPCB with LCSC components?

 

[sarieri]

Cool, I like it... I'll start tonight with your simulation. Let me know anything that comes to your mind regarding terminals and hardware.

What is LCEDA? do you mean LJPCB with LCSC components?

It’s a major pcb/smt factory in china. You can search for JLCEDA.

About the terminals, I think OP’s choice of terminals are pretty good. A set of XLRs for input/output and a set of female banana socket. The only drawback I can think of is that those female banana socket should be connected to the input of LPF and we need and additional terminal for gnd input to connect amp’s gnd to LPF.

 

[Hipocrates]

t’s a major pcb/smt factory in china. You can search for JLCEDA.

Got it, is JLCPCB (fabricator), LCSC (components), and EasyEDA (software), all part of the same group.
I'll make the fabrication files and BOM with them

PCB Prototype & PCB Fabrication Manufacturer - JLCPCB

Instant online PCB quote, get PCBs for only >. Check design with the online gerber viewer, Easy and quick PCB Price Calculator from JLCPCB

jlcpcb.com

 

[mcdn]

For the AUX0025, I’m more worried about its EMI sensitivity, I need to change its position every time to capture a FFT without much lower frequency noise, I’m suspecting that the inductance used is causing the issue.

I'm not sure why you're getting EMI issues. Is yours in a metal enclosure with the XLR connectors connected to both the case and the PCB ground? Here's what I get on a 128k FFT (Blackman-harris 7, no overlap) with 4 averages, using the Cosmos ADC 10V input, 32bit sample depth, and the Cosmos being powered by a battery. Dead silence, even with an active mains cable draped over the filter:

 

[mcdn]

Oh and the case does _not_ need to be fancy! Here's mine, still with the switches from before I removed the pad down resistors and diodes from the design.

I agree with @sarieri that the design would be improved by putting the pad down resistors and banana sockets before the filter circuit. Question for people, would you want the banana sockets to be in the circuit before or after the pad-downs? I'm thinking after as it's more flexible.

 

[sarieri]

I'm not sure why you're getting EMI issues. Is yours in a metal enclosure with the XLR connectors connected to both the case and the PCB ground? Here's what I get on a 128k FFT (Blackman-harris 7, no overlap) with 4 averages, using the Cosmos ADC 10V input, 32bit sample depth, and the Cosmos being powered by a battery. Dead silence, even with an active mains cable draped over the filter:

View attachment 350138

I’m testing everything bareboard. By no input, do you mean no voltage output from DAC or just ADC connect to LPF and no connection physically from DAC to LPF?

 

[mcdn]

I’m testing everything bareboard. By no input, do you mean no voltage output from DAC or just ADC connect to LPF and no connection physically from DAC to LPF?

The second option . The input to the AES17 filter is unconnected, i.e. has no cable attached. The output of the AES17 filter is connected to the Cosmos ADC.

The reason you are having issues is the lack of a case. When you do put the board in a case, it's best if the connectors are the ones specified in the BOM, with the spike that connects the connector pin 1 to the chassis when they are screwed in. Also make sure the case is all metal, and all parts of the case are electrically connected to each other. It's common for people to use e.g. aluminium cases and find say the front panel doesn't actually have a good connection to the top or bottom of the case.

 

[sarieri]

The second option . The input to the AES17 filter is unconnected, i.e. has no cable attached. The output of the AES17 filter is connected to the Cosmos ADC.

The reason you are having issues is the lack of a case. When you do put the board in a case, it's best if the connectors are the ones specified in the BOM, with the spike that connects the connector pin 1 to the chassis when they are screwed in. Also make sure the case is all metal, and all parts of the case are electrically connected to each other. It's common for people to use e.g. aluminium cases and find say the front panel doesn't actually have a good connection to the top or bottom of the case.

Right. But I guess I won’t bother to proceed on this anymore. Ivan’s LPF does the job for me without any case.

 

[mcdn]

Right. But I guess I won’t bother to proceed on this anymore. Ivan’s LPF does the job for me without any case.

Ivan's board seems great for DAC testing, but remember that it's a completely different purpose than an AES17 filter for class D amp testing. My board design is hugely overbuild for low power testing - it's good for 140volts into a short circuit for brief periods! Anyway I'm glad you have what you need!

There are probably members here who would love to take the boards you don't want off you for shipping and cost of your time

 

[sarieri]

Ivan's board seems great for DAC testing, but remember that it's a completely different purpose than an AES17 filter for class D amp testing. My board design is hugely overbuild for low power testing - it's good for 140volts into a short circuit for brief periods! Anyway I'm glad you have what you need!

There are probably members here who would love to take the boards you don't want off you for shipping and cost of your time

I tested ncx500 and 3E Audio's TPA3255 using Ivan's LPF without any problem. The DIY AUX0025 has much steeper attenuation at higher frequency so it's probably better for a larger range of testing. Modern Class D seems to have rather mild out of band noise though. I do think it's really important to design a voltage divider with this LPF though. And the H3 distortion there is not so great. Although it's probably not gonna matter for most class D amps, it will matter if measure smth like Nilai500 which has extrermely low H3.

It's better if we could try different inductance and capacitance on a bread board before put it in production.
I buy bunch of stuff from China monthly, PCB and parts from JLCEDA are stellar quality and always authentic, CNC drilled ALU case will be less than $30 if built with SMT parts. But if we build this with metal case for best performance, that's gonna be a pain. I searched around for metal case before, not much nice option there.

 

[mcdn]

I tested ncx500 and 3E Audio's TPA3255 using Ivan's LPF without any problem. The DIY AUX0025 has much steeper attenuation at higher frequency so it's probably better for a larger range of testing. Modern Class D seems to have rather mild out of band noise though. I do think it's really important to design a voltage divider with this LPF though. And the H3 distortion there is not so great. Although it's probably not gonna matter for most class D amps, it will matter if measure smth like Nilai500 which has extrermely low H3.

It's better if we could try different inductance and capacitance on a bread board before put it in production.
I buy bunch of stuff from China monthly, PCB and parts from JLCEDA are stellar quality and always authentic, CNC drilled ALU case will be less than $30 if built with SMT parts. But if we build this with metal case for best performance, that's gonna be a pain. I searched around for metal case before, not much nice option there.

I'm not saying anything bad about Ivan's 2nd order filter, just that it isn't a drop-in replacement for an AUX0025. And you are 100% right, with modern assembly services the case drilling and assembly work begins to dominate not just the cost but the time and effort. It would be bloody amazing if JLC offered cheap drilled cases!

Also maybe I'm wrong but I don't think you've shown any H3 distortion that isn't just the source DAC distortion? If we can find a protocol to replicate the issue you are seeing then we can improve the design. Voltage divider at the front end is super important I agree, I'm updating the design today, hope to publish something in the next week or so, including moving the test points for the banana sockets to the front end of the filter.

 

[sarieri]

I'm not saying anything bad about Ivan's 2nd order filter, just that it isn't a drop-in replacement for an AUX0025. And you are 100% right, with modern assembly services the case drilling and assembly work begins to dominate not just the cost but the time and effort. It would be bloody amazing if JLC offered cheap drilled cases!

Also maybe I'm wrong but I don't think you've shown any H3 distortion that isn't just the source DAC distortion? If we can find a protocol to replicate the issue you are seeing then we can improve the design. Voltage divider at the front end is super important I agree, I'm updating the design today, hope to publish something in the next week or so, including moving the test points for the banana sockets to the front end of the filter.

It’s clear that with APU and scalar, the H3 is about 20dB higher. Without APU and scalar, the H3 are about the same with Ivan’s LPF. I don’t know if the low input impedance of ADC is kicking in here. I will test again today with Scalar only at 0dB gain level just to bridge impedance better.

 

[mcdn]

It’s clear that with APU and scalar, the H3 is about 20dB higher. Without APU and scalar, the H3 are about the same with Ivan’s LPF. I don’t know if the low input impedance of ADC is kicking in here. I will test again today with Scalar only at 0dB gain level just to bridge impedance better.

Going back to your measurements in Post #52, it's interesting that the second set (without the APU and scaler) are all basically the same at 20.5 bits of resolution, and HD3 of -133.3 (direct), -137.4 (Ivan's LPF), and -136.1 (DIY AUX0025). This corresponds well to the known performance of the D6S and Cosmos ADC.

The first set of measurements though, with the APU and scaler in the chain, all seem to exceed the native performance of the devices at the ends of the chain, at 23.2 (no filter), 23.2 (Ivan's filter) and 21.8 (DIY AUX0025) bits. Also the noise floors look quite different between the first set and second set. I'm having trouble understanding how this is happening if the FFT settings are the same. The APU on its own can't be removing noise, it's just a very damn good notch filter! It would be good if you could share the detail of the FFT settings used in both sets of measurements.

Also you make a good point about the ADC input impedance. I recommend setting it to the highest level of 10k if the filter is the last item in the chain. Maybe try putting the filter between the DAC and APU to take advantage of the high input impedance of the APU?

 

[sarieri]

Going back to your measurements in Post #52, it's interesting that the second set (without the APU and scaler) are all basically the same at 20.5 bits of resolution, and HD3 of -133.3 (direct), -137.4 (Ivan's LPF), and -136.1 (DIY AUX0025). This corresponds well to the known performance of the D6S and Cosmos ADC.

The first set of measurements though, with the APU and scaler in the chain, all seem to exceed the native performance of the devices at the ends of the chain, at 23.2 (no filter), 23.2 (Ivan's filter) and 21.8 (DIY AUX0025) bits. Also the noise floors look quite different between the first set and second set. I'm having trouble understanding how this is happening if the FFT settings are the same. The APU on its own can't be removing noise, it's just a very damn good notch filter! It would be good if you could share the detail of the FFT settings used in both sets of measurements.

Also you make a good point about the ADC input impedance. I recommend setting it to the highest level of 10k if the filter is the last item in the chain. Maybe try putting the filter between the DAC and APU to take advantage of the high input impedance of the APU?

-133dB H3 at 5Volt is not complying with Amirm’s measurement. He measured -140dB H3 at 4V output. H3 at 5V should be even lower. The difference of measurements between AP and E1DA may have something to do with AP’s autorange. The kink in Amirm’s SINAD and measured level graph is AP adjusting it’s input sensitivity. I found that once it reaches around 120 THD+N, APU+ADC always gets better measurements than Amirm’s AP.

 

[sarieri]

Maybe try putting the filter between the DAC and APU to take advantage of the high input impedance of the APU?

APU will behave crazy if there is HF noise feeding into its input. That’s pretty much why Ivan came up with that LPF design in the first place.