I am trying to measure the distortion of a breadboarded amp I am working on at the moment and would appreciate some practical advice on procedures for these measurements. I would be happy to discuss this on the forum or privately by email, as I could envision the possibility the subject would not be of general interest.
Background on the amp, if it matters, is that it is configured as 1:5 input transformer->6SN7 PP or differential->12B4 PP or differential->8.6k output transformer.
The equipment I have available to make the measurements includes a few oscillators, a scope and a wave analyzer. The measurements I have made so far seem to show some degree of reproducibility, which I take to be a good sign, but I am unsure of how to determine possible systemic error, etc.
Well, you're using the right equipment. A spectrum analyzer is a good thing to have, too, but the waveform analyzer will handle most of those chores.
When you're measuring something like a tube amp with a lot of distortion and limited bandwidth (in a relative sense), you can also gin up an instrumentation amp to compare the (scaled) output to the input. Looking at the distortion residual with a scope can be enlightening.
Also, I haven't rigged Speaker Workshop up for amplifier distortion measurement yet, but it seems pretty straightforward to do. And it's a pretty cheap way to go. It's limited in how low in distortion you can measure, but you're measuring stuff which is quite a bit bigger.
methods of measurement
Yeah, I am waiting for a nice spectrum analyzer to fall into my hands on the cheap. The waveform analyzer does seem to do the trick, but I don't know how to verify it's performance. Do you have any suggestions?
By looking at the residual on the scope, do you mean by normalizing the two channels to the same height and then adding them (with inversion if necessary) and looking at that output? I have done that in the past, but quantitation on a scope is touchy stuff and it still doesn't show me the order of the distortion.
I have Speaker Workshop. There does seem to be a bit of a learning curve involved in using it, though that might be the best solution in the end. I think I bought most of the components to make the test jig, but I haven't built it yet. The biggest problems with that method are my great fear of smoking my computer with an inadvertantly clipped lead on the breadboard and the fact that the computer is in a different room and not handily moved back and forth.
A couple more questions, when expressing the distortion quantity as a percent or dB amount, is that calculated from the amplitude of the fundamental or the amplitude of the entire signal? Darn, I forgot the other one.
In theory, distortion is calculated as a % of the fundamental. In practice, it is measured w.r.t. the fundamental. Unless the distortion is appallingly high, the two are pretty much the same.
You need to use your wave analyser to determine how much distortion your chosen oscillator produces (and also the analyser's resolution). Once you know the absolute limits of measurement, the other thing that can cause you problems is earthing.
A good book is the "Audio Measurement Handbook" by Bob Metzler, published by Audio Precision. Understandably, the book is biased towards using AP, but the fundamentals are there.
With power amplifiers, you start by measuring distortion at 1kHz. Then you try to minimise the deterioration in distortion at 20Hz by using a better output transformer. Still at full power, you observe the shape of the output waveform as you increase frequency, if it starts turning noticeably triangular within the audio band, it's time to attend to the driver stage.
Thanks for the info and the reference. I will try to find it.
You touched on the question I forgot in the last post. How do you correct for the distortion of the source? What would the math for this be?
I had been using an old HP oscillator, but the it's distortion was rather high when I measured it. I tried a test disc with a 1 kHz sine wave played on a Sony amp/CD player and the distortion from that, hooked up to the amplifier under test as when testing, seems to be in the area of 0.037% 2nd, 0.026% 3rd and on to 6th which measured 0.04 mV and was at the limit of detection on the settings needed to test the 380 mV input. Do these seem like reasonable numbers?
I then tested the amplifier with no changes except the location of the wave analyzer leads, which were moved from the primary side of the input transformer to the secondary side of the output transformer 16R tap (loaded with a 10W 16R resistor). The output was 4.4W and previous tests had shown that this setting was just under the output where visible clipping of the waveform showed up on the scope. Does that seem like the right sort of procedure? Anything else I should account for?
One of the reasons I would like to have an accurate and repeatable procedure for this is so that while I am on the breadboard I can quickly gauge the changes in absolute and relative distortion amounts caused by changes in operating points, component changes, tubes with different degrees of mismatch between sections, etc.
I will make a table of the distortion numbers i have now if anyone is interested, but my main interest is learning how circuit changes can affect distortion and by listening before or after the measurements, how my ear perceives the any differences in musical quality.
i was stunned that my ancient Heath HD-1 distortion analyzer came close to the results from my HP339a -- I had "recapped" the HD-1 and changed the potentiometer to a multi-turn variety. The HD-1 was difficult to null at low frequencies, and a bit tricky to use at high frequencies, but around 1kHz it was very good.
you can make a very low distortion oscillator as Cyril Bateman did for his capacitor series of articles in Electronics World -- or you can use a 1kHz square wave and a couple of Linear Tech or Maxim switched capacitor filters to get down to a gnat's eyebrow of distortion -- this was described in AudioXpress a couple years ago.
for a notch filter, there's a nice easy to implement design on TI's website.
oh, I bought a Krohn-Hite for cheap on the Bay -- easier to use than the HP.
There is a method to find out the real amplifier distorsion by changing phase between the source and the amplifier and finding the maximum point and the minimum point and then calculate the real amplifier distorsion this assumes that the generator distorsion is less then the amplifier distorsion. this method is unpractical at audiofrequencies due to the problem to find variable delay lines.
The cheapest way to get a spectrum analyser is to download SW from the internet and use the PC with an soundcard to measure. depending on the soundcard quality it is possible to measure down to around -110 dB which should be enough for most purposes.
I am using this SW "Audiotester" http://www.audiotester.de/mainE.htm
which I found is quite good, it cost some money to buy the unlimited version but the demoversion works for a few minutes and then you have to restart it.
Re: methods of measurement
Second, the best way to get the residual is to use an instrumentation amp, and adjust the attenuation on one side of it to the best null. This won't work for super-low measurement, but seems fine for the sorts of stuff you get from tube amps.
I would recommend this free and excellent program:
Right Mark Audio Analiser
I use it extensively and it produces quick and accurate results. All you need to measure an amplifier with it is a reasonable quality soundcard and a variable divider (i.e. volume control pot) on the card input to reduce the level and match it to the card's full scale value.
Hey, thanks guys, your a wealth of information.
Jack, I have an HD-1 too, but I have not tried it yet. I don't have a manual and I really didn't know how to use it, but now I think I can figure it out. I have an HP333A also, although I am not sure of its accuracy, I have only tested it once. I go to a university surplus store and so I can pick up lots of old test equipment for little dollars. The question is whether it is worth even the few dollars I pay?? Right now I am using an HP302A wave analyzer. It is an all transisitor design that was introduced in 1959 as I understand it. It does have 3 or 4 crystal oscillators in octal based envelopes though. Seems to work very well from what I have been able to determine. What sort of Krohn-Hite did you get? I have a pair of tubed bandpass filters of theirs. Also have not tried them. As you might guess, I am just getting into trying measurements on my amps. I have done quite a bit of work in this area on my speakers though.
"the reason is that the total measured distorsion is the vector sum of the distorsion from the signal generator and what is produced by the amplifier"
Thanks for this tidbit, it answers some of the questions I was coming to when I was using the higher distortion source. Also very much in the line of information I was hoping to receive when I posted this thread. I expect there are a lot of points in making and interpreting these measurements which are, to me at least, rather non-obvious. I have tried the shareware version of audiotester and it seems pretty nice. I fired it up again after I read your post and it doesn't want to let me use line in for some reason, but I expect a cure for that can be found.
OK, OK, I'll get off my lazy butt and build the jig for Speaker Workshop. I can see a tilt in these posts to using the computer for analysis and measurement. Maybe I will have to try to scrimp and save and get an older computer I can dedicate to testing so it is more convenient and relieves me of the worry about toasting my main machine.
Does anyone have suggestions for the best inexpensive sound card to use for computer analysis of audio?
I used an older version of the Audio Analyzer program a year or so ago and I had just gone and downloaded the latest yesterday. There seem to be quite a few improvements and it would be a very quick test to run. Information about higher harmonics is not available as I see it though. Does that seem right?
Thanks again for all the information. I can see there are a lot of tricks of the trade, so to speak, in doing this kind of thing. So all help is appreciated.
|All times are GMT. The time now is 06:21 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio