I bought a distortion and level meter, old model, but can measure up to 0.03 % (only at 1KHz), much more of what I need for tube amps. Today I played a bit with it, with some puzzling results, so I am not sure I am using it in a right way and if I am following the right measurement methodology.
My goal was to check the effect of different tweaks on the amp distortion. My test bed, is the Abhelah design 6N1P + KT88(triode mode). I started with the configuration as it is posted in the original design(there is an entire thread on it) and I tested the following:
1) led instead of cathode resistor for the driver.
2) CCS as cathode load. (cascode IXYS and LM317) with various current settings.
3) changes on value of the KT88 cathode resistor.
I was using a 4 ohm resistor as a load and a 5K to 4ohm OPT(James).
As these modifications affect the overall gain of the amp, I was changing the input signal, in order to have always 4V rms as output. So 4W.
I measured 3.2 % distortion with the original configuration. Mods 1 and 2 did not improve much. And I was very surprise.
Changing the cathode resistor of the KT88 had a measurable effect, but only changing from 560ohm to 220 ohm. It was very hot, not really good for normal operation.
One thing that surprised me is that swapping two matched tubes (svetlana KT88) the distortion changed from 3.2 % to 4% !
The 6N1P(winged C) were more consistent.
Am I doing things right ?
I am trying to setup my PC to do some FFT, but I am never sure if I am seeing the distortion of the input stage of the sound card clipping, or the amp.
Please comment,
Davide
My goal was to check the effect of different tweaks on the amp distortion. My test bed, is the Abhelah design 6N1P + KT88(triode mode). I started with the configuration as it is posted in the original design(there is an entire thread on it) and I tested the following:
1) led instead of cathode resistor for the driver.
2) CCS as cathode load. (cascode IXYS and LM317) with various current settings.
3) changes on value of the KT88 cathode resistor.
I was using a 4 ohm resistor as a load and a 5K to 4ohm OPT(James).
As these modifications affect the overall gain of the amp, I was changing the input signal, in order to have always 4V rms as output. So 4W.
I measured 3.2 % distortion with the original configuration. Mods 1 and 2 did not improve much. And I was very surprise.
Changing the cathode resistor of the KT88 had a measurable effect, but only changing from 560ohm to 220 ohm. It was very hot, not really good for normal operation.
One thing that surprised me is that swapping two matched tubes (svetlana KT88) the distortion changed from 3.2 % to 4% !
The 6N1P(winged C) were more consistent.
Am I doing things right ?
I am trying to setup my PC to do some FFT, but I am never sure if I am seeing the distortion of the input stage of the sound card clipping, or the amp.
Please comment,
Davide
I would have expected that the CCS load could have reduced distortion a little. It is however possible that a more linear driver gives less distortion cancellation between driver and output, and you end up with about the same THD, but possibly different mix of 2'nd and 3'rd.
LED bias will probably not affect measurable results.
Interesting the variation between KT88 tubes, if they were operating at the same current.
That is quite easy;
1) Do a loopback of the soundcard to check its noise floor and distortion (if any), and FR.
2) Always keep the signal into the soundcard below its max level (0dB digital). The limit is fairly easy to spot since all harmonics go up sharply.
The Spectrum analyser in ARTA is very good for such measurements, and shows clearly the digital clipping level in red.
The FFT has the advantage over a THD meter that it shows the spectrum of harmonics.
Svein
LED bias will probably not affect measurable results.
Interesting the variation between KT88 tubes, if they were operating at the same current.
That is quite easy;
1) Do a loopback of the soundcard to check its noise floor and distortion (if any), and FR.
2) Always keep the signal into the soundcard below its max level (0dB digital). The limit is fairly easy to spot since all harmonics go up sharply.
The Spectrum analyser in ARTA is very good for such measurements, and shows clearly the digital clipping level in red.
The FFT has the advantage over a THD meter that it shows the spectrum of harmonics.
Svein
My experience is with solid state, but matched can mean many different things. IMHO, there are no identical tubes or transistors. Similar, and matched at one or two points on a curve, but not identical. It's no surprise to me that swapping the tubes changed the THD number, as it also does for solid state output devices. Agreed that the SA can be much more revealing of what's going on, but I still like a dedicated THD meter as a reality check.
CH
CH
Your meter may need calibration. Try measuring the distortion of a piece of wire as a base measurement.
Once your distortion analyzer is working (which model you have?) you may be able to take the residual signal (distortion only) and FFT analyze that on your PC. This may allow you to resolve the distortion spectra a little better but also magnifies the noise.
The secret of measurement is having good control measurements to compare against. That and consistency... And keeping good records...
Cheers,
Michael
What are you using for signal generator? The distortion of a normal function generator is typically on the order of a few percent. A real audio generator should be well below that (mine is 0.002 %).
As others suggest - hook up the distortion meter directly to the signal source and measure the combined THD of the generator and the analyzer frond end. No matter what you do, you won't be able to measure distortion below this number.
~Tom
As others suggest - hook up the distortion meter directly to the signal source and measure the combined THD of the generator and the analyzer frond end. No matter what you do, you won't be able to measure distortion below this number.
~Tom
So I have a Panasonic VP-7201A as a signal generator and a Leader 1701 as a distortion meter. They are both Japanese product, pretty difficult to find documentation on them. The equipment was tested for calibration and it performed OK. When attached directly to the distortion meter the signal generator gives 0.008 of distortion. I have a long way to go before (as Morgan Jones says) my amps will "challenge the author testing equipment"
About the results with the KT88, I think that reducing the cathode resistor increased the current. I expected a reduction of distortion, but I did not think I had to push the tube so hard to see it. The valves are "matched" in the sense that they have similar gm, but other specs obviously are not. I was wondering why I should get crazy to decrease distortion of 0.5 % if just replacing the valve can increase it sensibly.
I am going to assemble a potential divider (actually I am copying the input stage of the Pete Millett interface) to make measurements with the PC. The think is that my laptop have only mic input, I am thinking about buying an external sound card. I am very interested in seeing the harmonic components of the distortion.
Another thing: the distortion meter has two features that I did not understand how to use. The first is marked "Judgement" (Dist only), it's a trimmer with a button that can be set in measurement mode and in preset mode.
The second one is the S/N ratio. it has two buttons set and meas and a trimmer set level.
Best Regards,
Davide
About the results with the KT88, I think that reducing the cathode resistor increased the current. I expected a reduction of distortion, but I did not think I had to push the tube so hard to see it. The valves are "matched" in the sense that they have similar gm, but other specs obviously are not. I was wondering why I should get crazy to decrease distortion of 0.5 % if just replacing the valve can increase it sensibly.
I am going to assemble a potential divider (actually I am copying the input stage of the Pete Millett interface) to make measurements with the PC. The think is that my laptop have only mic input, I am thinking about buying an external sound card. I am very interested in seeing the harmonic components of the distortion.
Another thing: the distortion meter has two features that I did not understand how to use. The first is marked "Judgement" (Dist only), it's a trimmer with a button that can be set in measurement mode and in preset mode.
The second one is the S/N ratio. it has two buttons set and meas and a trimmer set level.
Best Regards,
Davide
A dummy load with voltage divider works fine, even to the mic input of a notebook.
The back-to-back zeners are for protection of soundcard input.
The interface I use has voltage dividers, about 1/500 for mic level output and 1/20 for line level.
A switch between 4 and 8 ohm load can also be practical.
SveinB
The back-to-back zeners are for protection of soundcard input.
The interface I use has voltage dividers, about 1/500 for mic level output and 1/20 for line level.
A switch between 4 and 8 ohm load can also be practical.
SveinB
I am building something like this, but I left out the 8 ohm resistor, so I can use it also on the output of the driver stage, and added a capacitor on input to filter DC.
Today I did more measurements: actually I was measuring the wrong thing. Putting the distortion meter at the grid resistor of the KT88 I was measuring only the performances of the driver stage. Here the difference is obvious. I could go from 1 % of distortion with 10 V swing to 0.2 % adding the CCS. I was passing 15 mA with the plate at 250 V.
Now after the optimization of the driver stage I measured the output again, and here I could not see much difference again. The distortion of the output is dominating. Now I wonder, what's the point of reducing the distortion in the driver stage if the output waste all my effort ?
I could not listen to it (it's late now here), I'll do tomorrow. I also plan to check the harmonic components.
Davide
Today I did more measurements: actually I was measuring the wrong thing. Putting the distortion meter at the grid resistor of the KT88 I was measuring only the performances of the driver stage. Here the difference is obvious. I could go from 1 % of distortion with 10 V swing to 0.2 % adding the CCS. I was passing 15 mA with the plate at 250 V.
Now after the optimization of the driver stage I measured the output again, and here I could not see much difference again. The distortion of the output is dominating. Now I wonder, what's the point of reducing the distortion in the driver stage if the output waste all my effort ?
I could not listen to it (it's late now here), I'll do tomorrow. I also plan to check the harmonic components.
Davide
Regarding CCS anode load; It will give a flat static DC loadline. For the actual dynamic loadline you will still have the input impedance of the final to consider (grid-leak + Miller + measuring probe).
Complementary unlinearity of driver and final can considerably reduce the total distortion, especially 2nd. 3rd may then become more dominant - and possibly less ear-plesing.
Svein
Complementary unlinearity of driver and final can considerably reduce the total distortion, especially 2nd. 3rd may then become more dominant - and possibly less ear-plesing.
Svein
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.