lock-in amplifiers

I know a fair bit about PLLs but nothing of lock-in amplifiers until I looked it up. It appears it is just a phase detector ie multiplier like you have in some analogue PLLs. Why call it lock-in in the first place if it is just a phase detector? Of course if it is used as part of a larger system with a frequency reference then fair enough, but that is getting back to a PLL. It is like they invented the wheel.


Hardy

On 15.03.14 07.55, gyansorova@gmail.com wrote:
> I know a fair bit about PLLs but nothing of lock-in amplifiers until I looked it up. It appears it is just a phase detector ie multiplier like you have in some analogue PLLs. Why call it lock-in in the first place if it is just a phase detector? Of course if it is used as part of a larger system with a frequency reference then fair enough, but that is getting back to a PLL. It is like they invented the wheel.

AFAIK a PLL is only needed for technical reasons. You can do lock in 
even without any PLL by filtering the reference signal until it is 
reasonably close to a sine wave and shifting the result by 90 deg in 
phase. Then you need to multiply the sin/cos reference with the input 
signal. So I would basically call it a multiplier.
Of course, instead of complex conditioning of the reference you could 
also use a PLL to create a new clean reference, especially if you need a 
large multiple of the reference frequency as DAC clock. The digital 
implementation is just simpler. Additionally it might work for non sine 
references too by using a cross correlation.


Marcel

Marcel M&#4294967295;ller <news.5.maazl@spamgourmet.org> wrote:
> On 15.03.14 07.55, gyansorova@gmail.com wrote:
>> I know a fair bit about PLLs but nothing of lock-in amplifiers 
>> until I looked it up. It appears it is just a phase detector 
>> ie multiplier like you have in some analogue PLLs. 
>> Why call it lock-in in the first place if it is just a 
>> phase detector? Of course if it is used as part of a larger 
>> system with a frequency reference then fair enough, but that 
>> is getting back to a PLL. It is like they invented the wheel.

A lock-in amplifier is a piece of laboratory equipment, amplifiers,
phase detector, reference oscillator (but you can use an external
reference, too), adjustable low-pass filters, and usually also
meters to display the result.

http://elektrotanya.com/princeton-applied-research_5204_lock-in-analyzer.pdf/download.html

> AFAIK a PLL is only needed for technical reasons. You can do lock in 
> even without any PLL by filtering the reference signal until it is 
> reasonably close to a sine wave and shifting the result by 90 deg in 
> phase. Then you need to multiply the sin/cos reference with the input 
> signal. So I would basically call it a multiplier.

All lock-in amplifiers that I know of use a square wave instead
of sine. I am not completely sure why, though.

> Of course, instead of complex conditioning of the reference you could 
> also use a PLL to create a new clean reference, especially if you need a 
> large multiple of the reference frequency as DAC clock. The digital 
> implementation is just simpler. Additionally it might work for non sine 
> references too by using a cross correlation.

I used to use them in the analog days. I haven't tried a digital 
one yet. Well, putting a digital voltmeter on the output doesn't
count. 

-- glen

glen herrmannsfeldt wrote:
>[snip]
>
> All lock-in amplifiers that I know of use a square wave instead
> of sine. I am not completely sure why, though.
>

Two images to consider:
   1. synchronous rectifier
   2. what happens when you multiply by 
+1,-1,+1,-1,+1,-1,+1,-1,+1,-1,+1,-1,+1,-1...

Richard Owlett <rowlett@pcnetinc.com> wrote:
> glen herrmannsfeldt wrote:
>>[snip]

>> All lock-in amplifiers that I know of use a square wave instead
>> of sine. I am not completely sure why, though.

> Two images to consider:
>   1. synchronous rectifier
>   2. what happens when you multiply by 
> +1,-1,+1,-1,+1,-1,+1,-1,+1,-1,+1,-1,+1,-1...

Square wave lets odd harmonics through, but that normally
isn't a problem. The whole idea is that the signal is synchronous
with the reference and noise isn't. Then a sharp enough low-pass
filter filters out most of the noise.

The one in the link works from 0.5Hz to 100kHz. The internal 
reference oscillator has to lock over that range. 

There is also a 2f option where the input reference is doubled
such that you can read the second harmonic. 

-- glen

On Fri, 14 Mar 2014 23:55:53 -0700 (PDT), gyansorova@gmail.com wrote:

>I know a fair bit about PLLs but nothing of lock-in amplifiers until I look=
>ed it up. It appears it is just a phase detector ie multiplier like you hav=
>e in some analogue PLLs. Why call it lock-in in the first place if it is ju=
>st a phase detector? Of course if it is used as part of a larger system wit=
>h a frequency reference then fair enough, but that is getting back to a PLL=
>. It is like they invented the wheel.

The output of a PLL is just a local phase-locked frequency source.
The amplitude information in the input signal is used only to provide
feedback to the local oscillator, or software equivalent thereof.  The
only component of the detected signal that is needed is the 90 degrees
out of phase component, because that is what provides the feedback.

A lock-in amplifier based on a PLL has all the same stuff, but in
addition to the 90 degrees out of phase component of the detected
signal, it also generates the in-phase component of the detected
signal.  This detected signal is the output of the lock-in amplifier.
The in-phase detection is not needed in a pure PLL, but it is needed
in a lock-in amplifier.  That is the main distinction between the two.

Robert Scott
Hopkins, MN

Robert Scott <no-one@notreal.invalid> wrote:

(snip)
> The output of a PLL is just a local phase-locked frequency source.
> The amplitude information in the input signal is used only to provide
> feedback to the local oscillator, or software equivalent thereof.  The
> only component of the detected signal that is needed is the 90 degrees
> out of phase component, because that is what provides the feedback.
 
> A lock-in amplifier based on a PLL has all the same stuff, but in
> addition to the 90 degrees out of phase component of the detected
> signal, it also generates the in-phase component of the detected
> signal.  This detected signal is the output of the lock-in amplifier.
> The in-phase detection is not needed in a pure PLL, but it is needed
> in a lock-in amplifier.  That is the main distinction between the two.

Some years ago, I was working on a system to measure the IV
characteristics of an electrochemical system. The system could
be consisdered as either a series or parallel RC circuit, measured
as a function of a slowly varying (DC) voltage. 

In the parallel case, you measure the AC current as 
a function of DC voltage, with a small (25mV, so 
about kT) AC component. The device under test is at the 
end of two long (about 2m) coax cables, connected
between the center conductors. One supplies the input voltage
(DC plus reference) the other measures the current across a small
(1, 10, or 100 ohm) resistor. Interestingly, the cable capacitance
doesn't bother the measurement. The in-phase signal is then
proportional to the conductance, the quadrature to the capacitance.

For the series RC case, you measure the voltage across the device
with a small AC current (in addition to the DC offset) through the
device. Again, the AC voltage should be less than kT (25mV).
The current is generated by supplying a voltage to a large 
(maybe 100M ohm) resistor close to the device. Now the voltage
is the sum of in-phase resistance and quadrature 1/C. 

The important parts of the lock-in amplifier, in addition to the
phase sensitive detector, are the amplifiers and filters. The signal
coming in has an interesting AC component of about 25mV. The
noise might be larger, as long as it is small enough near the
reference frequency. Near being the width of the low pass filters
on the output.  

Usual might be a 10kHz reference, and 1s time constant. 

-- glen

On Sat, 15 Mar 2014 09:41:57 +0000 (UTC), glen
herrmannsfeldt <gah@ugcs.caltech.edu> wrote:

>Marcel M&#4294967295;ller <news.5.maazl@spamgourmet.org> wrote:
>> On 15.03.14 07.55, gyansorova@gmail.com wrote:
>>> I know a fair bit about PLLs but nothing of lock-in amplifiers 
>>> until I looked it up. It appears it is just a phase detector 
>>> ie multiplier like you have in some analogue PLLs. 
>>> Why call it lock-in in the first place if it is just a 
>>> phase detector? Of course if it is used as part of a larger 
>>> system with a frequency reference then fair enough, but that 
>>> is getting back to a PLL. It is like they invented the wheel.
>
>A lock-in amplifier is a piece of laboratory equipment, amplifiers,
>phase detector, reference oscillator (but you can use an external
>reference, too), adjustable low-pass filters, and usually also
>meters to display the result.
>
>http://elektrotanya.com/princeton-applied-research_5204_lock-in-analyzer.pdf/download.html
>
>> AFAIK a PLL is only needed for technical reasons. You can do lock in 
>> even without any PLL by filtering the reference signal until it is 
>> reasonably close to a sine wave and shifting the result by 90 deg in 
>> phase. Then you need to multiply the sin/cos reference with the input 
>> signal. So I would basically call it a multiplier.
>
>All lock-in amplifiers that I know of use a square wave instead
>of sine. I am not completely sure why, though.

The original analog lock-ins used square waves, because they
were easy to implement:  Just switch between the signal and
its inverse.  They responded to all the odd harmonics in the
square wave, but often times that wasn't an issue.

Back in those days, there wasn't really a good
low-distortion way to multiply analog signals. The earliest
lock-ins to use sine wave multipliers used pulse-width
modulation... ugh!

But now, I think that everybody uses DSP methods and true
sine waves.  (Check out the SRS offerings, for example.)

Back to the OP's question:  The PLL has always been a
standard feature of lock-ins... even when it wasn't needed
and was a detriment to performance.  The basic lock-in, as
noted, is just a phase-sensitive detector that multiplies
the input signal times a reference.  The problem is that the
phase of the signal may be different than the reference, so
you need a quadrature reference and two multipliers, etc.
The PLL is used to convert a single reference into
quadrature.

That sort of approach is needed for experiments where the
reference is something you have no control over.  But in
many (most, in my experience) cases you provide the
reference as a driving or stimulus signal, and measure the
response.  In this case it is totally crazy to use a PLL
since it adds lock-time issues.  Instead, the reference
should just be generated in quadrature to begin with.  

Since modern lock-ins include signal generation, this should
at least be an option.  Maybe it is, by now, but the early
digital units (maybe 20 years ago?) had a single internal
generator feeding a PLL to get the quadrature references.
There wasn't even an option to provide your own quadrature
references... you *had* to wait on the damned PLL!  The
manufacturers that I talked to were not interested in
quadrature generation, apparently because nobody else was
asking for it.

Anyway, synchronous waveform averaging (coherent averaging)
turns out to be a much better approach than a lock-in for
most things, since you can recover the full response time
waveform while still reducing the noise... like 1024 (or
whatever) separate lock-ins, one for each waveform time
point.  And you can do it on the cheap, often with the sound
card you already have!

Best regards,


Bob Masta
 
              DAQARTA  v7.50
   Data AcQuisition And Real-Time Analysis
              www.daqarta.com
Scope, Spectrum, Spectrogram, Sound Level Meter
 Frequency Counter, Pitch Track, Pitch-to-MIDI 
   FREE Signal Generator, DaqMusiq generator    
          Science with your sound card!

Bob Masta <N0Spam@daqarta.com> wrote:

(snip, I wrote)
>>A lock-in amplifier is a piece of laboratory equipment, amplifiers,
>>phase detector, reference oscillator (but you can use an external
>>reference, too), adjustable low-pass filters, and usually also
>>meters to display the result.

>>http://elektrotanya.com/princeton-applied-research_5204_lock-in-analyzer.pdf/download.html

(snip)

>>All lock-in amplifiers that I know of use a square wave instead
>>of sine. I am not completely sure why, though.
 
> The original analog lock-ins used square waves, because they
> were easy to implement:  Just switch between the signal and
> its inverse.  They responded to all the odd harmonics in the
> square wave, but often times that wasn't an issue.

Yes. Well, maybe more important, one could accurately generate the
appropriate square wave. The first harmonic response, then, can
be accurately determined, and second harmonic appropriately low.
 
> Back in those days, there wasn't really a good
> low-distortion way to multiply analog signals. The earliest
> lock-ins to use sine wave multipliers used pulse-width
> modulation... ugh!
 
> But now, I think that everybody uses DSP methods and true
> sine waves.  (Check out the SRS offerings, for example.)

Also, the ones I used to know offered F and 2F as options,
so you could measure the second (and other even) harmonics.
They now have F, 2F, 3F, 4F, and many higher multples. Easy
in digital, hard in analog.
 
> Back to the OP's question:  The PLL has always been a
> standard feature of lock-ins... even when it wasn't needed
> and was a detriment to performance.  The basic lock-in, as
> noted, is just a phase-sensitive detector that multiplies
> the input signal times a reference.  The problem is that the
> phase of the signal may be different than the reference, so
> you need a quadrature reference and two multipliers, etc.
> The PLL is used to convert a single reference into
> quadrature.

Yes. It seems to me not so easy to genrate the reference sine,
along with in-phase and quadrature square (or sine) waves 
appropriately accurately using analog circuitry. Maybe not so
hard for one frequency, but all the ones I know of allow a wide
frequency range.
 
> That sort of approach is needed for experiments where the
> reference is something you have no control over.  But in
> many (most, in my experience) cases you provide the
> reference as a driving or stimulus signal, and measure the
> response.  In this case it is totally crazy to use a PLL
> since it adds lock-time issues.  Instead, the reference
> should just be generated in quadrature to begin with.  

Do they still allow for a continuous adjustment of reference
frequency? 

Also, the ones I remember were about equal internal and external,
but it might have been that the internal generator wasn't as good
as it could have been. Also, they tended to have low output power,
where an external generator might have more power (and lower
impedance).
 
> Since modern lock-ins include signal generation, this should
> at least be an option.  Maybe it is, by now, but the early
> digital units (maybe 20 years ago?) had a single internal
> generator feeding a PLL to get the quadrature references.
> There wasn't even an option to provide your own quadrature
> references... you *had* to wait on the damned PLL!  The
> manufacturers that I talked to were not interested in
> quadrature generation, apparently because nobody else was
> asking for it.

I don't remember having a long wait, but I usually ran at a
high enough frequency. As they go down below 1Hz, I can imagine
long lock times in that case.
 
> Anyway, synchronous waveform averaging (coherent averaging)
> turns out to be a much better approach than a lock-in for
> most things, since you can recover the full response time
> waveform while still reducing the noise... like 1024 (or
> whatever) separate lock-ins, one for each waveform time
> point.  And you can do it on the cheap, often with the sound
> card you already have!

-- glen

On Monday, March 24, 2014 1:55:30 AM UTC+13, Bob Masta wrote:
> On Sat, 15 Mar 2014 09:41:57 +0000 (UTC), glen
> 
> herrmannsfeldt <gah@ugcs.caltech.edu> wrote:
> 
> 
Do they normally have a hard limiter before the inputs to the phase-detector? Otherwise the loop will respond to amplitude as well as phase variations