Simple BPF VS Cavity Filter

[Fovakis]

Hi to all,

i have a SISO system with microwaves devices. The 16-QAM information is about 20KHz , upconverted at 30 MHz and then at 2462MHz.At the receiver the same procedure, downconverting. After the first LNA i have a BPF fc=2430MHz which has large Bandwidth so it does not protect me from interferense signals. It has 200MHz BW. Although, i have a BER about 10-4 maybe 10-5 if the channels is clear (more often at evening).I did this setup in a Laboratory and behind us there are a lot of with WiFis transmitting all day etc...

i did a purchase of Cavity Filters with center freq at 2462MHz and BW=25MHz. i replace the BPF with the cavity and........the result is worse!!! i have 10-3 maybe 10-4 BER :-o:-x:-o:-? Can you please tell me some ideas to fix this problem?

https://i1284.photobucket.com/albums/a578/fovos1/receiver_zps552b2e4a.png

 

[BigBoss]

If you have placed this cavity filter before LNA, the Noise Figure will be increased by the insertion loss of this cavity filter.Let say the insertion loss is 2 dB, overall Noise Figure of the system will start from 2dB.If you take the other receiving block into account, this Noise Figure will much higher.This is the error you did..
Instead, you should place this cavity filter after LNA since you don't worry about the power levels of interferer signals.( I presume they have low power levels).
Another solution is to design an "Frequency Selective" or "Narrowband LNA".by doing this, out of band interferers will be less disturbing.But there is nothing to do for in-band interferers except digital tricks.

 

[Fovakis]

If you have placed this cavity filter before LNA, the Noise Figure will be increased by the insertion loss of this cavity filter.Let say the insertion loss is 2 dB, overall Noise Figure of the system will start from 2dB.If you take the other receiving block into account, this Noise Figure will much higher.This is the error you did..
Instead, you should place this cavity filter after LNA since you don't worry about the power levels of interferer signals.( I presume they have low power levels).
Another solution is to design an "Frequency Selective" or "Narrowband LNA".by doing this, out of band interferers will be less disturbing.But there is nothing to do for in-band interferers except digital tricks.

Hi and thanks for the reply. If you can see the image i have the FILTER after the LNA.

The interferer signals are in low levels that's right.

First time with a filter of 200MHz and the second with a cavity of 25MHz. Why in first case the system work better? At both cases the interferences signals boosted from the LNA.

Maybe is there any problem of matching of the cavity filter? Also the cavity has two ports. I don't understand which one is the input and which one is output?Is there any difference with that?

 

[vfone]

2462 MHz is Channel 11 in WiFi.
There is no RF filter in the world that can help you rejecting a co-channel interferer at the receiver input and let through only the desired signal.
The co-channel interferer rejection can be improved only in the DSP of the digital system.

If you have access to the local WiFi interferer system just change its channel to a different one.

 

[Fovakis]

2462 MHz is Channel 11 in WiFi.
There is no RF filter in the world that can help you rejecting a co-channel interferer at the receiver input and let through only the desired signal.
The co-channel interferer rejection can be improved only in the DSP of the digital system.

If you have access to the local WiFi interferer system just change its channel to a different one.

I think that you don't understand what i am saying. I will try again.

BPF filter has better results that cavity. i tried with 2465mhz and 2462mhz with both filter.but the BPF filter with the huge bandwidth had better results.
why is this?

Also about the wifi i don't think that the signals are higher that my desired signal's power which is about -50dBm at the receiver's antenna!

 

[BigBoss]

I think that you don't understand what i am saying. I will try again.

BPF filter has better results that cavity. i tried with 2465mhz and 2462mhz with both filter.but the BPF filter with the huge bandwidth had better results.
why is this?

Also about the wifi i don't think that the signals are higher that my desired signal's power which is about -50dBm at the receiver's antenna!

Have you ever measured the insertion losses of these filters ?? BPF may have lower insertion loss compare to cavity.Cavity may have have lost its tuning or something is wrong with this filter??

 

[RCinFLA]

If you placed the cavity after the LNA then it is likely matching on the cavity. A lot depends on the output impedance transformation from the LNA. In filter design 1+2 does not equal 3. The first pole of the filter is the LNA output match. You may even caused the LNA to become unstable due to the pure reactive loading the cavity will have on the LNA beyond the cavity's nose bandwidth.

The better the unloaded Q to loaded operating Q of the filter elements, the more critical the termination loads are on the filter. In purest terms, a filter input and output impedance is not matched to (in conjugant match terms). A filter is "terminated" or "loaded" with its design impedance, which must existence at least across its two to three db down bandwidth in order not to effect the filter's nose response and loss.

You did not say what the LNA gain was or mixer input noise figure, but try dumping the LNA and just try running the cavity in front of the mixer.

 

[biff44]

QAM modulation is sensitive to the type of filter you put in the system. If it has too much phase or amplitude non-linearity vs frequency, then the BER will get worse just from the filter itself. You need a narrow filter to reject the interference, but not so narrow that u end up causing intersymbol interference.

 

[Fovakis]

Have you ever measured the insertion losses of these filters ?? BPF may have lower insertion loss compare to cavity.Cavity may have have lost its tuning or something is wrong with this filter??

The insertion loss of cavity is really greater but only 1 to 1.5dB. I don't think is so critical. I don't see any great difference with the specifications of the filter..

- - - Updated - - -

If you placed the cavity after the LNA then it is likely matching on the cavity. A lot depends on the output impedance transformation from the LNA. In filter design 1+2 does not equal 3. The first pole of the filter is the LNA output match. You may even caused the LNA to become unstable due to the pure reactive loading the cavity will have on the LNA beyond the cavity's nose bandwidth.

The better the unloaded Q to loaded operating Q of the filter elements, the more critical the termination loads are on the filter. In purest terms, a filter input and output impedance is not matched to (in conjugant match terms). A filter is "terminated" or "loaded" with its design impedance, which must existence at least across its two to three db down bandwidth in order not to effect the filter's nose response and loss.

You did not say what the LNA gain was or mixer input noise figure, but try dumping the LNA and just try running the cavity in front of the mixer.

As i understand the are reflected waves from the cavity. LNA gain is 15dB and NF=0.76dB , mixer conv.loss is 8dB and so the noise also.

check the mixer here http://www.minicircuits.com/pdfs/ZFM-4212+.pdf

I think the most critical is what biff44 says.about the ISI cause of the great group delay the cavity filter inserto to my 16QAM signal.....?

- - - Updated - - -

QAM modulation is sensitive to the type of filter you put in the system. If it has too much phase or amplitude non-linearity vs frequency, then the BER will get worse just from the filter itself. You need a narrow filter to reject the interference, but not so narrow that u end up causing intersymbol interference.

You mean that the group delay of the cavity causing the problems right? what kind of filter design are the cavitys? Butterworth
Chebyshev ? or it depends?

 

[RCinFLA]

With a signal occupied bandwidth of only 20KHz for your 16QAM it is doubtful a filter with a 25 MHz bandwidth, even poorly terminated, will have any impact on variance in group delay across any 20 KHz segment of the cavity filter response.

 

[biff44]

With a signal occupied bandwidth of only 20KHz for your 16QAM it is doubtful a filter with a 25 MHz bandwidth, even poorly terminated, will have any impact on variance in group delay across any 20 KHz segment of the cavity filter response.

yes i would agree, a 20KHz signa in a 25 MHz bandwidth will not be a problem. But, i wonder then what he means by a "cavity" filter? Is it really a flat 25 MHz bandwidth?

- - - Updated - - -

With a signal occupied bandwidth of only 20KHz for your 16QAM it is doubtful a filter with a 25 MHz bandwidth, even poorly terminated, will have any impact on variance in group delay across any 20 KHz segment of the cavity filter response.

yes i would agree, a 20KHz signa in a 25 MHz bandwidth will not be a problem. But, i wonder then what he means by a "cavity" filter? Is it really a flat 25 MHz bandwidth?

 

[RCinFLA]

At 2.46 GHz a 25 MHz bandwidth filter is quite respectable and would require filter resonators with very high Q, a.k.a. cavity, likely meaning high quality quarter wave resonators in silver plated cavity cells.

 

[jiripolivka]

At 2.46 GHz a 25 MHz bandwidth filter is quite respectable and would require filter resonators with very high Q, a.k.a. cavity, likely meaning high quality quarter wave resonators in silver plated cavity cells.

From all above I think it is not correct to judge filters and other system components by BER only. BER is affected by many system parameters.

If you consider the inteference to be strong in your environment, you should first use a spectrum analyzer and a directional antenna to see such signals, their amplitude, spectrum shape and point of origin.
The out of band interferers can be suppressed by the filters as you intended but some stay strong. Using directional antennas can help too. If nothing helps, it is time to start uing a new system e.g. in 5.8 GHz band.

In-band interferers usually cannot be suppressed at all.


Good luck in the overcrowded 2.4 GHz band...

 

[Fovakis]

yes i would agree, a 20KHz signa in a 25 MHz bandwidth will not be a problem. But, i wonder then what he means by a "cavity" filter? Is it really a flat 25 MHz bandwidth?

- - - Updated - - -



yes i would agree, a 20KHz signa in a 25 MHz bandwidth will not be a problem. But, i wonder then what he means by a "cavity" filter? Is it really a flat 25 MHz bandwidth?

This is the band pass that i use **broken link removed**

and this is the cavity filter i use **broken link removed**

So, you don't believe the problem is about group delay... it make sense, to be true, because indeed the bw of information is so "short"

 

[Fovakis]

Hi again. Basically, i am almost sure that you are all wrong about why this problem came up. Except RCinFLA ! He was right. The problem was with the matching at the input and output ports.

 

[RCinFLA]

When putting RF passive filter between active devices there are three primary effects that are of concern. These are filter response changes due to termination matching across the bandpass of filters, noise figure degradation due to improper optimal noise figure impedance, and intermod performance degradation due to filter shirt impedance.

Common mistakes in stage segmented design are input impedance for optimum noise figure and IM performance. For example, in this case the narrower filter may raise the RF amp's out of bandpass termination impedance, raising its gain in the out of cavity bandpass frequencies and thereby raising the level of IM products within the RF amp that, in-turn, create greater interference of IM produces within the passband of interest. You can select matching configurations that create and 'impedance inversion' at out of band frequencies to lower the out of passband impedance instead of raising it. A tapped capacitor output matching configuration is the simplest matching configuration that can help in this situation. You can also end up with RF amp instability if the output load is being relied upon to keep the RF amp stable.

If the input impedance of a mixer or RF amp require a non-conjugate match for optimum noise figure performance then by definition the output of the filter will not have the proper load termination, and the active device will not be getting the optimum noise figure input impedance termination. You would need an isolator or balanced rat race active device design to present the proper load to the filter. It may be better to sacrifice a couple db of gain to get better alignment of optimum input noise figure matching and conjugate match (low reflected power match).

Two topic related common myths taught in school:

1) Casade equations for IM analysis always work.
They rely on no phase shifting between all frequency components between successive stages which often happens as intermod components are put through a filter.

2) Mixers always have a 3 db degradation in noise performance due to the RF image noise contribution.
Just like an optimum input impedance termination for desired passband noise figure there can be an optimum image frequency termination to minimize image noise contribution. Just adjusting the 50 ohm line length between the filter and the mixer input can improve mixer noise figure by minimizing the image noise.