Active filter with built-in rectifier ? ? ?

[mags]

Hello everyone, I'm new to the forum.
Hopefully this question will be simple for an analog expert.

I have a project where i need to isolate a certain frequency and measure its strength.
I plan to use a Sallen-Key bandpass filter, rectify the output from it, and feed that onto a leaky capacitor (integrator) where i can then measure the voltage from time to time.

I could simply place a rectifying diode between the filter and the integrator, but then i will lose any signal smaller than 0.7 V.
What i would prefer to do is this: (see first image SK1)


As the diode is inside the feedback loop then its voltage drop should be compensated for (?).
My question is, would it work, or will the diode upset the workings of the filter?
I've seen some circuits with the diode within the feedback loop, e.g. here:
http://www.tu-varna.bg/tu-varnaetm/images/IE_EN/L11_Rectifiers.pdf
But those aren't filters.

Failing that, i could compensate at one of the inputs, like this: (see second image SK2)

The trouble being that if the two diodes don't have exactly the same Vf then it will introduce an offset error. But i should be more able to live with that.

I want a simple solution that adds the minimum of extra components, as space will be tight on the board.
It doesn't have to be really accurate, provided i don't lose those weak signals!
I don't yet have much experience in electronics, so maybe i overlooked some obvious solution. Any help would be really appreciated.

I'll be out for the next week, so happy Christmas all. 

[ejeffrey]

So if you google for precision rectifier you will find a variety of circuits that work like your first one.  They use an op-amp to compensate for the diode drop.  They tend to be fairly slow.  The problem is that when the diode is not conducting the op-amp has no feedback and the output is all the way at the rail.  When the input changes sign, the output has to race all the way back to where it is supposed to be.  Also, I am not sure your circuit as drawn will work. I think the diode will interfere with the operation of the bandpass filter.  It would be much easier to use a second opamp (use a dual opamp package) as the rectifier.

If your signal is very narrow in frequency and you have a reference signal that is at the same frequency and in phase with your signal you can use a lock-in technique to recover the signal.  That is just an analog multiplier that multiplies your signal with the reference followed by a low-pass filter.  The desired signal is mixed down to DC and then smoothed by the filter.

[slateraptor]

Just a heads up: your posted pic is not Sallen-Key. What you have is multiple feedback topology, which is clearly identifiable by C1 and R3 in the negative feedback loop, as opposed to S-K, which exploits positive feedback.

Other than that, I'm pretty much in agreement with ejeffrey. Active filters function on the principle of feedback. The inherent problem I see is that, by incorporating a diode into the filter, you've adulterated the system's feedback loop in a non-linear fashion.

Think about what characterizes a LTI system: sinusoidal input yields the same sinusoidal output where the only properties that the filter can modify are amplitude and phase (spectral content remains the same).

Now think about the consequences of incorporating a diode (non-linear component) into the feedback loop: sinusoidal input yields rectified output (implies a change in spectral content), which is fed back into the system to drive transients.

Without proper analysis and simulation, it's difficult to predict the outcome, but as it is, prospects don't look very promising.

As noted by ejeffrey, it would be much simpler to cascade a super diode as its own stage:

input --> [BP] --> [super diode] --> [integrator] --> output

...or perhaps more advanced super diode topologies that can be exploited; or if you're up to the challenge (and budget), a PLL-based sol'n.

[mags]

Hi, I'm back now.
Thanks to you both for your suggestions.
I'ts clear to me that i shouldn't have confidence in my first idea.
Using a precision rectifier or supersdiode is something i'm sure would work. However i didn't explain that i need to detect at least 3 different frequencies, each with its own filter. As far as I'm aware the precision rectifier requires an additional op-amp stage, and i really don't have much space.
Using a PLL did cross my mind, but that would be well beyond my experience. I'd need to find a way to get it to measure the strength of the AM signal as well as simply detecting it. Also i'd need to continually retune it to measure each of the 3 frequencies (don't think i can stretch to 3x PLLs). Does retuning and acquisition take time?

I will probably now settle on my second solution using an offset on the input side, to precompensate for the diode after the output side.