[help]Is it possible to realize this freq. response?

[surfing]

Dear all

Here are freq. response curve provided by our customer, is it possible to realize?
How to design Filter type to guarantee the amplifier's curve.
Can some experienced one give me some suggestion.

Thanks much

 

[FANT]

Being the bandwidth quite wide, I suggest the coupling of a low pass + high pass.

Mandi

 

[surfing]

high pass(left end) seems to be realized easily.
but low pass(right end) is too sharp??
bcs delta freq.=50MHz.
can we design one filter response as: in less than 50MHz bandwide, acutely descend more than 25dB@near 900MHz.
can somebody give me any suggestion about the low pass(right end) filter?
anyway, the products is in low cost

 

[BigBoss]

Like this??( better vision)

 

[surfing]

Dear Bigboss
thanks
your filter seems reach our requirement
q: could you guarantee 470Mhz, 860MHz are within the bandwidth?

Could you give me some ref. design about this UHF filter?

 

[BigBoss]

Connect part_1 and part_2 in series.

 

[surfing]

Thank u, bigboss
could u send me project file in @ds
via following mailbox:
surfingyee@21cn.com
ycl@philex.com.cn

 

[djalli]

I think BigBoss showed the schematic. You want the file?

By the way the schematic is using ideal lumped elements. At that frequency you can not realize this filter using lumped elements.

Schematic must be transformed into distributive elements (stubs etc). Then why you need the BigBoss ADS schematic file for? Student?

 

[gecky]

Agree with djalli.... end of the day, you got to implement the resonator and J/K-inverters in ML....

Besides, try the filter design guide in ADS. If i'm not wrong, BigBoss used that to gen. the required lumped filter.

 

[djalli]

I did not catch it why J/K inverter is needed here?

 

[BigBoss]

By the way the schematic is using ideal lumped elements. At that frequency you can not realize this filter using lumped elements.

Schematic must be transformed into distributive elements (stubs etc). Then why you need the BigBoss @DS schematic file for? Student?

Why this filter is not used at that frequency and must be transformed into distributed components???? I have used with/without standart components in many times to implement that kind of filters.

If you look carefully, the low cut-off frequency is around 470MHz and you know stubs and distributed components will be very large...So, distributed components approximation does not give a solution.

 

[djalli]

BigBoss you caught me I never saw that. Damn wavelength(~0.65meters if I am wrong I flunked in algebra once) always is fogotten by us to do a simple division.

Yes you are right, you can use lumped elements in this circuit. Thank you BigBoss.

 

[biff44]

I am going to vote with Djali on this one. How are you going to realize those parts? L = 22.83255 nH ?

I suppose you could laser scribe your own custom capacitors to +/- 0.2 pF in production, but you are going to have to buy those inductors, probably in steps of 1, ie 22 nH, 23 nH.

Also, if you put in realizable Q values, Qcapacitor=500 and Q inductor=40, you will find the loss increases to 2.3 dB and the 25 dB rejection is no longer met.

I am not sure how I would do this filiter, but lumped elements does not sound too producible.

Maybe high Er ceramic resonators with variable capacitor tweak tuning for the poles, and lumped elements for the non-resonant components?

 

[g579]

At the first post there were no insertion loss requirements.
And with simulating with the "real" models of the actual "lumped" Ls and Cs , and optimizing the transfer function to be similar to the first post's response, there is no problem realizing such filters with 1 - 2 dB insertion loss, with commercially available cheap components.
Bigboss just showed a very good starting point for the required filter.
g579

 

[biff44]

You can optimize all you want, I still say it will not meet spec with that few a number of elements, and really is a problem if there is a 5% tolerance on parts.

 

[sinatra]

Hello all.
In this frequency range it should be no problem to apply distributed element technique.
In this range combline filters are still realisable and not so big, still competitive against lumped element filters. FR4 would be a good substrate since it has a high epsilon (4.4) and it would help to make the filter smaller.
The problem is another. The bandwidth desired is quite large. Normally available syntesis equations such like those found in G.Mathei book are only valid for smaller bandwidths, i.e., 5% to 20%.
Probably using a 20% design as an initial shot and running an optimizer it could be realized.

Of course a lumped parammeter filter would be much smaller. But it would have other drawbacks. You will have to characterize all components you will want to use and then include all parasitics and losses in a simulation and optimize the circuit to get the desired response. The final implementation of the filter would require that some capacitors be adjusted to get the desired response.....etc...

My main question would be what would be the best technique/strategy for mass production?

S.

 

[g579]

Hi biff44,

1. I agree on the first part of your reply. Probably more elements (higher degree filter) are needed.
2. I do not agree with the second part. With 5% elements it can be done more or less easily. Naturally the design should be a little bit tighter in response compared to the needed, to allow for the 5 % tolerances effects.
g579

 

[BigBoss]

In fact I agree with all fears more or less but in the real life all components will have a tight or loosen tolerances and another mission of an engineer to develop and compansate to all these restrictions with a suitable and robust solution.
For instance to realize 16.3nH inductor, it may be tought that to add 1.3nH microstrip inductor to 15nH inductor, so that will give us totally 16.3nH. In that case while doing pcb , all compansating distributed elements are simulated in a EM simulator and then they can be optimized in a certain way...

Or to use adjustable components...etc.

There will always be a way to implement any circuit and our mission to do that...

 

[djalli]

Yes that is way to go BigBoss, you find compromises to solve the problems.

 

[surfing]

Thanks all of you