Low distortion oscillator?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

rjm

Member
Joined 2004
Paid Member
There was an application note I read once that dealt with measuring the distortion of op amps, possibly it was part of a datasheet. Linear Tech or Burr Brown most likely. Anyone remember it? I can't find it anymore.

My existing tools are fine for frequency sweeps, but none of my function generators will not cut it when trying to determine distortion below 0.01%.

I'm thinking I just need a fixed frequency oscillator circuit for around 1kHz and around 0dB (1 V rms) output. For what I intend to do that would work fine ... but I'm unsure about the best approach.

Advice appreciated.
 
One of the most interesting articles is in Jim Williams book Analog Circuit Design: Art, Science and Personalities. If you go to Google Books, the chapter on Hewlett Packards oscillator and JW's efforts to do a solid state version is available to read, or at least most of it. I think it starts around page 50.

There's also a similar application note, AN-43 that covers a lot. That should be available on the LT site.

I'd also read about the Cordell distortion analyzer, as the signal generator portion of that is extremely good. You'll find that on Bob Cordell's site.
 

rjm

Member
Joined 2004
Paid Member
My soundcard has a THD of about 0.003% at 2496 - round trip, output looped to input.

I found the datasheet I was thinking of earlier about low distortion sinewave generator: it was the Linear Tech LT1007/1037. (Thanks Frex for jogging my memory.)

LT1007 - Low Noise, High Speed Precision Operational Amplifiers - Linear Technology

So the question is it realistic to imagine I might be able to better my soundcard sinewave distortion by about 10dB using the osicallator circuit? Anything less and it wouldn't be worth the trouble...

(..OK datasheet says "parts per billion" - but real world performance?)
 

Attachments

  • Onkyo SE-200PCI 2496 loop 110501 RJM Audio.png
    Onkyo SE-200PCI 2496 loop 110501 RJM Audio.png
    4.4 KB · Views: 675
The LT AN67 oscillator perform better THD performance of any sound card available. At my knowledge, no commercial hardware are capable to reach this level of THD at this frequency .
Remember that it's frequency is 10kHz, not at 1kHz as you find in all THD tests.


Frex
 
The LT AN67 oscillator perform better THD performance of any sound card available. At my knowledge, no commercial hardware are capable to reach this level of THD at this frequency .
Remember that it's frequency is 10kHz, not at 1kHz as you find in all THD tests.


Frex

I sit by the mailbox awaiting Jim Williams new design of which we got a tickle last week courtesy of EDN. I couldn't measure any distortion in the AN-67 oscillator I built, but I now have somewhat better instrumentation.

10kHz or 20kHz is probably much more useful than 1kHz -- at least according to the estimable Mr. Cordell, so the AN-67 design, tweaked to 20kHz by changing cap values is "vaut le voyage"

The Krohn-Hite 4400 and Tektronix SG-505 are excellent LD oscillators.
 
Jackinnj,

I'm waiting too the next EDN issue, i'm very curious to read it !
And, the french expression "vaut le voyage" is it fluently use in USA ? ;-)
It's your french touch !


Conrad,

First, i don't know how you can see THD with an oscilloscope, except if it very high !
What do you see exactly ? Maybe you could post screenshoots/pictures...
Regards.

Frex
 

rjm

Member
Joined 2004
Paid Member
Soundcards should perform to their specification, and as you can see in the FFT I posted above are pretty decent.

Perhaps your test methodology was at fault: running into 600 ohm or 50 ohm inputs, or strange ground loop developing?

@Frex

Nice work. Been reading the threads, interesting.
 
Jackinnj,
I'm waiting too the next EDN issue, i'm very curious to read it !
And, the french expression "vaut le voyage" is it fluently use in USA ? ;-)
It's your french touch !

French was the preferred "foreign language" studied in the US until the 1980s, and it was the one course that undocumented, imported collegiate hockey players had no difficulty with.
 
distortion at harmonic frequencies - nearly inaudible, IMD - highly audible

there's no need for a super low distortion oscillator in audio amp testing - THD is pretty useless by itself

while Cordell's 20KHz THD is a more useful metric than 1KHZ it is less useful than IMD tests which can be done at very high resolution with moderate performance oscillators/signal sources

if you really want to use distortion measurements to improve your amplifier you should want several socillators to look for Intemodulation prodcuts caused by amplifier nonlinearities

by resistively summing 10KHz + 11KHz from separate souncard channels and amplifying the amplifier's input error voltage I have resolved down to -160 dB at the 1KHz 2nd order difference frequency

a op amp summing circuit and a few fixed frequency low distortion filters to attenuate excitation frequencies and pick small IMD difference frequencies out of multi-tone excitation signals, sweeps will give resolution well below the audible noise floor with just soundcard or universal player for multitone source and moderate performance soundcard ADC

you will need some filter in any event - you have to knock down the excitation frequency even with a perfect single tone source if you don't want to be limited by your soundcard's ADC nonlinearity/input

another technique that saves on filters: The EDN designer's companion - Google Books
 
Last edited:
re: noise level on my Emu- It certainly has low THD, but the problem is broadband noise and maybe some PC related clock junk. I'd expect it to be better, being an external box, but the output looks terrible. The input, when recording from good mics, is clean as can be. I'll hook this up again sometime soon and see about specific numbers and pictures.

When I want a clean signal I use my Cordell analyzer, or for higher distortion but no HF peaks, an HP204. Otherwise I use a Wavetec, but because that derives the sine from a triangle, it does have tiny peaks top and bottom.
 
I can't Graeme's "selective amplification" technique in the 2nd part of the article online - but the basic idea is to float the amp so that you can measure the diff input error signal directly between the top of the signal souce and the inverting amp input with a single ended ADC input
 
Last edited:
[--]I was thinking of earlier about low distortion sinewave generator: it was the Linear Tech LT1007/1037.

LT1007 - Low Noise, High Speed Precision Operational Amplifiers - Linear Technology

(..OK datasheet says "parts per billion" - but real world performance?)

Hi,
I made this little oscillator, you need:
  1. a very slow bulb, long time constant, that is 1 watt & high voltage, rather 110 volt than 50 volt; I have 35 volt (best I had)
  2. make the fixed resistor variable (the 430 ohms): you nee to tweek it a bit to get lowest distortion and/or have the oscillation start at al: this depends on bulb.
  3. I added a low pass filter with a second opamp, to reduce the harmonics and act as buffer
  4. good practice is to force the opamp in class A with a 3-5 kohm reistance from out to minus.

With that I got less than 0,02 % distortion from the oscillator, I do have an analogue meter that then barely moves two lines. I could improve maybe but thought this OK.
It took me about 2 hours work to have it running.
albert
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.