Alternative to Decoupling Cap for LF

Why do you need to decouple? What are the specs?

Not enough info.

Cheers

Phil Hobbs

You might be better off with a bandpass filter instead.

Insufficient data to make sensible suggestions.

The question was theoretical.

Is it possible to use another arrangement of components apart from a large cap (in this case) to decouple the signal path into and out of a two-pole active filter.

If it's a "no" there is no point going any further.

I will try the bandpass suggestion, just to see what it does.

Maybe making the input cap function as part of the filter will solve half the problem.

That's a start.

Thanks,

Kevein Foster

Theoretically... transformer resistors/diodes as a dc shifter opamp dc level shifter or just a dc coupled system optoisolator

NT

Well you could do it all digitally...

Does the active filter itself not have big caps too?

Well, you can use a small cap, a big resistor, and an op amp. There'll be a bit of a noise penalty, of course.

Cheers

Phil Hobbs

You're talking about start-up issues.

Are you using a single supply, or +/-?

A balanced supply might work better, or avoid decoupling issues in general.

For single supply, you might decouple at a different point in the signal path, to reduce distortion effects but the natural time constant will still be there at sub hz frequency.

You might force settling time at a signal null point, or servo a node dynamically using a nonlinear response, but this would result in larger earlier disturbances at start-up.

RL

I think you may have a differnet definition for the term "decouple".

Are you talking about power rail bypassing or something else?

Decoupling caps usually refer to power rail bypass caps.

If that is what you mean, active power supply regulation would be your solution.

Mark

Transistor, collector fed with a current source, bias DC current to base from negative fed back opamp with LP filter? Would work (or beter: might work, as I haven't played with this for a long time) if the DC (or average) voltage swing's frequency is much lower than that of the signal.

joe

An LPF passes DC so you don't need any.

If you were actually making a bandpass, that would explain the long settling time!

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

I need to decouple a two pole active LP filter situated between two op amps. The signal is a 1 Hertz (yes) arbitrary waveform.

If I use large caps on each side (1000uF), it introduces distortion and has quite a long settling time prior to first charge.

Generally speaking, is there another decoupling method without these drawbacks that could be applied?

Kevin Foster

I think the OP is talking about the coupling capacitors not the supply decoupling.

Oh, you said "decouple". What you mean is "couple" or "AC couple". You only need caps as large as your bandwidth (and phase) suggests. Capacitive coupling is essentially a single-pole low-pass filter. The way to use smaller caps is to use larger resistors (for the same bandwidth). However, since you're having start-up time problems it suggests that your low frequency roll off is too low.

It *is* part of the filter.

Yes, absolutely. For starters, instead of a series cap you could use an inductor in shunt. Of course, in reality it'll be even bigger than the cap...

If you wanted a real-world solution then you could use a gyrator in shunt (a gyrator is an op-amp circuit that pretends to be an inductance to ground). With the right op-amp and capacitor you could make it smaller than a "large cap".

However, you'd probably be as well off to choose a smaller cap, then choose the appropriate resistance to ground to get your pole where you wanted it, then follow that with an op-amp voltage follower. The smaller the cap the higher the resistance you need, this will lead to more voltage noise both from thermal effects and from input noise current on the op-amp, and to more stringent requirements on the op-amp's leakage current.

I know it's anathema, but at the frequencies you're interested in you may actually get superior noise performance in less space than a capacitor with a really good ADC, a microprocessor, and a really good DAC, all connected up to act as a gyrator. It's a very twisted way to do things, and DC coupling is probably miles better for most things, but I've seen roughly similar sorts of things done with infra-red imaging systems that worked quite well.

It's not breaking any physical laws, so it's not "no". Just "is it wise?"

Yes, of course. A transformer. Or, inductor or small capacitor (you have an op amp, use large resistors and buffer the output). The problem, is that such couplings have dynamic range, and frequency limitations, and we don't have ANY numbers, except one frequency (not a range) and one capacitance (1000 uF) which is referred to as 'large'.

Large compared to what?

One can also decouple by doing A/D conversion, subtracting the unwanted signal, and doing D/A conversion. Quantization noise might mean that no benefit is conferred, though.

Use a dual rail supply and DC couple all of it.

Power supply start up can be tweaked and select a good set of OP-AMPS to use that have good power supply rejection ratio and make sure the supply rails are symmetical in level. I normally use dual channels on the scope or a diff probe to track that.

Jamie

Thanks to everyone for their numerous helpful suggestions.

I will try a few on the bench this weekend.

However, a coupling transformer is not an option post filter due to the sub-audio frequency.

Kevin Foster

at lower voltage a transformer will work to proportionately lower frequency.

Transformer does not do too well as frequency decreases... ...and is an absurd "solution" at 1Hz.

Its not a good solution in most cases, but it is possible. A 240v 50Hz transformer is as happy on 5v 1Hz. (In practice more).

NT