A few years ago, I did many mods and tests to several THD instruments. The best I found is the ShibaSoku 725 "D" model. accurate to a fraction of a db down to -160dBv. A true laboratory standard instrument.
In so testing, i used Victors and highly modified KH and Davada's and ShibaSoku's generator and others. Including AP 2722.
Right now it is the generator which limits me.... measuring into the -140's.
JD -- the tracking notch is a great asset. Combined with Davada's variable freq gen makes a nice variable freq T & M system for Ultra low distortion studies. .
THx-RNMarsh
In so testing, i used Victors and highly modified KH and Davada's and ShibaSoku's generator and others. Including AP 2722.
Right now it is the generator which limits me.... measuring into the -140's.
JD -- the tracking notch is a great asset. Combined with Davada's variable freq gen makes a nice variable freq T & M system for Ultra low distortion studies. .
THx-RNMarsh
Last edited:
I just did a quick test with my Victor oscillator cold. 50 X 128 during the biggest drift. No signs of issues with it tracking. This is with the Lynx.
Different install for Windows. The other one crashed. I'll know if this survives the latest update soon.
Attachments
Maximum output level means 3-4V? The LM4562 has the lowest value at this point. It is also advisable not to set the level by means of AGC, but with a voltage divider at the output.
Markus which PCBs? You man the tracking notch? That's not finalized yet, I hope to be able to wring a few more dB out of it.
Also the comments given so far give me some more ideas ... ;-)
Jan
Thanks Monte, that's a good point. The max FFT sample rate of my AP is 32k, but I can set the # of samples up to 4M, which is what I used. There is a small improvement when going from say 64k samples to 4M but probably not enough to warrant waiting 20 minutes for the results.
I will try with more averages, that's faster than more samples. I can use synchronous sampling when the AP is the source but not with the Viktor, of course. That may also explain some of the difference.
BTW I used the Flat Top window.
Jan
I have suggested that to Viktor a few times. Maybe he needs some more encouragement
Jan
Monte's referenced article can be found here:
https://www.edn.com/design/analog/4400342/Injection-lock-a-Wien-bridge-oscillator
Nice article.
Jan
https://www.edn.com/design/analog/4400342/Injection-lock-a-Wien-bridge-oscillator
Nice article.
Jan
Last edited:
In my distortion compensator I fortunately do not have to fight stable frequency (same clock for DAC and ADC), but stable level for calibration is important. It takes the soundcard (a plastic enclosure sprayed inside with conductive paint) about 20 minutes to warm up to stable output. My table is next to window and it makes a difference if the window is covered with curtain in the evening - just like your random air flow experience.
Also the capacitor in the LPF discriminator takes about 10 minutes to charge up (or warm up?) until the LPF output level stabilizes down to 0.001dB/minute. Then the capacitor is quite stable even if disconnected from the signal for a while.
One thing I haven't tried yet.
Normally, with the AP2722, the FFT bandpass tracks the generator or the sweep frequency (with the sweep driving the generator).
But for external sources there is also the option to have the FFT track the external source frequency by setting the bandpass to 'Counter tuned'. I'll see if that makes a difference.
Jan
Normally, with the AP2722, the FFT bandpass tracks the generator or the sweep frequency (with the sweep driving the generator).
But for external sources there is also the option to have the FFT track the external source frequency by setting the bandpass to 'Counter tuned'. I'll see if that makes a difference.
Jan
Last edited:
Yes, some features I would like to have/see
- balanced operation (may 2 circuits of them) & none balanced mode
- target tracking notch frequency range 0.1kHz .. 20kHz ??
- Switch to shorting the notch for base signal calibration purpose
(shorting and not a relays switching, while the relay contributes with THD)
- signal operation range 0.1V..10V ??
- providing noise & THD figures of the tracking notch
- notch Q without 2. Harmonic attenuation's
while currently have a 1kHz balanced bootstrapped T-Notch with high noise contribution, while using 10nF & 16k resistors.
Attn Victor:
Additional the victor may improved using lower resistor values but may increase THD's. The deal may is to lower the resistors (may half) and double the cap's and using single opamp's of a dual, due the opamp chip cross talk.
my 2 cents
Hp
Yes, some features I would like to have/see
- balanced operation (may 2 circuits of them) & none balanced mode
- target tracking notch frequency range 0.1kHz .. 20kHz ??
- Switch to shorting the notch for base signal calibration purpose
(shorting and not a relays switching, while the relay contributes with THD)
- signal operation range 0.1V..10V ??
- providing noise & THD figures of the tracking notch
- notch Q without 2. Harmonic attenuation's
Hp
Well, that's not going to happen - at least not from me.
Currently the notch tracks 1kHz +/-5Hz. The tracking input must be very non-intruding so as not to introduce distortion. It is the exact same reason why Viktor's oscillator can not be tuned over a wide band.
Shorting and balanced operation is possible, and operating range is limited by noise at the bottom and distortion at the top. But there's always my Autoranger ;-)
For balanced operation I would propose an balanced input and a single-ended output.
BTW, I was just offered a Keysight Technologies N5225A Network/Spectrum analyzer, for less than half price! Only $ 135,000. Jeez.
Jan
Last edited:
Well, the point for full balanced operation..
OK, may needs 2 of them while even good "sound cards" have only a balanced input as Lynx L22, RME ADI 2 Pro. Otherwise in none balanced mode, we deal also with hum contribution to the hole setup.
my 2 cents
Hp
Don't forget you are now operating at about -100dB and balanced will not have such an impact if any. Having two identical notches in parallel is another can of worms I think. But I will think about it.
The graphs I posted yesterday are se in to se out. BNC, no RCA, no ground loops due to SilentSwitchers ;-)
Jan
Does the second line need to be notched too? How about notching just the hot line and connecting the cold line directly to DUT's output ground (fake unbalanced -> balanced connection)? IME it helps notably with suppressing the measurement ground loop noise of an unbalanced DUT.
Even fetching the ground to another unbalanced ADC channel and subtracting the two channels in software (poor man's fake balanced input) helps Virtual balanced in/out from regular soundcard in linux - results but gentlemen in this thread are not poor and have proper balanced inputs.
Even fetching the ground to another unbalanced ADC channel and subtracting the two channels in software (poor man's fake balanced input) helps Virtual balanced in/out from regular soundcard in linux - results but gentlemen in this thread are not poor and have proper balanced inputs.
