Amp measurements
I've now been able to replicate SC's results, using my new dummy load / voltage divider. Here's the new box:
Using that, I'm able to comfortably test at 100W. So, some results! First, a 'loopback' test with the dummy load, but no amp. Here's the RMAA report:
http://jhoward.fastmail.fm/images/DiyAudio/Spect/Loopback_RMAA.htm
And here's a selection of spectra of the loopback through the dummy load:
1kHz
4kHz
7.5kHz
And now, the RMAA report for the amp (including the dummy load):
http://jhoward.fastmail.fm/images/DiyAudio/Spect/SCULD_RMAA.htm
And spectra for the amp (all @ 100W):
1kHz
4kHz
7.5kHz
Finally, a chart of the THD for each channel, plus for reference SC's reported THD:
This is using the same sound card as my last set of tests, so please look back a couple of pages in this thread for details about that, and the software I am using, since I've already provided a lot of detail about that.
The 'loopback' results are all using the same cables, sound card, and dummy load as the amp tests.
I've now been able to replicate SC's results, using my new dummy load / voltage divider. Here's the new box:
Using that, I'm able to comfortably test at 100W. So, some results! First, a 'loopback' test with the dummy load, but no amp. Here's the RMAA report:
http://jhoward.fastmail.fm/images/DiyAudio/Spect/Loopback_RMAA.htm
And here's a selection of spectra of the loopback through the dummy load:
1kHz
4kHz
7.5kHz
And now, the RMAA report for the amp (including the dummy load):
http://jhoward.fastmail.fm/images/DiyAudio/Spect/SCULD_RMAA.htm
And spectra for the amp (all @ 100W):
1kHz
4kHz
7.5kHz
Finally, a chart of the THD for each channel, plus for reference SC's reported THD:
This is using the same sound card as my last set of tests, so please look back a couple of pages in this thread for details about that, and the software I am using, since I've already provided a lot of detail about that.
The 'loopback' results are all using the same cables, sound card, and dummy load as the amp tests.
jp_howard said:
JP, have you been watching the thread
http://www.diyaudio.com/forums/showthread.php?s=&threadid=142166&perpage=25&pagenumber=9
Are you intending to build another of these amps. It would be interesting to compare the one you have built with one of the ones from the other thread. Are you interested.
Terry
THD+N and THD measurements
Hello Jeremy
Can you tell me Jeremy to achieve 0.01% THD how much db down all the harmonics need to be, assuming the reference level is 0dB as in the case of the 7.5KHz (final spectram shown).
In case you dont know its -80dB your graph shows 15kHz (3rd Harmonic) at -70dB an yet the THD is 0.0015% for 7500Hz input clearly there is a problem with your THD measurements. Looks like the free software has problems calculating THD. Can you tell me also what all the harmonics need to be down to achieve 0.0015% THD in this case.
Jeremy do you know the difference between THD and THD+N measurements, google it . Can you guess what the Audio Precision measures and what the SC measure with.
Regards
AR
Hello Jeremy
Can you tell me Jeremy to achieve 0.01% THD how much db down all the harmonics need to be, assuming the reference level is 0dB as in the case of the 7.5KHz (final spectram shown).
In case you dont know its -80dB your graph shows 15kHz (3rd Harmonic) at -70dB an yet the THD is 0.0015% for 7500Hz input clearly there is a problem with your THD measurements. Looks like the free software has problems calculating THD. Can you tell me also what all the harmonics need to be down to achieve 0.0015% THD in this case.
Jeremy do you know the difference between THD and THD+N measurements, google it . Can you guess what the Audio Precision measures and what the SC measure with.
Regards
AR
Sorry for the unclear results, Arthur. The issues you mentioned are all due to me, in my hurry to show results, not setting the FFT and sampling parameters optimally. The mismatches created many off-harmonic "distortions" in the graphs. For instance, the 15k "harmonic" you mention is actually at 14,700Hz, and is due to this mismatch. I've now re-done all the charts and measurements, and also switched from THD to THD+N for the final comparison. So first, here's the "loopback" (NB: all measurements and charts now shown using Virtins Multi-Instrument Pro, FFT 32k samples, windowing Kaiser7, 176.4kHz sample rate, 20Hz-40kHz THD bandwidth):
1kHz
4kHz
7.5kHz
And the amp measurements (in my higher-distortion channel):
1kHz
4kHz
7.5kHz
Finally, the comparison (using THD+N)
The SC article showed both THD and THD+N charts, which were almost identical. I'm finding the same thing - the higher THD results in my earlier post were due to the measurement problems I mentioned.
An externally hosted image should be here but it was not working when we last tested it.
1kHz
An externally hosted image should be here but it was not working when we last tested it.
4kHz
An externally hosted image should be here but it was not working when we last tested it.
7.5kHz
And the amp measurements (in my higher-distortion channel):
An externally hosted image should be here but it was not working when we last tested it.
1kHz
4kHz
An externally hosted image should be here but it was not working when we last tested it.
7.5kHz
Finally, the comparison (using THD+N)
An externally hosted image should be here but it was not working when we last tested it.
The SC article showed both THD and THD+N charts, which were almost identical. I'm finding the same thing - the higher THD results in my earlier post were due to the measurement problems I mentioned.
I am interested, but I have a lengthy list of projects ahead in my queue. First is to complete the case for this amp, then I have 3 different chip amps I want to build, along with 3 different types of volume control. I also have the Pluto speakers to build. And, I want to compare all these both with listening tests and analytically...pheonix358 said:
If anyone in Melbourne makes a "next gen" SCULD and would like me to test it, please let me know. I'm in Port Melbourne. Of course, the results I'll get with my very basic rig aren't as accurate as what Arthur can get with his AP machine, but it's better than nothing.
I've managed to get the noise floor down, by reconfiguring Vista to allow 196kHz sample rate, and by rerouting some cables. I've redone the most recent set of charts and updated them in place. The results have not changed much, but because of the lower noise floor they're now more accurate, so the final comparison has a shape more in line now with what you'd expect.
Unmodded results
Well this is very interesting - I have tested the THD of the SCULD without the LTP or biasing mods, and I get nearly exactly the same THD results as SC:
The "No LTP/bias" line, as you can see, is the same as SC's result, except for the 10k measurement. I don't know why SC's 10k measurement is so high... but except for this one sample I'm seeing exactly the same results as them.
So this explains a lot - it looks like the amp that SC built is exactly as described in the article. If only they had got the bias right, they could have boasted about a THD level 8x lower!
The "No LTP" line shows the amp results without the LTP trimpot adjustment (i.e. with the trimpot at 0 Ohm). I've also done a complete RMAA report showing how these three configurations compare on IMD, dynamic range, etc: http://jhoward.fastmail.fm/images/DiyAudio/Spect/Comparison_RMAA.htm (NB: I couldn't get RMAA to run stably at 192k, so this report uses 96k sample rate. However all other charts and comparisons use Multi-Instrument Pro, @192k.)
Well this is very interesting - I have tested the THD of the SCULD without the LTP or biasing mods, and I get nearly exactly the same THD results as SC:
The "No LTP/bias" line, as you can see, is the same as SC's result, except for the 10k measurement. I don't know why SC's 10k measurement is so high... but except for this one sample I'm seeing exactly the same results as them.
So this explains a lot - it looks like the amp that SC built is exactly as described in the article. If only they had got the bias right, they could have boasted about a THD level 8x lower!
The "No LTP" line shows the amp results without the LTP trimpot adjustment (i.e. with the trimpot at 0 Ohm). I've also done a complete RMAA report showing how these three configurations compare on IMD, dynamic range, etc: http://jhoward.fastmail.fm/images/DiyAudio/Spect/Comparison_RMAA.htm (NB: I couldn't get RMAA to run stably at 192k, so this report uses 96k sample rate. However all other charts and comparisons use Multi-Instrument Pro, @192k.)
For those interested in how the modded SCULD compares to commercial amps, here's a comparison to a few other amps I have lying around the house, both in terms of THD and THD+N at 1k and at 10k:
I don't have any mega-$$$ audiophile amps to compare to - however this shows that on THD measurements the SCULD beats these commercial amps very easily indeed. I found it particularly interesting to see that the non-harmonic noise of all the amps was relatively high - here's the spectra of each at 1k so you can see what this noise looks like:
Rotel RA-920Ax
Denon UDRA-M10
Denon AVR-3806.
I don't have any mega-$$$ audiophile amps to compare to - however this shows that on THD measurements the SCULD beats these commercial amps very easily indeed. I found it particularly interesting to see that the non-harmonic noise of all the amps was relatively high - here's the spectra of each at 1k so you can see what this noise looks like:
Rotel RA-920Ax
Denon UDRA-M10
Denon AVR-3806.
Hi,
if you look at D.Self's published distortion results you will see that using his test procedure that they all have a rising trend as frequency increases. He discusses which type/source of these distortions and how they relate to the slope of the increase.
He also shows the bandwidth for his plotted results. The reason being to show when to disregard the roll off and apparently falling distortion as the test gear can no longer measure the HF harmonics.
You're testing bandwidth is so low that you cannot do any accurate HF harmonic measurements when the fundamental is >=4kHz. It simply misses them out or under reads. It looks like your present bandwidth limits you to ~<=3kHz.
Please explain again what no LTP & no LTP/bias results are.
if you look at D.Self's published distortion results you will see that using his test procedure that they all have a rising trend as frequency increases. He discusses which type/source of these distortions and how they relate to the slope of the increase.
He also shows the bandwidth for his plotted results. The reason being to show when to disregard the roll off and apparently falling distortion as the test gear can no longer measure the HF harmonics.
You're testing bandwidth is so low that you cannot do any accurate HF harmonic measurements when the fundamental is >=4kHz. It simply misses them out or under reads. It looks like your present bandwidth limits you to ~<=3kHz.
Please explain again what no LTP & no LTP/bias results are.
"No LTP" means that I set the CM/LTP trimpot to 0Ohm (so it's the same as the SC article, which has a 0R0 jumper here), but left the bias trimpot at the level that provides 16mV (into 0R22). "No LTP/bias" means I also set the bias trimpot to 0Ohm, which provides about 2mV across the output resistors.
Andrew, my tests also show increases as frequency increases:
I'm afraid I don't understand how you can tell that my test gear can't measure the harmonics correctly for fundamentals >4k. Can you tell me more about this?
The Audio Precision test report for my sound card shows -1dB at 50kHz, and -2dB at 60kHz (using an AP SYS-2722, 192kHz sample rate, 24 bit), for the line-in. For the line-out, it's -1dB at 70kHz. Given this, I should be able to comfortably measure the first 4 harmonics over a 10kHz fundamental, right? Or am I missing something here?... Can you think of any other tests I could do to dig further into this?
Your comments would certainly explain why my results match SC's exactly for the measurements only up to 4k fundamentals. Although I don't understand then how I measured such high THD for the Denon UDRA-M10 @10kHz (>0.04%) - doesn't that show that my test setup can successfully identify harmonics over this fundamental?
Many thanks as always for your interesting insights, Andrew.
Andrew, my tests also show increases as frequency increases:
I'm afraid I don't understand how you can tell that my test gear can't measure the harmonics correctly for fundamentals >4k. Can you tell me more about this?
The Audio Precision test report for my sound card shows -1dB at 50kHz, and -2dB at 60kHz (using an AP SYS-2722, 192kHz sample rate, 24 bit), for the line-in. For the line-out, it's -1dB at 70kHz. Given this, I should be able to comfortably measure the first 4 harmonics over a 10kHz fundamental, right? Or am I missing something here?... Can you think of any other tests I could do to dig further into this?
Your comments would certainly explain why my results match SC's exactly for the measurements only up to 4k fundamentals. Although I don't understand then how I measured such high THD for the Denon UDRA-M10 @10kHz (>0.04%) - doesn't that show that my test setup can successfully identify harmonics over this fundamental?
Many thanks as always for your interesting insights, Andrew.
Hi,
could the relatively poor Denon result be due to high 2nd and 3rd harmonics at the 10kHz test frequency?
Look at the graphs posted by D.Self.
It clearly shows the roll off of measured distortion.
He often uses 80kHz bandwidth and those results start to curve the wrong way @~25kHz.
The 30kHz bandwidth inflexion occurs @~6kHz.
could the relatively poor Denon result be due to high 2nd and 3rd harmonics at the 10kHz test frequency?
Look at the graphs posted by D.Self.
It clearly shows the roll off of measured distortion.
He often uses 80kHz bandwidth and those results start to curve the wrong way @~25kHz.
The 30kHz bandwidth inflexion occurs @~6kHz.
No LTP/Bias = 2mV
chA = 16 mV
No LTP = 16mV
i.e. the only configuration in which the biasing trimpot was touched is the "No LTP/Bias" config. In "No LTP" the only difference to the modded SCULD is to set the current mirror/LTP trimpot to 0Ohm. "Ch A" is the SCULD with all the mods I've done (the same config as all the earlier, more detailed measurements I've been showing).
chA = 16 mV
No LTP = 16mV
i.e. the only configuration in which the biasing trimpot was touched is the "No LTP/Bias" config. In "No LTP" the only difference to the modded SCULD is to set the current mirror/LTP trimpot to 0Ohm. "Ch A" is the SCULD with all the mods I've done (the same config as all the earlier, more detailed measurements I've been showing).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.