Active filter with gain

It's not a bad idea: if you do half sampling, then you only need to filter ~4 AA bands, which can be nuked with zeroes in the stopband. (Another advantage of the coupled resonator design: coupling between non-adjacent resonators provides zeroes.)

The filter doesn't need to be exceptionally sharp either, because you can clean up the passband in DSP. (Unless you need it sharp for dynamic range purposes.)

Tim

If you only need the phase, how about using a lock-in? A CMOS dual 2->1 MUX driven in quadrature with the signal, an RC lowpass, and then your delta-sigma. Moving the passband down to DC like that will make the filtering a whole lot easier.

Cheers

Phil Hobbs

But how do I lock in to a weak, noisy input signal? I mean, in a considerably simpler way than just extracting the signal with a good analog band-pass filter?

How is the mux supposed to work?

This is more or less what I want to do, but with an on-chip sampling mixer. The simplest possible experiment (connected a speaker) proved the approach is viable, setting the LO 1kHz below the input signal produces audible beeps. But the environment is so noisy that the beeps are there only if I disable most of my equipment. Hence the need for a sharp analog input filter. I know I can dig the signal out of the noise floor using the well known correlation-based detection techniques, but hey, why should I intentionally make the detector's life harder?

Best regards, Piotr

Use an ANALOG phase detector... aka a multiplier. Will extract signals buried beneath the noise.

...Jim Thompson

First of all, I'd like to thank all of you for your input.

Very tempting, but I don't have the necessary equipment to ensure the results are fine. In the case of the LT part it is LT that makes sure the device will work as expecteed. Obtaining tunable pot cores today would be pretty hard challenge too. Nobody sells them and even if they do, they prefer to deliver in quantity. Then the LT part which costs an arm and a leg looks pretty cheap...

But then wouldn't the result be isomporhic to what LT1562 does?

Best regards, Piotr

Believe me, a lot of things are simpler than the filter you're proposing. That thing will be a production headache of no small size.

You can't measure the phase modulation if you don't have a phase reference of some sort--CPU timing, clock recovery, a tank circuit, or something like that.

If you already possess the phase reference, then you can use a 1-phase lock-in amplifier (not necessarily a PLL).

If not, you can use an analog PLL whose bandwidth is much less than that of the signal to pull it out of the noise. Floyd Gardner's classic "Phaselock Techniques" is a really good read on that sort of stuff.

Cheers

Phil Hobbs

Would you consider an opamp switching between +1 and -1 gain in sync with the LO a good enough analog multiplier or nothing below a full-blown Gilbert cell counts?

Best regards, Piotr

I'm late to the conversation... what frequency are we trying to lock to? It's all about the speed of the OpAmp if you want to do the +1/-1 trick. I've done that at low audio years ago...

There will be no production, one specimen is fine. BTW, why?

I have an NCO with ~8uHz resolution and 50% duty, derived from a 24MHz 10ppm clock. It will be synchronized by a PID to the input signal. That's pretty close to what a PLL would do, so yes, this is my reference.

I have never done anything with lock-in amplifiers in practice.

Best regards, Piotr

"Phaselock Techniques", Agreed! Though Floyd himself told me once that my circuit couldn't be working... but it _was_ >:-} ...Jim Thompson

Precisely 77.5kHz.

It's not up to me, it is one of the operating modes of a ready-made Switched Capacitor/Continuous Time block of PSoc5LP:

They say it can go up to 500kHz in the +1/-1 mode (continuous-time up-mixer in their parlance, which is a kind of BS, as it can downsample as well, effectively convolving the input signal with the square wave LO).

So my definition of simplicity is embedded within this very specific context: use as many existing on-chip analog blocks as possible, without the need for external parts or even pins.

Best regards, Piotr

I have looked into that function before, for the PSoC 1, and AFAIR there is several application notes including measurements that confirm the function

That switched cap block is really nice

Cheers

Klaus

Although to my ear the sampled mixer *sounds* better than its CT counterpart. :->

Best regards, Piotr

That would be a "suitable transformation". :)

Give or take isolation, because you have voltage outputs there, whereas resonators just resonate. But that tends to make filter design easier, as you only need to consider pairs of poles, and as long as you have some means of coupling the poles themselves (which you do, it's general enough for that), you can realize poles and zeroes as desired.

Tim

that my circuit couldn't

You probably explained it wrong. ;)

Cheers

Phil Hobbs

It was a breadboard right in front of him, with attached 'scope probes ;-) ...Jim Thompson

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | STV, Queen Creek, AZ 85142 Skype: skypeanalog | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

I'm looking for work... see my website.

Thinking outside the box...producing elegant & economic solutions.

It is odd that Floyd Gardner's book doesn't refer to Jim's MC4024 and MC4044 circuits, though it describes exactly what the MC4044 did.

I've wonder why here before. May be Floyd Gardner found Jim as loveable as I do.

The last time I looked I found a stocks of a pot core and compatible tuning slugs - you screw them down a cylindrical hole in the centre of the pot co re to bridge the gap in the centre to an adjustable extent - on the web-sit e of a US broad-line distributor. John Larkin was the sceptic being aimed a t.

I don't think that anybody stocks equivalent parts for RM cores (if they ev er did) but there does seem to be a legacy market for the even older pot co res.

Should be.

One other approach that might be easier if you have a couple of timer outputs free is a dual-conversion superhet. Mix up to 455 kHz, filter with an inexpensive ceramic filter, then mix down to some very low IF, filter that, then digitize.

Cheers

Phil Hobbs

Who has to quit goofing off on usenet and get back to reading a lidar patent.