Building a +-15V dual supply for 32 instrumentation amplifiers

Thread Starter

fieryfire

Joined Feb 14, 2017
138
Hi all. I am currently using 32 dual channel instrumentation amplifier (INA2128) to amplify strain signals from bridges. Although this requires that I would need to power the in amps via dual line power supply. Could anyone suggest a suitable method on how i would be able to achieve this? Either by an IC or other means. I plan to use a 12 v 3Amp power supply through a 2.5mm jack. Although this isnt a hard requirement. Please suggest a suitable option for this application.
 

Reloadron

Joined Jan 15, 2015
6,140
You can likely buy a +/- 15 volt dual supply for less than you can build one. How much current do you plan to need, in other words +/- 15 volts how many watts supply? For an industrial app with 32 IAs I would likely buy two and place one on a shelf as a spare. Anyway you need to estimate your needed power and then allow some head room (maybe 25%).

Ron
 

AnalogKid

Joined Aug 1, 2013
9,414
How critical is it that the system be powered by a single supply? Turning a single +12 V source into +/-15 V isn't terribly difficult, but it will create significant analog noise that must be filtered at each inamp.

Does your circuit require +/-15 V, or can it run on +/-12 V?

If you are stuck with the +12 V source, consider purchasing a board-mount DC/DC converter module with +/-15 V outputs and filter the doors off of it.

ak
 

ebp

Joined Feb 8, 2018
2,332
I've used isolated DC-DC converters for a bunch of instruments. The ones I've used tend to allow high frequency spikies through from input to output, which is a potential issue with a switcher as the main supply. It is generally not desirable to put an inductor in the output common (not an issue if the outputs are 4-terminal with separate 0V for each output), so I recommend inductance in both input connections (full LC filter). A "bad" common mode choke with some normal mode inductance can work well. Beware of the fact that some of these converters don't like a lot of capacitance directly on the outputs, so LC pi filters are generally most suitable. Also beware of making high-Q filters and ringing noise spikes down into the bandwidth of interest. There are numerous ways to spoil the Q of the filter (resistance in the inductors comes for free with the right inductor, series RC across the main Cs, etc.) This is a good thing to fiddle with with simulation. I find the easiest thing to do is use "ideal" Ls & Cs and discrete resistances rather than adjusting the resistances of more real L & C models. With good quality real parts you can get alarmingly high Q, and if you can't use it to your advantage it may be much to your detriment.

Of course not allowing more bandwidth than is useful in amplifiers is helpful noise-wise, but that can be tricky with IAs and make things worse if you aren't very careful.

In much of the most recent work I did, the excitation for the bridge consumed more power than the amplifiers and I was using multiple amps per bridge because of a temperature compensation signal from the bridge.

Watch the current limit on power plugs and jacks. A lot of them aren't rated for more than an amp or so.
 

ebp

Joined Feb 8, 2018
2,332
On caps for power supply filters, in case you haven't run into it:
I strongly recommend staying away from most of the high dielectric constant surface mount ceramic types. The negative voltage coefficient of capacitance of those things is abominable. Some of them are down to 20% or less of nominal capacitance if you operate them at anywhere near rated voltage. The X-something-something types (eg X7R) aren't free of the problem, but are much much better. A 4.7 µF X7R might actually yield more capacitance at working voltage than a 22 µF Z5U that is the same physical size. They're horrible, I tells ya, horrrible!
Some mfr's data sheets are much better than others when it comes to this nasty detail.
 

Thread Starter

fieryfire

Joined Feb 14, 2017
138
You can likely buy a +/- 15 volt dual supply for less than you can build one. How much current do you plan to need, in other words +/- 15 volts how many watts supply? For an industrial app with 32 IAs I would likely buy two and place one on a shelf as a spare. Anyway you need to estimate your needed power and then allow some head room (maybe 25%).

Ron
Hi Reloadron, I estimate my total power between 3 and 4 Amperes. Could you provide me a link or direct me to where i can find a +/-16V power supply?
 

Thread Starter

fieryfire

Joined Feb 14, 2017
138
How critical is it that the system be powered by a single supply? Turning a single +12 V source into +/-15 V isn't terribly difficult, but it will create significant analog noise that must be filtered at each inamp.

Does your circuit require +/-15 V, or can it run on +/-12 V?

If you are stuck with the +12 V source, consider purchasing a board-mount DC/DC converter module with +/-15 V outputs and filter the doors off of it.

ak
Well i have done various simulations of my circuit on difference rail to rail voltage.

+/-5V seems to simulate just fine.
Although if i use +/10V and +/-12V, the simulation cannot run.

Again on trying +/-15V the circuit works just fine.

It is really strange indeed. I would highly recommend a suggestion for a noise free power supply of +/-15V. Most of the power draw comes from the bridges themselves. 64 bridges consume about 7.2mA each. The INA2128 probably doesnt consume much.
 

Thread Starter

fieryfire

Joined Feb 14, 2017
138
I've used isolated DC-DC converters for a bunch of instruments. The ones I've used tend to allow high frequency spikies through from input to output, which is a potential issue with a switcher as the main supply. It is generally not desirable to put an inductor in the output common (not an issue if the outputs are 4-terminal with separate 0V for each output), so I recommend inductance in both input connections (full LC filter). A "bad" common mode choke with some normal mode inductance can work well. Beware of the fact that some of these converters don't like a lot of capacitance directly on the outputs, so LC pi filters are generally most suitable. Also beware of making high-Q filters and ringing noise spikes down into the bandwidth of interest. There are numerous ways to spoil the Q of the filter (resistance in the inductors comes for free with the right inductor, series RC across the main Cs, etc.) This is a good thing to fiddle with with simulation. I find the easiest thing to do is use "ideal" Ls & Cs and discrete resistances rather than adjusting the resistances of more real L & C models. With good quality real parts you can get alarmingly high Q, and if you can't use it to your advantage it may be much to your detriment.

Of course not allowing more bandwidth than is useful in amplifiers is helpful noise-wise, but that can be tricky with IAs and make things worse if you aren't very careful.

In much of the most recent work I did, the excitation for the bridge consumed more power than the amplifiers and I was using multiple amps per bridge because of a temperature compensation signal from the bridge.

Watch the current limit on power plugs and jacks. A lot of them aren't rated for more than an amp or so.
This is exactly what i am worried about. The introduction of the high voltage spikes to my analog signal conditioning chain would be a big problem. Could you guide me on how exactly i would be able to approach a +/-15V power supply like the way you designed. It would be much appreciated.
 

Thread Starter

fieryfire

Joined Feb 14, 2017
138
On caps for power supply filters, in case you haven't run into it:
I strongly recommend staying away from most of the high dielectric constant surface mount ceramic types. The negative voltage coefficient of capacitance of those things is abominable. Some of them are down to 20% or less of nominal capacitance if you operate them at anywhere near rated voltage. The X-something-something types (eg X7R) aren't free of the problem, but are much much better. A 4.7 µF X7R might actually yield more capacitance at working voltage than a 22 µF Z5U that is the same physical size. They're horrible, I tells ya, horrrible!
Some mfr's data sheets are much better than others when it comes to this nasty detail.
Completely agree with you there bro. And yes thank you for the advice. This could explain the noise elements in my previous designs. X7R capacitances are the way to go.
 

Reloadron

Joined Jan 15, 2015
6,140
Hi Reloadron, I estimate my total power between 3 and 4 Amperes. Could you provide me a link or direct me to where i can find a +/-16V power supply?
Depending on your location will determine where you find a power supply. When I did projects like this which were work related I preferred linear DC supplies over the less expensive switch mode designs. That said I liked brand names like TDK Lambda, Crouzet and Sola to name a few. Linear supplies while not inexpensive worked best in my applications. Using a SMPS as everyone has mentioned, you need adequate filtering, especially in sensitive circuits. Also your reference voltage for your bridges needs to be clean and very stable.

Looking around this morning linear supplies are not as abundant as they once were. :(

Ron
 

Thread Starter

fieryfire

Joined Feb 14, 2017
138
That totals 0.5 A for the bridges and less for amps. Where is the rest of the "between 3 and 4 Amperes" used?

ak
I would say that 3A is a bit of an overkill. Since we will be using an acquisition card which takes about 500mA itself(LABJACK T7). But this would be powered separately. Besides this bridge and the amplifier. There are also 64 stages of active filtering from a 2nd order butterworth filter from an Op Amp each. I am unsure of the power draw from the op amps at peak load. but i would assume that, the total power consumption would be typically under 1A.
 

Thread Starter

fieryfire

Joined Feb 14, 2017
138
Depending on your location will determine where you find a power supply. When I did projects like this which were work related I preferred linear DC supplies over the less expensive switch mode designs. That said I liked brand names like TDK Lambda, Crouzet and Sola to name a few. Linear supplies while not inexpensive worked best in my applications. Using a SMPS as everyone has mentioned, you need adequate filtering, especially in sensitive circuits. Also your reference voltage for your bridges needs to be clean and very stable.

Looking around this morning linear supplies are not as abundant as they once were. :(

Ron
Ahh perhaps i might have been unclear. What i was looking was for a breadboard option to power up the signal conditioning chain. I was looking for a split rail power supply like the LTC3265, which gives about +/-12V output from 15V input. I could perhaps buy an ac adapter with output at 15V and design my circuit around the LTC3265. the total power it can give is about 50mA, which is sufficient to power up 33 INA2128 in amps. provided that each INA2128 consumes a maximum of 1.4mA
 
Top