Maker Pro
Maker Pro

lock-in amplifier

S

scot

Jan 1, 1970
0
Anyone have experience with lock-in amplifiers? I am thinking of purchasing a used one but don't know enough about them. I know they are extremely effective at finding small signals buried in lots of noise but what are the downsides to a lock-in besides expense?

Also, this is probably over my head but I have read about the AD630 chip from Analog Devices and supposedly that can be used to make a very low cost lock-in amplifier. Anyone ever try that? Was it worth it for the money? Ishould mention a friend of mine has a couple of these AD630 IC's so I would be willing to solder something up pretty quick if it's not a waste of time.

Scot
 
G

George Herold

Jan 1, 1970
0
Anyone have experience with lock-in amplifiers?  I am thinking of purchasing a used one but don't know enough about them.  I know they are extremely effective at finding small signals buried in lots of noise but what are the downsides to a lock-in besides expense?

Also, this is probably over my head but I have read about the AD630 chip from Analog Devices and supposedly that can be used to make a very low costlock-in amplifier.  Anyone ever try that?  Was it worth it for the money?  I should mention a friend of mine has a couple of these AD630 IC's so I would be willing to solder something up pretty quick if it's not a waste of time.

Scot


As Phil said, there are better IC's these days,

But for learning how a lockin works the AD630 will be fine.
There's at least two kinds of lockin's (all analog techniques can be
done in digital.) Multiplying and switched gain. (gain = +1/-1) The
AD630 is a switched gain lockin.

George H.
 
S

scot

Jan 1, 1970
0
Anyone have experience with lock-in amplifiers? I am thinking of purchasing a used one but don't know enough about them. I know they are extremelyeffective at finding small signals buried in lots of noise but what are the downsides to a lock-in besides expense?



Also, this is probably over my head but I have read about the AD630 chip from Analog Devices and supposedly that can be used to make a very low costlock-in amplifier. Anyone ever try that? Was it worth it for the money? I should mention a friend of mine has a couple of these AD630 IC's so I would be willing to solder something up pretty quick if it's not a waste of time.



Scot

Thanks Phil for the suggestion on the ADG1236. I read up on Lock-in's lastnight and learned a lot. I found this interesting article and it looks like the principles are pretty understandable.

http://physlab.lums.edu.pk/images/9/90/LIA.pdf

The article specifies a DG303A chip but I think the ADG1236 can be substituted instead. My main goal here is to learn about Lock-in's by building onemyself since reading only takes you so far. What do you think? Is this agood starting point?

Scot
 
B

Bill Sloman

Jan 1, 1970
0
Lock-ins are very useful for (slow) measurements in low-SNR
situations--they can pull signal out of some pretty grimy-looking
interference.

Building your own analogue lock-in is an excellent project.  I built my
first one when I was a postdoc (about 1988) for an atomic force
microscope.  More recently, I've built a few just like that for
customers, as part of the back ends of their measurement systems.  They
really are Good Medicine.  (I found the ADG1236 just the other day, when
I needed parts to make a baby-scale loop.)

There are two main dangers with lock-ins.  The first and more insidious
one is that they can tempt you to skimp on the front end of the
measurement, and so lose SNR for no particularly good reason.  The other
is that in the presence of drift or 1/f noise, narrowing the bandwidth
doesn't help, and may make things worse.

The last sentence should be wrong, if you are using lock-in right. The
basic idea is that you start off by modulating whatever it is you are
measuring at some convenient frequency, and the detect only the
modulation frequency in the output that you are looking at (usually
after amplification and a certain amount of filtering).

For added extra credit you can detect just some of the higher
harmonics of the modulation frequency, but it takes sophisitcated
experimental design to make this a useful approach.

One of the great virtues of this approach is that you can stay well
away from the frequencies where 1/f noise (and drift can be seen as a
form of 1/f noise) is a problem.
 
G

George Herold

Jan 1, 1970
0
The last sentence should be wrong, if you are using lock-in right. The
basic idea is that you start off by modulating whatever it is you are
measuring at some convenient frequency, and the detect only the
modulation frequency in the output that you are looking at (usually
after amplification and a certain amount of filtering).

For added extra credit you can detect just some of the higher
harmonics of the modulation frequency, but it takes sophisitcated
experimental design to make this a useful approach.

One of the great virtues of this approach is that you can stay well
away from the frequencies where 1/f noise (and drift can be seen as a
form of 1/f noise) is a problem.

I think Phil was talking about 1/f noise elsewhere in the signal
chain.
(I assume he knows how to pick a modulation frequency.)
But in answering for someone else I've likely stuck my foot in my
mouth. (again :^)

George H.
 
G

George Herold

Jan 1, 1970
0
Lock-ins are very useful for (slow) measurements in low-SNR
situations--they can pull signal out of some pretty grimy-looking
interference.

Building your own analogue lock-in is an excellent project.  I built my
first one when I was a postdoc (about 1988) for an atomic force
microscope.  More recently, I've built a few just like that for
customers, as part of the back ends of their measurement systems.  They
really are Good Medicine.  (I found the ADG1236 just the other day, when
I needed parts to make a baby-scale loop.)

There are two main dangers with lock-ins.  The first and more insidious
one is that they can tempt you to skimp on the front end of the
measurement, and so lose SNR for no particularly good reason.  The other
is that in the presence of drift or 1/f noise, narrowing the bandwidth
doesn't help, and may make things worse.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510 USA
+1 845 480 2058

hobbs at electrooptical dot nethttp://electrooptical.net- Hide quoted text -

- Show quoted text -

Re: front ends
With a nice front end I find a DSO with an average function, and sync
trigger is 'good enough' lots of times. (a poor man's lockin.)

George H.
 
G

George Herold

Jan 1, 1970
0
That's more of a poor man's signal averager, which is a much better
approach in the presence of 1/f noise.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510 USA
+1 845 480 2058

hobbs at electrooptical dot nethttp://electrooptical.net- Hide quoted text -

- Show quoted text -

Well it's not phase sensitive, but it does select frequency, so 128
averages is like a band pass with a Q of 128(?)

Ya know I've never seen a good article on the many ways to use the
average button on a DSO.

George H.
 
J

John Devereux

Jan 1, 1970
0
Phil Hobbs said:
The AD630 is an ancient chip that was never that great to begin
with. Nowadays you can do a great deal better with a low charge
injection mux, e.g. the ADG1236 dual SPDT. With a bit more work, you
can do better still with a couple of dual-gate MOSFETs, which have
almost no charge injection at all.

Could you expand on that, or suggest a favorite part?

I found the ADG1236 too for low charge injection. Also DG636.
 
M

Mr Stonebeach

Jan 1, 1970
0
Could you expand on that, or suggest a favorite part?

I found the ADG1236 too for low charge injection. Also DG636.

Same here. Unfortunately the DG1236 does not work at
cryogening temperatures, unlike e.g. the TS5 series
from Texas Instruments. I have been looking for a
low-injecting switch to operate in LHe, and Phil's
suggestion of using dual-gste MOSFETs is good. At least
such devices as the 2N7000 (from some manufacturers) and
the FDV301P do work, so I might try throwing in eg. the
BF998.

Regards,
Mikko
 
G

George Herold

Jan 1, 1970
0
Something like that.  It is phase-sensitive actually, because you're
triggering on some sync signal, which makes I and Q easy to tell apart.

The main thing is that if you're sweeping at 1 kHz, with 1000 data
points, anything below 1 kHz can cause only a baseline shift, whereas
the actual data is taken from a forest of comb filters, each centered on
a harmonic of the sweep frequency.  Almost all of that sensitive
bandwidth is well outside the 1/f noise of anything except a pHEMT.





You just need to get Teachspin to rebadge a Rigol or Owon scope, and
write an app note about it. ;)

Cheers

Phil Hobbs

(Just bought an Owon SDS7102, which should be here tomorrow.  It's
mostly for my son to use in his continuing firmware endeavours.)

Well let us know what you think of the Owon. We bought one (probably
a different model) years ago, but it now lives in Europe for trade
shows over there. I didn't like the screen, it had a very restricted
viewing angle. (It seemed a bit noisier than a low end tek too... but
no real data.)

George H.
 
J

John Devereux

Jan 1, 1970
0
Phil Hobbs said:
Phil Hobbs <[email protected]> writes:
[...]
Could you expand on that, or suggest a favorite part?

Last time I used it, it was a metal-can 3N201. ;) You drive the
source, and come out the drain. Connect G2 to the source, and put the
gate signal on G1. You can do somewhat different things with G2
depending on the application, but the low charge injection seems to
come from the top half of the cascode cutting off before the bottom
half, so that the gate charge comes back out the source instead of the
drain.

Unfortunately I no longer have the lab notebook that had the pictures in it.

OK, thanks.

Looks like something I will have to breadboard some time.

I seem to recall someone (John L?) talking about using opto-fets as zero
charge injection switches. But I wonder are they really? They have fets
inside, with gates that charge and discharge presumably. Hmm, not sure I
understand how they work, now I come to think of it. Photovoltaic mode
photodiode(s) connected to the gate I guess.
 
S

Spehro Pefhany

Jan 1, 1970
0
Will do. This one has an 800x600 display, and otherwise similar specs
to the Rigol DS1102E, for just about the same price ($400). Saelig was
out of stock of the Rigol, so I switched my order.

It's probably going to spend most of its time looking at SPI busses and
so on, at least for the first while.

I'd be interested to know what you think of it. The display is a lot
bigger and higher resolution than my Tek, and some of the models have
an interesting 16-channel LA function at an interesting price point
( < $1K ).
 
J

John Devereux

Jan 1, 1970
0
Jim Thompson said:
Phil Hobbs said:
On 03/07/2013 01:40 AM, John Devereux wrote:
[...]

The AD630 is an ancient chip that was never that great to begin
with. Nowadays you can do a great deal better with a low charge
injection mux, e.g. the ADG1236 dual SPDT. With a bit more work, you
can do better still with a couple of dual-gate MOSFETs, which have
almost no charge injection at all.

Could you expand on that, or suggest a favorite part?

Last time I used it, it was a metal-can 3N201. ;) You drive the
source, and come out the drain. Connect G2 to the source, and put the
gate signal on G1. You can do somewhat different things with G2
depending on the application, but the low charge injection seems to
come from the top half of the cascode cutting off before the bottom
half, so that the gate charge comes back out the source instead of the
drain.

Unfortunately I no longer have the lab notebook that had the pictures in it.

OK, thanks.

Looks like something I will have to breadboard some time.

I seem to recall someone (John L?) talking about using opto-fets as zero
charge injection switches. But I wonder are they really? They have fets
inside, with gates that charge and discharge presumably. Hmm, not sure I
understand how they work, now I come to think of it. Photovoltaic mode
photodiode(s) connected to the gate I guess.

Well? You know how reliable Larkin's advice is >:-}

Not even sure it was him.

But do *you* know how they work? Is there charge injection?
 
J

John Devereux

Jan 1, 1970
0
Jim Thompson said:
Jim Thompson said:
On Thu, 07 Mar 2013 15:26:59 +0000, John Devereux


On 03/07/2013 01:40 AM, John Devereux wrote:

[...]

The AD630 is an ancient chip that was never that great to begin
with. Nowadays you can do a great deal better with a low charge
injection mux, e.g. the ADG1236 dual SPDT. With a bit more work, you
can do better still with a couple of dual-gate MOSFETs, which have
almost no charge injection at all.

Could you expand on that, or suggest a favorite part?

Last time I used it, it was a metal-can 3N201. ;) You drive the
source, and come out the drain. Connect G2 to the source, and put the
gate signal on G1. You can do somewhat different things with G2
depending on the application, but the low charge injection seems to
come from the top half of the cascode cutting off before the bottom
half, so that the gate charge comes back out the source instead of the
drain.

Unfortunately I no longer have the lab notebook that had the pictures in it.

OK, thanks.

Looks like something I will have to breadboard some time.

I seem to recall someone (John L?) talking about using opto-fets as zero
charge injection switches. But I wonder are they really? They have fets
inside, with gates that charge and discharge presumably. Hmm, not sure I
understand how they work, now I come to think of it. Photovoltaic mode
photodiode(s) connected to the gate I guess.

Well? You know how reliable Larkin's advice is >:-}

Not even sure it was him.

But do *you* know how they work? Is there charge injection?

It definitely takes a change of charge on the surface of the channel
to turn the FET on or off.

Yes, I imagine an isolated little battery connected between source and
gate, magically charging itself up. But does that inject charge out the
drain, say? Must do I think.

But I'm unsure of the construction of
"opto-fets". Anyone have a part number and I'll look them up.

The datasheets I have seen don't show anything much. I use a lot of
LCA710 (for power switching, so probably not the best choice for this).

Actually I think it might have been a Jim Williams app note, not
JL. Sorry to be so vague; really can't remember.
 
J

John Devereux

Jan 1, 1970
0
Phil Hobbs said:
Phil Hobbs said:
On 03/07/2013 01:40 AM, John Devereux wrote:
[...]

The AD630 is an ancient chip that was never that great to begin
with. Nowadays you can do a great deal better with a low charge
injection mux, e.g. the ADG1236 dual SPDT. With a bit more work, you
can do better still with a couple of dual-gate MOSFETs, which have
almost no charge injection at all.

Could you expand on that, or suggest a favorite part?

Last time I used it, it was a metal-can 3N201. ;) You drive the
source, and come out the drain. Connect G2 to the source, and put the
gate signal on G1. You can do somewhat different things with G2
depending on the application, but the low charge injection seems to
come from the top half of the cascode cutting off before the bottom
half, so that the gate charge comes back out the source instead of the
drain.

Unfortunately I no longer have the lab notebook that had the pictures in it.

OK, thanks.

Looks like something I will have to breadboard some time.

I seem to recall someone (John L?) talking about using opto-fets as zero
charge injection switches. But I wonder are they really? They have fets
inside, with gates that charge and discharge presumably. Hmm, not sure I
understand how they work, now I come to think of it. Photovoltaic mode
photodiode(s) connected to the gate I guess.

I did a bit of digging. NXP has discontinued all of their
enhancement-mode dual gate FETs, including the cute ones like the
BF1100, which had an interesting feedback FET mechanism to improve the
linearity of AGC.

The best-looking one for sampling appears to be the BF998, which has
reasonable I_D with V_G2S = 0.

OK, nice, thank you.

I just went on digikey and I see that is the same one I already clicked
on, when I idly looked one up after you mentioned them!

Ordered some for a rainy day.

Thanks,
 
M

Marte Schwarz

Jan 1, 1970
0
Hi John & Jim,

Just think a little before flaming someone else, please.

opto-fets can never have a charge injection, because the charge they use
to load the gate comes from the chanel that the charge flows in. so the
sum will allways be zero. I would say, this is as obvious as Kirchhoffs
law can be.

Regards

Marte
 
J

John Devereux

Jan 1, 1970
0
Marte Schwarz said:
Hi John & Jim,


Just think a little before flaming someone else, please.

Flaming? Really? First time I've been accused of that!

It was an honest question. Just thinking aloud.
opto-fets can never have a charge injection, because the charge they
use to load the gate comes from the chanel that the charge flows
in. so the sum will allways be zero. I would say, this is as obvious
as Kirchhoffs law can be.

Well, I don't think it is that obvious.

Even Jim "charge is always conserved" Thompson doesn't seem to think so
:)

(And please, folks, we dont really need to go there again!)
 
M

Marte Schwarz

Jan 1, 1970
0
Hi John,
Flaming? Really? First time I've been accused of that!

Well? You know how reliable Larkin's advice is >:-}
It was an honest question. Just thinking aloud.

What a question?
Well, I don't think it is that obvious.

Where do you expect the injected charge comming from?
Even Jim "charge is always conserved" Thompson doesn't seem to think so
:)

May be Jim woul tell his statement, too.

I don't see any source for charge-injection.

Marte
 
J

John Devereux

Jan 1, 1970
0
Marte Schwarz said:
Hi John,


Well? You know how reliable Larkin's advice is >:-}

What a question?

Well you snipped it didn't you?
Where do you expect the injected charge comming from?

The voltage at the gate changes, relative to the other terminals. There
is a capacitance between the gate and the other terminals. dQ=C.dV.
May be Jim woul tell his statement, too.

I don't see any source for charge-injection.

The photodiode? (or whatever equivalent generates the change in gate
voltage).
 
G

George Herold

Jan 1, 1970
0
If they're just photoconductors, you have to build up enough
photocarriers to get a decent ON resistance.  To reduce the amount of
light required, you want a long carrier lifetime (which is easy to get
in silicon, of course) but then you have to wait for them to recombine.

Photoconductors also have twice full shot noise in the photocurrent
portion--both the generation and recombination are stochastic.

Ahh, (a light dawns) you've mentioned the double shot noise in
photoconductors, but I never got it.

I've got this 'weird' dark noise from the reverse biased LED's I've
been playing with. I'm pretty sure it's after pulsing of the
avalanche. Which they attribute to some sort of traps in the avalnche
region.
There's an observable 'bunchiness' in the counts.


George H.

(The
 
Top