And Polypropylene caps have a significant tempco. In "the old days" you could get resistors with the matching tempco to compensate for the thermal drift. That is all a lost art. The best compromise seems to be Polystyrene with a significantly better tempco which are harder to come by. In manufacturing for film type caps the machines seem to be limited to 2% or so accuracy and you need to select to get any better. They well may be more uniform but not the value you want.
To get the best null on the notch you will need to trim or adjust the resistors and match the caps pretty precisely. If making a number of these provide for trim caps to pad to the target value and trimmer pots for the resistors. Select caps and resistors a few percent smaller than the target and adjust with parallel caps (an assortment of 10% parts would give the necessary range) and the trim pots for the resistors.
I really appreciate Victor's patience in component selection for his oscillators. They are built with the attention to detail of a primary standard. And way beyond anything I have the patience to do.
The other approach is to just attenuate the fundamental 20 or 40 dB and let a more conventional analyzer do the rest. 40 dB fundamental attenuation plus a -100 dB analyzer will get you into the SOTA range. A degraded notch with some simple tuning will get there and the autotuning of the analyzer does the rest.
Keep in mind a passive notch will limit its usefulness if you are looking for impedance interactions and distortion. I use the Shibasoku on the input of amp circuits to see the impact of the input nonlinearity. Its high impedance and low cap modulation makes this easy and you see lots of unexpected results.Looking at the output makes it harder to separate the effects.
To get the best null on the notch you will need to trim or adjust the resistors and match the caps pretty precisely. If making a number of these provide for trim caps to pad to the target value and trimmer pots for the resistors. Select caps and resistors a few percent smaller than the target and adjust with parallel caps (an assortment of 10% parts would give the necessary range) and the trim pots for the resistors.
I really appreciate Victor's patience in component selection for his oscillators. They are built with the attention to detail of a primary standard. And way beyond anything I have the patience to do.
The other approach is to just attenuate the fundamental 20 or 40 dB and let a more conventional analyzer do the rest. 40 dB fundamental attenuation plus a -100 dB analyzer will get you into the SOTA range. A degraded notch with some simple tuning will get there and the autotuning of the analyzer does the rest.
Keep in mind a passive notch will limit its usefulness if you are looking for impedance interactions and distortion. I use the Shibasoku on the input of amp circuits to see the impact of the input nonlinearity. Its high impedance and low cap modulation makes this easy and you see lots of unexpected results.Looking at the output makes it harder to separate the effects.
I have just designed a small temperature controlled 'oven' for 4 PP caps for a filter. The temp. stability of 0.001degC (or better, this is how deep I can look) reduces the caps 250ppm/C to an equivalent 0.25ppm/C. No longer a factor.
Jan
You can get Vishay precision resistors with 0.2ppm/C tempco. So that is in the same order as the ovenized caps, for a freq. drift of say < 0.5ppm/C of the whole assembly.
Jan.
Jan you could frequency lock the oven control. If you have a universal counter you could use the output from that for super fine frequency locking.
Thermal control is slow by comparison to other control. Don't need to sample often. Even a pic could achieve the resolution over time.
Just a thought.
Thermal control is slow by comparison to other control. Don't need to sample often. Even a pic could achieve the resolution over time.
Just a thought.
Like 0.00009% v. 0.00006% you mean? They are essentially linear for all practical purposes.
Just FYI, that is a typical value, and typical is not typical--if you refer to S102K series. If you want 0.2ppm you need to measure and select. That is not easy at the 0.2ppm level. I'm putting four S102k together randomly and measuring 0.3-0.5ppm. Of course, if you put your resistors in the oven, as well, that will probably do it, regardless.
For dependable tempco characteristics, Susumu URG series are good.
My interest is in metrology, btw.
C0G are also much more consistently manufactured than film caps.
I also think it matters which have been tested, and maybe also the size and voltage rating. Newer C0G caps have BME with Calcium Zirconate dielectric. Older PME types may perform differently.
I'm tempted to ask about frequency, current, etc. but I might as well come up with an experimental method myself and write up the results.
I was looking at the Vishay Zfoil resistors:
https://www.mouser.be/datasheet/2/428/zseries-1022194.pdf
Jan
Yes capacitance, Tan delta, leakage current and distortion, other than that little effect!
Take a few and run them at 1,000 hertz from an audio power amplifier. Do keep the total capacitance below 20 uF and the voltage about 1/2 rated.
Last edited:
What type of capacitors you have tested in this way? Metallized film, film/foil or other? What kind of dielectric?Yes capacitance, Tan delta, leakage current and distortion, other than that little effect!
Take a few and run them at 1,000 hertz from an audio power amplifier. Do keep the total capacitance below 20 uF and the voltage about 1/2 rated.
I can make some guesses. Putting a voltage on a capacitor causes the plates to attract each other, so the cap will tend to contract. I would imagine putting AC on a cap as described could make the thing generate audio (and with no DC bias I would imagine it would have with a rather strong second harmonic - a low efficiency but operational electrostatic speaker!). I'd limit the current (and thus frequency) to something that doesn't generate significant heat. For voltage I'd go with a peak value of 90 percent of rated voltage. I'd measure parameters before doing any "breaking in." then after a few seconds, minutes, hours, days and weeks. This will give a good idea how much things will change over time.
Long term effect might be permanent "compaction" and thus slightly higher capacitance. It's hard to imagine how other parameters might change, but if you push a capacitor near its limits doing this type of "break in," the parameters will change less during the capacitor's (remaining) lifetime.