Greater than 12 bit resolution with gain stage and two ADCs

Log in or register to post comments
Go To Last Post
13 posts / 0 new
Author
Message
mojo-chan
mojo-chan's picture
Level: Raving Lunatic
Joined: Thu. Jan 24, 2008
Posts: 2300 View posts
#1
Posted by mojo-chan: Mon. Aug 19, 2013 - 08:47 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I am working on a project that requires sampling noise over a large dynamic range. The ADC's 12 bit resolution doesn't seem to be enough so we are looking for a low cost way to increase it.

The obvious answer is a log amplifier but they do add cost. As an alternative has anyone tried using two ADCs sampling simultaneously, one with gain and one without?

Our signal is low pass filtered to 5KHz. The plan would be to sample it twice simultaneously with the ADCs triggered by an event. One ADC would have a gain of 1, the other a gain of 64. For large signals the high gain ADC would be saturated and the unity gain ADC would be used for the reading. For small signals the high gain ADC would be used.

We have a 2.048V reference so get 0.5mV/LSB from a unity gain ADC. A 64 gain ADC would be able to measure down to 7.8nV/LSB.

Has anyone tried this? Is it madness? :-)

Tags:

AVR Microcontrollers, AVR and XMEGA
Top
John_A_Brown
John_A_Brown's picture
Level: Raving Lunatic
Joined: Sat. Sep 20, 2003
Posts: 6219 View posts
Location: Underneath the Cotswolds escarpment, England.
Posted by John_A_Brown: Mon. Aug 19, 2013 - 09:15 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I would guess that you will get discontinuity problems at the point where you change horses, you might have to "blend" the two reading together over some range of voltages to try and compensate for this.
My thinking would be to incorporate both readings in to some sort of formula whereby, at the low end of the "take over" range the high gain would be weighted accordingly, at and the high end of the "take over" region the low gain would be the overriding input. I think I have done something along these line before to try to resolve discontinuities caused by external differences.

Four legs good, two legs bad, three legs stable.

Top
mojo-chan
mojo-chan's picture
Level: Raving Lunatic
Joined: Thu. Jan 24, 2008
Posts: 2300 View posts
Posted by mojo-chan: Mon. Aug 19, 2013 - 09:58 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I agree that it will be a pain to get right. I should clarify by saying that absolute readings are not all that important as we are converting the reading a to pseudo-dB scale using a lookup table. It's replacing a previous project that used pseudo log amps and needs to have more or less the same scale.

Top
electronic.designer
electronic.designer's picture
Level: Hangaround
Joined: Thu. May 27, 2010
Posts: 488 View posts
Location: The land of Cyrus the Great
Posted by electronic.designer: Mon. Aug 19, 2013 - 12:45 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

mojo-chan wrote:
The ADC's 12 bit resolution doesn't seem to be enough so we are looking for a low cost way to increase it.

For higher resolution, this may help:

www.atmel.com/Images/doc8498.pdf

Ozhan KD
Knowledge is POWER

Top
erg144
erg144's picture
Level: Hangaround
Joined: Thu. Dec 29, 2011
Posts: 145 View posts
Location: Pittsburgh, PA
Posted by erg144: Mon. Aug 19, 2013 - 04:12 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have tried the oversampling and seemed to work ok in my small case (14-bit).

Top
mojo-chan
mojo-chan's picture
Level: Raving Lunatic
Joined: Thu. Jan 24, 2008
Posts: 2300 View posts
Posted by mojo-chan: Mon. Aug 19, 2013 - 04:36 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I'm going to need at least 16 bits, preferably more than that... There is another option I thought of though. I could make a sample-and-hold system and then sample it twice with different gain factors as a kind of per-sample AGC.

Top
erg144
erg144's picture
Level: Hangaround
Joined: Thu. Dec 29, 2011
Posts: 145 View posts
Location: Pittsburgh, PA
Posted by erg144: Mon. Aug 19, 2013 - 08:01 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

>16 bit? Have you thought of an external ADC (e.g. a 24 bit adc?).

Top
Bronze Marktheusch
theusch's picture
Level: 10k+ Postman
Joined: Mon. Feb 19, 2001
Posts: 40591 View posts
Location: Wisconsin USA
Posted by theusch: Mon. Aug 19, 2013 - 08:09 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Besides having "permanent" high and low ranges, you could also think about "auto-scaling" that switches offset (and/or gain) through different bands until the signal can be read. This works better when the signal isn't fast-moving--for example, temperature sensor.

You can put lipstick on a pig, but it is still a pig.

I've never met a pig I didn't like, as long as you have some salt and pepper.

Top
mojo-chan
mojo-chan's picture
Level: Raving Lunatic
Joined: Thu. Jan 24, 2008
Posts: 2300 View posts
Posted by mojo-chan: Tue. Aug 20, 2013 - 08:06 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I am experimenting with auto-scaling at the moment. Having some issues but I think they are hardware related.

There are no cheap >=16 bit ADCs capable of 5KHz unfortunately.

Top
mojo-chan
mojo-chan's picture
Level: Raving Lunatic
Joined: Thu. Jan 24, 2008
Posts: 2300 View posts
Posted by mojo-chan: Tue. Aug 20, 2013 - 03:59 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A little update. Using two ADCs works quite well. The PGA is surprisingly good for such a cheap little device. Of course, it draws twice the current of one ADC.

I'm going to try and do a kind of AGC instead, since hopefully our signal should vary fairly slowly in overall amplitude, even if it the signal itself is noise.

Top
zajc3w
zajc3w's picture
Level: Rookie
Joined: Sun. Aug 11, 2013
Posts: 42 View posts
Location: Barking, Essex, Engand
Posted by zajc3w: Tue. Aug 20, 2013 - 11:48 PM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Is it one off project or mass production? If one-couple then there are free 24bit ADC's - free samples from ti.com ;P .
For mass production You can get 24 bit 64ksps ADC for under 5USD @ 1k units(ads131e04,digikey).

Top
mojo-chan
mojo-chan's picture
Level: Raving Lunatic
Joined: Thu. Jan 24, 2008
Posts: 2300 View posts
Posted by mojo-chan: Wed. Aug 21, 2013 - 08:09 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

This is mass production. $5 is a bit expensive (more than the MCU), but something like that might be an option. We have really limited PCB space as well (dimensions are not determined by us and the customer just said we need to make it slightly smaller) so we would have to use a part with fewer pins.

Top
mspeng
mspeng's picture
Level: Rookie
Joined: Thu. Jul 14, 2011
Posts: 38 View posts
Posted by mspeng: Wed. Aug 21, 2013 - 08:22 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have used auto-scaling and it works. You might get discontinuities where you switch over, but if your application can handle some non-linearity it is the way to go because you don't have to burn the power of 2 adcs every sample.

Top