Following this thread got me inspired to finish building the ultrapure oscillator from AN67. I used Frex's board, powered by an old Hp 6205B power supply (eg, nothing special for the power supply). Took a bit to get it up and running, but the results are quite good. With the output at approx 1.5 Vrms I measured the 2nd H at approx. -126 dB. Tested it with Dick's Twin-T, nulled the fundamental to about -46db. The pic from left to right are the board in the enclosure, full scale measurement, and a zoomed in measurement (on this one ignore the values in the boxes as they aren't appropriate for a zoom in like this - have to re-write the VEE and am to lazy).
Ken
Ken
Attachments
Hi Dick,
I probably mis-stated the fundamental, it's at 10.04kHz, so the second is at 20.08khz.
Ken
Last edited:
Sorry guys, 
I went back this morning and looked at where the 10.4k came from. It's left over in the VEE programing from when I used to use my HP8903A as a source. I changed the programing this morning to read the left meter on the HP8903A to the output fundamental box. So, ignore the fundamental output box and know that the fundamental was at 10.040kHz.
I had previously changed the VEE to read the right meter to report the Vrms, but had forgotten to change it to read the left meter as well.
Ken
So, if you believe my numbers, the AN67 oscillator is quite good yes? But, from the looks of it Victor's is the better deal...
I went back this morning and looked at where the 10.4k came from. It's left over in the VEE programing from when I used to use my HP8903A as a source. I changed the programing this morning to read the left meter on the HP8903A to the output fundamental box. So, ignore the fundamental output box and know that the fundamental was at 10.040kHz. I had previously changed the VEE to read the right meter to report the Vrms, but had forgotten to change it to read the left meter as well.
Ken
So, if you believe my numbers, the AN67 oscillator is quite good
yes? But, from the looks of it Victor's is the better deal...
@Ken -- The results from the AN-67 unit are hard to understand given the general belief that this is a truly great oscillator with THD in ppb. There are many possible sources of harmonic intrusion that are not intrinsic to the oscillator.
I couldn't see any recognizable higher harmonics in the noise, but an expanded view may be warranted. Compounding the issue is that we really don't know what the residual of the Active Twin-T is -- I have yet to find a way to get at that, and was hoping the ppb oscillator would be helpful.
I couldn't see any recognizable higher harmonics in the noise, but an expanded view may be warranted. Compounding the issue is that we really don't know what the residual of the Active Twin-T is -- I have yet to find a way to get at that, and was hoping the ppb oscillator would be helpful.
Hi Dick,
I'm not running it on batteries, that might help, but, who knows when I will get to that.
kind of a catch 22 - as you say, maybe the 2ndH we're seeing is a contribution of the active twin-t...
I was thinking this morning that I should try to "zoom - in" on each of the the harmonics, seems the hp3562 resolution gets better when the fundamental (null of course) is not in the view. Is that what you would like to see?
For me, it's certainly clean enough...
Ken
I forgot to mention, Frex did a really thoughtful job of layout and enhancement to the schematic. Such things as parallel resistors that will allow fine tuning of the frequency, fixed resistors around some of the trimmers to allow finer trim. Good job Frex!
Ken
p.s. If anyone does intend to build one of these, let me know and I can post a bit more elaborate adjust proceedure. It took a couple of hours to get it running consistently.
Ken
p.s. If anyone does intend to build one of these, let me know and I can post a bit more elaborate adjust proceedure. It took a couple of hours to get it running consistently.
Thanks for the offer to post your adjustment proceedure. I'm ordering two boards form Frex today.
A couple of notes on parts, I found Farnell Uk carries polystyrene caps in tighter tolerances. So the 10,000pF, 2.2nF and 47pF all came from there. The other polystyrene caps I ordered tripple the quantity from mouser and hand selected with a LC meter. But you might just want to order all at 1% tolerance from Farnell.
Ken
More testing of the AN67 tonight - I zoomed in a bit more on the target harmonic frequencies. Other than 2nd H all were burried in the noise floor of the signal analyzer. The frequency rose slightly through the test proceedure from 10,032 to 10,036. It was a bit higher yesterday, and maybe a bit warmer in my test room.
Oh and the test frequency is now the reading taken from the hp8903a via HPIB.
Ken
Oh and the test frequency is now the reading taken from the hp8903a via HPIB.
Ken
Attachments
KH4400 Measurements
This is more of a demo of too much resolution. It the KH4400 adjusted per spec. but only modified with the injection locking scheme I mentioned. I locked it to the Boonton's PLL stabilized 1 KHz output. I'm injecting 1V at 1 KHz but I could see no difference in distortion with and without.
The measurement is a 16,000,000 point fft at 192KHz/24 bit (with the AK5394a demo board for capture) If it was not locked I could not get the measurement, the frequency would drift too much for this in the 20 minutes or so it took to make the measurement. I will do another pass with a different windowing function. The Key differences I see is the shape at the base of the measurement with different windows.
I don't think 16M points offers much beyond 512K unless you need to see something at the noise floor.
This is more of a demo of too much resolution. It the KH4400 adjusted per spec. but only modified with the injection locking scheme I mentioned. I locked it to the Boonton's PLL stabilized 1 KHz output. I'm injecting 1V at 1 KHz but I could see no difference in distortion with and without.
The measurement is a 16,000,000 point fft at 192KHz/24 bit (with the AK5394a demo board for capture) If it was not locked I could not get the measurement, the frequency would drift too much for this in the 20 minutes or so it took to make the measurement. I will do another pass with a different windowing function. The Key differences I see is the shape at the base of the measurement with different windows.
I don't think 16M points offers much beyond 512K unless you need to see something at the noise floor.
Attachments
DAC based low thd oscillator.
Yes, possibly. Or a really good 24 bit could do really well with a tracking filter. Or, 1 bit. It would have the advantage of any freq and fast freq changes and multiple tones. --RNM
Yes, possibly. Or a really good 24 bit could do really well with a tracking filter. Or, 1 bit. It would have the advantage of any freq and fast freq changes and multiple tones. --RNM
A DAC as a source for ultra low distortion will be elusive still. Several limitations make it hard to do. First the max sample rate (192 KHz in practice) second the linearity is not that good, well below the performance of these oscillators, with only the ES9012 coming close, and finally keeping the digital junk away from the analog output (plus isolation issues) is a very difficult thing to do.
I think for now analog source and digital capture(enhanced with analog front end magic) is as good as is possible.
I think for now analog source and digital capture(enhanced with analog front end magic) is as good as is possible.