It's a puzzle to me, but maybe it's obvious to some of you!?
Here is a situation: THD+N at ~ 4% before 10 kHz drops to below 0.1% shortly after 10 KHz.
Here is a situation: THD+N at ~ 4% before 10 kHz drops to below 0.1% shortly after 10 KHz.
What are you using to make these measurements? If the measurement system doesn't have response to at least 50 KHz it's not responding to enough harmonics to adequately measure THD above 10 KHz. If all of your THD is third harmonic, then your measurement system needs to be flat out to 30 KHz. Some sound cards have low pass filters at 20 to 22 KHz, so that all of the harmonic information above about 11 KHz is lost. In this case a 10 KHz square wave will show a low THD.
What are you using to make these measurements? If the measurement system doesn't have response to at least 50 KHz it's not responding to enough harmonics to adequately measure THD above 10 KHz. If all of your THD is third harmonic, then your measurement system needs to be flat out to 30 KHz. Some sound cards have low pass filters at 20 to 22 KHz, so that all of the harmonic information above about 11 KHz is lost. In this case a 10 KHz square wave will show a low THD.
It's the cheap Analog Discovery 2, from Diligent. With the BNC board.
And it's quite possible I didn't set it up right. 🤔
So, I verified it's not the measuring device or the test set-up.
I swapped the tube amplifier with a solid sate amplifier and THD+N was below 0.1% between 20 Hz and 20 kHz for the solid state amplifier.
Furthermore, I also tested both channels and got the same odd result with the tube amplifier, that is an order of magnitude drop in THD+N past 10 kHz.
I swapped the tube amplifier with a solid sate amplifier and THD+N was below 0.1% between 20 Hz and 20 kHz for the solid state amplifier.
Furthermore, I also tested both channels and got the same odd result with the tube amplifier, that is an order of magnitude drop in THD+N past 10 kHz.
The very regular scalloping in the response below 10kHz and the manner of the transition says that some signal is aliasing into the FFT at low frequencies. Calculate the bandwidth between the scallops points to determine the base frequency of the aliasing. Then use a narrow bandwidth sweep to look for the low frequency signal.
The very regular scalloping in the response below 10kHz and the manner of the transition says that some signal is aliasing into the FFT at low frequencies. Calculate the bandwidth between the scallops points to determine the base frequency of the aliasing. Then use a narrow bandwidth sweep to look for the low frequency signal.
Thank you for this. I'll try and parse this tomorrow.
But it works correctly on a different amp. So your suggestion isn't the issue.The very regular scalloping in the response below 10kHz and the manner of the transition says that some signal is aliasing into the FFT at low frequencies. Calculate the bandwidth between the scallops points to determine the base frequency of the aliasing. Then use a narrow bandwidth sweep to look for the low frequency signal.
Jan
My money would be on a level and/or frequency dependent oscillation.
If oscillartion is there, it can look like a high THD (the 'THD meausurement' takes everything excepot the fundamental).
Use the scope function of the AD2 and see if there's an oscillation at the amp output.
Or try the sweep at a different level.
Did you use a load on the amp?
Jan
If oscillartion is there, it can look like a high THD (the 'THD meausurement' takes everything excepot the fundamental).
Use the scope function of the AD2 and see if there's an oscillation at the amp output.
Or try the sweep at a different level.
Did you use a load on the amp?
Jan
A vacuum tube amplifier distortion and noise drops off sharply above 10kHz. That is a mystery.
1. But the real mystery is what is the amplifier model, what is the topology, what is the schematic, etc.?
Is it just possible that there is something that makes the sharp drop off.
First, check the frequency response of the amplifier.
If the frequency response is good, then how about some information on the amplifier (like in 1. above).
Just my curiosity.
1. But the real mystery is what is the amplifier model, what is the topology, what is the schematic, etc.?
Is it just possible that there is something that makes the sharp drop off.
First, check the frequency response of the amplifier.
If the frequency response is good, then how about some information on the amplifier (like in 1. above).
Just my curiosity.
I would guess its the sampling rate of the analyzer. If its set to 48KHz then anything above about 20kHz (second harmonic) will be removed by the ant-aliasing filter.
What is the FR of the tube amplifier? Its upper cutoff? Maybe the other amp can play higher than the tube amp?
If the solid state amp already has a low distortion below 10KHz and can't drop lower at 10KHz and higher because of the measurement device, it may stil point to the measurement device not being oké.
Shouldn't the thd rise towards 20KHz anyway?
It's just because of the high harmonics of the tube amp that the low thd at 10K and higher stand out.
Shouldn't the thd rise towards 20KHz anyway?
It's just because of the high harmonics of the tube amp that the low thd at 10K and higher stand out.
It is possible that the amp itself has a sharp drop in frequency response at 20 KHz due to its OPT or other factors. If this is the case, then your results are somewhat normal as the frequency response rolloff chops off any harmonic distortion energy above 20 KHz.
The first test one must make when setting up a new test station is the loopback test. Connect the input of the Discovery's "scope" to the output of the "generator" and run a frequency response test at a reasonably strong signal level. This must be flat to 50 KHz or so to be useful for testing audio amps.
If the loopback test reveals a good response, insert the SS amp and run it. Then test the tube amp.
I have an Analog Discovery 2 and a Digital Discovery here that were used in a music synthesizer design workstation, but due to a massive teardown and upgrade session for nearly every dedicated PC in my lab, that station is inoperative for now. I don't think I ever tried to test a tube amp with it though, I have a different PC with a Focusrite Scarlett Solo being set up for amp testing. The original Windows 7 PC in that station died triggering the consolidate and upgrade cycle. Most of my dedicated use PC's are 7 to 10 years old.
The first test one must make when setting up a new test station is the loopback test. Connect the input of the Discovery's "scope" to the output of the "generator" and run a frequency response test at a reasonably strong signal level. This must be flat to 50 KHz or so to be useful for testing audio amps.
If the loopback test reveals a good response, insert the SS amp and run it. Then test the tube amp.
I have an Analog Discovery 2 and a Digital Discovery here that were used in a music synthesizer design workstation, but due to a massive teardown and upgrade session for nearly every dedicated PC in my lab, that station is inoperative for now. I don't think I ever tried to test a tube amp with it though, I have a different PC with a Focusrite Scarlett Solo being set up for amp testing. The original Windows 7 PC in that station died triggering the consolidate and upgrade cycle. Most of my dedicated use PC's are 7 to 10 years old.
Thanks all for the help!
I will consider each point in turn later today.
One important clue (I think) is that distortion before 10 kHz is level-dependent (I can make it increase by increasing level), while it drops below 0.1% after 10 kHz regardless of level.
I will confirm that observation as well.
I will consider each point in turn later today.
One important clue (I think) is that distortion before 10 kHz is level-dependent (I can make it increase by increasing level), while it drops below 0.1% after 10 kHz regardless of level.
I will confirm that observation as well.
What is the FR of the tube amplifier? Its upper cutoff? Maybe the other amp can play higher than the tube amp?
There is roll-off towards 20 kHz, no doubt, but I’ll check response to 50 kHz.
My money would be on a level and/or frequency dependent oscillation.
If oscillartion is there, it can look like a high THD (the 'THD meausurement' takes everything excepot the fundamental).
Use the scope function of the AD2 and see if there's an oscillation at the amp output.
Or try the sweep at a different level.
Did you use a load on the amp?
Jan
Thanks! Yes, there is a load on the amp.
OK, walked the signal path...and...it seems the pot introduces the distortion somehow.
Signal is fine before the pot, from the wave generator.
Signal is fine at the pot entry.
Signal has 5% distortion at the wiper exit.
Now, the pot common pin is connected to star ground, which is connected to earth ground, so it could be coming from there...umm...
Signal is fine before the pot, from the wave generator.
Signal is fine at the pot entry.
Signal has 5% distortion at the wiper exit.
Now, the pot common pin is connected to star ground, which is connected to earth ground, so it could be coming from there...umm...
The first test one must make when setting up a new test station is the loopback test. Connect the input of the Discovery's "scope" to the output of the "generator" and run a frequency response test at a reasonably strong signal level. This must be flat to 50 KHz or so to be useful for testing audio amps.
This checked out OK, FR response pancake flat to 50K in a loopback test.
- Home
- Amplifiers
- Tubes / Valves
- Puzzle: Order of magnitude drop in THD+N around 10 kHz