Is isolation my solution? Advice requested

DaveC wrote:

You could isolate the MUX and InAmp in the DAC domain, but that won't help if it's the DAC that's causing the damage.

If I do this and I still get failure then the problem is not coming from the boost converter, right?  If I isolate in this way I would have at least narrowed down where to look for problems, right? 

Do you have any suggestions for a particular component or class of components to do this?

Yes, but I'd need to see at least a partial schematic showing the various interfaces.

DaveC

Regarding your 60V supply.

Using a multimeter to monitor the supply is inadequate. It could have a 10V ripple and you wouldn't notice...

ESD: In general external connections should be protected from ESD. I'm assuming that this circuit is connected and disconnected from the resistor network that you are measuring, so you should definitely look into that.

The diode would not be across the high side shunt, but from the AMUX inputs to ground. You said that the input voltage to the AMUX is only on the order of 100mV even though your sketch doesn't make it clear how the muxes are actually connected to the circuit...

We could help you much better if you could provide a more detailed schematic.

Klaus

Thanks again.

KRZ wrote:

Regarding your 60V supply.

Using a multimeter to monitor the supply is inadequate. It could have a 10V ripple and you wouldn't notice...

True, I have not considered this.

I'm assuming that this circuit is connected and disconnected from the resistor network that you are measuring

correct

KRZ wrote:

The diode would not be across the high side shunt, but from the AMUX inputs to ground. You said that the input voltage to the AMUX is only on the order of 100mV even though your sketch doesn't make it clear how the muxes are actually connected to the circuit...

It is sub 100 mV on the low side but on the high side it could be up to 60 V.

here is a better representation of how the mux is connected

and here is how I interpret your suggestion with the diodes

The reason I have the mux inputs on either side of the high side shunt is because any of the DAC outputs can be ground-60V or float.  I may want to alternate these outputs between high and ground.  This probably sounds odd, so let me explain what the resistor network is. 

I mentioned that I know the layout of the resistor network but not the values.  This is because the resistor network is actually a microfluidic network, and I am transporting fluid in the network via electroosmosis by applying an electric potential between fluidic inlet and outlet ports.  In the above example I show two DAC outputs.  If I alternate these outputs between high and ground then I will transport fluid back and fourth in the microfluidic network. 

As one fluid in the network displaces another, the conductivity of the channel will change and the measured electrical current will change.  I can use the known properties of the channels along with the change in electrical current to determine flow rate and velocity. 

OK, that makes it clear.

The reason why you are killing the parts is because the mux inputs are directly connected to the DAC outputs and therefore the DAC will try to drive them to 60V. Before that happens the body diodes of the AMUX will turn on and the DAC will drive as much current as it can into the positive supply of the AMUX (through the diodes inside the AMUX).

Most of the time the AMUX will go first, but sometimes it's the DAC output that dies first...

Your best fix would be to add series resistors into the leads going to the AMUX plus the diodes I suggested right at the pin. The resistors should be large enough to not load the DAC output too much as they are effectively grounded by the diodes that I suggested. However they can't be too large either so as to not have an undue effect on the INAMP's ability to measure the differential voltage.

Klaus

EDIT: 600k resistors would load the output with 200uA (2*100uA assuming your shunt is relatively low resistance). The DAC output is rated at 1mA, so your fluid's resistance should be higher than 75k when driven with 60V. Your instrumentation amplifier should have a much higher input impedance 1Gig or more. Also any capacitance on the leads after the series resistors will add to your settling time and limit your switching frequencies.

The reason why you are killing the parts is because the mux inputs are directly connected to the DAC outputs and therefore the DAC will try to drive them to 60V. Before that happens the body diodes of the AMUX will turn on and the DAC will drive as much current as it can into the positive supply of the AMUX (through the diodes inside the AMUX).

So just for clarity on my part, you would expect this to happen even though I am only activating the AMUX switches on the low side.

Most of the time the AMUX will go first, but sometimes it's the DAC output that dies first...

This would agree with my observation.  The multiplexer dies quicker than the DAC.

assuming your shunt is relatively low resistance

The shunt resistors are 500 ohm.

The DAC output is rated at 1mA, so your fluid's resistance should be higher than 75k when driven with 60V.

If my memory serves, driving water through a 2 cm long channel at 60 V will give me about 80 uA.

Your instrumentation amplifier should have a much higher input impedance 1Gig or more.

I am using a MAX4238.  The datasheet does not say what the input impedance is...

Also any capacitance on the leads after the series resistors will add to your settling time and limit your switching frequencies.

I am not too worried about the switching frequency.  To pump fluid from point A to point B takes around 10 seconds to several hundred seconds depending on the length of the channel, fluid, and voltage.  Thanks for pointing this out though.

Thanks everyone for your responses.

I will order some new ICs and update this post with the results

I am actually surprised that it worked even just briefly.

The DAC should not be able to drive its outputs much beyond the 5V supply of the AMUX before going into current limit.

Maybe your measurements were actually done at much lower drive voltages than you thought...

The MAX4238 is not an instrumentation amplifier. It is just a low offset high impedance OPAMP that can be used as a building block for an instrumentation amplifier. To use current limiting resistors like I suggested, you will need a proper instrumentation amplifier with high impedance inputs. You can do this by buffering the signals coming out of the AMUX and then apply these to a difference amplifier.

However you usually get better results by just buying an instrumentation amplifier, like an AD8237 or AD8236.

The former has lower offset voltage but higher bias current and the latter has the opposite characteristics.

You need to figure out which error is easier to calibrate out in your setup.

Klaus

I am actually surprised that it worked even just briefly.

If you look back to the diagram in my first post, I show a 3 resistor network.  This is the network that I have spent the majority of my time testing.  Most of my testing has been with actual resistors and not with fluids.  The example I gave is also accurate to what I have been doing.  Which is to say, one of the DAC outputs is left unconnected, another is always high, and the other two alternate between ground and float.  With this experimental setup I did not bother to wire the AMUX to the shunt on the high DAC output.  This is why both DAC and AMUX last for a while at least.

Maybe your measurements were actually done at much lower drive voltages than you thought...

As I mentioned before, when the floating DAC output is connected my current measurements do not make sense given the known resistance values of my network and the DAC output states. As a result I constantly measure (with a multimeter) the output of the DAC (ground, high, and floating) to make sure I am getting what I have programmed.  I also measure the voltage on either side of all resistors in the network including the shunts to see what is going on.  If I simply disconnect the floating DAC output from the circuit (which makes sure that it is truly floating), the current measurements as well as the voltage drops throughout will be exactly as Ohm's law would predict.  And when I say current measurement, I am not measuring current with the multimeter.  I am referring to the calculated current based on the voltage at the ADC and shunt resistance. 

The image below shows current measurements sent from the microcontroller when the floating DAC output is disconnected and known actual resistors (not fluids) are used.  Each increase in current corresponds to a 5 V increase in voltage at the high output DAC starting with 0V.  The spike at the last transition is smaller than typical.  These spikes show up once every 20 transitions or so.  I gave my explanation for these spikes in a previous post (though I could be wrong as to the cause).

The MAX4238 is not an instrumentation amplifier.

Well that is embarrassing...

Thanks for your INAMP recommendations

I thought about this some more and came to the conclusion that my idea to use protection diodes won't work for you. The reason is the same as why you have trouble with floating outputs and that is leakage current.

The outputs of the DAC have a typical leakage current spec of 10uA and it is not production tested, i.e. you don't know what the actual value is and it is likely to change over time and operating conditions. That's why your measurements are off with a floating DAC connected.

Adding the series resistors and diodes will add add even more leakage to undriven legs of your network. The only upside is that this leakage will be better controlled and could conceivably be compensated out. However it is pretty large compared to your measurements.

I think the correct solution for your application might be mechanical relays. Sounds like last millennium technology, I know, but you can't beat the guaranteed low leakage they offer.

If this is a one off instrument then I would recommend you looking into getting a commercially available relay switch matrix. HP made them for the 75000 series instruments. National Instruments also has them for PXI and in other form factors. You could get DAC and multimeter cards as well and build the whole instrument from off the shelf cards...

However, if this is an actual product then you could try PhotoMOS relays such as Panasonic 's AQY210KS which have a maximum guaranteed leakage of 1uA for a more compact solution if that amount of leakage is acceptable.

Klaus