PXIe-4150 vs. PXIe-4139 Pre-Sales Question, Testing Inrush Current of High Capacitance (10mF...30mF) Loads.

[andrew_psemc]

Hello, apologize in advance if this has been asked, I didn't see it.

This is a pre-sales question for an automated system intended to replace bench tests. The products being tested have a relatively large input capacitance (10mF...30mF) on the power input and relatively low steady-state current draw. There is little no current limiting series resistance, mainly limited by power supply foldback. In some cases power can be hard-switched on (e.g. via electromechanical relays / EMR or field-effect transistor / FET). In other words a delta T of perhaps 200us (200 microseconds) to 1ms (1,000 microseconds). The power being switched in may be 5V, 12V, or even as high as 30V. Using a basic I = C*dV/dt equation it's clear the theoretical inrush current blows up to an unrealistic number (because of course the source and path will have some resistance, the capacitors will have some ESR, etc.)

I = C * dV / dt

 = 0.03 F * 30 V / 200E-6 s

 = 4,500 Apk

Realistically, in some cases we have measured 20Apk to even 70Apk, which gives one an idea of how much foldback / equivalent series resistance is present in the system (I'll spare you that math 🙂 ).

The existing equipment we're attempting to replace for this (and other) measurements is a Tek scope with a hall effect current probe (model numbers aren't really important). We've found some clever ways to get around overloading the equipment, but essentially we need two different measurements:

• one inrush measurement at fast sampling rates and low Amps/div setting on the scope, and 
• one long-term measurement at slow sampling rates and high Amps/div setting on the scope.

Going forward with automated equipment, I think it's reasonable we don't need to switch as hard (test equipment can reasonably limit inrush), but I need to have a comparative uncertainty analysis to be sure we can still test and reasonably detect an off-nominal input capacitance condition. 

For example, knowing an inrush current is 10Apk for the system, should be fine. Let's say capacitor limits are +/-1%, we have 100x better basic accuracy within those limits using a 16-bit DAQ sampling at 1MS/s across a sense resistor. (No, we are not going to use an LCR meter for this application, I'll spare you the details.)

I am planning to have NI switches in the circuit to switch among multiple units under test, probably an EMR for this high inrush test, but I'd rather put the work on the supply or our own PMOS FET rather than wear out the EMR switch by hot-switching. As such, I'm not satisfied with the detail in the 4150 and 4139 specifications and I'm asking for additional info. For more background, 

• In the 4150 configuration an example of the system is something like PX6771950, and
• In the 4139 configuration an example of the system is something like PX6772373 (see PXI Advisor.)

PXIe-4150 questions:

1. Is an external supply (AUX POWER IN) really required to achieve performance in Figure 1 of the specs?
   • My initial reading made me think "no" because I saw no such note in the specs or datasheet, however...
   • In the PXI Advisor it specifically says so (see image below)
   • If required, the external supply is crazy expensive (almost as much, or as much) as the 4150 itself... just an observation. (I guess that was passive-aggressive for any Sales people reading.) I'm tempted to just plug in a $200 benchtop supply instead.
2. More driver question than anything - see table 11. I would like to just get a 30V step with maximum output current capability for >=5ms. What is the current range (is it the max 10A) if I configure a 30V voltage-mode step?
   • If so, is there any idea how much foldback (excess current for a brief <=5ms period) there would be? (10Apk for 10A range, 20Apk for 10A range, etc.)
   • My planned method is to use a sense resistor and Multifunction DAQ to determine current at turn-on.
   • Option 1 Voltage mode: As long as I know that output current capability max of the 4150 is 10A or so, I may be able to detect deviations in peak current and reliably assume the deviations is due to the product, not the test equipment.
   • Option 2 Current mode: Alternatively, if I can't get enough current (10A or so) in voltage mode, perhaps I need to configure current mode and do some math to estimate how it will affect voltage on each individual product. Seems riskier... 
3. Are there any max load capacitance needs of the 4150? I only see a note about 1uF on a noise specification.
   
   PXIe-4139 questions
4. Figure 1 of the specifications gives a fairly clear picture of what this instrument can do at 10A. It shows 10Apk for 1ms...10ms at 30V. Is there any way to exceed (in the driver) 10Apk for shorter (lets say 1ms) pulses?
5. More a driver question than anything - see Figure 1 - is the 1ms...10ms driver limited somehow (e.g. will throw an error if exceeded) or is it just a hardware limitation that shows up without driver error? I'm pretty sure it's the later, because if it's the former, I don't know a method to avoid errors at 10Apk, various supply voltages, other than "guess and check" each pulse width. 
   • If hardware as I suspect, could I get any characterization / finer granularity on pulse widths (e.g. short to ground) for these cases? See also image below, mark-up of Figure 1
     • 10Apk 3.3V output
     • 10Apk 5.0V output
     • 10Apk 12V output
     • 10Apk 20V output
     • 10Apk 30V output 
6. Same as 4150 question - are there any max load capacitance spec of the 4139? I see no notes.
7. Per note 11, I don't see the article "Examples of Determining Extended Range Pulse Parameters and Optimizing Slew Rate using NI SourceAdapt" when I search for it on ni.com or on another search engine.
   • Maybe this is now SourceAdapt Custom Transient Response Ranges ?
   • In any case, this post PXIe-4139 Output Stage Shunt Resistance from Jared R. in 2016 does a great job of identifying the real resistance involved in these kinds of power questions. I simply want to raise awareness I can't find the article shown in the specs.

Thanks, 

Andrew

[ChrisR.]

Hi, Andrew.  You have a lot of helpful detail in your question, but I’m confused about the big picture.

What exactly do you need your automated test system to do?  On the one hand it sounds like you want it to power your DUTs and measure their current draw, but on the other you say you’re replacing a scope plus current probe, which obviously can’t supply any power to the DUT.  Do you want the system to supply power, or just measure current from a separate power supply?

Assuming you need the system to supply power, how much power do you need to supply?  You mention measuring inrush with “low Amps/div” and long-term current with “high Amps/div” which sounds backwards to me.  What is the continuous current draw of your DUT?  If the power requirement is low enough, you should be able to use a PXIe-4139.

I’m also unclear on if you actually want to measure a high inrush or if you just want to be able to infer the bulk capacitance from the inrush you see, in which case having the SMU limit the inrush shouldn’t be a problem.

Then you throw in measuring the current with a DAQ board and a shunt when you’re already sourcing with an SMU, which can directly measure the current it’s supplying.

So you can see I’m a bit confused.  I will say that extended range pulsing on the 4139 probably doesn’t work like you think it does.  Exceeding the standard power limit is done by generating a pulse in the driver, which checks the requested V and I setpoint and limits vs. the duration of the requested pulse.  So it is driver-limited.

I don’t know that much about the PXIe-4150, but I do know it requires an external power supply of some sort.

Anyway, if you can clear up my confusion a bit, I may be able to be a bit more helpful.  

[andrew_psemc]

"What exactly do you need your automated test system to do?"

Both supply power and provide measurements.

The current system and procedure correlates 3-4 independent benchtop instruments (voltage, current probes, scope, and power supply) that I believe can all be done with one SMU.

"Then you throw in measuring the current with a DAQ board and a shunt when you’re already sourcing with an SMU, which can directly measure the current it’s supplying."

Assuming the SMU can't measure the inrush current in the way I expect, maybe I'll need to add a DAQ and a sense resistor, but I doubt that will be the case (more details below.)

"Assuming you need the system to supply power, how much power do you need to supply?"

Depends on the time frame, obviously, and I'm suggesting we need to soft-switch for production test. To do so, I'm suggesting we break apart the tests into two tests - 1 for inrush and 2 for during steady-state - using an SMU such as the PXIe-4139.

Test 1 - set current pulse value and duration, measure dV/dt, calculate capacitance. Because of the 10Apk limitation of the PXIe-4139, this doesn't turn on the DUT, it only characterizes the input capacitance. I argue this is OK and meets the intention of what we were already doing.

Test 2 - set output voltage and wait for DUT to fully turn on, then measure steady-state current.

Assuming I lose this argument (break apart the existing test into two separate tests) I think I'd need to use something like the PXIe-4150 supply instead of the PXie-4139 SMU.

"You mention measuring inrush with “low Amps/div” and long-term current with “high Amps/div” which sounds backwards to me."

Yikes, I said that?! Sorry about that! You're correct.

"What is the continuous current draw of your DUT?"

<100mA

"Exceeding the standard power limit is done by generating a pulse in the driver, which checks the requested V and I setpoint and limits vs. the duration of the requested pulse.  So it is driver-limited"

I think this answers my question #4, which I pretty much expected, so I'm more preferring the additional information I requested in queestion #5 (specifically, I'm looking for better resolution on the number of ms allowable for each of the 10Apk pulses I mentioned.)

"I don’t know that much about the PXIe-4150, but I do know it requires an external power supply of some sort."

Hmm... ok, that's disappointing. Both because the external supply pretty much doubles the cost, and also because it's not stated in the specifications very clearly (e.g. the 4150 graphs.)

So I'm trying to keep focused on the high-level - does this make more sense to align with my previous questions?

[ChrisR.]

First, I'd like to make sure you realize there are two versions of the 4139: the original 20W version and a newer 40W version.  Their pulsing specs are different, and the pulsing graphs you pulled are for the 40W version.  The 20W version didn't merit a graph, I guess, since its specs are pretty straightforward.  Let's start with that one.

That 0.2 J energy spec is the one you need.  It tells you how much energy you can get from a single pulse.  So taking your 20V example, 20V x 10A = 200W.  0.2J / 200W = 1 msec.  Similar math for all the other voltages.

The "maximum cycle average power" spec is stating that if you're pulsing 200W, the downtime after your pulse has to be long enough to keep the average output power to 10 W.  If the output power during the off time is 0, that means you'll have to wait 19 times as long after the pulse as the pulse was on.  e.g. 1 msec on, 19 msec off.  The "maximum duty cycle" spec means that 5% is the maximum allowed duty cycle for an extended-range pulse even when the power levels are lower.

If you're not familiar with how pulsing is specified in NI-DCPower, I suggest you take a look at https://www.ni.com/docs/en-US/bundle/ni-dcpower/page/pulsing.html, which might clarify some terminology.  One thing that might not be obvious is that an SMU can do pulsing without using the "pulsing" API as long as the requested levels and limits fit within the DC power capabilities of the device.  The pulsing API is primarily to get access to pulsing beyond the DC limits.

The specs also contain the information you requested for the 40W 4139.  Look for the piece of the specs that starts like this:

I won't try to work through the equations here, but I think you should be able to figure them out.

One thing you might consider trying is to create a simulated 4139 and see what the software allows you to do.  I'm no expert on how to set that up, but it should be pretty straightforward.  It used to be done in MAX, but now I think there's a newer tool to do it.

I now understand better what you're trying to do, but I'm still a bit confused, partly because of comments like this: "Because of the 10Apk limitation of the PXIe-4139, this doesn't turn on the DUT, it only characterizes the input capacitance." coupled with the statement that the steady-state current draw is under 100 mA.  Why not just power the DUT with the 4139 with some moderate current limit that lets you sidestep the whole pulsing issue?  Is there some reason you have to characterize your DUT input capacitance at 10 A rather than at some lower current?

As this thread grows, so does the probability that I've missed some key question of yours.  Please repeat it if that's the case.

[andrew_psemc]

"First, I'd like to make sure you realize there are two versions of the 4139....and the pulsing graphs you pulled are for the 40W version."

I'm exclusively referring to the 40W version of the PXIe-4139, sorry I didn't explicitly state that.

"That 0.2 J energy spec is the one you need." (or 0.4 J for 40W)

I did see that math and ran the numbers, but nevertheless I was hoping for a graph with more detail for the 10Apk case at the voltages I provided simply to say "hey, here's an official NI graph that shows us what we'll get." 

Also, I asked about the more precise pulse times because I saw this note in the spec and I wasn't able to find the article it refers to. Maybe it's deprecated? Maybe it was renamed to the "pulsing" article you pointed me to ?

"One thing you might consider trying is to create a simulated 4139 and see what the software allows you to do."

If I need to generate it on my own I suppose I can. As far as creating a MAX simulated device, I'm not opposed to it, but it's a pre-sales question so I was hoping to get by without going through the exercise of downloading / installing all the software. IT tightly controls installed software, NI or otherwise, and I'm evaluating the case for whether we want to go that far.

"I'm still a bit confused, partly because of comments like this: "Because of the 10Apk limitation of the PXIe-4139, this doesn't turn on the DUT, it only characterizes the input capacitance." coupled with the statement that the steady-state current draw is under 100 mA"

Sorry if I'm somehow making this sound more complicated than it is. Here's an LTSPICE drawing of a basic system that operates this way. There is a large input capacitance, however the device doesn't turn on until over 20V is applied.

Existing tests supply >20V with a huge inrush current (>20Apk...70Apk) but with a very low steady-state current (100mA)

My proposal with the SMU is to break apart the tests. First, characterize the capacitance with a known current pulse, which won't exceed 20V and therefore won't turn on the device. Second, use a 3A DC (smaller) current source to turn on the device (>20V) and continue functional testing.

"As this thread grows, so does the probability that I've missed some key question of yours.  Please repeat it if that's the case."

OK, see below. I think we are 3 for 7, with 2 of 7 contingent upon my getting approval for / downloading NI software for / creating a simulated device for the answer on my own.

PXIe-4150 Questions

1. Answered Is an external supply (AUX POWER IN) really required to achieve performance in Figure 1 of the specs?

I think your answer is "yes, it's required". I think I already reacted that it's surprising to me because it's not stated in the specifications, and adding this external supply can double the original cost of the PXIe-4150.

2. Pending More driver question than anything - see table 11. I would like to just get a 30V step with maximum output current capability for >=5ms. What is the current range (is it the max 10A) if I configure a 30V voltage-mode step? If so, is there any idea how much foldback (excess current for a brief <=5ms period) there would be?(10Apk for 10A range, 20Apk for 10A range, etc.)

If we need to just plan to download and install and try with a simulated device, then I'll let the parties signing off on the sale that's it's a technical question / risk until we do so.

3. Not Answered Are there any max load capacitance needs of the 4150? I only see a note about 1uF on a noise specification.

I don't think there's been any attempt to answer this, correct me if I'm wrong.

PXIe-4139 questions

4. Not Answered Figure 1 of the specifications gives a fairly clear picture of what this instrument can do at 10A. It shows 10Apk for 1ms...10ms at 30V. Is there any way to exceed (in the driver) 10Apk for shorter (lets say 1ms) pulses?

I think this answer here is "no" just looking for confirmation, maybe you have already.

5. Pending More a driver question than anything - see Figure 1 - is the 1ms...10ms driver limited somehow (e.g. will throw an error if exceeded) or is it just a hardware limitation that shows up without driver error?

This is the most recent question we've discussed, and I think there's a missing reference in the specification. I have an answer for whether I get an error if I exceed the equations or if it's just a hardware limitation. 

Seems like right now this still stands where #2 stands, it's a risk until we get IT approval, install software, and see what the driver does to a simulated device.

6. Not Answered Same as 4150 question - are there any max load capacitance spec of the 4139? I see no notes.

I don't think there's been any attempt to answer this, correct me if I'm wrong.

7. Not Answered Per Note 11, I don't see the article "Examples of Determining Extended Range Pulse Parameters and Optimizing Slew Rate using NI SourceAdapt" 

Same thing as the other point I made about note 7 above, I can't find the "Pulse On-Time vs Pulse Current and Pulse Voltage" and I'm wondering if these notes were removed / renamed.

-Andrew

[ChrisR. · Accepted solution]

i.  Regarding graphs with more detail, sorry.  The graphs that are there are just to give a rough idea of what's possible, with the intention that a little math is required if you want more detail.

ii.  "Pulse On-Time vs Pulse Current and Pulse Voltage" is in fact the title of the very graph that you wish had more detail.  It's not an article.  It's supposed to give a quick idea of the range of values to expect, and which equation in the spec to focus on first.

iii.  Regarding simulation, I thought maybe you already had NI software installed.  If not, it may not be worth it.  You'd need to install our actual driver to test what its limits were.  My point is that you can do that without having the hardware.

iv.  The drawing helps a lot, as does knowing your DUT doesn't turn on till at least 20V.  What I still don't get is what the advantage is of breaking things into two tests.  What if you just set your 4139 to source 30V with a 1A limit?  It would charge your 10-30 mF in a second or less, and you could determine the actual C from the dV/dt.  You could then run all the other tests once the SMU leaves current limit, meaning the C is charged.  Is the issue that you want the test to go faster than that?

Anyway, on to the 7 questions:

4150:

1. Answered.  99% sure an external supply is required.

2. I suspect the 4150 behaves like our SMUs in that it will current limit at whatever limit you specify, which has to be within the selected range.  It may be possible to go a few % above max range, but not beyond that.  You will definitely not hit the specified <5msec voltage settling time to 30 V with a 20mF load.  ...  And I'm just now recalling that you may in fact be planning to switch among multiple DUTs, presumably with the power source already set at the desired voltage, e.g. 30V.  In that case, you will get some extra current, but any overrange runs the risk of the power supply or SMU self-protecting.  I'll address this use-case after I get through the rest of the questions.

3. There are no capacitive load limits.

4139:

4. 10A is the maximum current the device can source and/or measure.

5. The driver will limit your extended range pulse according to the equations in the specs.

6. There are no capacitive load limits.

7. Clicking on the "Examples of Determining Extended Range Pulse Parameters and Optimizing Slew Rate using NI SourceAdapt" link in the specs just takes me to a section further down in the spec sheet that gives examples of how to use the graphs and equations we've already discussed.

Back to the DUT-switching issue.  An SMU (or power supply) will limit its current as it changes from one voltage to another.  So starting at 0V and moving to 30V, the SMU will never exceed its programmed current limit, e.g. 1A.  That may mean that it takes a while for the output to get to its setpoint, if the load is capacitive.  Or it may mean it never gets there, if the load is sufficiently heavy, e.g. a short circuit.  If you want to switch between DUTs, the safest thing to do would be to return the SMU voltage to zero between DUTs.  If you leave the SMU at 30V and then switch to another DUT that is sitting at 0V, there will be a current spike that will likely exceed the SMU current range.  This should be fine, as the SMU should immediately lower its voltage to limit the current, but that does take time, and so there's a big current spike.  As you stated, this would be hard on any relays involved, and it means there's some period of time where you don't know exactly how much charge/energy you've dumped into your DUT, since you exceeded the SMU measurement range.  There's also a small risk that if for some reason the SMU is not able to bring its current under control quickly enough, it will trip an error to protect itself.  So switching at 0V is just an easy way to reduce the likelihood of hot-switching any relays or tripping any errors.

If you really want to charge up your 20mF DUTs as quickly as possible, you could stick a 100mF cap on the output of the 4139.  That would give your DUTs lots of inrush current, destroying any electromechanical relays and forcing you to your FET solution.  The 4139 would be isolated from that inrush current, although it would have to recharge the 100mF, which would have been discharged by 20% or so.  That recovery would take just as long as charging the 20mF would originally (same amount of charge to deliver), but maybe other tests could be happening during that time.

[andrew_psemc]

"What I still don't get is what the advantage is of breaking things into two tests.  What if you just set your 4139 to source 30V with a 1A limit?"

I can think of two advantages off the top of my head, but there may be other concerns that come up in review.

1. Stress testing components to weed out manufacturer defects that would cause early mortality under hard-switching cases.
   • This is not a requirement, but a sustaining function
2. Testing at a similar level (10Apk) to what is already done manually (20Apk...70Apk) looks better on paper to show the equivalency of automated vs. manual stimulus level and response. For example, consider ESR differences in the input capacitor:
   • 1/2x...1/7x source current difference with 10x better measurement resolution (as in the 10Apk SMU case when compared to manual tests) is easier to justify the test will reveal the same off-nominal behavior of the input capacitor.
   • 1/20x...1/70x source current difference with 10x better measurement resolution (as in the 1A DC SMU case when compared to the manual tests) is more difficult to justify the test will reveal the same off-nominal behavior of the input capacitor.

Hopefully that helps clarify. I may request the other engineers reviewing the automated test plan consider what you say - just source 1A DC instead of 10Apk - and we'll see what they say.

"Is the issue that you want the test to go faster than that?"

No, I don't think that would provide a benefit. Just the two potential benefits above (early mortality of components, ESR measurement at equivalent stimulus/measurement TAR.)

" And I'm just now recalling that you may in fact be planning to switch among multiple DUTs, presumably with the power source already set at the desired voltage, e.g. 30V."

No, I'll be explicitly (that is, hard-coded) avoiding that case. In no case can I think of an advantage to hard-switching while voltage is already present. The algorithm will always be:

1. Switch in untested DUT
2. Turn on Supply/SMU
3. Run Test
4. Turn off Supply/SMU
5. Switch out tested DUT / Switch in untested DUT 

Nevermind, it looks like you already mentioned this later:

"If you want to switch between DUTs, the safest thing to do would be to return the SMU voltage to zero between DUTs."

I don't see any advantage to hard-switching with large load capacitances in automated test - only potential problems - so I'm saving that for bench tests if anyone wants to do it. In other words, I only want to know maximum capacitance for steady-state or transient cases the supply/SMU itself presents (the switches will not create additional transient behavior.) More specifically, is the SMU / Supply stable with up to 30mF as if it were not connected through a switch at all? It looks like you are answering this as follows:

"6. There are no capacitive load limits."

"5. The driver will limit your extended range pulse according to the equations in the specs."

Ah, ok. I thought this may be the case... so those spec equations are hard-coded into the driver somewhere to throw an error if exceeded. It helps to know that because of how I need to plan for catching / handling errors.

I find this interesting in your response:

"...This should be fine, as the SMU should immediately lower its voltage to limit the current"

So the SMU is voltage-controlled all the time, but the current is in the feedback loop for the voltage control?

But other than that (somewhat tangential question), I think you've answered all my questions, thank you for the responses!

[ChrisR.]

So the SMU is voltage-controlled all the time, but the current is in the feedback loop for the voltage control?

In voltage mode, the SMU is a current-limited voltage source.  In current mode, it's a voltage-limited current source.  You can also program an output impedance if you want something other than 0 (voltage source) or infinity (current source).  Of course the speed of the SMU is finite, which means the current or voltage limits don't operate instantaneously.  It also means the programmed output impedance, including zero and infinity, apply only at DC.  More info here: https://www.ni.com/docs/en-US/bundle/pxie-4139/page/output-impedance.html

[andrew_psemc]

Thanks for that Chris. I'll mark your most recent long answer as the "solution" since perhaps it'll be the most useful to someone else with questions in the OP.

In case anyone else wants a graph like I described with more resolution than Figure 1 from the specifications offers, here you go. I tried to cobble together something that looks similar to what NI's Figure 1 by bolding the numbers that define the <=1ms and >=10ms lines on Figure 1 from the specifications. Disclaimer: Obviously, this is not from NI.

If you want the 0.2 J (20W) numbers just change this cell A2 from 0.4 to 0.2

[ChrisR.]

I noticed that your very nice shaded table doesn't highlight that in-range pulsing is of unlimited duration.  I'm terrible at Excel, but I had a crack at fixing that, as well as creating a simple table for the 20W board, which has not only half the available out-of-range pulsing energy, but also a fixed 1msec maximum out-of-range pulse time.  In the following tables, I arbitrarily assigned a value of 999 msec to the in-range region, but of course there's no limit, since those VxI combinations are within the DC limits of the device.

And I rather sheepishly attach my hacked version of your spreadsheet below.