Low-distortion Audio-range Oscillator

Jump to Latest
#5,801
Here is some more links relevant to windows and spectrum. (You can think of windows as similar to the bandpass skirts on a swept analyzer).

http://www.keysight.com/upload/cmc_upload/All/6C06DATAACQ_WINDOWS.pdf
https://www.dsprelated.com/showarticle/155.php (Note he says flat top should not be used for signals close to the noise floor which matches my experience)
https://kluedo.ub.uni-kl.de/files/4293/exact_fft_measurements.pdf Good plots of window shapes
Window Functions in Spectrum Analyzers | Tektronix Describes a number of the common windows and their errors
https://ccrma.stanford.edu/~jos/sasp/sasp.html If your in the wrong forum and know a lot more than Ido.
 
#5,802
1audio said:
https://www.dsprelated.com/showarticle/155.php

Ido.
Click to expand...

I like the guy who you linked to there, in the dsp related link.
It's non other than Richard Lyons. He's a helpful thoughtful
guy who will answer questions for you from his books etc.

Just like the good folks that hang out here too. Just good people.

Demian, I couldn't find who Ido was though.

Thanks for the links Lots of good information and variation.
 
#5,803
Hello,

To avoid misleading amplitude of FFT, a possible way is to lock the oscillator frequency
to the multiple of ADC sampling rate (So FFT bins are exactly on oscillator frequency and it's harmonics).
That can be done by taking the ADC clock and passing through a divider to get the desired frequency,
and then sending a very low portion of this in the oscillator loop.
I've made this some time ago on my oscillator with success, without compromising the THD product.
Anyway, i must say that the frequency band where the frequency can be correctly locked is very narrow (few Hz).
I don't have read all the post here, so if i repeat something, apologize me..
Regards.

Frex
 
#5,804
I have worked with injection locking as well. It works across a narrow range and starting with a clean sine wave should not add harmonics. It does bring another connection with its grounds and coupled noise to sort out

Sent from my LG-H811 using Tapatalk
 
#5,805
Frex said:
Hello,

To avoid misleading amplitude of FFT, a possible way is to lock the oscillator frequency
to the multiple of ADC sampling rate (So FFT bins are exactly on oscillator frequency and it's harmonics).
That can be done by taking the ADC clock and passing through a divider to get the desired frequency,
Regards.

Frex
Click to expand...


I think some of the new commercial systems might do this. Do you know, off hand, which analyzers use this method?


THx-RNMarsh
 
#5,811
SyncTronX said:
David, Just do it, take the plunge. You will be happy that you did.
Especially now that you do have the time,

It's good stuff, it's worthwhile and benefits other people
as well as you.

Cheers,

The smiley face guy on the front panel!
Click to expand...

I agree, David. Just go for it. Once you have the design and the figures you want to show in the article, its not really that hard to write it up. There is always some iteration; as you write up the design you may think of a few more measurements to take or even tweak the design a bit due to things you think of while you're writing it up.

Your posts here have been quite articulate, so I think you'll have no difficulty producing a good, informative article.

I'll be happy to give it a proof read for you.

Cheers,
Bob
 
#5,813
Bob Cordell said:
I agree, David. Just go for it. Once you have the design and the figures you want to show in the article, its not really that hard to write it up. There is always some iteration; as you write up the design you may think of a few more measurements to take or even tweak the design a bit due to things you think of while you're writing it up.

Your posts here have been quite articulate, so I think you'll have no difficulty producing a good, informative article.

I'll be happy to give it a proof read for you.

Cheers,
Bob
Click to expand...

Thanks Bob.
 
#5,816
Cordell THD analyzer back-to-back

I just decided to take a close look at my old THD analyzer I designed in the early '80s in back-to-back mode. I ran the analyzer at 1kHz at a level of 1V rms.

I connected the residual output to my HP3580A analog spectrum analyzer, set for center frequencies of 2kHz, 3kHz, 4kHz and 5kHz with resolution bandwidth of 1Hz to see well under the noise. The analyzer still uses 5534A op amps.

The second harmonic was a good value at -124 dB - but not spectacular.

3rd was -150 dB - very good, in my opinion.

4th was -153 dB

5th was -155 dB

With the 3rd that low, I'm a little surprised that the 2nd is quite a bit higher. I always have a gut reaction that 2nd will usually be easier to get way down. This could possibly be due to a JFET AGC element or the AGC control voltage. It is possible that a better tweak of the JFET feedback ratio on one or more of the JFETs (e.g., one in the oscillator and 2 in the analyzer) could reduce the 2kHz residual distortion number. Just speculation at this point. The very low values of the 3rd and higher harmonics could possibly be due to fortuitous harmonic cancellation, but a much more sophisticated setup would be needed to separate this out - perhaps by running the osillator output through an extremely low distortion bandpass filter to make it more pure.

I might try making the measurements at a different analyzer input attenuator setting so as to operate the analyzer internal signal levels at a different value.

Cheers,
Bob
 
#5,819
1audio said:
John curl suggested loading the oscillator with a cap to reduce the harmonics. You should see a change in levels of harmonics at least plus a phase shift. Not conclusive but an indicator.

Sent from my LG-H811 using Tapatalk
Click to expand...

This is a good point, and I have thought of doing that.

I did just now play with some different operating levels in the analyzer and at the output of the oscillator, looking closely at the resulting changes in 2H and 3H. Two things I am convinced of with these experiments.

First, that the 2H residual distortion is occurring in the analyzer, not the oscillator or its output amplifier.

Secondly, the various changes in operating levels did not change the very small level of 3H. This suggests that the low value of 3H is not the result of harmonic cancellation. It also possibly suggests that the small value of 3H comes from the oscillator, but this is fairly speculative at this point. I think the idea of putting a capacitor across the 600-ohm output of the oscillator would help further figure this part out.

BTW, THD+N in these tests was about 0.0004% (-108 dB).

Cheers,
Bob
 
You must log in or register to reply here.
Top Bottom