Pressure sensors

[vne147]

Hello everyone. I've started a project is which I plan to measure a differential pressure between 0-1 psid, amplify it, and feed it into the ADC pin of a micro-controller. The differential pressure sensor I'll be using must be a wet/wet type capable of handling either liquid or highly humid gas on either side. After looking around a bit, the sensors that seem to be the cheapest while still meeting the requirements are made by Honeywell. The 24PC01SMT and the 26PC01SMT both look like they would work well with the only difference between the two being that the 26 is temperature compensated and calibrated and the 24 is not.

Temperature compensation for my project would be ideal because the sensor may experience temperatures anywhere between 32F - 100F. Unfortunately, the temperature compensated model is significantly more expensive than the other one ($26 as opposed to $13).

So here's my question. If I used the non temperature compensated sensor, could I build temperature compensation into the circuit instead of having it built into the sensor? How would I go about doing this? Would this be difficult? More or less expensive? Also, if anyone knows of a different sensor that might fit my needs, please let me know. Thanks in advance for any help.

 

[crutschow]

A compensation circuit would require a temperature sensor mounted on the pressure sensor. The temperature signal would then be used in the compensation circuit to provide analog compensation, or by the micro to provide compensation in software. I doubt that you can make such a compensation circuit cheaper then just paying the extra $13 for the compensated device.

 

[vne147]

A compensation circuit would require a temperature sensor mounted on the pressure sensor. The temperature signal would then be used in the compensation circuit to provide analog compensation, or by the micro to provide compensation in software. I doubt that you can make such a compensation circuit cheaper then just paying the extra $13 for the compensated device.

OK, thanks for the input. I had another idea. The datasheet says that the offset null drift over the entire temperature range is only ± 1mV typ. The span shift over the temp range is 5% of the span.

The typical full scale span is 45 mV. So, over the entire temperature range I could expect:

45*1.05 + 1 = 48.25 mV

So the µcontroller could see as much as 48.25 mV for the full scale span depending on the temperature. If I design the amplifier to take an input differential of say 50 mV and output 1 - 4 V then I would be able to measure the entire span regardless of what temperature the sensor was experiencing.

I could then have the µcontroller read a temp sensor mounted on the body of the pressure sensor and correct the output of the ADC accordingly. I would base the correction factors on testing I'd have to perform to measure the drifts over the full temperature range. I would lose some resolution in my analog to digital conversion but it might be the best compromise all things considered.

It would result in a few more lines of code but would only increase the cost of the project by whatever a temperature sensor costs and I know I have seen pretty cheap ones ($1 or less) before.

I hope I explained that clearly. What do you think about that idea?

EDIT: Also, I forgot to ask about the op amp I want to use. I found a very useful application note on the Freescale website that used an OP496 precision op amp for amplifying pressure sensor outputs. I am planning on ordering all my parts from mouser and they don't carry the OP496. Can someone suggest an equivalent part? I can go through data sheets as good as anyone and compare specs but I haven't found a part yet where all the specs line up and I'm uncertain which parameters are critical that I pay attention to in this instance and which ones don't really matter that much.

EDIT AGAIN: Would any of these be a suitable replacement for the OP496?

 

[crutschow]

Likely the most important parameters for your application are rail-to-rail outputs, low input offset voltage, and low input offset voltage drift.

That's a great deal of work to temperature compensate the sensor just to save a few bucks, especially if you factor in the time/cost of calibrating each sensor over temperature. Why the great concern about cost?

 

[Boncuk]

Hi vne147,

why don't you use a Motorola/Freescale pressure sensor type MPX2010DP?

It's a laser trimmed differential pressure sensor with temperature compensation.

The MPX201DP has a range of 10KPa and should be well within the range you want to measure. 1PSI=6.97475793KPa

Here is a circuit with in- and output offset correction and adjustable gain with factors 50 to 1,000.

The schematic uses the MPX2050DP (range 0 - 50KPa), but works with any of these sensors out of the MPX-series.

Regards

Boncuk

 

[vne147]

Likely the most important parameters for your application are rail-to-rail outputs, low input offset voltage, and low input offset voltage drift.

That's a great deal of work to temperature compensate the sensor just to save a few bucks, especially if you factor in the time/cost of calibrating each sensor over temperature. Why the great concern about cost?

Carl,

The reason I'm so concerned about cost is I'm trying to develop something that I plan on one day selling. A production cost difference of ~$10 would be significant I think. Especially when the whole thing will probably only cost ~$20 or so.

Correct me if I'm wrong but when I start making these things, I'm going to have to calibrate each one. I'll have a few pots in the circuit that I'll have to adjust. One of the pots will set the null voltage and the other the amplifier gain. So, the amplifier will output 1 - 4V from 0 - my max psid respectively. Once I do that, I would think that the drift from temperature would be approximately equivalent for all sensors of that model. So, the software compensation should work fairly well without me having to repopulate a look up tabe for each sensor. Does that sound reasonable? If there is a better way to do it short of coughing up the extra money for the temperature compensated sensor, please let me know.

I ran some numbers and even if I don't compensate for temperature at all, it would only affect the results by roughly 5%. I might be ok with that error. I'm going to have to do some experimenting to find out what works best.

As far as op amp selection, I think I'm going to go with the MCP619. It's got rail to rail output, 150 µV input offset voltage max (as opposed to 300 µV with the OP496) and only 2 µV/C input offset drift with temperature. The OP496 is 1.5 but I think 2 will be close enough. The input signal I'm measuring is 4 orders of magnitude greater than the temp drift so I think I'll be OK there.

Thanks for the input.

Hi vne147,

why don't you use a Motorola/Freescale pressure sensor type MPX2010DP?

It's a laser trimmed differential pressure sensor with temperature compensation.

The MPX201DP has a range of 10KPa and should be well within the range you want to measure. 1PSI=6.97475793KPa

Here is a circuit with in- and output offset correction and adjustable gain with factors 50 to 1,000.

The schematic uses the MPX2050DP (range 0 - 50KPa), but works with any of these sensors out of the MPX-series.

Regards

Boncuk

Boncuk,

Hey there. Thanks for the suggestion. I actually already looked at the MPX series which included ordering a few and testing them out. I was pretty excited that I thought I had found a suitable cheap sensor. But I wanted to make sure that the sensor would work in my application where it would come in contact with fluid and/or humid gas so I emailed Freescale tech support about it and I got this response:

Unfortunately, we do not recommend all our sensor (MPXV2010DP) to become in
direct contact with any kind of liquid substance which would damage the gel and
cause the sensor to become out of specification.

The strain gauge and the electronic circuitry for calibration and compensation
are protected by a nitride layer but the aluminum bonding pads which provide
electrical connections between the leadframe and the gauge are not protected, in
order to make the bonding feasible. The complete die is also protected with a
silicone gel. This gel is not fully hermetic, although we use much better gel
for our newer types of pressure sensors, water or any other fluid can penetrate
the gel and can reach the die. When the sensor die is in contact with water
e.g., oxydoreduction reactions between Al/Al3+ and water start as soon as the
sensor is biased. After some working hours, or maybe days, the aluminum pad of
the supply pin is definitely destroyed, so that an open circuit on the Vcc pin
occurs. This is an aluminum corrosion phenomena. But the corrosion phenomena is
stopped when the sensor supply voltage is switched off. There are also some
other failure causes like galvanic corrosion, but the Mean Time To Failure
(MTTF) due to these other causes much longer than the MTTF caused by
electrocorrosion of the Vcc pad. Therefore electrocorrosion is the major failure
cause, and they are permanent.

So, that's why I decided to go with the more expensive Honeywell sensor. It's advertised as a wet/wet sensor so I hope it will work.

Thanks for the circuit. I'll be sure to study it before I design my amplifier.

I haven't begun to layout my circuit yet but this is what I'm thinking so far.

The power source will be two AA and I'll need two voltages in the circuit. 10V for sensor excitation and 5V for everything else. I'll have a boost converter stage to make the 10V. I haven't determined yet if the boost converter output will be stable enough for me to power the sensor directly or if I'll have to bump it up to maybe 13V and then use a 10V linear regulator. I know that would be wasting some power though.

I'll then either use a second boost converter for the 5V or a linear regulator with the 10V output from the first boost converter as its input. Once again a waste of power though.

Is there a better way to do it than that?

I'm thinking about using the MC34063ABN for the boost converter stage. The data sheet says that at 12V input, 28 V output, and 175 mA load the output ripple is 300 mV. That's kind of large but I may be able to reduce that with a few additional caps and or a higher switching frequency.

In addition to the power supply portion of the circuit, I'll have the sensor, the op amp IC, the micro controller, a small LCD, and a few buttons.

Once I design the circuit I'll post is back here for you guys to tear apart and help me improve it.


Thanks for all the advice.

EDIT: Boncuk, can you please repost the schematic at a higher resolution? I'm having trouble seeing the component values and IC names, etc. Thanks!

 

[bobledoux]

Eliminate the op amp. I'm using an ADS1100 AD converter which is 16 bits. It costs $4.50. It has a resolution of .1mv per bit. It also has a prescaler that will multiply your input by 8. Feed your pressure sensor into into the AD and then to the microcontroller. With proper design you eliminate all those Op amp trimmers and the calibration process. If you need to you can drive the AD on one input and use a "Zero adjust" trimmer on the other input for simple calibration. This gives you 15 bits of resolution.

Even if you only need 8 to 12 bits of resolution, having 15-16 allows you to select a narrow range of the AD span--no need for op amp scaling or offset.

Also look at the ADS1110 converter.

 

[vne147]

Eliminate the op amp. I'm using an ADS1100 AD converter which is 16 bits. It costs $4.50. It has a resolution of .1mv per bit. It also has a prescaler that will multiply your input by 8. Feed your pressure sensor into into the AD and then to the microcontroller. With proper design you eliminate all those Op amp trimmers and the calibration process. If you need to you can drive the AD on one input and use a "Zero adjust" trimmer on the other input for simple calibration. This gives you 15 bits of resolution.

Even if you only need 8 to 12 bits of resolution, having 15-16 allows you to select a narrow range of the AD span--no need for op amp scaling or offset.

Also look at the ADS1110 converter.

I thought about using a dedicated ADC instead of an amplifier and the built in µ-controller ADC but I don't think it will work well for my purposes for the following reason. The 24PC01SMT sensor has a full scale span anywhere between 25 - 65 mV. So depending on the span of the specific sensor, a reading of 25 mV could mean 100% (for the case where the full scale span is only 25 mV) or ~38% (for the case where the full scale span is 65 mV). If I went with the ADS1100 or similar IC, I would have to change constants in my code, recomplie, and re-program the µcontroller for each build to account for the differences from sensor to sensor. That seems like a lot more of a headache than just adjusting a pot to make sure the voltage output full scale span is the same regardless of a specific sensor's full scale span. That's why I need a continuosly adjustable gain and not the set programmable gains of the ADC IC.

If I'm missing something obvious that can be done to account for the differences in full scale span from sensor to sensor please let me know. But, the only workaround I see right now is software based.

Also, the op amp is about 25% the cost of the ADC IC.

Thanks.

 

[Reloadron]

If I'm missing something obvious that can be done to account for the differences in full scale span from sensor to sensor please let me know. But, the only workaround I see right now is software based.

I may be off with this as their data sheet is a clear as mud to me on the sensitivity but I see it this way. If you use 10 Volt excitation on the bridge the F/S output will be 45 mV. The output of the bridge will be a function of your excitation voltage. When using a circuit like this, it is very important that your excitation voltage be stable.

Typically with a bridge transducer the output is specified as mV/V at F/S out. If for example I have a 1 PSID transducer and apply 1 PSI to the high side port I will get a F/S output based on my excitation voltage. The sensitivity could be specified as 2 mV/V so with 10 V excitation at full scale I would get 20 mV out from my bridge.

Looking at their spec it confuses me because where they say typical and mention 10 V it would tell me that at 10 Volt excitation the output would be 45 mV which is 4.5 mV/V. The max excitation is 12 V but they say:

Supply Voltage 10.0 Vdc typ., 12.0 Vdc max.

Followed by:

Sensitivity 45 mV typ.
Full Scale Span 25 mV min., 45 mV typ., 65 mV max.

I would think at 12 Volt excitation you would get 54 mV out for F/S?

Since these things are pretty cheap I would order a few and experiment a little. Apply a precise 1 PSI to the high side port and see how it behaves with various excitation voltages.

Ron

 

[Oznog]

You DON'T need to "excite" a sensor with the rated voltage, unless it's got a built-in amp.

It's just a differential Whetstone Bridge of 4 resistances. The voltage rating is only a limit where the I^2*R heating begins to affect calibration. You can use it on a 2v Vdd if you like.

Note that this is a differential signal, and the differences are small. Most PICs don't have a true differential ADC, and only 10 bits. The offset error due to temp can be very significant if your signal is small. The offset of a differential amp (and the offset may vary with temp) may not only be an accuracy problem, it can cause the output to saturate at + or - too soon and not achieve the intended scale.

 

[Boncuk]

Hi vne147,

normally a pressure hose is connected to the sensor's port. The liquid, even sulfuric, won't enter the diaphragm chamber since you only compress air within the pressure hose and the chamber.

Just take care the sensor is elevated above the max liquid level.

You are going to measure static pressure - no dynamic pressure!

The air trapped in the system prevents acid to enter the pressure chamber.

Regards

Boncuk

 

[Reloadron]

I never said the sensor needed to be excited at its rated or max voltage. I did say the higher the excitation voltage to the bridge the higher the output for a given pressure input within the sensors rated limits. I did say the excitation voltage needed to be a stable voltage. I don't see this project going well as to accuracy for the reasons mentioned by Oznog and the fact that the sensor being considered is not a very good sensor if this needs to be accurate. Sometimes it is hard to get accurate and inexpensive in the same project.

Ron

 

[Acetronics]

The air trapped in the system prevents acid to enter the pressure chamber.

Regards

Boncuk

Hi, Boncuk

Halas, no ... Acid vapours will be mixed with the air ...

( air will become saturated with acid vapour - see your physics lessons ! )

Alain

 

[vne147]

I may be off with this as their data sheet is a clear as mud to me on the sensitivity but I see it this way. If you use 10 Volt excitation on the bridge the F/S output will be 45 mV. The output of the bridge will be a function of your excitation voltage. When using a circuit like this, it is very important that your excitation voltage be stable.

Typically with a bridge transducer the output is specified as mV/V at F/S out. If for example I have a 1 PSID transducer and apply 1 PSI to the high side port I will get a F/S output based on my excitation voltage. The sensitivity could be specified as 2 mV/V so with 10 V excitation at full scale I would get 20 mV out from my bridge.

Looking at their spec it confuses me because where they say typical and mention 10 V it would tell me that at 10 Volt excitation the output would be 45 mV which is 4.5 mV/V. The max excitation is 12 V but they say:

Supply Voltage 10.0 Vdc typ., 12.0 Vdc max.

Followed by:

Sensitivity 45 mV typ.
Full Scale Span 25 mV min., 45 mV typ., 65 mV max.

I would think at 12 Volt excitation you would get 54 mV out for F/S?

Since these things are pretty cheap I would order a few and experiment a little. Apply a precise 1 PSI to the high side port and see how it behaves with various excitation voltages.

Ron

Ron,

I agree that the data sheet is pretty horrible but I'm not sure I agree with your interpretation of the min, typical, and max span values. I understand the ratiometric relation between supply voltage and output voltage span but I'm not sure that is what they are trying to say in the data sheet. I'm thinking that because the sensor is not calibrated, there is a larger range of spans that could result. When I look at the datasheet for the 26PC01SMT (the temp compensated and calibrated version), it's range of spans is much much smaller.

But, I do agree with your comment about just buying a few an experimenting. That's exactly what I plan on doing and I'm about to order a bunch of stuff including the sensors from Mouser today.

You DON'T need to "excite" a sensor with the rated voltage, unless it's got a built-in amp.

It's just a differential Whetstone Bridge of 4 resistances. The voltage rating is only a limit where the I^2*R heating begins to affect calibration. You can use it on a 2v Vdd if you like.

Note that this is a differential signal, and the differences are small. Most PICs don't have a true differential ADC, and only 10 bits. The offset error due to temp can be very significant if your signal is small. The offset of a differential amp (and the offset may vary with temp) may not only be an accuracy problem, it can cause the output to saturate at + or - too soon and not achieve the intended scale.

Oznog,

According to the data sheet the offset error due to temp is ± 1 mV. The worst case would be if the sensor had the minimum span of 25 mV and then the offset drift could induce a 4% error. The differential voltage will be small as you said (between 25 - 65 mV) but that will be fed into a differential amplifier and not directly into the PIC. The output from the amplifier will be referenced to the circuit common ground so any run of the mill PIC with an ADC should be fine.

I'm also building the amplifier with a few pots so that I can adjust for the null offset and gain once built. I'll make sure that the amplifier output voltage range is set according so that regardless of the temp drift, I will not saturate it. I understand that I'll be sacrificing some resolution in the ADC with this approach because some of the input voltage range will be dead space but I don't have a requirement to be super accurate with this. If the final circuit had a resolution of ±1% of the total span, I think I'd be pretty happy with that.

Hi vne147,

normally a pressure hose is connected to the sensor's port. The liquid, even sulfuric, won't enter the diaphragm chamber since you only compress air within the pressure hose and the chamber.

Just take care the sensor is elevated above the max liquid level.

You are going to measure static pressure - no dynamic pressure!

The air trapped in the system prevents acid to enter the pressure chamber.

Regards

Boncuk

Boncuk,

I agree that the air trapped in the pressure hose will compress and should prevent liquid from entering the pressure chamber but there will be some evaporation into the trapped air that might affect sensor operation. Also, after repeated uses, cleanings, and the possibility that the device might be stored on it's side or up side down when not in use, there is the possibility that liquid could eventually work its way into the sensor.

Because of that, I'm going with the Honeywell sensor for the time being. If I could somehow get comfortable using the other sensor, I can always change but for now I'm playing it safe.




Thanks everyone for the input. Once I design the circuit, I'll post a schematic so you can critique it.

 

[Reloadron]

Yeah, I just don't know. The span of output just seemed pretty wide to me. Even for an uncalibrated unit but I just don't know. Yeah by all means get one or a few and see how they behave. Must be 20 years ago I got a few Honeywell (SenSym) SCX15AN 15 PSIA transducers and they have a similar spec sheet but much tighter specs. Anyway, once things get going please do post what you find. I would be interested.

Ron

 

[vne147]

OK, I finished laying out the schematic minus the PIC and display.

Basically what's going on is I have a DC-DC boost converter to make 12V from the 2 AAA batteries. That 12V then becomes the input to a 10V LDO voltage regulator (the L4940D2T10-TR) and an ordinary 7805. The 10V output from the first regulator is used to power the pressure sensor only. The 5V output from the 7805 is used to power everything else (op amps, temp sensor, PIC, etc).

The differential output from the pressure sensor is fed into the MCP619 op amp where I'm using 3 of the 4 available op amps. The pot VR2 is for offset adjustment and the pot VR1 is for gain adjustment. The output from the op amp IC1C will connect to an ADC input pin on the PIC.

Last but not least, there is a temp sensor (the **broken link removed**) that I will mount to the body of the pressure sensor. The output of the temp sensor will connect to a second ADC input pin on the PIC. I plan to use the temperature reading to apply a temperature correction factor that I'll have to determine at a later time after some testing.

Anyway, there's my first go at this. Please don't be shy and tell me what I did wrong and how I can improve the design.

Thanks!