Low Pass Filter Design

Hello Everyone,

I am finding difficulties to design a low pass filter that meets the following specs :

- F cutt-off = 657 KHz
- Ripple in-band < 1 dB
- Stop band attenuation = 60 dB (>30 dB @ 750KHz & > 60 dB @ 2 MHz)
- Group delay peak-to-peak < 1uS

I have been using AWR MWO Nuhertz and iFilter in order to help the design, but once replacing the real componentes (inductors and capacitors) by either s-parameters or SPICE models commercially available it seems impossible to achieve the specs above, most likely due to the low Qu. I also tried active design with Amp-Ops without success. I have been thinking to use toroids instead (in e.g. Fair-Rite material 61) in order to increase the Qu at the design, but I have no idea of what Qu it can be expected.

Any good ideas how to overcome the problem in order to achieve the specs above ?

Thank you

The group delay spec of 1us seems very tough. I would expect to see 5us group delay in a typical high order LPF design that tries to meet your stopband etc..

I did design similar filters many years ago and I used RM4 cores with the M33 ferrite material.

Also, if you can allow op amps then I would suggest you design a lumped elliptic filter and design it for a high impedance. eg 500 ohm as this will give inductor values that achieve high Q at 600kHz. You can achieve 50 ohm to 500 ohm transformation easily by using high grade op amps with the relevant termination resistances. The op amps need to be good quality in order to not add extra group delay so expect to pay good money for suitable opamps.

With RM4 cores and M33 material you can expect to see inductor Q of 250 to 350 for inductances in the region of 100uH at 650kHz. The other bonus of the RM4 cores is that they come with a tuning screw that allows a reasonable adjustment range so you can fine tune the elliptic filter.

However, I don't see how you can meet 1us pkpk group delay with your other requirements.

I suppose you could include a lumped compensation network after the LPF that attempts to cancel the group delay ripple?

Group delay peak-to-peak < 1uS

Click to expand...

is really tough. I used shunt RC compensation network, and get peak-to-peak <1.8us. maybe you use more complicated network can get better result.

tony_lth said:

is really tough. I used shunt RC compensation network, and get peak-to-peak <1.8us. maybe you use more complicated network can get better result.

Click to expand...

Hi guys thanks for the inputs.
I realised that I will need 7th or 8th order Elliptic low pass filter, and also cascade a phase equalizer (group delay flattening) circuit with the filter.

Meantime as I said, I tried to replace either the Op-Amp or lumped elements by SPICE or s-parameters, so in either cases it is really tough to achieve.

As Jeremy R. said I could try the RM4 cores and M33 material, in e.g. EPCOS. If the Qu stated is about to 250 then I guess it will be ok. Only detail is that at the phase equalizer circuit the iFilter states that it must have mutual coupling inductor which is not feasible to use pot cores, so perhpas toroids with a tap in the wounded wire would be OK.

All the best,
Enrico.

hai can anyone pls tell me how can we calculate the capacitor and inductor values for the 7th order elliptic low pass filter

The filter design tools mentioned in the initial post will do. Other references to filter design software, also free tools, can be found in various edaboard threads. Just search!

hai can anyone please tell me which simulator will be best for the elliptic low pass filter

---------- Post added at 11:09 ---------- Previous post was at 11:02 ----------

hai FvM, actually i want to redesign the elliptic lpf with the new cuttoff. i dont know how to calculate the values for that one. can u please help me for that one. and i heard that there is a simulator which will give the values bye giving the cutt off and db. if u know about this can u please help me.

just as the invention of meta materials has changed what it possible for small antennas (i.e. allowing you to violate what we used to think were the "laws of physics"), is there some non-standard way to make filters with much less group delay than their bandwidth implies? Some use of active negative inductors, etc?

As I suspected, there are microwave filters using metamaterials with " an LC ladder configuration for negative group delay (NGD) values". I suspect there are ways of getting the group delay you want with the narrow bandwidth you want too. Just not sure if anyone operating in the KHz region is looking in these areas.

Google

**broken link removed**
**broken link removed**