INA240: Ringing on output from low side current monitoring & Tina results

BP101,
I'm discussing this with our design team presently to see what they think about the causes of the ringing and how to mitigate it. I'll let you know the results of that discussion shortly.

Hi Jason,

Perhaps the INA240 was never tested low side PWM where some of the phase kickback (FET dv/dt) ends up going to ground through the shunt. The other part dv/dt goes to B+ at times so high side monitor would likely have similar issues.

It might behoove to install 240 in phase if it can tolerate half the 160v supply CMM. Seemingly that would put it in safe area +80v if lab testing confirms CMM is isolated in the switch branch. Only that gives little room for high voltage transient spikes know to exist in an inverter environment, capacitors might not smooth. 

Oddly the INA240 is not producing a symmetrical wave shape in both directions and seem to recall the 282 had the same issue looking at past captures. The wave shape the 240/282 produces has a sine wave top flag waving appearance but is missing the negative complement, both REF/2 or tied REFs.

Does anybody notice the very first ringing in 1st Tina waves posted looks like State of Texas?

We were just whiteboarding some ideas and we're wondering if something like this would work to curb the ringing?

I haven't built the sim myself (obviously from the drawing) but I thought I would share this idea with you, and you could let us know if you think it has merit.

Hi Jason,

A body diode is built into each FET but more surprising is my Tina analysis after removing the current sources inline shunt to ground. Then the analysis seems proper and gives very clean output signals that closely follow the shunts CMM. That seems to indicate there are issues with the INA240 differential amplifier front end as the simulation takes the shunt bias -200u. I did add a 470uf cap to V3 +24v yet the glitches still show up and were likely analysis error since I couldn't reproduce it again especially without the current sources below each shunt. Yes major ringing is present on the real INA240A2 output but it seems it should not be there at all.

The actual INA240 output signal looks nothing like it should according to Tina transient and steady state analysis simulations. Even testing  INA240A2 output 16 levels of harmonic indicate over 50% THD, so there is noise present in the Tina analysis. Notice how smooth and dv/dt free the signals on output are to what is actually produced.

Tina analysis 12.5Khz PWM injected directly into low side of shunt at INP:

Scope captures are missing the lower half cycle besides far to much noise is plaguing the output signal. Signal seems as if the input is disconnected on one side of the differential amplifier even though all 3 INA240 produce the very same noise.

Nice work Jason!

And after looking at the Tina simulations and real time PWM captures the real time does kind of take on the shape of computer simulation. Notice the -200u pulse period jotting below GND seem to be present in the yellow ovals if you discount all the PWM noise. Those real PWM captures are INA240A2 +3v3 REF/2 and probe AC coupling 10ksps deep scope setting. The Fourier THD was near 75% after again retesting 16 levels of harmonic distortion on the shunt inputs and INA240 output.

One thing I have discovered is the input of the differential amplifier gain seem 2x the actual designated value. The only way to reduce the output gain is the lower the shunt R value and divide the INA output gain resistively by almost 2.5 both 240/282. So 1.65v/2= 825mv roughly. Have considered a 200uohm shunt and INA240A2 captures are with 500uohm shunt.

Hey Jason,

When Tina INA REF1/2 are removed from the 1.65v (500m) reference and say REF1 is GND and REF2 set to VCC the output node voltage is not quite 1/2 the reference. The transient analysis output voltage scale seems off in that measure.

If  we put REF2 between a 10k to GND and 10K to +3v3  (1.58v REF/ 2 = 790mv) the INA output nodal voltage is roughly 808mv. Should the INA240 output voltage go above the REF2 pin level or stay just below it?

It seems 8.2.3.3 figure 29 INA240 output is above 1/2 the reference in Tina depending on the inductive current ramping from PWM modulation.

Is 810mv considered the full swing to rail of A2 output 8.4.3.3? The A2 would be 2.8amps peak at 10mv/amp but the current should go much higher with 500uohm shunt 50 amps load is 500uv CMM or 25mv full shunt scale * 50v/v = +1.25v full swing to rail but is that still true at 1/2 REF?

. 

BP101,

I am trying to understand what you're doing with the reference circuit - I did the two standard configurations we recommend in TINA and it looks ok to me.  See below.

Tied to Vcc and GND:

Tied to a 1V source (could be anything >GND and <VCC, just chose a unique value):

Also I've been asking around about your circuit and one colleague of mine suggested:

"There is very little resistance (damping) in the circuit so it rings before settling when it switches.  In other words, I think that the INA240 is measuring the true, ringing current at its input.  If the customer added series resistance to the circuit inductance, he could eliminate the ringing at the cost of significantly increased rise time of the current and power loss."

Do you concur with that assessment?

Hi Jason,

You didn't answer the question,  external REF in 8.4.3.3 of datasheet does not state if the REF/2 is the output rail top or bottom. Tina seems to believe sometimes the output into R10k is the bottom rail voltage and other times the window top. Again what is the definition of swing to rail in this 8.4.3.3? I'm thinking to reduce the dv/dt the differential amplifier REF must be handled with care.

The inductive ringing in scope captures above are caused by the FET diode dv/dt going to GND through the shunt and B+ which should not be showing up in the output. That real PWM circuit has +3v3 REF/2, perhaps the INA REF1/2 leading into differential amplifier R divider is picking up B+ PWM?  That seems the likely place it is getting into the INA rail if the input is filters dv/dt. Do you concur with such hypothesis and perhaps add some PWM feed back into the supply that feeds Tina VCC/VMID reference and see the results. What kind of REF voltage filtering was used in the lab bench testing? My though is by reducing the current into the INA REF inputs it will also reduce the ringing amplitude in the real world circuit.

Yellow box R5/6 +1.56v REF2 and is the output 808mv considered top of rail per 8.4.3.3 or the widow that moves up to half of REF2/2 nodal 790mv?

Jason Bridgmon said:

"There is very little resistance (damping) in the circuit so it rings before settling when it switches.  In other words, I think that the INA240 is measuring the true, ringing current at its input

This is what FET turn on di/dt and dv/dt look like when the FET avalanches current through an inductor.

Jason Bridgmon said:

found that I had issues with the reference voltage simulation in TINA on the INA240 unless I selected "Use initial conditions" in the transient analysis box.

Does this help?

That answers the question. Tina uses the parallel resistance formula divides product over the sum. Makes sense that a reference IC does not expose the INA240 REF inputs to a parallel resistance, rather an isolated voltage level.

Another oddity is the inductive current measures in series with 80v battery and one near bottom where PWM links to INA240, the shunt current tops at 3 amps. The same amps measure remains even at 162v. Perhaps transient simulation is not switching the inductors same as Matlab does and will try the other startup states. 

What I was trying to discover if REF inputs are prone to noise by reducing the output threshold isolating REF away from GND and VCC. Seems that the differential amp will gain any noise respectively to the REF threshold. So the summation seems to work as expected but gains up the output from the first INA240 feeding the REF's tied together. 

BP101,
A side note here, a new TI Design came out today about 3-phase motor control with low-side sensing using the INA303. It's TIDA-00778, if you're interested. The INA303 is a new part that came out just a month or so ago and has a pretty decent bandwidth and slew rate for this kind of application, plus the reference pin you need for your application.

BTW

Removing all the other INA240-X from the TA and leaving only one fixed the nodal DC current flow. Starting TA from initial state or zero captures an unrealistic startup current spike which pushes the INA output up to rail top. It seems that Tina has issues with multiple INA being in the same project and the shunts resistance where somehow being paralleled so the DC nodal probe indicated as being incorrect drop.

See the results:

   

Starting TA from operating point removes the turn on current spike. Notice the scale of PWM is more real looking and shows inductive ringing being picked up in the lower impedance shunts under 0.5m. It seems higher value shunts above 500u don't exhibit the same ringing pattern as the resistance scale changes on the INA differential inputs.

10 amps PWM inductive current flowing across 0.5m shunt.