How to test a DIY amp?

[mariner4]

Hey, guys!

Thanks to extensive searching of this site, I've put together a version of Andrea Ciuffoli's "no-compromise" headphone amp. This is my first electronics project, so I guess should be lucky that I hear sound and that I didn't get burned, cut or shocked in the process, but I'd like to figure out if it sounds right.

How do I test this thing? It sounds pretty good, but I have never heard a good headphone amp. Are there specific tracks I can play that will help illuminate flaws? Do I need to hook up a spectrum analyzer? (Things look fine on the scope.)

Thanks!

[Miles Prower]

Originally Posted by mariner4

How do I test this thing? It sounds pretty good, but I have never heard a good headphone amp. Are there specific tracks I can play that will help illuminate flaws? Do I need to hook up a spectrum analyzer?

I use a couple of tests on my DiY designs.

Twin-T Test:

For this, use either a load resistor at the output, or you can use the speeks, if the resulting noise isn't a problem. Between the output load and the o'scope input, connect a Twin-T notch filter. Adjust your signal generator for the deepest null of the fundamental. What remains on the o'scope screen is harmonic distortion. You can get an idea as to how its magnitude compares to the output magnitude.

Doing this test on an 807 amp running open loop showed a severely distorted residual whose amplitude came in at something like 0.9% of the output, being definitely in line with the spec sheet THD= 1.8%. The residual after nulling out the fundamental (1100Hz sine) showed lots of third and higher order harmonics.

Running 807s open loop does tend to sound nasty with some material (i.e. Bark at the Moon / Ozzy (CD) )

Doing the same test with 6BQ6s running open loop shows a higher magnitude for the harmonics (2.8% of output) yet it looks almost completely sinusoidal at three times the fundamental. Running open loop, 6BQ6s don't produce that pentode nastiness unless you're running them almost into clipping, and sound a bit "agressive" or "edgy", again agreeing with this test.

Another way to test is to use a differential amp to subtract the input signal and the output (a convenient place for this is to connect one end of the differential to the gNFB summing node. Adjust the input for the deepest null. A perfect amp would show a straight line from the differential, as output would be identical to the input. Doesn't work that way, though. Any residual that looks sinusoidal is pretty good.

The ultimate test is, of course, listening. I like to listen while running open loop for at least a week (two is better) every day to see what sonic flaws reveal themselves, before deciding on what needs to be done so far as adding NFB. A variety of different material from different sources is also a good idea as some material (notably early CDs) just doesn't sound good at all, and can even sound worse the better the amp that it's played through. (Sometimes CDs are actually mastered to sound good on lousy car audio and "Big Box" solid state equipment, and a good hollow state amp reveals just how horrible they are.)

It was obvious that those 807s needed help in the form of local NFB, as well as gNFB, so that was what was added to get them sounding really good.

6BQ6s didn't need that extra help, and just the gNFB was all that was required to take off the excessive edginess.

[mariner4]

Thanks! I will try the differential amplifier approach.

[ChrisA]

Originally Posted by Miles Prower

I use a couple of tests on my DiY designs.

Twin-T Test:

For this, use either a load resistor at the output, or you can use the speeks, if the resulting noise isn't a problem. Between the output load and the o'scope input, connect a Twin-T notch filter. Adjust your signal generator for the deepest null of the fundamental. What remains on the o'scope screen is harmonic distortion.

You can buy an HP334 distortion meter on Ebay. This is basically a twin-t filter that can be tuned over the audio range. So you can place youe generator dead on (say) 1KHz and the move the twin-t to null the signal. The HP334 has the ability to measure and read out in units of percent THD. It also has output that can connect to a 'scope. Heathkit made a distortion meters too. You can buy it for about $50 on eBay.

This is a very good way to get a feel for distortion but if you want better numbers out to three decimal places you can use a studio grade audio interface and a computer. Find one that des 96K samples per second and 24-bit samples. Then you use software create a spectrum analyzer. There will be noise but if the software is any good it can average readings for several seconds and produce very accurate results.

But be warned that in both cases the results are only as good as your signal generator. The generator needs to be able to produce a sine wave with THD at least an order of magnitude below the device under test.

Both of these methods can be used togood advantage for testing components and stages. Like for example measuring an outpuit transformer.

It is good to have a tunable twin-t filter so you can see how distortion varies with frequency