Band-Pass Filter Help

[EBrown]

So I am trying (very hard I might add) to design effective band-pass filters for a project. I need to filter audio into several channels, and so far I am having some issues designing my band-pass filters effectively.

I have read up on them (quite a bit, actually) and I know that there are several ways to build them, from creating the traditional Resistor-Capacitor band-pass filter, to creating a Resistor-Inductor band-pass filter, to even creating an Inductor-Capacitor band-pass filter. I have tried all of them, but I can't seem to get them within the required parameters.

The problem with using Inductors is that I need very large Inductors for the lower channels. (I.e. 20Hz) Inductors seem to be inadequate for use in circuits requiring band-pass filters lower than 1kHz. The other problem is that as I get larger Inductors, the resistance on them increases and causes signal drop problems and works against myself. (At least according to LTSpice graphs.)

The Resistor-Capacitor band-pass filter filters them fine, but I need to chain four of them together to get the drop I want, and they're not as narrow as I would like. (Graph Attached.)

I guess my question is: What is the optimum dB drop for audio filtering, and if I have achieved the optimum dB drop, how can I narrow them to get a finer tuned frequency?

Any help is greatly appreciated.

Thanks,
EBrown

 

[Harald Kapp]

Have a look at active filters to overcome the burden of obscenely large inductors.

What is the optimum dB drop for audio filtering

That's up to you to decide and depends on what you want to do. If you want to do audio processing, a rather high dB number may be desirable to minimize cross talk between different frequencies. If you build a lighting console, possibly a few dB (~12 or so) may be enough for a good effect.

For reference: A CD's dynamic range is 96 dB

if I have achieved the optimum dB drop, how can I narrow them to get a finer tuned frequency?

These parameters are not independent in most filters. You'll have to design the filter that it fulfills both requirements. Narrowing the filter pass band often (not always) goes hand in hand with increases steepnes of the slope at the pass band's edges.

 

[BobK]

Active filters are the only way to go. You get 12db / octave for each stage, and need only resistors and capacitors (and op amps) to implement.

Here is free software to design active filters.

http://www.ti.com/lsds/ti/analog/webench/webench-filters.page

Bob

 

[EBrown]

From the research I have done I need 0dB to 20dB signal for the chip I plan to use, so if I can get the dB below that point I have gotten rid of all the signal I don't need. (That's after I amplify it of course.)

I'm just not having any luck, even when I attempt active band-pass filters.

I tried a Sallen-Key with no luck.

Does anyone have any ideas on how I can filter it appropriately? Or what style of Active Filter I should use? Because it seems there are a lot. I'm probably approaching this whole thing wrong as it is.

Thanks,
EBrown

 

[Harald Kapp]

I tried a Sallen-Key with no luck.

What did you try exactly? Where did you fail (no luck)? Show us your design and your expected behaviour (bandwidth, center frequency, slope...). Normally a sallen key filter is rather reliable if designed correctly.

 

[EBrown]

What did you try exactly? Where did you fail (no luck)? Show us your design and your expected behaviour (bandwidth, center frequency, slope...). Normally a sallen key filter is rather reliable if designed correctly.

Well the problem I am having is tuning it to the right center frequency. I can't seem to figure out how to force it to a specific center. (The math just isn't making sense.)

I've attached a picture of my design, I just need to find the right values to put in for the various resistors and capacitors.

Thanks,
EBrown

 

[BobK]

It is pretty much impossible to do an active filter design by hand. Did you try the design software I linked to? It will give you a working design based on the parameters you input for center freq and bandwith.

Bob

 

[EBrown]

It is pretty much impossible to do an active filter design by hand. Did you try the design software I linked to? It will give you a working design based on the parameters you input for center freq and bandwith.

Bob

I keep getting an error because I am trying to use frequencies that are way too low. (Attached)

Thanks,
EBrown

 

[Laplace]

Rather than design a bandpass filter with a pass band of almost 6 octaves, a better approach might be a low-pass with a corner frequency of 1000 Hz followed by a high-pass having a corner frequency of 20 Hz.

 

[duke37]

There was a question on the homework section recently to evaluate a filter. I was interested since it was so simple. As shown it is a band pass filter passing audio frequencies. I did a simulation in 5spice.

I have 'Electronic filter design handbook' by Arthur B. Williams. This deals with passive and active fiters. I could extract from this if you could give me the specitications of the filter.

 

[EBrown]

There was a question on the homework section recently to evaluate a filter. I was interested since it was so simple. As shown it is a band pass filter passing audio frequencies. I did a simulation in 5spice.

I have 'Electronic filter design handbook' by Arthur B. Williams. This deals with passive and active fiters. I could extract from this if you could give me the specitications of the filter.

I'm trying to filter audio so I can throw it in a VU Meter and further analyze the signal. I want to build a hardware solid-state equalizer, which requires that I use a RC/LC/RL filter as opposed to digital filters. (Plus Analog Filters are much faster and have no processing time so to speak.)

And Laplace, I need to splice the audio signal into the various channels, not one huge channel.

Thanks,
EBrown

 

[BobK]

I think Laplace had the right idea. I tried another filter design tool and the problem is the large pass band you want. These 2nd order active filters fall off too fast to do that.

Here is a tutorial that shows how to use a low and high pass filter together to produce a wide passband.

http://www.electronics-tutorials.ws/filter/filter_7.html

Bob

 

[EBrown]

I think Laplace had the right idea. I tried another filter design tool and the problem is the large pass band you want. These 2nd order active filters fall off too fast to do that.

Here is a tutorial that shows how to use a low and high pass filter together to produce a wide passband.

http://www.electronics-tutorials.ws/filter/filter_7.html

Bob

The problem is that I don't want the large pass band. The one I have right now is far to large for the application I need it for. A 20Hz signal would bleed through to the 40Hz, 60Hz, 100Hz and quite possibly the 160Hz filters.

Thanks,
EBrown

 

[BobK]

Here is a filter design that meets your specs, designed based on the multiple feedback filter on the page I linked you to. It wasn't so hard after all. I have to wonder why the programs did not handle it. The specs I used are 20Hz center frequency 100Hz 3db point, and the graph looks like it did that successfully.



Note that is has a gain of -22db, which is necessary to get the broad passband. You will probably need an amplification stage after.

Bob

 

[BobK]

Well, why didn't you tell us that before I went off and designed one? It would really help if you would not mislead us.

So, what do you want the -3db point to be?


Bob

 

[EBrown]

Well, why didn't you tell us that before I went off and designed one? It would really help if you would not mislead us.

So, what do you want the -3db point to be?


Bob

I want the -3dB point to be roughly 10Hz both ways. I want the 20Hz filter to cover 10 to 30, the 40Hz to cover 30 to 50, etc. I guess I should have said that earlier.

And I do have an amplification stage after all the filtering. I knew I would need one as I only have up to .447 Volts input, so I needed an amplification stage to get up to the 10V that the top of my VU chip is at. (I think, as best I can see from the datasheet it operates the LED's on a scale from .708V to 10V)

Thanks,
EBrown

 

[duke37]

I did a Google, there is a circuit there using a LA3660 for a 5 band equaliser.

 

[EBrown]

So I've finally settled on what Band-Pass filters to use and where, the new problem I have is that my amplifier (An LM324 Op-Amp) keeps changing the frequency of the input.

Are there any Amplifier Circuits out there that don't have this effect on the output?

Or is the problem the factor by which I am amplifying the circuit. (1 + R1/R2 = roughly 750-780)

It seems that the center frequency has shifted slightly. (Albeit by a small amount. But shifted none-the-less. It becomes more pronounced as I try to get a higher gain or as the frequency get's higher.)

What can I do to alleviate this? I'm not even at the dB level I need to be at, I need to be at 20dB, but when I try to get even close it has way too much frequency shift. (Image attached)

Thanks,
EBrown

 

[Harald Kapp]

my amplifier (An LM324 Op-Amp) keeps changing the frequency of the input.

What do you mean by that? Your image doesn't tell us much without some information about the underlying circuit and the meaning of the different curves.

The LM324, although a bit old by today's standards, doesn't change the frequency of a signal. It may introduce noise and distortion above what you'd consider distortion free. But for audio applications it still is a usable part. A problem when using it in a circuit with high impedances may be it's comparatively high input current and limited gain. These parameters will shift the actual center frequency of a filter a bit compared to the theoretical ideal, but not by much in the range that's of interest here.

This tool takes into account the actual parameters of the opamp used. Unfortunately the free version includes only ua741 and lf411 and an ideal opamp. You may use this version anyway to evaluate the influence of a non-ideal opamp vs. an ideal opamp.

I need to be at 20dB

At which frequency do you need the -20dB? Before you mentioned only the -3dB point. You are aware that you may need more than one stage of filter if you want steep slopes, are you?

 

[EBrown]

What do you mean by that? Your image doesn't tell us much without some information about the underlying circuit and the meaning of the different curves.

The LM324, although a bit old by today's standards, doesn't change the frequency of a signal. It may introduce noise and distortion above what you'd consider distortion free. But for audio applications it still is a usable part. A problem when using it in a circuit with high impedances may be it's comparatively high input current and limited gain. These parameters will shift the actual center frequency of a filter a bit compared to the theoretical ideal, but not by much in the range that's of interest here.

This tool takes into account the actual parameters of the opamp used. Unfortunately the free version includes only ua741 and lf411 and an ideal opamp. You may use this version anyway to evaluate the influence of a non-ideal opamp vs. an ideal opamp.

At which frequency do you need the -20dB? Before you mentioned only the -3dB point. You are aware that you may need more than one stage of filter if you want steep slopes, are you?

I don't need -20dB, I need to amplify every frequency so that the center frequency is at roughly 20dB. Ignore the -3dB point for now, as I have overcome that issue.

Right now, the problem is that when I amplify my signal from ~-60dB to even ~-6dB the Op-Amp is shifting my higher frequencies by an increasing amount for each higher frequency. They are all shifting to become slightly lower frequencies. I need to get all the signals to ~-6dB right now, then I need to increase them again to 20dB. The -6dB signals are what will be sent to the audio output, the 20dB signals will be sent to the VU Meter which uses everything from -4dB to 20dB to display that particular frequencies strength. (Where 20dB is a .447V input on that frequency which the VU Chip would see as a 10V input on the Vin Leg.) However, as I try to amplify things from -60dB to -6dB I am getting a shift on the op-amp, or I suppose a low-pass filter as that web-page you mentioned states.

On the image above, the Pre-Amp Curve comes from the node attached to the + coming from R67, whereas the Post-Amp curve comes from the node attached to the tip of the Op-Amp. I don't think the image I attached is for that particular frequency, but the design is exactly the same. All the Op-Amps use the same resistance on both resistors. The only difference is that R67 is 4.02k.

My question is, how can I overcome the Op-Amps internal Low-Pass filter characteristics for my higher frequencies?

Thanks,
EBrown