How high? For low noise, you can use like 2k. Enough buffering for the source not to impact the notch.
Jan
It means that the notch and its surrounding circuitry is less clean than your source and analyzer. More work there to make it lower noise so the harmonics climb out above it.
How much gain do you have post-notch, and what is providing the gain?
Jan
With 2k you get 2 dB measurement error or more (e.g. with Victor's osc). You never know exactly what the DUT output impedance is. Do you measure every time?
Victor's oscillator has an exactly defined Zout.
But you need to measure the DUT output anyway to make your dB's meaningful ;-)
Jan
Did you use the one chip for the buffer before the notch and the post notch amplifier? Two pieces of the one package?
One dual opamp. Do you suspect crosstalk?
p.s. I will do the measurement with the 10nF cap this evening.
The whole party is meant to measure other DUTs and not just the oscillator and a coax cable. Therfore Zin>>Zout.
Since the line level follower problem is interesting, I made the same setup and got the results.
Also first about the post notch amplifier. Here are measurements without the input follower. 2V RMS at the passive notch input, and the same 40dB post notch amplification:
NE5532:
https://content23-foto.inbox.lv/albums/v/viccc/FollBuf/NE5532noB.jpg
LM4562:
https://content23-foto.inbox.lv/albums/v/viccc/FollBuf/LM4562noB.jpg
OPA2134:
https://content23-foto.inbox.lv/albums/v/viccc/FollBuf/OPA2134noB.jpg
So, the difference is negligible and no distortions are visible.
Now measurements with the follower before the notch.
NE5532:
https://content23-foto.inbox.lv/albums/v/viccc/FollBuf/NE5532.jpg
LM4562:
https://content23-foto.inbox.lv/albums/v/viccc/FollBuf/LM4562.jpg
OPA2134:
https://content23-foto.inbox.lv/albums/v/viccc/FollBuf/OPA2134.jpg
So, the LM4562 is not the best choice for high level signal non inverted follower, but the NE5532 is good in this place, but also is not good enough for ULD measurements.
Also I made the same measurements with the separate post notch buffer and got the similar results. Probably the crosstalk is not the problem.
You need to know what comes out of the DUT. For that you don't have to know its Zout. Zin>>Zout is not necessary for the measurement.
Distortion = fundamental out of the DUT/harmonics out of the DUT.
Jan
Last edited:
Very interesting. Did you or are you going to try the AD797 with the distortion reduction cap?
Jan
I use the AD797 often, because it has very low and very simple distortion (only 2nd and 3rd as far as I can tell).
The circuits I use never have noise gains high enough for the distortion cap to make any difference, so you might not find the need to use one. I think you need to be running well above 20dB of noise gain for it to matter. Of course, you can try it out, but I've never seen any difference with noise gains less than 10dB.
FWIW, my experience with non-inverting stages (followers) has not been good - I cannot find one that can provide distortion below about -140dBc or thereabouts, whereas the same chip as an inverter can often do 10-15dB better. Driving 2kΩ or even 4kΩ is pretty easy for any competent amplifier, and the extra noise of an inverter is often not crippling.
The circuits I use never have noise gains high enough for the distortion cap to make any difference, so you might not find the need to use one. I think you need to be running well above 20dB of noise gain for it to matter. Of course, you can try it out, but I've never seen any difference with noise gains less than 10dB.
FWIW, my experience with non-inverting stages (followers) has not been good - I cannot find one that can provide distortion below about -140dBc or thereabouts, whereas the same chip as an inverter can often do 10-15dB better. Driving 2kΩ or even 4kΩ is pretty easy for any competent amplifier, and the extra noise of an inverter is often not crippling.
Douglas Self has already examined the AD797 in his book and the results were unconvincing. The LM4562 performs much better.
Quantify "much", not that I care at this point in time the difference of 10 or 15 years of process technology is like eons.
I think at 10V, 20kHz and 500 ohm load, the distortion was twice as large.
Wow, horrifying 6dB in all that time. How about an oscillator driving a 10k load?
