Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Help designing customizable instrumentation amplifier board

Status
Not open for further replies.

Triode

Well-Known Member
I work in a lab at the School of Freshwater Science at UW-Milwaukee. I'm pretty experienced with digital circuits, and I can get an op-amp to work, but I definitely am not an expert on analog circuits.

The issue at hand is that we use lots of sensors with a differential output, the span is usually between 20mV and 500mV, all of our systems take 0-5v or more rarely 0-10 or 0-12. Most of the researchers here don't really do electronics and they have to buy instrumentation amps made to their purpose, and the prices are ridiculous for what you get.

What would be great is if we could make a PCB that would take some fairly standard components, some specially chosen resistors, and give you a custom amp, then we could just make a batch order of 50 of them, and of all the common parts and op-amp chips, and whenever we need to amplify a sensor we could produce the amp in no time.

So It would be great if some of the more analog savvy people could tell me (a mechanical engineer and programmer) where to start. For example, what do I need to look for if I'm amplifying a differential signal in this voltage range, and my power supply is single 12v? I've tried to find existing designs that will work by reading about op-amps and looking at diagrams, but I don't know what's up to date, and what will work for my purposes.

Main power supply - 12v from a li-poly battery pack, may be regulated down
Sensor outputs - differential, some sweep 20mV, some as much as 500mV
Outputs - 0-5v in most cases, 0-10, 0-12 in some
Frequency of desired signal is usually low, under 100 Hz
 
Last edited:
While this may not be what you are after or low cost I like an IA (Instrumentation Amplifier) like the Analog Devices AD524 for a number of reasons. Among the reasons is very easily programmable gain of 1, 10, 100 and 1,000 and a variable calibration gain can be easily added. Another feature is the very low noise and CMRR. There are plenty of less expensive chips out there but I happen to like this one. I would use a chip along those lines and add a dual rail +/- supply using either on board solution like a DC / DC supply or external supply. As to a DC / DC supply maybe something like this Cosel U.S.A. Inc. ZUW1R51212 . So it all comes down to what you want the cost per board to be?

Ron
 
The companies that make instrumentation amplifier boards which are specifically for attaching differential sensors to ADC converters tend to charge at least $120 a unit. So if we can make one using a chip that costs $18 that would be fine. Since the PCBs are tiny and we could do batches of 50 or so, I think they would only be a few dollars a piece. So with the other components that could be around $30 and probably under 2 hours of assembly time. Which would be a huge improvement. After all, the time cost of ordering a ready made one isn't actually 0, you have to give them the specifications and get the order approved since it's over $80, and the wait time is 2-4 weeks.

For a place to start, what do you think of a design like **broken link removed**? He isn't using a split power supply, what is the disadvantage of that?
 
What your asking for should be a relatively off-the-shelf item, not a "custom" item. Action Instruments, M-Systems and Analog Devices come to mind. Action instruments has amplifiers that are configurable with a PC.

A reminder that the case, power supply, connectors and real estate (The PC board) is expensive.

I see why you would be using a voltage output because of the battery supply, otherwise a current output would be the way to go.

IA is used below as Instrumentation Amplifier. If Isolation AMP is used below, it will be spelled out.

Questions/comments I have:
1. Is differential output important?
2. Is filtering/ bandwidth limiting required?
3. The 0-10 and the 0-12 will cost you with only a 12 V supply. You could purchase a small low-ppwer DC-DC converter per board to get a +-15 V supply rather than designing one. The switching supply will add noise.
4. The AD524 or 624 are very nice chips.
5. You might be able to get away with a few jumpers/trimmers and fixed resistors to accommodate what you need.
6, As with all IA's, you have to have a bias return path.
7. Also look at Analog Devices AMP01 and AMP02 IA's.
8. You may need a) an IA for the input, a filter and a differential output
9. Is isolation important?
10. DIN rail or waterproof/water resistant boxes?
11. How would you do calibration?
12. Computation of error budget and temperature effects.

With a +-15 V internal supply, you can accomodate what your after.

The 0-5, 0-10 and 0-12 could be another amplifier with a gain of x1, x2 and x? rather than placing all of the work at the input.

i.e. you could have a 50 mV, 250 mV Fs, 500 mV FS and 1 V FS trimmable or jumperable inputs. You may need different trimmer for each range, but they could be selectable with a jumper.

Some ideas., anyway.
 
In addition to the fine suggestions by KISS something I failed miserably to point out is the desire for a dual +/- supply. I work with a wide variety of assorted sensors but let's look at for example load cells. Many load cells work in Tension and Compression so for example a 1,000 Lb load cell may have an output of 2.0 or 3.0 mV/Volt so with 10 volts excitation the full scale output will swing from -20 or -30 mV to 20 or 30 mV. Applications like this work out well using an IA that can swing - to + depending on the input. I am well aware this type of signal conditioning can be done from a single supply op amp using a created virtual ground but prefer a good low noise dual supply op amp. Low noise is also very important depending on how accurate you want the results to be.

While I seriously doubt you will find a one size fits all for signal conditioning you can likely come close. It becomes like trying to define a wide range of sensors. The wider the range, the more little sacrifices we make along the way. So it becomes a matter of compromise.

Just My Take
Ron
 
Nice catch Ron. That adds.

Is it necessary to have the output offset?

There are two reasons to do so. One is the plus/minus that Ron pointed out and the quantization error.
An easy way to describe this is suppose that our input is 7 bits. 7 bits is 127, but for this illustration let's pretend that 7 bits represent 100.
If this fake A/D converter was 1 bit, the error is +-100%. When we are at full scale, the error is 1%.

So, if i had an A/D with 4095 as the full scale value (ignoring lots of stuff), I need to have a certain amount of signal, so I get a reasonable amount of resolution, thus 1-5 and 1-10 V is more attractive than 0-5 and 0-10 Volts.

I hope that makes sense.
 
I admire the concept of what you're proposing. There are some notable pitfalls to overcome.

And you can see by the previous posts that you've given yourself a difficult task.

Years ago I proposed an identical concept to the staff at the lab I worked for here in SC. I was even given the opportunity to pursue it (and I was fully qualified to perform the task).

And I'm totally empathetic with your disgust with the pricing for such commercial equipment (for instance, from Ωmega, and I considered them cheap, at that time). And believe me, I understand your motivation.

But, this is what I discovered:

In-house built instrumentation scares the begeebees out of senior scientific staff (if they didn't build it themselves): total hair-on-fire antics when faced with it. And, to a degree, not totally unrealistic. It's unseemly, but typical.

A primary issue with analog instrument amps, especially in scientific research and factory level production, is consistency of performance, particularly with very low level analog output sensor signal strength and an equally low level amplifier output to the ADC(s). Throw in a little outdoor, or wetlab usage and those environments can really dork with equipment performance (read verifiable data capture).

And if the equipment hasn't been designed and constructed with circuitry (CMR, PID, etc.) and components (often uncommon values) that meet rigorous standards , front end analog amplifier created data stream variants can render that numbers produced unstable and therefore, by definition, unusable.

Of course, in the real world there's tons of data slop. Can't be helped.

But no matter how well you can do this, you're putting yourself in the line of fire for criticism and worse, a solution, for any, ANY data blips. It won't matter that it isn't your fault - it was your equipment.

My two cents worth...
 
Thanks for all the helpful input. This is a small lab and I'm not too worried about the perception of these devices, for high precision systems people would know to use something certified and more expensive, but we make lots of custom built automated systems. In reality if I can get just the 0-5v range, or close to it, even with quite a bit of error it will be good enough for most of these sensors, many are just being used in student projects.

I realize that a decent instrumentation amp that has the correct filtering, band width tuning and protection for industrial use is quite an undertaking, but we can buy those. I've designed a quick and dirty circuit using an AD620 which works, but I figured even for $20 I could probably come up with something more reliable if I talk to some people who know what to consider with analog.

To answer specific questions,
1. Is differential output important?

The sensors we need these for all use Wheatstone bridges, I'm not sure if there is any other way to read them

2. Is filtering/ bandwidth limiting required?

The load cells, strain gauges and magnetic sensors I'm dealing with are all fairly low frequency, even a cut-off above 100 Hz would probably be good.

3. The 0-10 and the 0-12 will cost you with only a 12 V supply. You could purchase a small low-ppwer DC-DC converter per board to get a +-15 V supply rather than designing one.
The switching supply will add noise.

Really, a 0-5v output will cover most of the devices, so I think I'll stick to that to keep it simple.

4. The AD524 or 624 are very nice chips.

I'll look at them,

5. You might be able to get away with a few jumpers/trimmers and fixed resistors to accommodate what you need.

Possibly, I think most of these will be embedded, so just putting in the resistors to set it is ok with me.

6, As with all IA's, you have to have a bias return path.

I don't know what that means.

7. Also look at Analog Devices AMP01 and AMP02 IA's.

These are interesting, most diagrams I see have a lot of support components, would these work as easily as the data sheet makes it look?

8. You may need a) an IA for the input, a filter and a differential output

I hope it's not that complicated, but maybe, I would like to just have one chip with a few support components.

9. Is isolation important?

Usually no, none of these are attached to people and if the are near an inductive load that can be separate. In my uses they are usually battery powered.

10. DIN rail or waterproof/water resistant boxes?

This also depends on the use, some of our machines have big metal instrumentation canisters with O-ring end caps, sometimes we just dip parts in sealer.

11. How would you do calibration?

Since most of the devices are custom, that would depend on the design it's being used in, they might just program it into the device, in the case of a feeler I'm working on that uses a magnetic range sensor I will probably make a routine that calibrates it at the full forward and back position.

12. Computation of error budget and temperature effects.
 
1. Differential output. You didn't understand that one. Differential output usually has a output sense and a ground sense pin. It's another way to eliminate ground loops. One of the BEST ways, at least in factory automation is to use a current loop, e.g. 0-20 mA or 4-20 mA and a fixed resistor at the voltage input. Your battery power kinda nixes that.

6. Bias return path. Any amplifier has an input current. Lets call it 1 nA for fun. It has to be provided a place to go. This usually means a resistor to ground at the input.

===

Also check out the AMP04. I had plans on using some of the AMPxx parts, but it didn't happen.
 
Here is a bridge transducer from Action Instruments which probably meets your specs. **broken link removed**

Not sure if you realluy meant bridge or just a small input Voltage.
 
$400 is about typical of many modules we see. To give you some examples, on researcher is making a large model airplane into a flying drone with a camera, the budget is $4000, and he just needs to read an altitude sensor with a 0-40 mV signal. I have a sensor array for a robot that consists of a bumper, some accelrometers, a magnetic displacement sensor and two load cells at the base of the shocks, the budget is $1000 and it just needs to know if it hit into something and which side it happened on. We do sometimes make experimental setups which require very precise and accurate sensor readings, and for those we buy a good amplification module or have an expert design one. I'm just trying to cover the %95 percent of the cases where we just want to know if a valve is open or closed, if a motor is getting hot, if a robot hit a wall ect. with a cheap amplifier that gets the job done.

I don't know if bridge was the correct term, I'm talking about the output from Wheatstone bridges in many of these cases, so it seems like it would be.
 
Last edited:
I see your point. Using +- supplies will makes your job a lot easier. DC-DC converters such as these http://search.digikey.com/us/en/products/VWRAS2-D12-D12-SIP/102-1527-ND/1558922 are not too expensive.

You do have to deal with LOTS of ripple on the converters though.

So, I still think the block diagram of Differential input w/gain, Band width limiting filter, Differential output with (gain) wouldn't be too bad.
I don't know if you have to cover offset. You should cover zero.

If it's in the cards, you might be able to insert plug in SIPs or DIPS that configure for a certain range and have trims for the ranges. Or possibly just jumpers and trim pots.
 
Here are the parts I've picked out tentatively:

DC-DC converter +/-12V, 83mA, $8.26 Digikey# 102-2095-ND
Precision 2.5v shunt $0.68 Digikey# 296-9585-5-ND

I've read that these can have "total regulation" with the addition of a regulator on each of the output lines. By "total regulation" does it mean less ripple, or just that this would put it closer to exactly the specified output voltage? Wouldn't some large caps on the outputs effectively remove most of the ripple?
 
With the high frequency converters, a simple 3 terminal regulator just isn't fast enough. Note that with that the module I picked you will require an LC filter which is typical. There is also a maximum capacitance. Yes, you want low ESR and also want electrolytics designed for switching power supplies.

The one you picked out is called a Semi-regulated, so it would seem it wants help for regulation and over current protection. Chose your 3-terminal regulator carefully. You could also loose some headroom and you would probably want a low dropout design.

As an aside: a Tracking supply is better for instrumentation circuits than a bipolar one. Tracking assures that ground is half-way between the two supplies. There isn't a LOT of talk about tracking power supplies.

The OP amp itself will also reject the power supply ripple and so will the common mode rejection.
 
I see, the one you linked to has a full diagram on the data sheet.
(**broken link removed**)
I'm assuming the ones on the output are to reduce the ripple, is the input one for filtering or a buck-boost source?

Would these inductors be suitable?
Output inductors: http://search.digikey.com/us/en/products/11R472C/811-2022-ND/1998209
Input inductor:http://search.digikey.com/us/en/products/RL622-560K-RC/M9990-ND/946874

It says you can leave ctrl pin 3 open if you want the device on. Since I don't have a reason to use that is there any reason not to just leave it disconnected?

I notice it says "To further reduce output ripple, you may increase the external capacitor, choose a capacitor with low ESR" But there is a table which gives the Cout value as 330 uF for the +/- 12v unit, and a note that capacitance should not be "too high". so is 330 uF a good value to go with?
 
Last edited:
I would suspect that th eLC filter in in input has two purposes:
1) To prevent radiation of the switching frequency back into the DC supply.
2) TO basically add a bit of current limiting since the current through an inductor cannot change instantaneously.

There is some "brief" theory here: **broken link removed**

If the datasheet allows open, then leave it open. You can always just bring the point out to a test point or pad if someone might find a need for it.

-- Cap
My take is that C is the highest you should go. I just picked this one http://industrial.panasonic.com/www-cgi/jvcr13pz.cgi?E+PZ+3+ABE0006+EEFSX0D271ER+7+WW as an example. It's not 330 uF, but it shows you in the datasheet that the ripple frequency is 1 MHz. Digikey p/n: PCE4480TR-ND in case the link doesn't work.
 
For now I'm sticking with through hole components. How does this cap look? It has low ESR, I don't know of any reason it wouldn't work.

For the voltage reference I'm looking at the datasheet shows a trimmer pot on the diagram. In use are you just supposed to test the output and adjust the pot till you get 2.50v? Would it be a good idea to put a trimmer with my gain resistor as well to fine tune the gain?

If it helps, this is one of the most common sensor types for us to amplify:
Magnetic Displacement Sensor Digikey# 342-1008-1-ND
It goes from +60mV to -60mV, when fed a 5v input across the bridge, which is what we would probably be giving it. It has a amplifier circuit shown on its datasheet, one thing I notice is that it shows a 10K resistor on both leads from the bridge to the amplifier saying "The 10 kilo-ohm input resistors present a high impedance from the Wheatstone bridge" I don't know if I should do that in my circuit too, I guess I'll try it on the breadboard and see. Maybe it should be in the general amp design if its always going to be used on Wheatstone bridges.
 
Cap: OK
2.5 V ref: Probably no need, since the system likely has to determine the zero and can subtract
Gain trim: Probably

You can get better performance by using a reference on the bridge rather than the power supply. You could even exite the transducer with an amplified version of the reference.

The 10K resistor. That resistor is extremely important as I learned the HARD way. It allows a place for the bias current to drop across and two of them provides some bias current compensation. I'm not sure what they mean by the 10k presenting a high-Z?
 
As to gain I would allow for adjustable gain and maybe even within ranges. The thinking being that just because for example a transducer specifies its output as mV/V being for example 2 mV/V that is not necessarily true. If you want to actually calibrate a sensor applying for example a known load you want to be able to adjust the output for a specific gain.

Just My Take
Ron
 
I may have explained that oddly, the 2.5V reference will be attached to the REF pin of the AD620 instrumentation amplifier. I'll combine my circuit schematics and post them. Basically the 12 battery connects directly to the DC-DC converter inputs, then there is the filtering specified in the datasheet using the caps and inductors I mentioned above, the + and -12V supply Vins to the AD620, the inputs from the Wheatstone bridge on the sensor each go through a 10K resistor then into the AD620 inputs, and the LT1009 shunt connects to the AD620 REF pin. For the gain resistor, which should be near a value of 1210 ohms, I'm thinking I'll use a 1.1K resistor and a 200 trimmer pot in series.
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top