Amplifier Performance Testing

[mwl6m]

I have just built a tube buffered gain clone, and I wanted to do a few performance tests. I want to measure the distortion of each section (tube, chip amp, and entire amp), the frequency response for each, and any other performance tests that are necessary.

I have access to a digital 4 channel oscilloscope, function generator, and many dmm's. If any one has any good sites or have any suggestions of how to make a few performance measurements would be great.

I am including the schematic if it helps anyone out. Thanks ahead of time for any of the help.

[gtforme00]

I would be interested to know the answer to this, although I am not qualified to respond. Here's a friendly bump!

[daatkins]

If you have a respectable soundcard, freebie software like Rightmark can do this for you. Obviously, you’ll be limited by the distortion of your soundcard but this can be cancelled out by doing a calibration run beforehand. Besides, modern cards are pretty low distortion anyway (around 0.05% for a correctly setup SB Audigy and 0.002% for a semi-pro card).

Loop soundcard output to input and calibrate the soundcard using Rightmark to determine the optimum signal levels for lowest THD. Then simply connect the amp between the soundcard line-out and line-in but make sure you have an attenuator on the amp output when the voltage exceeds >3v.

http://audio.rightmark.org/index_new.shtml

Nice one,
David.

[mwl6m]

Thanks for your help. I will download the sw and try to see if I can get it working. I used to have a nice sb audigy external sound card, but don't have that anymore. Is there anyway you can use a standard sound card to obtain some results, or is this out of question?

[gootee]

Hi, mwl6m! I'm glad you started this thread.

I am not an expert (yet). But I've been doing some research, as I design and test (and then sometimes build) amplifier and power supply designs, often with the aid of the LT-Spice circuit-simulator (Highly recommended! See http://www.fullnet.com/~tomg/gooteesp.htm for the free LTspice direct download link, plus a direct link to the truly-excellent LTspice user/support group, plus my very useful downloadable "example" LTspice simulation files, AND a link to a huge list of links to OVER 20,000 free downloadable Spice component models! Good stuff.).

There are some "standard", well-known types of tests that are often used to judge and compare amplifiers:

There's THD (Total Harmonic Distortion), expressed as a percent. THD is usually measured by using as pure a sine input as possible and then measuring the THD of the output. It's something like the total power in the output harmonics divided by the total power in the input sine, times 100.

THD measurements must be specified by the frequency that was used. THD-20, i.e. THD at 20 kHz, is a major benchmark figure, but, unfortunately, soundcards are not capable of performing the 20 kHz measurement.

THD generally appears to be MUCH lower when lower frequencies are used. So the 1 kHz measurements that are often given are virtually meaningless, by themselves. THD "should" also be specified as to how many harmonics were included. For example, in LTspice simulations, the default number of harmonics used for THD calulation is (IIRC) nine.

THD also usually appears lower when an amplifier is running at higher power output. So any THD measurement should also specify what the peak-to-peak output voltage was, and what the impedance of the load was, and what the maximum continuous output power capability of the tested amplifier was.

If you really want to measure THD accurately, you'll want to use a real Distortion Analyzer. There are good used ones on ebay.com, often. I got an old HP 334A for about $150, in calibrated, working (and very good cosmetic) condition, just a few years ago. Its most-sensitive scale is only 0.1%, full-scale. But that lets me go down to 0.001% for half of the leftmost tic-mark, on the analog meter. Of course, there are MUCH better more-modern ones, for (much) more money.

You'll also need a low-distortion sine for the input, for making THD measurements. If you don't have a good signal generator (which you would typically then follow with a beefy multi-stage low-pass filter, to purify the sine as much as possible), maybe you could buy a "test signals" CD, or download some, from somewhere (Let us all know where, if you find a good download source, please.). But also note that most of the older HP (Hewlett Packard) sine generators are truly high-quality instruments, and can be found in good used condition for extremely low prices. Even the really ancient ones are usually very good.

Another "standard" test is TIM, i.e. Transient Intermodulation Distortion. I'm not an expert, as I said. But I think that one standard way of measuring this is to use twin tones as inputs, typically 19 kHz and 20 kHz, and then measure the total power present in any other frequencies that come out (some may be lower and some may be higher than the original frequencies), and take that as a percent of the original inputs' total power. But don't quote me on that one. Do a search, here at diyaudio.com, for "Bob Cordell twin tones" (and/or "TIM"), without the quotes, and you should find some more info. Bob's website probably has a lot more about that, and a lot of cool related stuff, too. It's at http://www.cordellaudio.com . He really knows his stuff, and is also a diyaudio.com regular. (When I'm simulating am amplifier circuit, I can just use ideal sources to inject the twin tones and then do an FFT (Fast Fourier Transform) of the output, which gives a plot of the output versus frequency, like a Spectrum Analyzer would give. [An aside, to all Spice users: Note that simulations can be very inaccurate if you don't model the parasitics of components and PCB traces or wires well-enough.])

For both THD and TIM, it might also be helpful to view the output on an instrument called a Spectrum Analyzer, which plots the amplitude or power versus frequency, for whatever its input is. (Spectrum Analyzers are way cool; and expensive.)

Another common test, related to stability margins, is to use a squarewave as the input. The actual repetition frequency probably isn't as important as the edge times. (Note that most amplifiers SHOULD have a low-pass filter, at their inputs, at the very least, to limit incoming signals to be within their bandwidth, and to keep things like RF (Radio Frequency) out.) At any rate, there are standards for how much "ringing" (bursts of decaying-in-amplitude high-frequency oscillation) is acceptable at an amplifier's output, due to the squarewave's edges. The real fun often begins when you try to meet the "capacitive load" performance standard, which is usually tested with something like a 2.2 uF capacitor in parallel with an 8 Ohm resistor (or whatever the designed impedance for the output is), on the output terminals. The ringing is usually supposed to die out within 5 cycles, IIRC. And there's some rule about its initial amplitude as a percentage of the total, I think, plus a few more rules of thumb.

Of course, you'll also want to be sure to (probably first) perform the even-more-common basic operational tests, such as checking the outputs' DC-offset voltages with the input shorted, checking the power supply rails' voltage levels AND their waveforms under various operating conditions, and determining the output power of the amp, and investigating its clipping behavior, etc etc etc.

You'll want to use a dummy load, rather than speakers, for most amplifier testing. There are really-great instruments called "electronic loads". But you can also make your own. Obviously, if you have a resistor for each output channel, that has the correct resistance and enough power-handling capability (think "heatsinks"), you can use those. But also remember that all resistors are not alike, and none of them are perfect. In particular, most of the high-power resistors that are very affordable are of the "wirewound" variety. i.e. They have inductance. Off the top of my head, I don't know if it's usually in a range that would affect audio testing, significantly. But I have seem quite a few types that say things like "non-inductive below 20 Ohms". If the inductance turns out to NOT be a problem, keep in mind the old amateur radio guys' trick of submersing a test-load resistor in a paint-can full of oil (not sure what kind, offhand; maybe mineral oil? maybe doesn't matter?), which acts as a multi-hundred-watt heatsink! You can use relatively low-power-rated resistors, that way, sometimes at way, WAY above their power ratings. But if you can limit your testing to average power levels of, say, 30 Watts, maybe you could use some nice Caddock 30-Watt (or other) 1% metal-film power resistors, which come in a TO-220 case, similar to a power transistor, ready to bolt to a heatsink. Unfortunately, for the Caddock resistors, I don't see values of 4, 8, or 16 Ohms listed. But there are many, many different types and sizes of power resistors (up to hundreds of watts) available at http://www.mouser.com . However, I recommend first ordering their free 1854-page paper catalog, since it makes it SO much easier to find things that you didn't previously know about, than using their website. The alliedelec.com catalog, which is even larger, is also really-great for that, especially since they recently changed it to have mostly color photos of everything.

Sorry to have blathered-on about all of that, for so long.

Have fun!

- Tom Gootee

http://www.fullnet.com/~tomg/index.html