Capacitive load impact on distortion

Jump to Latest
#21
jan.didden said:
Matt, thanks for that. But my purpose is a tracking notch that follows the oscillator as it wanders or drifts slightly. Say total tracking range of 6 to 10 Hz. For a nominal 1kHz oscillator that means it could vary between say 995 and 1005Hz and the notch, at nominally -40dB, will follow it and keep it centered. The goal, as you noted, is to decrease the ADC dynamic range requirements and thus its distortion.

A flat notch with a similar width would be nice but the ones I found were always 4 or more staggered notches and pretty complex. The tracking is not that hard to do, but the issue is the distortion from the tuning system. It is much easier though than the AGC of an oscillator because the control speed can be very slow (oscillators typically drift over 10's of seconds or even minutes) and there is almost total absence of oscillator signal in the control signal. The remaining non-linearity is the inherent non-linearity of a tuneable impedance.

Jan
Click to expand...

Hi Jan,

An alternative to the tunable Bainter notch would be a state variable filter notch that can be tuned a few Hz one way or the other without degrading the notch. This could be implemented by putting a variable resistance JFET in series with each integrator resistor, with its source at the virtual ground of the integrator op amp.

This JFET can still benefit from the use of 1/2 of the drain signal as feedback to its gate. In an application where you only need to move a few Hz, the voltage that needs to be across the JFET will be quite small, so very little distortion will be introduced.

Note that the gain of both integrators must be changed so that the gains of the integrators remain the same; else the notch will be degraded. Although the Bainter can in principle provide a deeper notch without tweaking, if you do not need a super deep notch, the SVF notch may work well for you.

Cheers,
Bob
 
#22
jan.didden said:
The author of that article is a member here and has already posted in this thread ;-)

I believe those resistors are for stability reasons. I have used my GP opamp implementations with wildly unequal input resistances and they were clean down to -160dBc.
There is something else going on here.

I will be away a few days but intend to test the cap to ground.

Jan
Click to expand...
I was not sure, if a person is real. Would you think, an author should instantly recognize his own creature? And if it didn't happen?
Regarding input matching, I'd refer to OP275 DS
https://www.analog.com/media/en/technical-documentation/data-sheets/OP275.pdf
Attention to Source Impedances Minimizes Distortion
Since the OP275 is a very low distortion amplifier, careful attention should be given to source impedances seen by both inputs.
As with many FET-type amplifiers, the p-channel JFETs in the
OP275’s input stage exhibit a gate-to-source capacitance that varies with the applied input voltage. In an inverting configuration,
the inverting input is held at a virtual ground and, as such, does
not vary with input voltage. Thus, since the gate-to-source voltage
is constant, there is no distortion due to input capacitance modulation. In noninverting applications, however, the gate-to-source
voltage is not constant. The resulting capacitance modulation
can cause distortion above 1 kHz if the input impedance is
greater than 2 k and unbalanced.
Click to expand...
Know, they are talking JFET, nevertheless BJT has similar C-B capacitance, and it's easier to cancel a variable in THD equation by matching impedance. Certainly, could be more than one reason for high THD level.
 
#27
scott wurcer said:
That would de-Qing the capacitor which should show up in a simulation.
Click to expand...

It does. Those 20R resistors decrease the notch dept from over 60dB to about 40dB. And it lowers the signal voltage at the output of the integrator by several dB. That could explain the difference in distortion, more than the 20R decoupling the capacitive load.

I was kind of stuck in this rut, I think I should stop worrying about that cap load and move on with the design.

That resistor R1 that I use for tuning is actually a trimmer on my protoboard. I will replace it with a fixed metal film to see if that makes a difference, but that will have to wait a few days as I am at the beach taking a Corona-break ;-)

Jan
 
#30
#32
I found a thread here about the Hall notch, where halfway someone promoted the Bainter over the Hall as much easier tuned. Go figure ;-)

Bainter with flat 2nd - 3rd response.

Jan
 

Attachments

  • flat2nd3rd.PNG
    flat2nd3rd.PNG
    29.6 KB · Views: 203
Last edited:
#34
jan.didden said:
but the issue is the distortion from the tuning system.
Click to expand...

You might think about the the CdS + LED (NSL-32S42) that Jim Williams used in his 2011 article--it works really well. The voltage across the CdS is very small (10-20 mV), the distortions of the oscillator are well under -140 dBC. Plus, the control is simple and 100% isolated--a DAC and opamp with 10-20 mA of drive capability could do the trick.
 
#36
Re -135 HD. We need to put this into perspective to see what could cause it. -135 dB is around 300 parts per billion, a very small number. for a 32 nF cap 1 ppm would be 32 fF so this is around 10 fF equivalent change in capacitance somewhere. Its very conceivable that the input C is modulating that much with even the small voltage at the inverting input. The input C change varies a lot in different amp circuits. A cascoded circuit will have less modulation. It usually shows as odd order (if I have it right) since its a single ended phenomena, except on complementary inputs where it could be symmetrical.


When you are looking at nonlinearity this low everything will show up. PCB, solder mask flux residue and component issues can all show up. Lower impedance helps until current effects show up. A classic demo with the CLT-1 is to take a pencil lead to a resistor lead on a resistor under test. it increases the distortion as the lead interacts with the magnetic field around the wire.
 
#37
1audio said:
Re -135 HD. We need to put this into perspective to see what could cause it. -135 dB is around 300 parts per billion, a very small number. for a 32 nF cap 1 ppm would be 32 fF so this is around 10 fF equivalent change in capacitance somewhere. Its very conceivable that the input C is modulating that much with even the small voltage at the inverting input. The input C change varies a lot in different amp circuits.
Click to expand...

An input charge error should show up in a simple resistive inverter circuit, the capacitor only shifts the phase 90 degrees. The composite amp IIRC has enormous AOL at 1kHz, I can only extrapolate from the paper but there should be <micro-Volts at the input.
 
#38
Yes. Samuel's best guess is that the distortion at 1kHz is about -200dB (his measurement limit was -180dB). The OL gain is gigantic at 1kHz. Microvolts at most as differential input signal. It find it very hard to imagine that this would modulate the input capacitance.

It will be some days before I am back home and can experiment again, but my suspicion now is the trimmer that is in R1 position.

Jan
 
#39
Last edited:
You must log in or register to reply here.
Top Bottom