Looking for ideas on electronically switching resistors. 2

[itsmoked]

I've got a regulator that controls to a specific operating point via current thru a resistor.

I want to be able to select between a couple of resistors to set, say, three operating points. My problem is that the resistors will have between 300mA and 1500mA running thru them and are somewhere around 0.1 ohms to 0.5ohms.

Digital pots are out cuz they don't do amps.

So I thought of using MOSFETs switching in resistors or perhaps shunting them. There are lots out there with sub 0.0x ohms ON resistance. But I'm thinking they'll be a problem since they all have fairly large thermal drifts. Just the thing to have around a thermally active regulator.. All their charts also seem to peter-out down around 1A so I'm not sure what happens when I'm only pushing 300mA thru them. The controller point looks to keep the top of the resistor at about 300mV which again isn't much of a source/drain voltage. Again that's off the charts.

Got any suggestions?

Keith Cress
kcress -

http://www.flaminsystems.com

 

[MacGyverS2000]

I always hate to suggest relays, but in this case...

Dan - Owner

http://www.Hi-TecDesigns.com

 

[Skogsgurra]

I assume that it is DC?
Yes, of course it is, how silly of me.

I don't have any low voltage MOSFETs within reach. Otherwise, I would have measured their performance on my Tektronix 575 transistor curve tracer (love it!). I am convinced that F. S. can do a quick chek on a few transistors to see how they behave below 1 A. Go ask them - they are in your neigborhood...

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[MagicSmoker]

I've got a regulator that controls to a specific operating point via current thru a resistor.

I want to be able to select between a couple of resistors to set, say, three operating points. My problem is that the resistors will have between 300mA and 1500mA running thru them and are somewhere around 0.1 ohms to 0.5ohms.

Are the set point resistors floating? If so, then the only practical option is a relay. If one end of the resistors is returned to either the positive or negative supply rail, then you can use a P or N type switch (MOSFET *or* bipolar) to select the resistor and simply read the voltage drop across the resistor alone, rather than the series combination of resistor and switch, to sidestep the issue with drift in channel resistance or saturation voltage.

 

[IRstuff]

I think the question was to use the MOSFET as the resistor.

It's possible to keep the device in the linear region, so long as Vgs > Vds. But, without running the transistor on a curve tracer, there's no definitive way to a priori determine or set the channel resistance. Once the saturation current characteristics are measured, it's theoretically possible to use the various equations to "set" a channel resistance.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[geekEE]

When you say a "specific operating point," do you mean a particular current? i.e. is the voltage fixed? If so, then how about a current source (or sink) circuit with feedback to set a specific current?

Glenn

 

[OperaHouse]

I would consider a small switch mode power supply either to add or subtract current to existing resistors or even placed in series as a voltage offset. Wouldn't take much power.

 

[ScottyUK]

How many of these things are you building, or how long do you need to do the tests? Programmable electronic loads occasionally turn up on ebay, and you could probably move it on that way too once you're done with it. Of course you might just come to love the versatility and hang on to it. :-D

 

[Skogsgurra]

I would. Have never been able to part with any equipment.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[MagicSmoker]

I think the question was to use the MOSFET as the resistor.

Nope... itsmoked said:

So I thought of using MOSFETs switching in resistors or perhaps shunting them.

So it's pretty clear he meant that he wants to use a power transistor to select a resistor

Thus my original suggestion applies.

 

[itsmoked]

I always hate to suggest relays, but in this case...

Yeah, I was horrified by this thought too.

575 transistor curve tracer

Wish I had one now!

Are the set point resistors floating? If so, then the only practical option is a relay. If one end of the resistors is returned to either the positive or negative supply rail, then you can use a P or N type switch (MOSFET *or* bipolar) to select the resistor and simply read the voltage drop across the resistor alone, rather than the series combination of resistor and switch, to sidestep the issue with drift in channel resistance or saturation voltage.

Hi Magic. (Had a Boa Constrictor named that.) The resistors are indeed ground referenced. Small mercies. So this all happens around 300mV from ground. I'm not following.. OH! Wait I see you're saying put the switch above(electrically) the sensing resistor so it's resistance/voltage drop is unmeasured. I love it! Thanks. This is exactly why I came to you guys. Get stuck in one mindset and miss the best solution.

I think the question was to use the MOSFET as the resistor.

It's possible to keep the device in the linear region, so long as Vgs > Vds. But, without running the transistor on a curve tracer, there's no definitive way to a priori determine or set the channel resistance. Once the saturation current characteristics are measured, it's theoretically possible to use the various equations to "set" a channel resistance.

Naw. Though the thought crossed my mind. I'm looking at maybe ten of these per unit. So curve tracing or characterizing individual switches.. no thanks. Not to mention the pwm the resistor will be subject to. Yikes.

When you say a "specific operating point," do you mean a particular current? i.e. is the voltage fixed? If so, then how about a current source (or sink) circuit with feedback to set a specific current?

Hi Glenn. I was thinking of that too, but since these switching controllers use the voltage across these sense resistors for each 300kHz cycle I was hoping to avoid the possible delays and stability issues I might have with an additional amplifier in the loop.

I would consider a small switch mode power supply either to add or subtract current to existing resistors or even placed in series as a voltage offset. Wouldn't take much power.

Ah.. Like what I see in Glenn's suggestion except removing any delays in the sensed signal. I believe this would work. This system would have widely varying supply voltages so I'd have to use current sources to provide added current thru the sense resistors.. hmmmm. I need to think about it some more. Thanks.

How many of these things are you building, or how long do you need to do the tests? Programmable electronic loads occasionally turn up on ebay, and you could probably move it on that way too once you're done with it. Of course you might just come to love the versatility and hang on to it. big smile

Hi Scotty! Yeah as mentioned, just above, I need ten or fifteen per product unit. I'd probably have to include shopping carts with the product. LOLOL

I would. Have never been able to part with any equipment.

Now we know why you live in the country. I bet you have a barn..

Many thanks for the fine participation here.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[IRstuff]

OK, I'm confused then. Are the resistors going to ground or floating? Actually, machts nichts. A sufficiently large Vgs is all that's needed to get the Rds to a sufficiently low value, isn't it?

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[itsmoked]

Are the resistors going to ground or floating?

The resistors are indeed ground referenced. So this all happens around 300mV from ground.

A sufficiently large Vgs is all that's needed to get the Rds to a sufficiently low value, isn't it?

True. And now with the actual resistance and voltage drop of the "switch" not being a part of the sense voltage the transistor needs only support the maximum expected current and be fully ON at logic levels. These will be set-an-forget so the switching speed, gate capacitance, etc., don't matter. Since the voltage above the resistor never exceeds 300mV the transistor doesn't even need any Vds ability. Schweet.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[itsmoked]

Dang!!

I don't see how MagicSmoker's idea is going to pan out now. Notice in B I can't get the paralleled voltage drop out to the regulator without getting the filthy transistors into the picture? Like picking off the return in A above the transistors.

Sounds like I'm back to Opera's idea.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[IRstuff]

I think you need to have the transistors and resistors swapped. I was imagining a single tie point with N resistors with their own pulldowns. So long as only one transistor is turned on, and turned on hard, only one resistor is sinking the current, and the control system is simply sensing between the tie point and ground. The only thing that you would need to do is to make sure that the control system has the correct gain term. Am I missing something here?

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[Skogsgurra]

Yes Smoked, I live in a barn. Some would say a large country - almost manor - house, but "a barn" describes it best.

Re the sense circuit. Use one resistor from top of R23, R21, R24 and let them land in a summing point (-ve of an opamp). Let R21 be your highest value resistor 500 mohms) and make the other resistors have values so that you get 0.3 ohms (my guess) and 0.1 ohms total resistance when switching Q4 and Q5.

The active path's voltage drops will add and those that do not carry current add zilch (resistor grounded).

You could, of course, also use a non-inverting opamp, but the math is somewhat less trasparant (the 1 added to Rb/Ra). And, finally - you can use same value resistors for R21 etcetera and use different values for the sense resistors.

PS I also have a 545A, a KEPCO BOP 72-6M and an adjustable Brandenburg 50 kV HV supply in my "barn". To mention a few of my favourites. And - yess, almost forgotten - an 8002A HP pulse generator with adjustable tr and tf. Those old instruments make life a lot more interesting. But I must admit that the modern LeCroy and Chavin-Arnoux scopes are a lot better than the 30 pound 545A. So the 545A stays at home when I do external troubleshooting. I go to the 545A and turn its knobs when I get nostalgic. They do not do that kind of knobs and switches any more.

See page 19 here:

http://www.google.se/url?sa=t&rct=j&q=&esrc=s&source=web&cd=3&ved=0CDkQFjAC&url=http://

http://www.elinor.se/index.php/Down...LMoNnV6fAp3NJ4g&bvm=bv.49967636,d.bGE&cad=rja

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[Skogsgurra]

Yes, IR. The R[sub]on[/sub] in the transistors. That is where it started - not knowing how R[sub]on[/sub] behaves "outside" the diagram. Putting switches in the feed-back circuit isn't commendable. Not when you are dealing with millivolt signals. You knew that!

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[IRstuff]

That would seem to be a matter of picking the right transistor. Keith's OP stated 0.1 to 0.5 ohms, so something like this:

http://ec.irf.com/v6/en/US/adirect/ir?cmd=catProductDetailFrame&productID=IRF6718L2TR1PBF

would be possibly even a bit of an overkill. The spec Rds is less than 1 milliohm with 61 A Ids.

There are lots of choices if a larger error is tolerable:

http://ec.irf.com/v6/en/US/adirect/...0+4294841672+4294837121+4294837206&numRecs=54

http://www.irf.com/product-info/datasheets/data/irf6712spbf.pdf

appears to be cheapest at $0.513 in 1000 qty and has less than 10 milliohm Rds. It looks like Rds will be limited by the non-channel fixed resistance portions of the circuit, as would be the case with the current sensing resistor connections.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[Skogsgurra]

1 milliohm - that's really low. Still, I prefer keeping switches out of the measurement path. All kind of switches.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[IRstuff]

Ideally, yes, but, there are those exceptions...

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529