Filtering out very low voltages outside a certain frequency

Thread Starter

Rock Slate

Joined Jul 22, 2015
27
Hi guys,

I have a 40Khz signal which is very heavily attenuated from a piezo sensor. I am thinking of using some amplifier(opamp) like LMP7702 to amplify these signals. However before I start I would like to know whether this particular opamp is a correct choice. The needs are:
- The amplitude of the signal will be only a few millivolts(5-10mV) due to attenuation
- The required frequency is 40Khz

Does anyone have any suggestions?
 

Thread Starter

Rock Slate

Joined Jul 22, 2015
27
Well, I was looking for the opamp to use which would meet these specs. Gain would be around 1000, the passband can be 38-42Khz. I still dont have the equipment to determine the sensitivity. I was looking for which opamp would meet these requirements
 

Papabravo

Joined Feb 24, 2006
21,159
A voltage gain of 1000 is 60 dB. You're unlikely to get that much from a single stage without problems. If you're not an experienced designer and layout person, I would start with something on the order of three stages of 20 dB each. Depending on the stopband attenuation you require, you could implement the filter on one, two, or all three stages. I'm not familiar with that particular opamp, maybe somebody else on this board is. Can you provide a link to a datasheet?
 

Papabravo

Joined Feb 24, 2006
21,159
I see the paper on the EEG application. I'll take you word for the similarity of the applications. I'm still worried about stability and oscillation with such a large gain in one stage. The link to the datasheet is missing.

OK - found the datasheet. Actually the HTML summary page.
http://www.ti.com/product/lmp7702
Look at the parameter called GBW. It stands for gain bandwidth product. For this part the number is 2.5 MHz. It means that for signal at 2.5 MHz. the maximum gain is 1. For 2.5 MHz / 40kHz = 62.5, it means the maximum gain for this part at that frequency is about 36 dB. You don't really want to push the limits like that so a gain of 20 dB leaves some margin for you. Three stages at 20 dB will get you 60 dB at 40 kHz.

The next thing to look at is slew rate. This is how fast the output can change in response to an input. This part has a slew rate of 1V per μsec @ a voltage gain of 1. If your output swing is 10V it will take 10 μsec to go from 0V to 10V and another 10 μsec to go from 10V back to 0V. 40 kHz is a period of 25 μseconds, so that means this part is just barely fast enough to keep up with the input.

This does not surprise me because nothing in the EEG application is as fast as 40 kHz. I'm afraid this one gets a vote of no confidence.
 
Last edited:

Thread Starter

Rock Slate

Joined Jul 22, 2015
27
Added the link(forgot). I can always add compensation netwroks and move the first pole farther away. But I am still worried as to whether this is a right opamp to choose
 

MikeML

Joined Oct 2, 2009
5,444
Spectrum of EEG is 50Hz and below. How is that relevant to 40kHz?

Highest gain you could get out of that amp is 2500kHz/40kHz = 62.5
 
Last edited:

Thread Starter

Rock Slate

Joined Jul 22, 2015
27
The filter pass band can be changed. I was looking for low noise since this is a major requirement for these sorts of sensors. The opamp I linked to has a unity gain bandwidth of 2.5Mhz
 

AnalogKid

Joined Aug 1, 2013
10,987
I think you should abandon the idea of using the LMP part for this project, for the reasons in posts 4, 5, 8, and 10. Also (piling on), There will be significant phase shift and stability issues when running right at an operating point that is directly on the open loop gain plot. And the amp will be running very nearly open loop, so there will be almost zero negative feedback. this means that all of the standard opamp operating assumptions about input impedance, output impedance, and distortion no longer apply to your circuit.

A current feedback video opamp might do what you want in one stage, but they usually have a relatively low input impedance and poor noise levels. Better to have an input stage that can be matched to the sensor output impedance and provide some gain, followed by another amp to make up the rest of the gain. You want each amplifier stage to have at least 20 dB of negative feedback at the frequencies of interest. Another option is an integrated instrumentation amp, a form of a multistage amplifier architecture.

ak
 

Thread Starter

Rock Slate

Joined Jul 22, 2015
27
Thanks for all the replies. So when I look at the opamp data sheet these are the things I should be looking for:
1- GBW
2- Input Impedance should match that of my piezo(I really dont know what the impedance of my piezo is, but I know that is is pretty high? I need to have the amp perform impedance matching also, but since most in-amps have very high i/p impedance I dont think this should be a major issue)
3- It should have low noise, with ability to reject common mode signals(CMRR should be very high)

Based on these specifications, I have identified the following:
http://www.linear.com/product/LT1028
There also seems to be the LT1128 which has 13Mhz GBW

I also found this:
http://www.analog.com/media/en/technical-documentation/data-sheets/AD8615_8616_8618.pdf. This has 65uV offset voltage with GBW 24Mhz

For better isolation , I though I should go in for an Instrumentation Amplifier followed by a gain stage:
http://www.analog.com/media/en/technical-documentation/data-sheets/AD8421.pdf

It looks to be an overkill. Does anyone have any suggestions?

I have looked at the direct integrated solutions like the MAXIM35101. But I need to deal with signals which are attenuated much more heavily. Thats why I need a more capable analog front end
 
Last edited:
Top