Need to measure either dissipation factor or Q at 1 MHz

[Freddybaby]

I am evaluating some RF capacitors and need a reliable and accurate method for Q's of 5,000 - 10,000. My LCR meter gives pretty unreliable results at these values and limited to 100KHz.

These caps are used at about 15KV, 1 MHz and a Df of .001 versus .0001 means a destroyed cap or a mini radiant heater...

 

[FvM]

State-of-the-art LCR meters, e.g. Agilent E4980A or Qualtech 4600 can measure at 1 MHz. The D respectively Q accuracy can be expected to depend also on the capacitance value.

 

[Freddybaby]

State-of-the-art LCR meters, e.g. Agilent E4980A or Qualtech 4600 can measure at 1 MHz. The D respectively Q accuracy can be expected to depend also on the capacitance value.

I have a Stanford research SR720, it bounces around about .0001 - .0006 and based on my RF heating tests at 6 and 9 KV pk-pk @ 1MHz there is a pretty narrow window of acceptance and this is in the error range. We don't want to purchase yet another high end LCR but rather a "simple" and reliable sorting method.

 

[FvM]

We don't want to purchase yet another high end LCR but rather a "simple" and reliable sorting method.

By buying a high end instrument, you pay also for the experience of it's designers (hopefully you get the value). To setup a "simple and reliable" measuring method on your own, you have to contribute your own engineering knowledge, or pay an expert to design it.

I don't see a simple alternative to build a resonant circuit with a stable, low loss air coil and calibrate it with a low loss capacitor (e.g. mica dielectricum). So far, the advice is for free. :smile:

 

[Freddybaby]

I do see your point and thanks for the free advice... My 20KV, 1MHz oscillator is in a sense "simple and reliable", just very time consuming waiting for the cap to heat up or not. The Df measurements don't seem to agree with my heating test in that several measure very low Df but still heat up and some do not.

I'm guessing the Df is also changing with temperature and probably not the same for each capacitor.. We are looking at some wideband impedance analyzers but the cost is ridiculous, and the frequency ranges are not ideal. Looks like the calorimetric method will have to do for now... Thanks again

 

[FvM]

My 20KV, 1MHz oscillator is in a sense "simple and reliable", just very time consuming waiting for the cap to heat up or not.

Yes, if you can utilize it for the test, it would be preferable.

I'm guessing the Df is also changing with temperature and probably not the same for each capacitor.

You didn't tell about the involved capacitor type. If it's HV ceramic, it possibly has also a voltage dependant loss factor. Ceramic dielectricum would also be my favourite candidate for a changing loss factor between production lots.

 

[Freddybaby]

Yes, if you can utilize it for the test, it would be preferable.

You didn't tell about the involved capacitor type. If it's HV ceramic, it possibly has also a voltage dependant loss factor. Ceramic dielectricum would also be my favourite candidate for a changing loss factor between production lots.

Yes these are multilayer 15KVDC rated parts but no AC current ratings. I did not know that the loss may be dependent on voltage. Is this DC bias or both AC/DC ? I have witnessed the cap's "runaway" though, as they heat up the loss goes up and pretty quickly I'm looking at a cracked part..

This particular test is very time consuming though, and we might go through 50-1000 caps in a month or so.

 

[FvM]

Basically all parameters of a ceramic capacitor are voltage dependant, because the dielectrica has a non-linear behaviour, particularly with high Er types, e.g. X7R. The question is to which extent. For MLCC, I would also expect a certain lot variation of the ESR, introduced by the firing process. Finally, excessive loss of assembled SMD capacitors may also indicate a problem of the soldering process.

 

[Freddybaby]

Excellent information.. BTW, Polyflon corp makes great Teflon HV caps, up to 50pF @ 30KV and 27.5A RMS current rating but they are huge, like a hockey puck.