Continue to Site

# Help with Active Bandpass Filter

Status
Not open for further replies.

#### Gandledorf

##### New Member
Well I've finally worked out what I think is the most appropriate bandpass filter for my 56kHz digital signal. It's an active state variable filter, with Q and G independantly settable.

Horowitz and Hill give the following equations for this filter:

f0 = 1 / (2 * pi * Rf * C)
Q = R1 / Rq
G = R1 / Rq
R = 10k (noncritical)

so for my filter I am planning to use:

Rf = 15k
R1 = 5k
Rq = 10k POT (set to around 100ohms)
Rg = 10k POT (set to around 5k)
C = 180pF
R = 10k

I'm assuming that because no one reponded, my design looks fine? I went ahead and made a few more revisions.

For one thing, I decided that it would be nice if my device could calibrate itself, as if I make the user twiddle with dials and knobs to calibrate, he probably won't get the best results, and this is supposed to be a fairly precise peice of instrumentation. Looking around, I found some digital potentiometers by catalyst semiconductor, they apparently come in 10k 50k and 100k models. This seems like a great solution, am I correct in this? Also, it doesn't list what the taps are (it only says there are 100 of them). Should I assume they are in a linear configuration, and thus 100ohms per tap?

Here is the data sheet, if anyone can glean something from it that I have missed.

Also, is it ok to wire the two POT's for Rf as I've shown? I am assuming the 5113 is accurate enough that this will allow me to twiddle the frequency on my device.

So, for my new values:

f0 = 1 / (2 * pi * Rf * C)
Q = R1 / Rq
G = R1 / Rq
R = 10k (noncritical)

Rq = 5113 10k POT (normally set to 100 ohms)
Rg = 5113 10k POT (normally set to 5k)
Rf = 5113 10k POT (normally set to 4.2k)
C = 680pF
R1 = 5k
R = 10k

Additionally, for the processing of my signal (an analog voltage level), is it best to use the built in gain stage, a pre-filter gain stage, or a post-filter gain stage? Possibly a combination of the three?

Final question, I promise How do I best select which chip to use for my op-amps?

The key to step size is "linear taper" (from the datasheet), which is a holdover from mechanical pots. Linear taper means that the slope of the ohms vs rotation angle graph is linear, so yes, in your case it means 100 ohms/tap. Keep in mind that all three of your adjustments are reciprocal functions. This means that they will have lots of sensitivity on the low end of the pot, and vice-versa.
The resistance tolerance of the pots is +/- 15%. The effect of the ganged Rf pots may depend on how well these two are matched.
Regarding your choice of op amps: What power supply voltage(s) do you have available?
I am still curious about what sort of modulation scheme you are using to put data on your 56kHz carrier. Care to share that with us?

Thanks for the clarification on the pots, these digital types are very interesting, opens up a lot of doors for computer controlled calibration.

My power supply is a battery, run through a +5V @ 500mA regulator. I'm unsure of which cell type I'll be using as of yet.

For this instance, there will be no data on the 56kHz line per se. Instead, an IR emitter will be producing the 56kHz carrier wave, and projecting it across a medium (turbid water). Aligned with it is a phototransistor, which feeds it's output into the bandpass filter. What I am looking for, is an analog voltage level, which should be roughly proportional to the % transference of the IR beam. This level is then used as an estimate for the turbidity of the water.

My basic circuit consists of the phototransistor, a bandpass filter, and then the uC, which has a built in ADC. I'm currently planning on software demodulation, unless an LM567 would work better? I've heard they can be used to demodulate. I have also debated on having a pre-filter gain stage, and/or a post-filter gain stage, but I am unsure of whether this would be wise, or beneficial.

With your predilection for digital design, I'm surprised you're not looking at a DSP chip for all this. You could do all your filtering and detection digitally, with no worries about drift of your carrier and/or your filter center frequency (I'm assuming your transmitter is close enough that you could drive it from the same system that is processing the received waveform).

For the op amps, I would look for a quad if possible, with rail-to-rail I/O, and a gain-bandwidth product (GBW) of at least 10MHz. This probably seems high, but remember that even the 3rd harmonic of your carrier is at 168kHz. If you had a GBW of only 1MHz (common), your open loop gain at that frequency would be 6. This is not enough for good filter performance. High gain settings also require high GBW. Also, make sure they will work on a 5 volt supply. Have a look at TI, Maxim, Linear Technology, National, and maybe Fairchild and On Semiconductor.
If you need more gain than the filter can provide, I guess I would put it in front of the filter, so the filter can operate on the noise that the amplifier generates or picks up.

The main reason I'm not using a DSP is because for this purpose, it's fairly easy to do the processing with a very inexpensive AVR. The instruments are for use mainly at universities and research labs, for which a low cost solution is the #1 priority.

I wasn't asking for more gain, per se, but whether it would be more effective, and less noisey to do a separate gain stage? I don't imagine we'll need a lot of gain, as we're only scaling up the highest signal to 5V.

It's been far to long since I've worked with filters, could you remind me the effect of the harmonics? I remember covering vaguely, but not enough to make sense of what you said.

I may have been a little off base with my comments about harmonics. I lost track of the fact that you are trying to get rid of noise, not harmonics. Nevertheless, my GBW recommendation stands. For some refreshers on harmonics, do a Google search on "'square wave' fourier" (without the quotation marks).

Thanks for all the suggestions Ron, I really appreciate them, I do have a quick question: Is there a table of common op-amps anywhere? I've been digging through data sheets, without much forknowledge of even the range of the values for the op-amp's I'm looking at, and it occurred to me that there is probably a better reference, is this true?

Also, how can I determine if something is "rail to rail I/O", for example, what in the following data sheet indicates whether this device has rail to rail I/O?

https://www.electro-tech-online.com/custompdfs/2004/02/242026TI.pdf

EDIT: Digikey had a somewhat decent search, and I found the following op-amp, is this appropriate?

https://www.electro-tech-online.com/custompdfs/2004/02/LT149820LT1499.pdf

I don't know of a "common op amp" table. I generally use the mfr's search tools. The LT1499 you found looks like a good candidate. I guess cost would be the parameter that might cause you to do a thorough search.

Rail-to-rail capability is usually mentioned in the blurb at the top of the datasheet. The one below does not have rail-to-rail capability. Look at the highlighted specs. If the input is rail-to-rail, the common mode range will include the supply voltages. If the output is rail-to-rail, the output voltages will be within a few tens of millivolts of the supply voltages.
You should take a look at Linear Tech's LT1568, which is an active filter chip. You can add 4 resistors and 2 capacitors, plus a couple of decoupling caps, and wind up with a 4th order bandpass. I haven't looked at it carefully. I just saw an app note in an advertisement in **broken link removed**. I doubt you will find the ad on their web page. You can get a free subscription if you qualify.

#### Attachments

• tlc274specs.gif
32.2 KB · Views: 805
Status
Not open for further replies.