About Elliptic Bandpass Filter Design

[ful24]

hi everyone:
i have got an assignment about designing an Elliptic bandpass filter, we dont have to build the actual circuit, all i need is: do the calculation, put the values in the simulating software to get the graph and a report. and the task is:

"To design an Elliptic bandpass filter for the output of an 849MHz 5W transmitter, with 50R input and output impedances. the filter is designed to reduce harmonic transmission to below -55dBc."

please help with extract information or specification out of the question. all the information i can get from the question so far is:
center frequency: 849MHz
55dB attenuation from center frequency

any more information i need to know to design this filter? any consideration or concerns?

thanks

 

[enjunear]

Here are the items that I glean from your specs:

• 
  Center freq = 849 MHz

• 
  Passband bandwidth, undefined (ask about what type of signal you are passing, that might give you some ideas... or if it's open to your choice, then pick some easy value out of the air, say 1% or 8.5 MHz)

• 
  Passband ripple, undefined (If not given by the definition of the signal passing through the filter, then pick something reasonable, maybe 0.1 dB for a circuit with ideal parts, maybe 0.5 dB for a circuit with non-ideal components)

• 
  Stopband frequency, at least 1698 MHz (lowest harmonic is the 2nd harmonic, so 849 MHz*2 = 1698 MHz)

• 
  Stopband bandwidth: something less than 1698 MHz, you get to pick.

• 
  Stopband rejection, 55 dB (if this is for a class, design the circuit for a few dB more, like 60. If you are really going to build this, then I'd design for least 10-20 db more than that; non-ideal component parasitic effects will eat into that margin if you really try to build the circuit)

As for fabrication techniques, you should have some design equations in your textbook, or filter prototypes that you can start from. Plug some of these values into those equations and solve for the real component values. Most of the texts I've seen use filter prototypes as a function of relative frequency, so f/f0, then everything scales up based on your center (or cut-off) frequency.

I hope that helps get you started in the right direction.

 

[LvW]

Hi ful24,

I think enjynear already gave you a good summary of the relevant specifications and a first approach for realizing.
One additional point: active or passive?
I think due to the required port resistances of 50 ohms (correct? Since you wrote: 50R) a passive realization with C's and L's is required.
As an additional hint you can assess the required filter order with the help of some filter design programs which are available for free to download.

 

[ful24]

thanks guys, this helps a lot.
and yes LvW, it is a LC filter, the tutor gave the tables of passband attenuation of 1dB, 0.5dB, and 0.1dB. with the summary enjynear gave, i think i can start the selection and calculation. the tutor suggested AADE software to simulate the result.

 

[enjunear]

thanks guys, this helps a lot.
and yes LvW, it is a LC filter, the tutor gave the tables of passband attenuation of 1dB, 0.5dB, and 0.1dB. with the summary enjynear gave, i think i can start the selection and calculation. the tutor suggested AADE software to simulate the result.

I've never used AADE before, so can't comment on it's usefulness. Most engineers I know started out on some flavor of SPICE, most of them PSPICE (but you have to buy it, IIRC). Linear Technologies (sell analog IC's) have a freeware app called LTSPICE that is pretty useful, and acts like a lot of other SPICE-variant programs (SPICE is a common language to describe a circuit, and set of algorithms to do DC and AC analysis of simple circuits... RLC's, BJTs, FETs, etc). You might try messing around with LTSPICE (or PSPICE, if you school has it available).

For industry-standard tools, most large companies use Agilent's ADS or AWR's Microwave Office, which are really overkill for what you're doing, but are good to play with in undergrad, so you have some familiarity w/ them when you get into industry.

 

[atripathi]

I agree with enjunear, quick spice simulation should be good starting point and in fact it may suffice your goal as well. You can also download the FREE evaluation version of PSpice from **broken link removed** This free version never expires and have very generous limits.

 

[LvW]

atripathi and enjunear,

for your information: AADE is not a circuit analysis program like PSpice, LTSpice,....
It is rather a tool to DESIGN passive filters up to an order of n=16. I really can recommend it.
And that is - I think - the tool ful24 needs at first.
Then - as a 2nd step - after the filter is designed he can use a simulation program for fine tuning, parts adjustment or something similar.

 

[FvM]

AADE Filter Design 4.5 is freeware.
Commercial tools like Nuhertz Filter Solutions can do the same, with a more intuitive design entry in my opinion.

 

[LvW]

AADE Filter Design 4.5 is freeware.
Commercial tools like Nuhertz Filter Solutions can do the same, with a more intuitive design entry in my opinion.

Yes, I agree. "Filter-free" (from filter-solutions.com) is easier to handle - however, the bandpass design is restricted to 3 pole pairs only (free demo version).

 

[ful24]

hi guys
i have done some calculations and simulation, i found out that base on the table the tutor gave us, i have to choose a bandwidth of 797MHz to achieve that stopband attenuation is >60dB. is this bandwidth reasonable? and the components are at low pF and nH, are those value realistic?
one more thing, that in the question, there is a "5W", is this 5W related to the filter design? how ?
PS: has anyone come across any transmitter that transmits at 849MHz-ish and 5W ? what kind of signal would that be? i have been asking my tutor this, but he wouldnt tell. =(
thanks

 

[enjunear]

hi guys
i have done some calculations and simulation, i found out that base on the table the tutor gave us, i have to choose a bandwidth of 797MHz to achieve that stopband attenuation is >60dB. is this bandwidth reasonable? and the components are at low pF and nH, are those value realistic?
one more thing, that in the question, there is a "5W", is this 5W related to the filter design? how ?
PS: has anyone come across any transmitter that transmits at 849MHz-ish and 5W ? what kind of signal would that be? i have been asking my tutor this, but he wouldnt tell. =(
thanks

Can you clarify some of those numbers? Are you saying that you have a 1 dB (?) passband from 450.5 MHz to 1247.5 MHz (849 - 797/2, and 849 + 797/2), then a stopband rejection of 60 dB at 1698 MHz? That's pretty wide, which makes your filter need to have very steep walls, which means more sections, high value for N.

I'm thinking about a filter bandwidth more in the tens of MHz, not hundreds. As for the parts values, I'd be expecting inductors in the nH range and caps in the nF range... not pF (maybe at the outermost sections of the filter, but not all of them).

As for the frequencies... skim through here and see if anything catches your eye.
Cellular frequencies - Wikipedia, the free encyclopedia

5W = 5 watts, as in, the power being delivered from the transmitter to your filter. When you design a real filter, you need to consider the power dissipation inside the filter and peak voltage/currents, so you can pick parts that can handle the levels without breaking down (e.g. 25V, 50V, 100V caps).

 

[FvM]

I'm thinking about a filter bandwidth more in the tens of MHz, not hundreds. As for the parts values, I'd be expecting inductors in the nH range and caps in the nF range... not pF (maybe at the outermost sections of the filter, but not all of them).

1. Yes. 5 to 10 or maximum 20 % bandwidth sounds reasonable
2. No. 1nF*1nH makes 160 MHz resonance, not 850 MHz. It's surely pF. According to the filter structure, you possibly get unpleasent L and C values, the problem has been discussed at Edaboard before. It may be necessary to chose the strcuture respectively.

However, no design constraints have been told in the initial post. In my understanding, it's an exercise problem. Implementation alternatives are probably addressed next.

Generally, low-passes are more popular for harmonic filtering, I think.

 

[ful24]

thanks again for the reply, guys. thats some very helpful information i will not get from the book.

and yes FvM, it is a exercise, we are not going to build it, only to get an idea how to use the transition formulas and simulation software.
i re-calculate the component values again using 1% BW = 8.49MHz, and i did get unpleasent L and C values. ( femto F, and milli H).

now i am up to change the BW wider to get reasonable component values.
BTW, the response curve i got from AADE software is totally different from what i expect (from the book).

 

[ful24]

another question, since i did get unpleasant L and C values, can i use stripe line or SAW filter instead of LC?

 

[FvM]

Some filter tools, e.g. Nuhertz Filter Solutions have an option to synthesize distributed (transmission line) filters. Unfortunately "unpleasant" component values seems to turn into unfeasible transmission line impedances using this approach...

For some LC filter types, you have alternative structures, that refer to a coupled resonator scheme, ending up in more reasonable component values. But as far as I'm aware of, there's no respective representation for the elliptic type.

https://www.edaboard.com/threads/211699/

 

[volker_muehlhaus]

since i did get unpleasant L and C values, can i use stripe line or SAW filter instead of LC?

SAW is for receive path only (low power).

I am still trying to understand why you want to design a band pass filter with small bandwidth. You wrote that your filter design is used in the output of a 5W transmitter, to reduce the harmonics. There is no need for a band pass, because any filtering close to the carrier should be done earlier in your transmitter, and the output stage is the wrong place for that. For the harmonics of n*849MHz (n=2,3,4 ...) you only need a simple low pass filter.

 

[FvM]

I am still trying to understand why you want to design a band pass filter with small bandwidth.

I guess, because the exercise problem requests it. Otherwise a 5th order elliptic or a 7th order chebyshev low-pass can achieve the specification with reasonable component values (nH/pF).

 

[volker_muehlhaus]

I guess, because the exercise problem requests it.

Maybe. Or maybe the instructions should have said elliptic low pass filter.

Technically, band pass after the PA makes no sense to me.
Why design a power amplifier and then loose most of its output power in a lossy narrow band BPF, where a low pass filter would reduce the harmonics with much less insertion loss to the transmitted signal?

 

[ful24]

I guess, because the exercise problem requests it.

exactly right. i think the whole purpose of this exercise is to get familar with the method of calculation and simulation software.

thanks all for the reply guys. it helps a lot with your experiances.