I did some testing to see. First off is a basic frequency response, which I like to do with Pink noise, because it's quick and painless. This is driving the amp board via USB connection, as setup automatically in W10.
There's something going on ~20K, so I did a closeup.
Ensuring that's not my soundcard as input, I checked that weird shape using my HP 33250's white noise output;
So something is going on with the amp / pink noise file at 19 kHz. It could be the file, but I doubt it... A quick spot check using the amp's analog input, driven by the function generator, did NOT show this anomaly.
Next I set the HP to sine, measured ~2.83 with my DMM and did a spectrum of that instrument's output. Just as a what's what reference.
Then, using its USB interface, whilst playing a 1kHz low distortion flac file and setting this little amplifier to appx the same level, I did two spectrums; one with the JBL HLS 610 speaker as load, one with an 8 Ohm, non-ind resistor as load.
Resistor load;
I used an old Gateway (Liteon) 19V laptop brick (6.3A) for DC Power. I believe both channels were driven.
There's something going on ~20K, so I did a closeup.
Ensuring that's not my soundcard as input, I checked that weird shape using my HP 33250's white noise output;
So something is going on with the amp / pink noise file at 19 kHz. It could be the file, but I doubt it... A quick spot check using the amp's analog input, driven by the function generator, did NOT show this anomaly.
Next I set the HP to sine, measured ~2.83 with my DMM and did a spectrum of that instrument's output. Just as a what's what reference.
Then, using its USB interface, whilst playing a 1kHz low distortion flac file and setting this little amplifier to appx the same level, I did two spectrums; one with the JBL HLS 610 speaker as load, one with an 8 Ohm, non-ind resistor as load.
Resistor load;
I used an old Gateway (Liteon) 19V laptop brick (6.3A) for DC Power. I believe both channels were driven.
Looks like the XY-AP100L is available on Amazon for three coffees, and no shipping fees for prime.
https://www.amazon.com/YUPVM-XY-AP100L-Amplifier-Dual-Channel-Stepless/dp/B09CYSJNB2
https://www.amazon.com/YUPVM-XY-AP100L-Amplifier-Dual-Channel-Stepless/dp/B09CYSJNB2
I guess they reverse engineered the TPA3116 chip to squeeze a few more dollars out of it. The pin count is different but arrangement is similar. So the power amp is separate - but that other chip that handles DAC/ADC/BT still does amazing things for being a sub component on a $4 amp.
Thanks for the measurements Jjasniew. Looking like a lot of higher order harmonic components. Possibly from the lack of a proper Class D filter. The measurements would indicate some nasty listening fatigue given the higher order distortion peaks. I wonder if this is an artifact of the the measurement somehow.
Thanks for the measurements Jjasniew. Looking like a lot of higher order harmonic components. Possibly from the lack of a proper Class D filter. The measurements would indicate some nasty listening fatigue given the higher order distortion peaks. I wonder if this is an artifact of the the measurement somehow.
Last edited:
I can try again. I recall seeing something similar on another class D amp I measured. It would be nice if I knew what it was supposed to look like - a "good one" - either from experience or results being reproduced in another's lab ;') I'm pretty much just getting started at this.
My sound card is USB connected. The noise floor is pretty low with no input, unplugged. I wonder if a USB isolation cable would help, or, not at all for frequencies above 10KHz?
I noted REW allows you to save a recording of the sound sample it uses to calculate the spectrum. Curious, does it do a spectrum from a previously recorded sound file then? Like, could I see how low distortion the "low distortion" 1KHz sound file I used actually is, with REW?
Ah, I see... Thanks for sharing that.
For starters, apparently my ADC doesnt have the herculean -140 db noise floor yours does.
This is with the coffee-money amp on, but signal from the function generator off going into it. If you turn the volume all the way down, it shuts down the output stages.
Next, I redid my HP function generator, producing the same voltage (~2.83) as the amps into my ADC / probe;
Next, I tried a small 6V6 amp I recently purchased from another member;
I can see it follows the function generator maybe even from 8K on up. So is it the ADC/ Probe, or what goes in must come out?
Then I tested a TDA7498 amp I picked up off ebay recently.
And a retest of coffee money.
I also tried 500Hz, which seems even messier.
I dont know two things;
1. How to subtract the distortion of the input signal from the result. Would I find this in "How to do distortion measurements using REW"?
2. If I got a better ADC, worked on better shielding for the probe got a -140 db floor; would the whole train of harmonics shown above between -70 and -80 stay at the same level, with the floor between them falling lower? I'd assume so.
The TDA7498 amp has "real" output inductors and film caps in its output filter, versus the little SMD things in this amp, along with MLC SMD caps. I did notice there's some unpopulated pads near the output connectors; maybe someone worked on what is showing up in the spectrums, but someone else decided "good enough for the price" to omit it.
For starters, apparently my ADC doesnt have the herculean -140 db noise floor yours does.
This is with the coffee-money amp on, but signal from the function generator off going into it. If you turn the volume all the way down, it shuts down the output stages.
Next, I redid my HP function generator, producing the same voltage (~2.83) as the amps into my ADC / probe;
Next, I tried a small 6V6 amp I recently purchased from another member;
I can see it follows the function generator maybe even from 8K on up. So is it the ADC/ Probe, or what goes in must come out?
Then I tested a TDA7498 amp I picked up off ebay recently.
And a retest of coffee money.
I also tried 500Hz, which seems even messier.
I dont know two things;
1. How to subtract the distortion of the input signal from the result. Would I find this in "How to do distortion measurements using REW"?
2. If I got a better ADC, worked on better shielding for the probe got a -140 db floor; would the whole train of harmonics shown above between -70 and -80 stay at the same level, with the floor between them falling lower? I'd assume so.
The TDA7498 amp has "real" output inductors and film caps in its output filter, versus the little SMD things in this amp, along with MLC SMD caps. I did notice there's some unpopulated pads near the output connectors; maybe someone worked on what is showing up in the spectrums, but someone else decided "good enough for the price" to omit it.
Last edited:
Thanks for trying all of this out! To get lower noise floor, try a higher sample rate on your ADC and select more data points on your sample (FFT length) and more averaging on the FFT.
What does a loopback on your audio interface look like? DAC out to ADC in? I wonder if there is something funny from your oscillator.
Loop back should only have maybe 2 or 3 harmonics.
What does a loopback on your audio interface look like? DAC out to ADC in? I wonder if there is something funny from your oscillator.
Loop back should only have maybe 2 or 3 harmonics.
Coffee Money Analyser...
Here's what the Lookback looks like on the device I used to take data. "JamLab USB", $20 - $30 on ebay.
This is after calibration and saving a cal file. The harmonics seem highly dependent on level coming out of / going into this thing;
This is for 1V RMS I/O, which is close to the level I setup when testing the amplifier. I figured any level below clipping, as shown on the REW oscilloscope would be fine;
Now here it is at the 250mV level, the level REW defaults to;
So, this suggests I better I/O at 250 mV, with the amps producing 2.83V.
I also have a Tascam US-122L, it behaves similarly at the ~1V RMS level;
One would think you could put a T connector on one input channel, have the amp output go into the other and REW would subtract the two, but I can see that it just doesnt work that way. I'll have to retest the amplifiers with an attenuation, such that the input level is 250mV, for 2.83V out.
Here's what the Lookback looks like on the device I used to take data. "JamLab USB", $20 - $30 on ebay.
This is after calibration and saving a cal file. The harmonics seem highly dependent on level coming out of / going into this thing;
This is for 1V RMS I/O, which is close to the level I setup when testing the amplifier. I figured any level below clipping, as shown on the REW oscilloscope would be fine;
Now here it is at the 250mV level, the level REW defaults to;
So, this suggests I better I/O at 250 mV, with the amps producing 2.83V.
I also have a Tascam US-122L, it behaves similarly at the ~1V RMS level;
One would think you could put a T connector on one input channel, have the amp output go into the other and REW would subtract the two, but I can see that it just doesnt work that way. I'll have to retest the amplifiers with an attenuation, such that the input level is 250mV, for 2.83V out.
Exploring further, (using laptop running on battery) I tried to see if there was any difference regarding the input level going into the ADC.
I tested the 6V6 amp using the -12 db level coming out of the DAC, but the full 2.83V going in. Here's that spectrum;
Then, I used a potentiometer to reduce the input signal to about the same level as the ADC would see doing the loopback at -12 db;
Even though the THD calculated by REW is very close, it seems looking at the spectrums the lower input level to the ADC gives a better looking measurement. FWIW.
Here's coffee money amp driven at the -12db level, attenuated to the -12db level, but actually putting out about 1 Watt as measured by my DMM;
With the full ~2.83V into the ADC;
Again, the THDs calculated by REW are the same, either way. Maybe the input attenuation isnt needed, as long as it's nowhere near clipping? Finally, here's that Xinrui TDA7498 amp @ 1W, voltage straight into the DAC w/o attenuation - it was $16, shipped, so coffee and a sandwich amp. No BT, USB to PC on that one though.
I tested the 6V6 amp using the -12 db level coming out of the DAC, but the full 2.83V going in. Here's that spectrum;
Then, I used a potentiometer to reduce the input signal to about the same level as the ADC would see doing the loopback at -12 db;
Even though the THD calculated by REW is very close, it seems looking at the spectrums the lower input level to the ADC gives a better looking measurement. FWIW.
Here's coffee money amp driven at the -12db level, attenuated to the -12db level, but actually putting out about 1 Watt as measured by my DMM;
With the full ~2.83V into the ADC;
Again, the THDs calculated by REW are the same, either way. Maybe the input attenuation isnt needed, as long as it's nowhere near clipping? Finally, here's that Xinrui TDA7498 amp @ 1W, voltage straight into the DAC w/o attenuation - it was $16, shipped, so coffee and a sandwich amp. No BT, USB to PC on that one though.
Last edited:
