free book on lock-ins

Sure there's no doubt a DSO is a poor lockin. (But lots of time's you don't need the big guns and it's an OK tool at the right place...the crescent wrench turns into a hammer sometimes.) I didn't get a chance to try the multiple function. But I'm going to guess that it won't multiply and then average. But only average first and then multiply. (which would be worthless in this case.)

George H.

What I was getting at is that sometimes one needs to find a needle in a haystack. A scope is the usual tool and 8 bits is all that's available. With averaging (or a lock-in, though I hadn't considered it) one can get find a known signal in a lot of noise. If multiplying it by the reference signal1 finds it faster than simply averaging, cool.

Absolutely. Horses for courses, and all that.

Cheers

Phil Hobbs

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

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

iplying ch-1 X ch-2 (which can also be done in the DSO)? Have never tried that.

e if that works as well.

onentially longer to get the same SNR.

is good enough (which it probably isn't) and the noise happens to be an ad equate dither source, you'll need to average 2**(N-6) traces to get N bits, i.e. it's exponentially slow.

The one time when we were really serious about this was with a stroboscopic electron beam tester. The electron beam we were using wasn't up to deliver ing much more than 1,000 electrons in a 20nsec interval - a 10nA beam - and mostly we worked with shorter intervals and at slightly lower current beam .

The square root of 1000 is close enough to 32 (2^5) so six bits was anythin g from fine to a massive over-kill (and we could use fewer). Photomultiplie r non-linearities weren't going to be big enough to matter at that level - I'd engineered the divider chain to make sure of that - so we could leave t he hardware to crunch data all day if we wanted to. The stroboscopic electr on beam tester was going to retail for a about $250,000 so we didn't expect that our customers were going to do that any too often, but we did put wid e enough counters and registers that they could.

The machine was the kind of stroboscope that can sample several times (up t o 1024 in our case) per cycle, so heavy number crunching was less time-cons uming that it might have been.

Absolutely.

Cheers

Phil Hobbs

Sounds like a phase-locked loop - what's the difference

Loads of free books here

I suspect they fell off the back of a lorry though.

More like correlation.

In lock-in detection, the reference doesn't follow the signal, i.e. you can easily detect a static phase shift.

Cheers

Phil Hobbs

A phase-locked loop generates a reference signal by observing a noisy input. A phase-sensitive amplifier determines, from a reference signal and noisy input, what the average amplitude is.

Main difference, is that a PLL generates a 'clean' frequency (often square-wave, of indeterminate amplitude), while a PSA determines an amplitude (volts, amps, lumens, or centimeters of tide...) of an oscillating signal.

You can use a PLL to demodulate some kinds of modulation on a signal. You can measure AM with a PLL much the same way as in a two-channel lock-in, but angle modulation will get modified by the loop characteristics. A PLL does a good job at demodulating PM at frequencies _outside_ its loop bandwidth, and FM _inside_. (You take the PM output from the phase detector and the FM output from the VCO input.) It does a lousy job at both, if the modulation frequency is too close to the loop bandwidth.

Provided you aren't using a limiting amplifier or a hard-switching phase detector, you can maintain lock when the input SNR goes below unity. (It can't go that low when measured in the loop bandwidth, of course.)

Cheers

Phil Hobbs