Passive Notch Filter For Thd Measurement

[fotios]

Hi to all

I would like to build a notch filter - with passive parts exclusivelly - for THD measurements with my DSO. The reason is that i don't trust the multi-instruments which based on the sound card of PC for accurate measurements. Of what order must be that filter so it can reduce in appropriate level the fundamental frequency to remain in its output only the part of harmonics produced? What is the type for calculating the values of parts of that filter for different notch frequencies?

Thanks for any help
Fotios

[Christer]

Rod Elliott has such a project on his site:
http://sound.westhost.com/project52.htm

[fotios]

Originally posted by Christer
Rod Elliott has such a project on his site:
http://sound.westhost.com/project52.htm

Hi Christer and thanks for the link which i knew already.
I have built for several years a notch filter with active parts - for use with microphones - of much higher quality than that proposed by ESP in his website. My question I think that was clear, but I place it again. I'm interested only for notch filters with passive parts exclusivelly.
Any other suggestion please?

Fotios

[analog_sa]

Originally posted by fotios


The reason is that i don't trust the multi-instruments which based on the sound card of PC for accurate measurements.

Are you seriously expecting a passive filter to deliver more accurate thd measurements? How steep does it need to be? How many components would you expect it to have for a single frequency?

[tschrama]

You can use the " parrallel -T " filter, that Rod Eliot uses, and than leave out the active components...

http://sound.westhost.com/project52.htm


I do that when I want extra resolution for THD measurements below -100dB.

few things :
1] the filter will attenuated 2nd Harmonic 10dB and 3rd Harmonic 5dB, but you can correct for that.
2] the output of the filter should be sampled with a high impedance equipment, of put a buffer after it..
3] use low value resistors to keep noise low.. somthing like 1K5 and 100nF..
4] use polypropylene capacitors to keep distortion down.. those tiny blue ceramics are horrible..

[fotios]

Originally posted by analog_sa



Are you seriously expecting a passive filter to deliver more accurate thd measurements? How steep does it need to be? How many components would you expect it to have for a single frequency?

I am seriously interested because my new project - a power amplifier - has lower distortion from my usb sound card EMU-0404. I have realy problem to estimate the distortion of my amp. And the worst case it is that i have the same problem for measuring also the TIM distortion!
How steep must be the filter? So much as it not attenuate the 2nd harmonic. I am not inderested for how many compos needed for building the filter. Also if they are resistors or inductors or caps. As for the frequency i am interested mainly for 1KHz, i thing only this is enough to the present.

Thanks
Fotios

[fotios]

Originally posted by tschrama
You can use the " parrallel -T " filter, that Rod Eliot uses, and than leave out the active components...

http://sound.westhost.com/project52.htm


I do that when I want extra resolution for THD measurements below -100dB.

few things :
1] the filter will attenuated 2nd Harmonic 10dB and 3rd Harmonic 5dB, but you can correct for that.
2] the output of the filter should be sampled with a high impedance equipment, of put a buffer after it..
3] use low value resistors to keep noise low.. somthing like 1K5 and 100nF..
4] use polypropylene capacitors to keep distortion down.. those tiny blue ceramics are horrible..

Thanks for your reply.

As i have seen the parallel "T" filters of esp are of 3rd order each one. What do you think? If we increase the order - let us to say in 4th - of each filter can be resolved the problem of the 2nd harmonic attenuation? And the phase shift also?

Fotios

[Christer]

Are you only interested in overtones, or will you measure IM etc. also? For overtones only, I don't think you need to use a notch filter and a simpler high pass filter will do. You still have to make the filter steep, an elliptical filter probably, but it is still simpler than an equally steep notch filter.

If you only want a filter for a single fundamental frequency, perhaps LC filters are a better option than RC filters, since you need fewer components for a given number of poles. On the other hand, coils are more expensive and probably hard to trim at audio frequencies.

[Elvee]

Originally posted by fotios


Thanks for your reply.

As i have seen the parallel "T" filters of esp are of 3rd order each one. What do you think? If we increase the order - let us to say in 4th - of each filter can be resolved the problem of the 2nd harmonic attenuation? And the phase shift also?

Fotios

A twin T filter has an infinite attenuation at Fo, provided the components have ideal values, so there is no need to increase the order. To solve the second harmonic problem, you need to increase the Q. This is classically done by incorporating the filter in the feedback loop of an op-amp, but in theory, there should be ways of doing it with passives only. This would require inductances however, and this could be a problem: I've made tests of various filters using state-of-the-art, telecom grade ferrite-cored inductances, and they always introduce small amounts of distorsion. The only distorsion-free option is to use air-core coils, but I think they would be too imperfect to achieve the required Q-factor.