For a project im working on i am using strain 350Ohm strain gauges in a bridge configuration to measure the pressure wave created in ballistic gelatin. I need an instrumentation amp capable of doing 0-500Khz as far as bandwidth is concerned.
I have tried my hand at building a classic 3 op amp instrumentation amp such as the one pictured here using lm318 high speed op amps.
http://en.wikipedia.org/wiki/Operationa ... _amplifier
I am running into alot of noise issues and having trouble getting everything balanced as far as gain and offset. I have built part of this circuit just for testing using regular lm741's to get rid of any noise associated with using hi bandwidth amps.
After this it is apparrent that most of the noise is bandwith related but im still having issues with getting everything balanced.
So far when using the high speed amps i have only added caps between the +/- V supply rails and ground. i have yet to try caps directly from the chips power supply pins to ground. I know i could also put caps from each amplifiers inputs to ground so that any frequencies higher that what i need are shunted to ground. As far as noise suppression that is pretty much my entire bag of tricks and if that doesnt work im not sure what will.
What i could use right now is some advice...
are there any tips or techniques for noise suppression that i may have missed or any advice in general when working with In Amps or hi speed op amps?
This circuit is proving to be quite bothersome am partly leaning toward using an IC Instrumentation Amp for simplicity if one exists that can meet the specs i mentioned before.... Does anyone know of any IC IN Amps that would work for this?
I am going to try to build this again from scratch starting with general purpose amps to get everything balanced properly and if that is successfull drop in some high speed amps so i know it's only noise issues i have to deal with.
Instrumentation Amp issues
-
- Posts: 98
- Joined: Thu Dec 06, 2001 1:01 am
- Location: ohio
- Contact:
I don't have a lot of advice for you as I have always found noise to be problematic in my analog designs. I'm learning but it's a tough road for me.
How are you prototyping this design? I think you'll need to build a real PCB with a real ground plane to start with. Point to point wiring can be made to work though I've had mediocre results. And you can about forget solderless breadboards (noise antenna arrays).
good luck. if you get any pointer fromelsewhere, could you post them here?
Phil
How are you prototyping this design? I think you'll need to build a real PCB with a real ground plane to start with. Point to point wiring can be made to work though I've had mediocre results. And you can about forget solderless breadboards (noise antenna arrays).
good luck. if you get any pointer fromelsewhere, could you post them here?
Phil
-
- Posts: 108
- Joined: Thu Nov 25, 2004 1:01 am
- Location: Buenos Aires Argentina
- Contact:
Re: Instrumentation Amp issues
Do you need the input buffer stages in your amplifier?unknown_entity wrote:For a project im working on i am using strain 350Ohm strain gauges in a bridge configuration to measure the pressure wave created in ballistic gelatin. I need an instrumentation amp capable of doing 0-500Khz as far as bandwidth is concerned.
I have tried my hand at building a classic 3 op amp instrumentation amp such as the one pictured here using lm318 high speed op amps.
http://en.wikipedia.org/wiki/Operationa ... _amplifier ....
With a 350 Ohm bridge, if the amp has even a moderate input impedance (10 k?) it will add no measurable error and it should reduce noise.
E. Cerfoglio
Buenos Aires
Argentina
Buenos Aires
Argentina
-
- Posts: 458
- Joined: Thu Feb 05, 2004 1:01 am
- Location: Minneapolis, MN
- Contact:
Um... where to start?
First, have you actually *looked* for an IN-amp with the specs you need?
ISTR Burr-Brown had some pretty fast ones. Even if I were building one from discrete opamps, I wouldn't start with a 741. At least start out with some fast precision amps.
Second, Analog Devices has an Analog Dialogue article from a few (20?) years back entitled something like "An Engineer's guide to Noise, Grounding and generally making things go right for a change" Check their site for back issues of AD
Third, you want to decouple the opamps right at their terminals. A few tenths of an inch of PCB trace can make a big difference. You are doing this on a PCB, or at least dead-bug style on bare copper clad, aren't you? Plug in breadboards are ABSOLUTELY USELESS (yes, I mean that for this kind of thing. Single point grounds where possible. Especially if there's anything digital in the power path, visualizing the flow of current can help avoid problems.
A lot of this stuff is just experience, but much of what you can read in a book will make a big difference right away. Get the A/D article I mentioned: it's gold!!!
First, have you actually *looked* for an IN-amp with the specs you need?
ISTR Burr-Brown had some pretty fast ones. Even if I were building one from discrete opamps, I wouldn't start with a 741. At least start out with some fast precision amps.
Second, Analog Devices has an Analog Dialogue article from a few (20?) years back entitled something like "An Engineer's guide to Noise, Grounding and generally making things go right for a change" Check their site for back issues of AD
Third, you want to decouple the opamps right at their terminals. A few tenths of an inch of PCB trace can make a big difference. You are doing this on a PCB, or at least dead-bug style on bare copper clad, aren't you? Plug in breadboards are ABSOLUTELY USELESS (yes, I mean that for this kind of thing. Single point grounds where possible. Especially if there's anything digital in the power path, visualizing the flow of current can help avoid problems.
A lot of this stuff is just experience, but much of what you can read in a book will make a big difference right away. Get the A/D article I mentioned: it's gold!!!
-
- Posts: 2277
- Joined: Wed Nov 24, 2004 1:01 am
- Location: ASHTABULA,OHIO
- Contact:
Entity
Your getting some pretty good advice here and maybe I could add a smidgeon. First, as mentioned, go with a later model opamp and forget about the buffer stages. One thing I dont see mentioned here is a small cap across Rfeedback. For your application this would be quite small (10 pf ? ) . Be careful here as you can sometimes cause more harm than good -Phase margins, etc. Have you made an attempt to identify your noise such as white noise or a spectrum speciic noise. What is the gain of your amp in relation to noise amplitude. Have you tried to determine location of noise entry point. Try shorting inputs,bypassing power supply close to opamp. Aside from the usual 10 mf and 0.1mf tandem capacitors, you may want to add an even smaller cap in parrellel with these ( 0.005 to 0.01 mf) which will do the job better at higher noise frequencys. As is always the case, Good construction practice has a direct effect on circuit performance and whole books have been written on this so I won't go into it here. As a matter of interest and after doing years of work in the VHF and UHF spectrum, I dont consider 500 Khz as being a very high frequency. I have built up many circuits on plug in bread boards higher in frequency than that. My plug-ins have separate power rail and ground buses and by taking individual grounds direct to that bus I have had no ground problems. I had assumed all plug-ins were configured the same until I began seeing a wide range of plug-ins on the market. "Whats in your Wallet"?
When I start having problems with my plugins is usually around 10 Mhz and for sure at 20 Mhz. However VERY HI GAIN will cap their useful limits at any frequency.
Your getting some pretty good advice here and maybe I could add a smidgeon. First, as mentioned, go with a later model opamp and forget about the buffer stages. One thing I dont see mentioned here is a small cap across Rfeedback. For your application this would be quite small (10 pf ? ) . Be careful here as you can sometimes cause more harm than good -Phase margins, etc. Have you made an attempt to identify your noise such as white noise or a spectrum speciic noise. What is the gain of your amp in relation to noise amplitude. Have you tried to determine location of noise entry point. Try shorting inputs,bypassing power supply close to opamp. Aside from the usual 10 mf and 0.1mf tandem capacitors, you may want to add an even smaller cap in parrellel with these ( 0.005 to 0.01 mf) which will do the job better at higher noise frequencys. As is always the case, Good construction practice has a direct effect on circuit performance and whole books have been written on this so I won't go into it here. As a matter of interest and after doing years of work in the VHF and UHF spectrum, I dont consider 500 Khz as being a very high frequency. I have built up many circuits on plug in bread boards higher in frequency than that. My plug-ins have separate power rail and ground buses and by taking individual grounds direct to that bus I have had no ground problems. I had assumed all plug-ins were configured the same until I began seeing a wide range of plug-ins on the market. "Whats in your Wallet"?
When I start having problems with my plugins is usually around 10 Mhz and for sure at 20 Mhz. However VERY HI GAIN will cap their useful limits at any frequency.
You won't find much output from a 350 ohm strain gauge at 500KHz. They're essentially small squeezeable resistors that don't respond well above 10's of kilohertz, if that. You need a pressure transducer that has more high frequency response to catch those high freqs. Check out-
http://www.vishay.com/brands/measuremen ... /guide.htm
http://www.vishay.com/brands/measuremen ... /guide.htm
Is 500 KHz response really needed? A strain gauge 1/4 inch long on a medium where the speed of sound was 6000 feet/sec would require about 100 KHz response.
What level signal do you expect from the strain gauges? If it is in the microvolt region, thermal noise may begin to be a basic limitation. Incidently, the LM318 is a rather old device and may not be designed for low noise. Newer devices sometimes give noise characteristics on the data sheet.
You might also consider using AC coupling. This will simplify your offset problems a great deal. If your medium relaxes between pulses, there may not be a DC component to ampllify.
What level signal do you expect from the strain gauges? If it is in the microvolt region, thermal noise may begin to be a basic limitation. Incidently, the LM318 is a rather old device and may not be designed for low noise. Newer devices sometimes give noise characteristics on the data sheet.
You might also consider using AC coupling. This will simplify your offset problems a great deal. If your medium relaxes between pulses, there may not be a DC component to ampllify.
-
- Posts: 98
- Joined: Thu Dec 06, 2001 1:01 am
- Location: ohio
- Contact:
Thankyou for the advice everyone, im an undergrad doing this as a small project for the physics department at my local college. I really haven't had to work with amplifying signals at such a low level while worrying about noise and bandwidth too much, so this is a learning experience/refresher for me on a few things.
Im working under a PhD. on this project so i was basically handed a strain gauge and told to make a circuit that can work up to 500 Khz. We both aggreed that the 500Khz bandwidth was a little more than needed but it will be needed later on for some other things we plan to do with the project.
For now this is just a proof of concept using only 1 active element in the bridge later designs will use 2 or 4 elements and im sure this will make the signal easier to work with.
I am going to work on this more friday and over the weekend.
And to resopond to everyones questions....
Earlier i thought about omitting the input buffers but i did not do so because i thought going straight into a subtractor/differential amp would degrade the CMRR because of the unbalanced input impedance inherent to differential amp circuits.
The output of the bridge is about .1 - .4 millivolts peak
As far as op amp choice i only thought of using a 741 out of frustration just to get everything balanced in the circuit and worry about the noise related to high bandwidth op amps later.
My reason for choosing the lm318 was more or less, what do we have in the lab thats fast enough.
Thanks again Everyone,
- James
Im working under a PhD. on this project so i was basically handed a strain gauge and told to make a circuit that can work up to 500 Khz. We both aggreed that the 500Khz bandwidth was a little more than needed but it will be needed later on for some other things we plan to do with the project.
For now this is just a proof of concept using only 1 active element in the bridge later designs will use 2 or 4 elements and im sure this will make the signal easier to work with.
I am going to work on this more friday and over the weekend.
And to resopond to everyones questions....
Earlier i thought about omitting the input buffers but i did not do so because i thought going straight into a subtractor/differential amp would degrade the CMRR because of the unbalanced input impedance inherent to differential amp circuits.
The output of the bridge is about .1 - .4 millivolts peak
As far as op amp choice i only thought of using a 741 out of frustration just to get everything balanced in the circuit and worry about the noise related to high bandwidth op amps later.
My reason for choosing the lm318 was more or less, what do we have in the lab thats fast enough.
Thanks again Everyone,
- James
In the 1960's, Tektronix made a Type Q plug in for their 500 series oscilloscopes. This was a strain gauge bridge using AC excitation rather than DC excitation. This allowed AC coupled amplifiers, which avoided the offset problems entirely. You might be able to use a similar approach with a carrier frequency in the MHz range to allow for the 500 KHz bandwidth.
This avoids problems with offsets and with low frequency noise, which tends to be higher than midband noise.
In this frequency range, subtraction and common mode rejection can be done with transformers and balance trims, which may be more stable than solid state amplifiers. Bifilar windings are a simple way to get good balance.
An arrangement similar to an "RF Noise Bridge" would give an output signal that is ground referenced and would not need a differential amplifier for further amplification. The noise bridge replaces two adjacent legs of a normal bridge with a balanced center-tapped transformer winding.
This avoids problems with offsets and with low frequency noise, which tends to be higher than midband noise.
In this frequency range, subtraction and common mode rejection can be done with transformers and balance trims, which may be more stable than solid state amplifiers. Bifilar windings are a simple way to get good balance.
An arrangement similar to an "RF Noise Bridge" would give an output signal that is ground referenced and would not need a differential amplifier for further amplification. The noise bridge replaces two adjacent legs of a normal bridge with a balanced center-tapped transformer winding.
-
- Posts: 458
- Joined: Thu Feb 05, 2004 1:01 am
- Location: Minneapolis, MN
- Contact:
One of the classic articles on noise was "Noise and Operational Amplifier Circuits" by Lewis Smith and D. H. Shiengold. This was originally published in the March 1969 issue of Analog Dialog (Vol 3, No 1). It was republished in 1991 in a paperback book called "The Best of Analog Dialog- 1967 to 1991". PDF files of that book are available on the Analog Devices web site (www.analog.com). The other articles in that book include several on layout, grounding, and shielding. Its worth looking at.
-
- Posts: 98
- Joined: Thu Dec 06, 2001 1:01 am
- Location: ohio
- Contact:
Who is online
Users browsing this forum: No registered users and 3 guests