Who can explain this FFT graph to me

[Bas Horneman]

I got my SE 300B amp measured. This was the result. But I did not get a lot of explanation about it. (There were other in the queue behind me so I had to move on)



I understand the 50Hz and100Hz thing. But what do the other spikes mean? The guy who measured it said. It was mostly 2nd harmonic. Someone else posted at the audioasylum that he was suprised at the amount of 3rd harmonic distortion (he suspected the pentode driver)

So who can teach me how to look at this graph?

[Colt45]

50hz, 100hz is PS noise. 2k, 3k, 4k, and so on, is harmonics
2k = 2nd, 3k = 3rd, etc.

the stuff between 100hz and 1khz is kind of voodoo... intermod or something? it's really low there anyways... not important.

[fernando_g]

The 50, 100, 150, 200 Hz, etc are your AC mains hum and its harmonics. Worst case being -70 db is not bad at all, and probably could reduced further with additional filtering and or shielding..

The concern I would have here is the 2k, 3k, 4k Hz, etc. That's harmonic distortion. Defenitively tweak your design to minimize that. Play a little with the biasing of your driver.
Now this is a SE amp after all, distortion will only go down so far after all. Your ears will be the final judge.

[Gerrit Boers]

There's one essential specification missing: What was the level for the 1Khz fundamental?
I'll presume that the levels are such that the fundamental would have been at 0dB.
If that's the case then the 2nd harmonic distortion is about 1% (-40dB), for the THD you have to add up all the harmonics of 1Khz. My guess is that you'll end up with about 1.2%.

The 50Hz can be hum or common mode noise (harmonics are all odd, 50-150-250 etc.)
!00Hz, as you probably know is from the power supply(ripple).

The peaks around 1Khz look like intermodulation products of the test signal with the 100Hz ripple. It looks as if the same thing happens with the harmonics of 1KHz.

[Bas Horneman]

Thanks all so far. Apparently there is a notch filter (in the measurement equipment) at 1k. And I saw some of the other guy's measurements where there was nothing at 1k.

How can I differentiate between even order and odd order on this graph?

[Bas Horneman]

If it helps...here is the thd vs output power graph.

[SY]

How can I differentiate between even order and odd order on this graph?

Even-order will be an even multiple of the test frequency (i.e., 2k, 4k, 6k...). Odd order will, unsurprisingly, be the odd multiples (3k, 5k, 7k...).

Anything that's not an integer multiple of the test signal is noise or power supply modulation.

[Gerrit Boers]

Apparently there is a notch filter (in the measurement equipment) at 1k. And I saw some of the other guy's measurements where there was nothing at 1k.

Correct, that's why it says THD+N on the graph. If you remove the fundamental before analysis, what you have left are the harmonic products of the the test signal and the noise of the amplifier under test.
So you have mostly 2nd, 3rd and 4th order distortion. The higher harmonics are 30-40dB below those (good thing). If you're not used to thinking in dB's: -40dB=1%, -60dB=0.01%, -80dB=0.001% etc.

How can I differentiate between even order and odd order on this graph?

Simple, even harmonics are 2KHz, 4KHz, 6KHz etc. Odd harmonics are 3KHz, 5KHz, 7KHz etc, It's all in the name.
Intermodulation products of the 1KHz test signal with the 100Hz ripple are the sum and the difference between the two (1000+100, 1000-100, 1000+200 etc.).

[rdf]

The Audio Precsion appears to have been set to notch out the fundamental test frequency of 1kHz for the harmonic graph. The 1 kHz residual may be related AC mains noise (AC filaments?) The spikes from 2 kHz and up are strong enough to be considered true distortion harmonics. Any harmonic at an even frequency is even (2, 4, 6, 8 kHz...), the rest are odd. That harmonic signature very much suggests to me a driver in distress, probably looking into a load it can't handle. This circuit doesn't use global negative feedback?

edit: eerrr, like everyone else said while I typed.

[7N7]

Hello Bas.

I think that RDF has hit the nail on the head. You mentioned a pentode driver? This means (without seeing the circuit of course) that the output resistance of the driver will be approximately equal to the load which means it would have a tough job driving the capacitance of your output valve.

In your place I would insert a cathode follower or perhaps a transformer.

7N7