Problem with amplifying a pressure sensor for measuring blood pressure

I'm building a system to measure the real-time pressure in a cuff of a blood pressure system.

I'm using the Omron SMPP-O2 sensor, amplified with an INA122 and sampling with a Teensy 3.0. I have it wired up but the datasheet is not very clear to me, and I'm not entirely sure how to provide it what I think it needs.
I'm getting what appears to be erratic ADC values and don't seem to change with pressure.

It appears I need to deliver a constant current (100 uA), and I'm not sure what I should use to do that.

I'm attaching my breadboard connections with Fritzing, and the relevant datasheets. If anyone would care to take a look. Thanks.

And this is the teensy 3.0 pinout


BloodPressureBreadboard_bb.pdf (682 KB)

ICpics.pdf (47.7 KB)

2SMPP_1112.pdf (498 KB)

ina122.pdf (745 KB)

I am not familiar with the Teensy boards, but looking at your breadboard picture makes me wonder: Are you shortcutting Pin12 + Vbat + 3.3V ? And also Reset + pin13 ?

Your sensor is this, http://www.omron.com/ecb/products/sensor/21/2smpp-02.html
For accurate pressure measurements, it should should be powered with a constant current of 100uA.
You use it with 5V and I think the connections in your breadboard drawing are okay. It should work, but you won't get the accuracy it is designed for.
The datasheet show how to use a opamp as current source.
Sensor seems to be 4x20k serial/parallel (I'm not sure), and with 100uA, the sensor needs about 2V (but I'm not sure).
With 2V, the outputs should be around 1V with a little offset.

In the datasheet the absolute maxium for the supply current is 130uA. With 5V you might have damaged the sensor already.

If you remove the INA122 and use a potmeter for the analog input, are the readings normal ?
Could you measure the sensor output with a multimeter to see if that is normal ?

I'm only using 3.3V, not 5V. I don't know how much current would come from that so I may have destroyed the sensor. So, I should use an op-amp to get a constant current source? I'll look over the example again, as I didn't realize that. I'm pretty naive to the electronics but I'm slowly getting there. It was confusing on the datasheet, cause I think it wants 2.5V for power, but the mouser page had it listed as 5V. And then it mentions current, so I assume it's somewhere in that range but current needs to stay constant.

Good point, I should check output directly from the sensor.

You are suggesting that I put a potentiometer and take the power source and divide the voltage and read the inputs? I think I checked the analog inputs on another sensor, but I should double-check that cause I can't quite remember. I'm pretty sure I get 0 connected to A-REF, and a 65535 connecting to 3.3V power pin.

Currently my analog input is hovering around 20,000, which should be at about 1V ((65536/2000) / 3.3).

Thanks.

Perhaps the sensor is damaged and not accurate anymore but not broken yet. Perhaps you can continue to use this one for testing.
If you are going to buy a new one, you could buy a sensor for a 3.3V supply and single linear voltage output. Many sensors have the opamp inside. Or a sensor with a digital output.

Some (or a lot) noise is normal with these sensors.
You could add 1nF or 10nF capacitors from the analog sensor outputs to ground.
After that, also take the average of about 5 readings.
The result should be good and stable after that.

I bought several more sensors when I ordered (somewhat expecting something to go wrong) and my reluctance to pay more for shipping than items. I'll try and figure out as much as possible with the potentially broken one. It should be a good learning experience.

Ok, finally had some time to sit down and check this out again.

It was definitely good to measure each item separately and check with a voltmeter.

I was probably wiring a couple of things wrong, and luckily it didn't seem to hurt anything.

The sensor was outputting -4mv with a null signal. Then the signal turns '+' for most of the output range. The datasheet mentions -40mv at resting, but in other places I've seen -4mv.

I only then connected up the INA122 and got up to nearly the 3.3V output. I was worried about the -4mv being amplified, but it seems to start from zero and go '+' from there. So, it looks like this combo works pretty well. Signal seems stable with me just blowing pressure into the tube.

Now, I'll pneumatically connect up to a blood pressure cuff and I'll post a pic of the pressure signal when I do. Cool!

I'll also update with my breadboard, I think I had some things wrong.

Thanks for your help.

I haven't calibrated the values yet so the values are approximate, but the output looks pretty good. I changed the resistor to 10Kohm.
Attached is a picture of the graphed output and the indications of what is occurring.

bloodPressureReadings.pdf (27.5 KB)

That is very cool.
But you need to reduce the noice. Both in hardware and software.
Did you find a way to supply 100uA ? That is the only way to get accurate values for the pressure.

I haven't taken any measures to reduce noise except to use the internal teensy 3.0 feature to output the average of multiple samples (ie 16).

As it is, I think the noise is probably ok. The initial noise with the pump is the nature of the erratic piezo pump in getting to max pressure.

Google on constant current ic.

Thx. I didn't quite get find anything that was super clear if there was a single IC that I could get to do this for me.

It looks there's a typical op-amp setup to do this, but requires several resistors and an op-amp.
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=10&cad=rja&ved=0CHUQFjAJ&url=http%3A%2F%2Fwww.physics.byu.edu%2Ffaculty%2Fpetersonb%2Fphys240%2FConstantCurrentSource.pdf&ei=5EItUfvUB5HSiAKbooHIDQ&usg=AFQjCNEegrXZxq3JjrtV9dBqsFn3UaezcQ&sig2=o5itmazSvhixlP5Yr0JmiA&bvm=bv.42965579,d.cGE

I did find this:
http://www.reuk.co.uk/LM317-Current-Calculator.htm

Could I just use an LM317 with a single 33kOhm resistor?

Or what if I combined this with a voltage reference IC, or is that too demanding?

That LM317 could do it. But I'm not sure that 5V is enough, I don't know for sure if the sensor needs 2V and I don't know for sure the 5-2 = 3V is enough for the LM317.

This is the same circuit with a different explanation:
http://users.telenet.be/davshomepage/current-source.htm

He mentions

When it's impossible to lower the voltage of the powersupply and the current is high maybe a switched current source will be better.<> I will explain this in the future.

Call me "the answer is out there somewhere" Mulder. Just trying to help!

Here's the updated breadboard in case anyone wants to wire it up.


https://docs.google.com/file/d/0B_XDOJc4fAYxWk9ON1pLVTFDdHM/edit?usp=sharing

edit: The 2SMPP-O2 ic is upside down, so it probably looks wrong. It only comes in a SMD and I had to drill a hole in a DIP adapter for the pressure port, and put it upside down. So it goes like this in pic:

1 2 3 -
6 5 4 -

Also, do you consider this very noisy now that I've zoomed in? Attached pdf.

BTW, it's also sampled at 100Hz in the picture.

bloodPressureReadingsZoom.pdf (18.3 KB)

I still think it is somewhat noisy.
A hardware filter, like a capacitor over the opamp or at the sensor input or both should make a smooth line.

I have a few 0.1uF caps. Will that work?

So where there is voltage output of the sensor, I bridge it to ground with a capacitor?
And where there is voltage output coming out of the opamp, I bridge that to ground too with a capacitor?

If you are looking for a 2 terminal current source try a CC100 diode. It's a specially constructed Jfet that is a 2 terminal constant current diode (@ 100uA). It can be done with a Jfet and a resistor but I suspect that this is a little beyond your current grasp of theory. There are extensive design notes available in the App note but I doubt they are available any longer... Siliconix was one Mfr of CC diodes...
The app notes I refer to are from the National Semiconductor analog design app notes, Vol 2.

Bob

I'm sure you're right that I won't comprehend it, but I can get some assistance from others who do. So I'll look into it.

I do see this which may be what you were talking about:

I think part of my noise problem is the resistor I'm using now for the instrumentation amplifier.

I don't know the tolerances on the one I have but it's probably not good. The one I was expecting to use, until it was apparent the output of the sensor was 10x what the datasheet said, was very good. Now I'll just need to find one with better tolerances.

I hooked up the resistor with lower tolerances and the signal looks pretty clean with me just blowing pressure in the tube to the sensor.

Words can be such a pain. Lower tolerance is the opposite of +/- smaller percent. Did you really mean that using a lower tolerance resistor made it work when the higher tolerance resistor did not? If so then the resistance you really want is not the one calculated but closer to what the LT resistor actually is.

A 5% tolerance resistor is higher tolerance than a 10% tolerance resistor. The higher the tolerance, the higher the accuracy must be to pass. It's intuitive if you make things to tolerances but otherwise not.

I think that they test the things every so many and if within 10% then that part of the run goes in the 10% bin even as the process is tweaked for higher accuracy. Still there's nothing saying that +/- 10% can't be within 1%, just that's not likely given consistent manufacture.