free book on lock-ins

Thanks, John. I have downloaded and am looking forward to reading it.

John S

Yep looks interesting, I never did figure out how lock in's work, makes me feel better knowing there is a whole book written about them though :)

cheers, Jamie

It's really simple. Given a noisy signal from some process, multiply it by a synchronous reference signal, and average, filter, or integrate the product. The coherent part of the signal, the part that aligns time-wise with the reference, builds up faster in the output than the noise does.

It behaves like a narrowband filter, only better. It can dig a signal out of so much noise that you can't see the signal on an oscilloscope.

It's a lot like doing a controlled experiment many times. Take that pill on alternate months and record how you feel, and look at the long-term correlation.

That's the key fact that a lot of folks don't get: a single-channel lock-in works exactly like a narrow bandpass filter followed by a phase sensitive detector. You'll detect noise, signal, radio signals from little green men, anything that's there and isn't in exact quadrature with the reference.

Well, so can a DVM. ;)

Cheers

Phil Hobbs

Not when the "noise" includes DC offset!

Hi John,

Thanks for the book reference. Lock-in amplifiers are great tools. They can give poor results if unwanted signals happen to be coherent with the reference. Randomizing the phase of the reference signal helps eliminate this problem. If the reference is a square wave, then it's direct-sequence-spread-spectrum.

ChesterW

So with a DSO, you can send the signal into chan 1. and the reference into chan 2... trigger on 2 and average.. a poor man's lockin.

George H.

I've done that for >30 years. I thought JL was talking about multiplying ch-1 X ch-2 (which can also be done in the DSO)? Have never tried that.

SO if you have a DC signal with noise, a lock-in can clean that up? Or is it liminted to AC signals?

I'm thinking along the lines of cleaning up noisy feedback signals.

Cheers

If the input signal has a DC offset, and you run it through a lock-in, the DC gets multiplied alternately by +1 and -1 (assuming a square-wave ref input to the multiplier) so the average output will be zero.

The trick is to have a signal and a coherent phase reference. That's done, for example, by modulating light with a chopper wheel, where the chopper driver or a pickup on the wheel provides the phase reference. Very weak, noisy signals form some photodetector can then be pulled out of the noise with a lock-in.

A balanced mixer, in an RF system, is pretty much the same thing.

The first version generates Lissajous figures, which can be informative.

What John Larkin was talking about would be closer to multiplying ch-1 by the trigger channel. The catch is that the output ought to be low-pass filtered after multiplication, which DSO's aren't really designed for.

t.

kes me feel better knowing there is a whole book written about them though :)

it by a synchronous reference signal, and average, filter, or integrate the product. The coherent part of the signal, the part that aligns time-wise w ith the reference, builds up faster in the output than the noise does.

ck-in works exactly like a narrow bandpass filter followed by a phase sensitive detector. You'll detect noise, signal, radio signals from little green men, anything that's there and isn't in exact quadrature with the reference.

l on an oscilloscope.

No.

Entirely.But if you can chop your desired DC signal in a controlled way, an d detect the content in what you look at that is at the chopping frequency, that can work. That's exactly what "chopper amplifiers" do.

That can be tricky. The process of averaging and filtering the detected des ired component while getting rid of the noise put a nasty lag in your feed- back loop.

You have to design the feedback algorithm to cope with it - PID with lots o f integral term in most cases.

Cute control schemes use tricks like second harmonic generation to get the signal that they monitor at a frequency that easy to select. ,Phil Hobbs pr obably has a chapter on that kind of trick.

Hmm, I've think one of our 'scopes does the multiply thing. I'll see if that works as well.

George H.

(To Bill S. The average function on the DSO will do the job of the low pass filter.)

The problem is the 6-8 effective bits.

Cheers

Phil Hobbs

On Thursday, August 7, 2014 9:11:00 PM UTC-7, Bill Sloman wrote: [about lockin amps and oscilloscope visualization of correlations]

Not true low-pass filtering, but they can do IIR (infinite impulse response) ; really, just summing several sweeps together. It takes a bit of searching through the menus for math mode, sometimes, to find this option.

But if you've already got (uncorrelated) noise...

...you can take exponentially longer to get the same SNR.

Cheers

Phil Hobbs

Not if you only have 6-8.

gnal out of so much noise that you can't see the signal on an oscilloscope.

nce into chan 2... trigger on 2 and average.. a poor man' lockin.

Sustained bad logic here. If your initial signal is noisy enough that the r ounding error introduced by only digitising it is smaller than the noise in the original signal (it's an rms sum, so any rounding error smaller than a bout half the original noise doesn't make much difference) the number of ob servations that you have to integrate over to get a particular signal to no ise isn't much affected.

Of course, if the noise on the original signal isn't big enough to dither out the rounding error, no amount of averaging is going to move you away fr om the most-probable digital output (the particular step on the stair-case where your signal finds itself) unless you are clever enough to add your ow n dithering.