Filter Design - no response on the GPB

[brmadhukar]

Filter Design

Hi,
I am designing a new passive filter. This is for the Maxim GPS IC. Impedance 4K,
Frequency 15.42MHz,
Bandwidth is +/- 8MHz.

The simulation works fine but when I get it assembled on the GPB, I find that there is no such response. Every thing is wrong. Should I take care of something which I have not. The soldering is perfect, the components are from Gowanda.

TIA

 

[multanova51]

Hi,

Difficult to say on the info you give below. Did you take into account the unloaded Q of the coils ? Also , you need to take coils with self resonances well out of your required freq. range. Otherwise you have responses that are completely out of the calc.

Usually, I have good agreement between calc. and actual circuits.
Can you give more details on your design parameters and on the type of coils you actually used for the trial.
What simulator did you use ?

Regards

Mike

 

[flatulent]

terminating impedances

Did your original design use 4000 ohms for source and load? Is your physical filter terminated in the 4000 ohms at each end when you measure it? At these frequencies you can use ordinary signal generators and oscilloscope. If you are using network analyzer, you will need minimum loss pads to satisify the analyzer 50 ohms and the filter 4000 ohms.

 

[toonafishy]

Does the mixer want to see a 4k load? Just because the output impedance of the mixer is 4k, it doesn't imply that the load is also 4k. It may not want to be conjugately matched but instead it may want to see a lower impedance. Maxim should know if this is the case. Also, it might be 4K real in parallel with some pF of capacitance.
High impedance filters are much harder to design and make than 50 ohm varities. They are much more suseptible to strays and parasitics.

 

[flatulent]

previous post

The previous post has a good point. What is the load the mixer requires? If it is a lower impedance, you can use a filter shunted at both ends with twice this impedance. Design the filter to go between these two impedances. In the passband the mixer will see both resistors in parallel.

 

[brmadhukar]

Hi,
I am using PSPICE for simulation of this circuit. The circuit is very simple. The mixer may be leftout for time being.

INPUT OUTPUT
R(4K)--------------C(2.7pF)----L(39uH)---|--R(4K)------GND
| |
L(39uH) C(2.7pF)
| |
GND GND

I am using 50ohms source (50ohms << 4K) and measuring output in a spectrum analyzer using a CRO (High impedance) probe.

Since the frequency is low, I do not expect any problem with PSPICE or for that matter any other tool.

The Gowanda inductors used have the following parameters.
Q < 15 at 15 MHz and I am sure this is the problem but I have used the Coilcraft Inductors also, for which I do not have data. May be it also hasthe same problem, but I am not sure of it.

Are there any layout issues to be considered while assembling on a GPB.

Thanks all for the info.

 

[flatulent]

parasitic capacitance

You are probably being affected by parasitic capacitance. One construction method that works well for preliminary measurements is to wire the components together by soldering their lead together up in the air and totally avoiding a board. Lay the "rats nest" on an insulating surface far away from any metal sheets, such as a metal table top.

It would help if you would describe the measurements in terms of center frequency and -3 dB band width.

When you do your simulation put a series resistor with the inductors equal to the reactance/Q. You also need to find the self resonant frequency of the inductors and put a capacitor across them which resonates the coils at their self resoanant frequency.

There is a reason that 50 ohms is a standard impedance level for circuits. Above this parasitic shunt capacitances ruin things, below this series parasitic inductances ruin things.

 

[Element_115]

At 15MHz those 2.7pF cap's are like open circuits.
Try removing the 2.7 pF cap in series, or use a 1uF cap.
You can alway put jumpers for each component, then one by one place them on and see which one is the bad seed.

PS Coil craft has the data in the cover lid of thier coil kits (if you have the kit)

Good luck.

 

[brmadhukar]

Hi,
I guess it should be a good idea to use the "Rats Nest", I have not yet tried it. I shall do it. Since I could not find the data for coild craft, I would like to know in general has anyone used inductors of >5.6uH-47uH, in range 8-30MHz. What is the Q/SRF of such inductors.

brmadhukar

 

[brmadhukar]

Hi Element,
It is a good idea to such jumpers. I will use it in all other protos. Thnx. I guess it can be a standard procedure to develop filters, by ptting jumpers to short Series elements and opening for shunt elements.

 

[multanova51]

Hi Brmadhukar

I found the coils on the internet and compared also to types of Toko and TDK. Iattach some examples. I think the inductance value you have is pretty high (also due to the 4K).
I designed a slightly different filter with lower inductances though. See filter.doc attached.
Perhaps you can give it a try.

Mike

 

[brmadhukar]

Hi Mike,
I tried the circuit you hav suggested first on Psice and am sure that it shall work. But the bandwidth is 8M I need 16M. Coupled reonators can be built upto 10M BW at 15.4M, but over that it is impossible. I am unable to find any inductor with acceptable Q ( > 20 ) at 8-24M over 10u. 1) Will I be able find them?
2) Is a passive filter of 16M possible at 15.4M?
3) Is an active filter possible.
4) what would anyone do when confronted with such a thing?
5) For reducing circiuit Z I may be able to change input impedance but I shall not be able to change the output impedance. Can it still help.

TIA

 

[flatulent]

tamden two filters

Your bandwidth is about equal to the center frequency. One standard way to do this is to use a high pass filter and a low pass filter in a row. Isolate them by an amplifier stage.

 

[brmadhukar]

Hi,
Is it the only possible way, i.e., combining HP and LPF's. Inserting an amp is difficult but I can afford to use a an Active filter what is your opinion on this

 

[flatulent]

lack of interaction

What you are looking for is lack of interaction between the two filters. They depend on a resistive load and source over all frequency for their operation. You can do many things to keep the interaction down. One way is to put an attenuator between them. Your making one active will also work.

If your simulator has an optimizer function, you can just string the two filters together with a low value attenuator (3 dB or so) and have the LC values adjusted for the response you want.

If you are economizing, try a LC low pass after the mixer and make the high pass from restricting the low end gain of the following amplifier stages.

As you discovered in your first design, bandpass filters with wide BW/CF ratios produce extreme value parts which are sensitive to parasitics.

 

[multanova51]

Hi
I fully agree with Flatulent. It looks only feasible with a separate LPF and HPF filter. I didnt notice first that you needed 16 MHz bw in total.
Fitting an amp in between may be quite simple. Even a simple BJT might do the job because you need only isolation, not gain. Because you are talking of a filter between mixer and IF, I think you cannot afford to loose too much S/N there by just putting a resisitive divider.
Making a filter with an opamp is not advisable because you need opamps with extreme high freq. which are more expensive.

Rgds
Mike

 

[brmadhukar]

Hi,
I am now trying a LPF and HPF combination. For practical values I had to configure the filter as

4K -- LPF (4K -> 500 ohms) -- HPF (500->4k) -- 4K

I have few questions on this.
1) Why should I use a Amp between LPF and HPF as they are matched. I have tried without amp. The response is bad. In such a case what is the imp. of impedance.

2) If the total filter has to be differential. What amp should I use.

 

[brmadhukar]

Hi,
Is there a way of switching the bandwidths of passive filters at 15.42MHz. I want to change the filter BW to 2MHz, 4MHz, .... 16MHz.

TIA

 

[strabush]

HPF + LPF design

I did design a number of times passive LC HPF cascaded with LPF with direct connection between them without any problem.
Your error is that you designed a different impedance at the connection of the filters.
You have to understand that each filter does not "feel" the second filter at its cutoff frequency, the second filter is just transparent (at least if the two cutoff frequencies are far enough) so you have to design each filter to be terminated with the end termination impedance as if the other filter does not exist. After that you can optimize the design using an optimizing software (like genesys from eagleware that I use).
Other two possible error sources are whether the termination impedances you use are really the impedances that the filter really see, and secondly, notice that using high impedance termination means high inductance and low capacitance which is making the filter sensitive to parasitic capacitances everywhere. check that the parasitic resonance frequency of the inductors is at least twice the cutoff frequency of your design.

 

[brmadhukar]

Hi Starbush,
Thnx for your inputs.

The RFIC requires see a high impedance ( output impedance is 400 ohms) of around 4K. This requires very high value inductance and I had used 500 ohms output to reduce the Inductance value. This was only the simplest I could think of at that time, is there any other method to increase impedance without increasing inductance.

brmadhukar