EDAboard.com | EDAboard.de | EDAboard.co.uk | WTWH Media

DC background subtraction

Ask a question - edaboard.com

elektroda.net NewsGroups Forum Index - Electronics Design - DC background subtraction

Goto page Previous  1, 2, 3, 4


Guest

Wed Jan 09, 2019 3:45 pm   



On Thursday, January 10, 2019 at 12:56:53 AM UTC+11, George Herold wrote:
Quote:
On Wednesday, January 9, 2019 at 8:09:07 AM UTC-5, Phil Hobbs wrote:
On 1/8/19 7:56 PM, George Herold wrote:
On Tuesday, January 8, 2019 at 5:10:44 PM UTC-5, John Larkin wrote:
On Tue, 8 Jan 2019 10:03:59 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

On 1/8/19 8:51 AM, George Herold wrote:
On Monday, January 7, 2019 at 8:44:16 PM UTC-5, Phil Hobbs wrote:
On 1/7/19 8:27 PM, George Herold wrote:
On Monday, January 7, 2019 at 8:01:28 PM UTC-5, bitrex wrote:
On 01/07/2019 07:52 PM, George Herold wrote:
On Monday, January 7, 2019 at 7:37:48 PM UTC-5, bitrex wrote:
On 01/07/2019 07:23 PM, George Herold wrote:
On Monday, January 7, 2019 at 2:34:58 PM UTC-5, Phil Allison wrote:
George Herold wrote:


Perhaps a silly simple question. I need to subtract off a 2.5V
background from a few mV signal. I first think about a difference amp,
or instrument amp.
But I wondered if there is any advantage to a simple
opamp summing amp. (I'll add in -2.5 V)



** Considered using a capacitor ??




.... Phil

Like a sample and hold? It needs to hold the offset
sorta forever, so a pot called for.

George H.


DC blocking capacitor. Y'know, like a high-pass on the input.

https://www.dropbox.com/s/he7zy2rdhomfveh/IMG_20190107_193244554_HDR.jpg?dl=0

Oh no we are talking slowly varying signals. Minutes to
an hour maybe..

So it's my bosses idea, you put a hall probe sensor inside
a High Tc super conductor toroid. Induce a super current.
(by cooling down with permanant magnet in the toroid.)
And then observe the field go away as you warm it up.

George H.


Got it, I figured you had not forgotten how to use a capacitor to block
DC but just so long as we cover all the bases here and the signal period
wasn't mentioned explicitly ;)

Oh dear no, big C, big R and fet opamp.. about 1 second is the longest
TC I can wait for. (10uF and 100k or 1uF and a meg.)
I was doing that today to look at power supply noise.
I think most of the circuits I build are 'scope preamps. :^)
I've never done the feedback cap thing, (you posted previoulsy),
looks like a Sallen-Key.

George H.


If you want to amplify the signal by N times and the offset by 1x (so
that you're still biased above ground), that's easy to do accurately
with a garden-variety noninverting amp.

I don't really care where the signal level is, I just want to look at
(small) changes. I'm trying to picture how a non-iverting amp
does this, but can't. Can you give me more of a hint?

George H.

A noninverting gain of N amplifier is also an inverting gain of N-1
amplifier, and the 1 is very accurate when the loop gain is high. Thus
if you have an offset of 2.50000V, connecting the input resistor of the
op amp (on the inverting side) to your offset generator and the
noninverting input to signal+offset, you get

Vout = (N+1)(Vos + Vsig) - N Vos = Vos + N Vsig.

This is just like the usual split-supply, single-ended noninverting amp,
except with the ground reference relabelled. Resistor inaccuracy causes
gain inaccuracy, but the offset remains correct.

This move is super common in single-supply data acq circuits, for instance.

Cheers

Phil Hobbs

That's a nice concept: get some single-ended gain first, then take out
the common-mode.

https://www.dropbox.com/s/8aw9g0sd5tcf3s1/Diffamp_2-stage.JPG?dl=0
Huh, OK. It's probably that I'm a circuit wusse, but inverting gain
w/o a ground reference bothers me. I can try it. One could HP
filter the output.

The gain is all noninverting. At the output, both signal and offset get
multiplied by positive numbers. The numbers are different, is all.

There's no difference whatsoever between this and a normal split supply
noninverting stage except that a different node is labelled 'ground'.

Right... my comment was a little tongue in cheek.. I should have added
the smiley face.


(In my case)
It looks like there is so much 1/f noise in the sensor that nothing
else maters too much. (Lasse's back to back seems to work fine,
'in theory' with equal noise 'densities' I should get a sqrt(2)
improvement in S/N. That has to apply to 1/f noise too? No?

If the origins of the noise are independent, sure. Air currents or
power supply junk, no.

In one of my early consulting gigs, I was building stabilized lasers for
laser gravimeters down in Albuquerque. They had to be very stable over
periods of hours and days. I had a brass plaque made for my cube wall
that said:

DC
The Final Frontier

Yeah it's a little hard to measure the lower frequencies of 1/f noise,
things below 1 Hz.


Not hard, just tedious. Data loggers make it easier, but you do have to accumulate data for minutes or longer.

--
Bill Sloman, Sydney


Guest

Wed Jan 09, 2019 5:45 pm   



On Tuesday, January 8, 2019 at 5:47:45 PM UTC-5, Phil Hobbs wrote:
Quote:
On 1/8/19 4:57 PM, bloggs.fredbloggs.fred_at_gmail.com wrote:
On Tuesday, January 8, 2019 at 2:09:04 PM UTC-5, Phil Hobbs wrote:
On 1/8/19 1:53 PM, bloggs.fredbloggs.fred_at_gmail.com wrote:
On Monday, January 7, 2019 at 7:52:59 PM UTC-5, George Herold wrote:
On Monday, January 7, 2019 at 7:37:48 PM UTC-5, bitrex wrote:
On 01/07/2019 07:23 PM, George Herold wrote:
On Monday, January 7, 2019 at 2:34:58 PM UTC-5, Phil Allison wrote:
George Herold wrote:


Perhaps a silly simple question. I need to subtract off a 2.5V
background from a few mV signal. I first think about a difference amp,
or instrument amp.
But I wondered if there is any advantage to a simple
opamp summing amp. (I'll add in -2.5 V)



** Considered using a capacitor ??




.... Phil

Like a sample and hold? It needs to hold the offset
sorta forever, so a pot called for.

George H.


DC blocking capacitor. Y'know, like a high-pass on the input.

https://www.dropbox.com/s/he7zy2rdhomfveh/IMG_20190107_193244554_HDR..jpg?dl=0

Oh no we are talking slowly varying signals. Minutes to
an hour maybe..

So it's my bosses idea, you put a hall probe sensor inside
a High Tc super conductor toroid. Induce a super current.
(by cooling down with permanant magnet in the toroid.)
And then observe the field go away as you warm it up.

George H.

There should be some handheld VOMs that will do statistics on signals like this. And they produce downloadable acquisition files for USB. Probably something available at HomeDepot :-)


Big waste of bits though. A nice noninverting chopamp with high DC
open-loop gain will do an excellent job of preserving those bits. Two
30-dB gain stages, or a 60-dB composite amp with a nice fast 40-dB stage
inside the loop is good medicine for that sort of job. An OPA188 for
the outer loop and an LM6171A inside would probably be a decent choice,
and yield a bandwidth of around 200 kHz. Using the composite amp
improves the low frequency characteristics on account of eliminating the
loading on the outer amp and increasing the DC loop gain. (Making the
inner stage a lead/lag amp, i.e. putting a capacitor in series with the
feedback resistor, will greatly increase the DC gain again.

That's ridiculous. This product, which is very economically priced, gets 12-bit resolution on differential analog input (multiple channels), and has analog output channels which can provide a programmable reference for the differential conversion. Sample rate max's out at 50ksps and can be streamed into the laptop via USB. It does a bunch of other stuff too, looks like a good catch-all gadget for the lab. Input range seems a bit limited for this application, but he should be able to figure something out. No waste of bits here.
https://labjack.com/products/u3

I'm a fan of Labjacks--I have three U6 Pros from various projects.

The waste of bits is that the signal is hundreds of times smaller than
the offset. With a 2.5 mV signal on a 2.5V offset, your 12 bits just
turned into four. Using its PWM outputs to subtract off the pedestal
isn't much of an improvement if you don't know how big it is *a priori*.

Just a noninverting amp solves that problem completely for a buck.


There may be only 4 bits left for the residual but the ultimate resolution of the original input signal is [about] 50 ppm, which is pretty darn good. The reason is because of fairly precision offset reference and the conversion of the differential input make this a subranging A/D [without the auto-ranging part]. If he uses an IA with gain x20 to drive the LabJack inut, he gets 2.5 ppm etc. Not sure of the physics involved or the computational characterization of the phenomenon they're looking at, but that sounds like pretty good resolution.
This looks like something that would come in handy on occasion, and it's cheap. https://www.digikey.com/products/en?keywords=EVAL-INAMP-82RMZ-ND

Quote:

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

hobbs at electrooptical dot net
http://electrooptical.net



Guest

Wed Jan 09, 2019 5:45 pm   



On Tuesday, January 8, 2019 at 7:18:55 PM UTC-5, George Herold wrote:
Quote:
On Tuesday, January 8, 2019 at 4:57:59 PM UTC-5, bloggs.fred...@gmail.com wrote:
On Tuesday, January 8, 2019 at 2:09:04 PM UTC-5, Phil Hobbs wrote:
On 1/8/19 1:53 PM, bloggs.fredbloggs.fred_at_gmail.com wrote:
On Monday, January 7, 2019 at 7:52:59 PM UTC-5, George Herold wrote:
On Monday, January 7, 2019 at 7:37:48 PM UTC-5, bitrex wrote:
On 01/07/2019 07:23 PM, George Herold wrote:
On Monday, January 7, 2019 at 2:34:58 PM UTC-5, Phil Allison wrote:
George Herold wrote:


Perhaps a silly simple question. I need to subtract off a 2.5V
background from a few mV signal. I first think about a difference amp,
or instrument amp.
But I wondered if there is any advantage to a simple
opamp summing amp. (I'll add in -2.5 V)



** Considered using a capacitor ??




.... Phil

Like a sample and hold? It needs to hold the offset
sorta forever, so a pot called for.

George H.


DC blocking capacitor. Y'know, like a high-pass on the input.

https://www.dropbox.com/s/he7zy2rdhomfveh/IMG_20190107_193244554_HDR.jpg?dl=0

Oh no we are talking slowly varying signals. Minutes to
an hour maybe..

So it's my bosses idea, you put a hall probe sensor inside
a High Tc super conductor toroid. Induce a super current.
(by cooling down with permanant magnet in the toroid.)
And then observe the field go away as you warm it up.

George H.

There should be some handheld VOMs that will do statistics on signals like this. And they produce downloadable acquisition files for USB. Probably something available at HomeDepot :-)


Big waste of bits though. A nice noninverting chopamp with high DC
open-loop gain will do an excellent job of preserving those bits. Two
30-dB gain stages, or a 60-dB composite amp with a nice fast 40-dB stage
inside the loop is good medicine for that sort of job. An OPA188 for
the outer loop and an LM6171A inside would probably be a decent choice,
and yield a bandwidth of around 200 kHz. Using the composite amp
improves the low frequency characteristics on account of eliminating the
loading on the outer amp and increasing the DC loop gain. (Making the
inner stage a lead/lag amp, i.e. putting a capacitor in series with the
feedback resistor, will greatly increase the DC gain again.

That's ridiculous. This product, which is very economically priced, gets 12-bit resolution on differential analog input (multiple channels), and has analog output channels which can provide a programmable reference for the differential conversion. Sample rate max's out at 50ksps and can be streamed into the laptop via USB. It does a bunch of other stuff too, looks like a good catch-all gadget for the lab. Input range seems a bit limited for this application, but he should be able to figure something out. No waste of bits here.
https://labjack.com/products/u3
I like my lab jack... I mostly just use it as a data logger,
which is a waste of a lot of nice kit.


Looks look like one of those tools that pays for itself a couple of times over the first time you use it, saving on all that proto labor.

Quote:

George H.



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

hobbs at electrooptical dot net
http://electrooptical.net


Goto page Previous  1, 2, 3, 4

elektroda.net NewsGroups Forum Index - Electronics Design - DC background subtraction

Ask a question - edaboard.com

Arabic version Bulgarian version Catalan version Czech version Danish version German version Greek version English version Spanish version Finnish version French version Hindi version Croatian version Indonesian version Italian version Hebrew version Japanese version Korean version Lithuanian version Latvian version Dutch version Norwegian version Polish version Portuguese version Romanian version Russian version Slovak version Slovenian version Serbian version Swedish version Tagalog version Ukrainian version Vietnamese version Chinese version Turkish version
EDAboard.com map