high side current monitoring

[NateS]

I've been looking a lot at high side current monitoring for a project. I thought that the AOE fig 7.27 circuit was clever and cheap until I did the math on the precision resistor count to null out 470V of common mode voltage, and the Zetex ZXCT1009 started looking a lot more attractive. I had another slick math solution until I realized that CMV was eating up all but the last bit of precision.

Anyone have experience with either these or the ACS71x series hall effect sensors? I was looking at the ZXCT1009 and extending the range per AN45, but the degree to which my choices of HV PNP transistors is limited suggested I might be getting further off the beaten path than i wanted to. (Incidently, after a couple times choosing a wierd/odd part/method too early in the design process, I'm finding the qty available at DigiKey almost as valuable as any other indicator of 'warning you are headed into left field')

[Enzo]

Hi. Just a thought. You might provide us a link to the fig 7.27 schematic or scan it and post. And links to data sheets for ZXCT thingie. AN45? Application note? On the chip data sheet? Somewhere else?

Otherwise we have to do a bunch of research just to find out what you want to discuss.

[NateS]

Perhaps becase the bigger question basically amounts to 'practical methods of high side current measurement'.

"art of electronic" "figure 7.27" "common mode"
zxct1009
acs712
an45

ZXCT1009 appears to be a clever little device. See figure 5 in the ap note above:

[Enzo]

I'll try to articulate this idea without having a specific circuit in mind. If 470v is hanging you up, how about making an artificial "ground" for the circuit at something like 450v. That way 470v only looks like 20v to the monitor circuit. Since all you care about is the drop across the sense resistor, the circuit can still work perched above ground like that. Then implement some sort of level shifter to translate the output signal to a more useful level.

Analogous to this idea would be the control circuits of the primary side of typical SMPS. The control circuit "ground" is really the -170vDC rail. Meaning that a +12v supply for the IC is really just -158vDC with respect to earth.

An optocoupler could serve similarly to act as a level shifter.

[jazbo8]

Similar idea with LT6101:



Jaz

[NateS]

I looked again at the IL300/PCNR200/LOC110 PVI optos with twinned outputs for feedback. Supplying the error amp on the high side adds fuss. Assuming the PVI type isolators are linear enough I could just take all of screen current through the driver side of the opto - several of these will tolerate 20-100mA forward current. That's the simplest, but it likely wouldn't be any cheaper. PVI's are pricey - but so far I'm only using them as switches or inside feedback loops - and Panasonic's APV1121S is about $2/ea qty 1.

Here's my sketch of the circuit above:
DigiKey BOM Qty 1 = $2.39, including the circuit protection

[J M Fahey]

Just curious, what are you trying to do?
What would be its practical purpose?

[NateS]

Its an amp project that got consumed by a subproject.

Arduino that controls soft start, current limits, biasing, and has programmable limits for overvoltage, overcurrent, dissipatiion. I have two high side current monitors, this one for the screen tap, and another for the heaters which are elevated. So far i have simple low side current sense on B+ and the cathodes, and voltage on B+ at the plate tap. Splitting out screen current separate isnt critical, but I wanted to be able to throw a fault for screen overcurrent/dissipation.

While the amp itself isn't like to make radical tube changes, i was trying to design the controller so it will work for most any tube choice. Especially if you're chosing to push the limits, (ie, experimenting with AB2) having and idea that she canna take it much longer Cap'n seems like a noble goal.

[R.G.]

The ZXCT1009 looks like fun, as does the LT device mentioned, or any similar. Once you have a linear current output, you can run it through a PNP or P-channel cascode to ground.

It's possible, especially with P-channel MOSFETs to do multiple cascode stages, since the lack of control node conduction prevents changes in the current. A couple of ZVP0545s should give you the ability to get up to 800V or so with impunity. You'd stack the power supply zener with three resistors in series, dividing the remaining voltage to ground from the supply to get equal parts across each device, tying the gates to the taps, and worrying about oscillation and gate protection as appropriate. This ought to transfer the current to a ground reference with much of the accuracy preserved.

Opto's more direct, obviously. So is floating a small uC like one of the 10F or 12F series on the line and using it to opto a digital signal back to the main controller.

ZVP0545 at Diodes, Inc.

[NateS]

Something that I'm still puzzled by though is where does the measurement current output come from? I'm limiting the base current to 450uA. Obviously it has to steal it from the sense resistor, and the designed current reading output for 100mA measured would be 1mA, which is in line with the app note designs. 1% doesn't really matter much to me - I'm using 1% in the sense and gain resistors, and in this case it's literally "close enough for rock n roll", but intuitively that seems like an not insignificant poll tax for a circuit that's supposed to be measuring current.

(*100mA is just where I picked for full scale reading - just to avoid having any and every overcurrent peg the meter. The software limit is more like 40mA / 16W for an EL34 duet). And oh yeah I just noticed the 1009 is backwards.

[J M Fahey]

A humble 58 cent ZTX560 (500V PNP) will do all the sensing you need, if you split that screen resistor in two, with the sensing one wired BE and with such a value as to drop 650 mV at the required trigger current.
The ZTX collector can go to ground through a 1M+10K divider.
When screen current surpasses your trigger value, you'll get around +5V across the 10K resistor, a nice "logic 1" for your Arduino (or whatever you fancy) processor.
http://www.mouser.com/ProductDetail/...smuJmF5H4UE%3d

[NateS]

On some good meds right now... I was more thinking about the current measurement impacting the measurement, rather than being wasteful or stealing from the screens, but that's not the case. I was also thinking 'devices designed to measure current are typically very careful about how much they waste. If you're measuring, you probably care, and if you care about current, you likely care squared. Doh. I promise it'll be looked over plenty of times after I'm off the meds.

As far as having a logic input - part of the project is having configurable limits.

[R.G.]

A humble 58 cent ZTX560 (500V PNP) will do all the sensing you need

Probably. That's where I started, but then I thought:

I worry a lot about measuring 470V with something that dies at some unspecified point over 500V. Users sometimes do pull all the tubes and turn on the power. That's really why I was calling for something with 800+V (from two 450V+ devices) instead of using one.

And a good high gain PNP will use less than 1% of its emitter current to turn the device on in most cases, but HFE falls off a lot at low currents for bipolars, and even with a few percent used up in resistor tolerances and such, you can have the error budget get blown quickly if HFE falls enough.

That's what pushed me to recommend cascoded and stacked MOSFETs instead of bipolars. The MOSFETs have other errors, but the source current is exactly the same as the drain current, and you can stack cascodes of MOSFETs without causing more errors on the sensed current.

In a fully developed setup I'd probably put a limit zener on the bottom sense resistor to keep the input to the control electronics actually sensing what happens within bounds so that glitches don't kill the control electronics.

Nate's working on a fairly fully developed sensing and monitoring system. As he notes, it helps to worry about end results, error budgets, voltage limits and so forth in the initial design.

[Gingertube]

I do high side current monitoring in a mosfet laser diode array driver in the day job.
Here is the part I use to do the shift to 0V referenced signal.
INA117P
http://www.ti.com/lit/ds/symlink/ina117.pdf
+/- 200V common mode range with protection of inputs to +/- 500V
Cheers,
Ian

[NateS]

Re: unloaded over voltage You know that's a good point.

I eyeballed the setup to see what I could do about it, this is kinda cool:
I already have a divider feeding an ADC pin to monitor B+ voltage, and everything going into the micro is zener clamped to 5.1v unless the line just can't ever go out of range. However.. shave a diode drop of the B+ voltage monitor clamp, feed it to a garden variety bjt. And now - not only is the micro protected but it creates an over voltage line that can be used to 1) scale back the current limit, and 2) sink some of that unused current.

Once the zener begins to conduct, it will drive Q3 into conduction. Invert that signal, and drive the gate of the MOSFET shunting B+. Not shown is using another BJT to pull the gate down on the current limiter while this happens. Aside from the OCD aspect of it all, it means you can relax some of the paranoia elsewhere if you can guarantee B+ is never going into Tesla coil territory. (Apologies for the whacky numbing - I didn't feel like creating an excerpt this time.)