Low Noise Voltage Source 4

[paulpoosk]

Hi, All:

I am monitoring the change in resistance of a load (~50 Ohms) with respect to a varying applied magnetic field (changes R_load by fractions of an Ohm). I would like to bias a constant voltage across this load to monitor these changes.

I am having difficulty maintaining a low-noise DC voltage bias across the load. Previously, I tried designing the circuit based on a noninverting opamp configuration, with the load at the inverting input and grounded. However the voltage noise across the load was near 100mV. Which is an issue since I will need to monitor changes in voltage of less than an ohm.

Any circuit ideas?

Thanks,
Paul

 

[melone]

Can you use a low ripple power supply?

 

[paulpoosk]

No, I don't think a low ripple power supply is an option.

 

[nbucska]

What is the freqency range of the magnetic field you want to monitor ? Is the purpose is measurement or signal
pick-up? 100 mV noise is a lot, you can do much better
you just must know what really do you need.


<nbucska@pcperipherals.com>

 

[OperaHouse]

I'm having trouble believing that 100mv noise is the best you can do with either a constant current or voltage power supply. You never mentiond how much voltage is on that 50 ohm load.

 

[paulpoosk]

I was incorrect at first...the noise is 10-15mV. I am applying 0-2V to 50 ohm load, with a need to detect changes in resistance of less than 1 ohm.

The idea is to retrieve an M-R transfer curve by monitoring the load.

 

[melone]

Does it have to be 2V?

 

[nbucska]

If it is for lab measurement just integrate it for a longer time. If there is a dominant noise frequency, integrate
it for integer periodes of this.

<nbucska@pcperipherals.com>

 

[paulpoosk]

It does not need to go all the way up to 2V.

 

[Wedwin]

Can You not use a bridge such that You reference voltage follows the measured voltage and thus cancel out the noise and just get the difference?

Widely used in strain measurements and such.

 

[IRstuff]

It appears that you are essentially doing a Hall measurement using a varying magnetic field. Presumably, since this is an experiment, you have control over the magnetic field input. You should be using a lock-in amplifier; which then allows you to synchronously time average the result.

This technique is very old and very good. See equipment from Signal Recovery (Princeton Applied Research), Boston Electronics, Stanford Research:

http://www.signalrecovery.com/lockinde.htm

http://www.boselec.com/products/siglimwhat.shtml

http://www.srsys.com/html/sr830.html

TTFN

 

[IRstuff]

As for the source, you should consider a prepackaged solution such as :

http://www.analog.com/Analog_Root/p...evel3%3D%252D1%26resourceWebLawID%3D0,00.html

see the other voltage references by Analog Devices:

http://search1.analog.com/searchProxy.asp?queryText=voltage+reference

You should be able to achieve noise in the low microvolt range for relatively low frequency bandwidths. For high frequencies, you'll have to use a lock-in amplifier

TTFN

 

[jtsamuels]

If you do not want to design/build a lock in amplifier, then you should at least consider taking your measurements differentially such as using an instrumentation amplifier. Again, either analog devices, analogic, national semiconductor or Maxim should have one of those devices. They also offer samples. This will reject any common mode noise that is present on your load. The PSRR (power supply rejection ratio) of the devices are usually over 110 db which will also help your measurments. Good luck.

 

[paulpoosk]

IRstuff, thanks a lot for your suggestions. However, my goal is to perform this experiment without a lock-in amplifier.

jtdsamuels, do you have any links to these sites that you mention? How about any circuit schematics of an instrumentation amplifier?

Thanks.

 

[IRstuff]

The links in my postings lead to the Analog Devices website. Do a search there for instrumentation amplifiers.

However, I would not bother with the notion of trying to build your own instrumentation amp. It can't be done without low leakage, high-performance transistors, which are integral to the design.

TTFN

 

[m3]

Paul,

For clean and stable DC bias use a battery and precision voltage reference (ex.:REF02 from Burr-Brown). This best way for low noise DC bias. Burr-Brown and Analog devices have a instrumentation amplifier (Ex.: INA101, INA102,INA105, AD628, AD620, etc).

Regards,

M3

 

[zimbali]

Hi there,
First of all if your sensor is not in a bridge configuration, PUT it in a bidge format; it is not a professional idea to use a precision sensor in a potentiometric format. Secondly, pick up the out of balance voltage and then amplify it, say 20 or 50 times with a good FET input op-amp like CA3140. Then add your dc signal to that. Now your measured signal is 20 or 50 (the more you amplify the better) times more and the error coming from the power supply will be 20 or 50 times less, as it is constant but your signal is more powerful!! then you can monitor your signal, however I also think you can acieve some 2mv ripple in your power supply.
If you have very much digital (computer or processor) stuff in the lab and really intended to have a virtually zero error, then do some DSP and filter out your signal with digital filters; very silly though!!!!
Battery as a power supply was a very nice idea; consider it.

I hope it would help
Zimbali

 

[paulpoosk]

Could someone elaborate on &quot;bridge configuration&quot;? Links?

Thanks everyone,
Paul