# Is it possible to filter a frequency?

I have built and tested an active band pass filter, with the intention of creating a speed detector, whose lower and upper cut off frequencies are 48 Hz and 1591 Hz respectively. During testing I confirmed attenuation of voltages out-with the bandwidth above and subsequent amplification. This was using a signal generator where of course, I could send the desired frequency through the circuit.

When I began using the radar module, transmitted frequency is 24 GHz (https://shop.bb-sensors.com/out/media/Datasheet_RADAR-IPM-165_DBE_Rev01.pdf), the received frequency was around 80 MHz and of course, voltage at this range even after amplification was extremely small.

Is there any way I can control the received frequency so I can analyse the frequency? If this isn't something that is possible, is there any other way I can analyse the received frequency?

I'm completely running out of ideas and thought this almost feels like last chance saloon.

I've added pictures of the circuit, radar input and circuit input and signal generator input and output.

As can be seen from the circuit diagram below, the radar is the input to the entire circuit. Both oscilloscope images display inputs and outputs. The outputs appear to be similar when a very large frequency is applied to the circuit, much larger than that of the upper cut off frequency.

• It sounds like you're putting a very high frequency into a low-pass filter, which attenuates it... If you're trying to detect doppler shift from the radar module, seems like something is missing from the system, don't you need a mixer to bring down the 24GHz carrier to baseband? Maybe a direct conversion or superheterodyne receiver? – MarkU Apr 12 '18 at 21:43
• Take a look at this application note: innosent.de/fileadmin/media/dokumente/Downloads/… – MrGerber Apr 12 '18 at 21:53
• @Jandy12 (1) Your scope says 40MHz on it. What do you think that means? (2) The limpkin site discusses opamps with only 1MHz gain bandwidth. Did you notice? (3) Either the short moments I spent reading weren't enough, or else I don't see a design that follows what I read as good advice in that whitepaper. So I think one of us has a reading problem. I'm willing to grant it is me. Perhaps someone else will point out the obvious. I'm too tired tonight to retrace my steps. But I do believe their product is good. – jonk Apr 13 '18 at 8:05
• I see a a 90 Hz sine wave with spikes and possible sampling errors, noise but I don’t see a block diagram and clear question showing a block diagram. Which part of the application note Do you not understand – Sunnyskyguy EE75 Apr 13 '18 at 13:44
• @Jandy12 The DC amplification comment they made is because they say the offset voltage might be near $200\:\text{mV}$ (in fact, they say you can use this fact to detect that it is ON) and that their signal will ride on that as much as $300\:\mu\text{V}$. We're talking 1000:1 here, ignoring noise issues. They are suggesting that you first do a simple DC gain of about 10 or so (which will amplify both the offset and the signal from $200\:\text{mV}\pm 300\:\mu\text{V}$ to $2\:\text{V}\pm 3\:\text{mV}$) before applying a blocking capacitor. The signal will be $3\:\text{mV}$ by then. – jonk Apr 13 '18 at 18:47

Any filter can be designed in theory to any practical specifications.

No filter can be designed with inadequate specifications.

A filter must not distort the signal amplitude in the passband (PB) with gain
But must also attenuate noise in the stopband (SB).

Thus a minimum of 5 parameters are needed for a LPF.

A full answer is not possible, because you MUST define signal spectral density ( f vs amplitude )and noise spectral density or amplitude an PW50 ( 50% pulse width ). An optimal filter may be selected by matching filter response to signal response and inverse of noise response.

There may be other specs from group delay.

Hypothetical             Spec
======================== ==========
Av (gain)                1000  2%
Ripple (dB)  in PB gain  0.5dB max
BWp of PB (Hz)           1591 Hz
BWs of SB (Hz)           5000 Hz  ( knowing SB/PS = -6dB/oct per order)
SB rejection at BWs (dB) -60dB


I see in the pictures that you have a lot of high frequency noise coming from radar output to the OA input.

It's true that overall the HF attenuation might be correct but at the U1 output the noise is still amplified. Over 0.5V noise with 30x U1 gain will lead to over 15V at U1 output making U1 and U2 working in non linear region.

Of course, if the frequency is high the noise will be filtered by R4 and C2 at the output but no LF small signal will survive.

MrGerber's app note shows two capacitors added in parallel with the feedback resistors (R3 and R6 in your schematic) to early filter the HF noise. In your case the capacitor value should be ~4.7nF for 600Hz (human move) or lower for higher object speed.

• Thanks for your help thats made a bit of sense to me. So, you're suggesting a capacitor in parallel with the feedback resistor at each stage, while the rest of the circuit should remain the same? Or using the capacitor in parallel with the feedback resistor and removing the low pass filter that's after the output of U2? – Jandy12 Apr 13 '18 at 11:32
• The output filter has a cutoff frequency of 1.5 khz, it makes no difference, let-it there until you figure out what happens. Later you can remove-it. The cutoff frequency for OA filters with 4.7nF capacitors is more like 1khz than 600 Hz. Everything else is good as is. – Dorian Apr 13 '18 at 11:44