EEVblog #102 - DIY dummy load - higher current?

[Eliminateur]

Hello,
this episode has been proven extremely useful to me, can't believe i'd already watched it before and forgot about it.

Anyway, sine i need to set it up for far more current than the example i setup to do some simulations as the parts can get quite expensive and wanted to try some variables...

first thing, Dave, where did you pull out that Vgs to Vs graph from?, i've tried to find it in the datasheet and couldn't find anything, the closest is the transfer characteristic(but doesn't has much detail on the interesting part we care about)
http://www.datasheetcatalog.org/datasheets/90/489299_DS.pdf

So, i set up the circuit on multisim, changing the opamp to a rail-to-rail output one(shouldn't affect it much) and went to test it...
And found some oddities, specially, the output to hte gate of the mosfet reads "7v" at full power when power is 5V , way to go multiFAIL!(unless i'm doing something WRONG)...


other than that, the circuit behaves as expected drawing the same ammount of current as the voltage in the divider.... so far, so good.

So i went about to change components, first a higher power transistor in a TO-247 package(i need to check the PSU parts bin in my office to see if i can scrounge something from there!). i found Vishay IRFP048 device to comply(and far exceed) my power needs http://www.vishay.com/docs/91198/91198.pdf
i reran the simulation and all values remain(gate voltage again returns whatever crazy value it can come up with, 7.95v this time)

Then a r-t-r opamp that can take >12V (since the gate voltage would probably far exceed 5V), i'm having some trouble here as a first model using LM6142 started oscillating wildly -suspect the gain bandwidth is too high-, so i changed to an LMC6462 with 15.5vcc max (i have to see if i can get this component here) but i can't get the simulation to output more than 4.93A at 15Vcc.

Revising the parametrics in natsemi i can't find something with low BW gain, dual/quad package and high VCC with r-t-r output and DIP package(the 4MHz gain ones are soic)..

but oh wait, the LM324D can get to 30V, so if i supply the system with +12V.... and yes.. looks like it could work!


am i missing something, could this work?

[alm]

Revising the parametrics in natsemi i can't find something with low BW gain, dual/quad package and high VCC with r-t-r output and DIP package(the 4MHz gain ones are soic)..

You could limit the bandwidth by bypassing the feedback resistor with a small cap. Limiting bandwidth is much easier than increasing bandwidth .

[tyblu]

what are you trying to do? Do you need fast response time/large bandwidth?

[jahonen]

I think the current controller error amplifier gain should be rolled off suitably so that it stops the oscillation. Try putting a 10k resistor in between of shunt resistor and inverting input, and 1k resistor and 100n capacitor in series between inverting input and opamp output. Then experiment with resistor/capacitor values, if you get acceptable transient response.

Regards,
Janne

[Eliminateur]

Thanks for the tips guys.
Tyblu: i'm trying to build a much higher power dummy load(5A-20V), i don't require anything remotely fast speed no transient responsive, in fact, the slower the better.
Janne: i've tested the simulation with a much more normal "off the shelf" op amp, a LF353 and i haven't noticed any oscillations in simulation.
After narrowing down components, i figured i could do without rail to rail just by using a higher Vcc (what worries me a little is the bottom rail, but i know there are some "half rail" opamps..., or i could offset the zero somehow).
Also, check my last line, with the LM324D i could make it work as well.

Anyway, another thing i've just noticed is Sense resistor troubles, with 5A the sense resistor will sink 25W, even with 10 resistors in parallel, i can't get any supplier that has 10R 3W 1% resistors, it's 5% everywhere.
of course i could go with 15 or 20 resistors but that gets bulky pretty fast, and i'd still need 1.5W to 2W precision 20/15R resistors....

So i was thinking, what about going to .5R?, i halve the power that way to 12.5W with 10 x 5R but i still need 5R-1.5W resistors(crud, i'm pretty sure i can't get that either).
If i go with 15 x 7.5R then it's 333mA per part... still 1W
and again we end up with a 20 resistor bulk of 10R resistors for 625mW each  , argh this is getting ridiculous...

What would be the problems of lowering the sense resistor even more?, do i risk losing resolution at lower currents due to opamp offsets?(meaning i'd need to use precision opamps?), or by multiplying the sense voltage(with another opamp maybe?, or i can do it in the loops itself, like Janne suggested?) i'd be safe?

[johnmx]

I recommend you to use a small shunt resistor. In fact I’m building one current load (10A max) with a 0.005Ohm shunt resistor. I’m planning to use the MAX430 (chopper-stabilized opamp).

As soon I finish my project (maybe before the end of the week) I’ll post everything about it, construction plans, schematic, problems detected, etc.

[Eliminateur]

Awesome John, i'm looking forward to your design.

How do you deal with Rds on active mode in the MOSFET with such a small shunt?.
Also, how do you reliably get such a small shunt value?(i'm guessing you're not using resistors either, am i right?, as even a 1% tolerance would throw the results all over the place)
I didn't knew about chopper-stab opamps, just did a quick research and looks like a very interesting device indeed(specially for very low level signals), bad thing is availability....

I recommend you to use a small shunt resistor. In fact I’m building one current load (10A max) with a 0.005Ohm shunt resistor. I’m planning to use the MAX430 (chopper-stabilized opamp).

As soon I finish my project (maybe before the end of the week) I’ll post everything about it, construction plans, schematic, problems detected, etc.

[Eliminateur]

umfortunately i can't, not here at least(i need to try and use local sources as much as possible to reduce costs, and we're in the stoneage of electronics with a total lack of any precision components) and it's not viable to pay 100$ of shipping from digi-key for a 2~3$ component

[Time]

You use the real time simulation feature for NIs EDA package?  I always prefered transient analysis, personally.

[johnmx]

How do you deal with Rds on active mode in the MOSFET with such a small shunt?.
Also, how do you reliably get such a small shunt value?(i'm guessing you're not using resistors either, am i right?, as even a 1% tolerance would throw the results all over the place)

I will use a BJT Darlington transistor (BU941ZP). Why? Just because I have one that can dissipate 155W and the most powerful MOSFET that I have can only dissipate 80W. At first glance I see a limitation in my circuit, it will only operate at full current above 2.5V. But I don’t plan to use this current load on sources lower than 3V.

About the shunt resistor, I will use the one included in the ammeter.
Let me say that this project is meant to be done at the lowest cost as possible, that’s why I’m using some used and old parts. The total cost of this project is 1.5€, spent on the small wood box.

The box and the power section are done, check the picture above. Now I need to design the control circuit and try if it works

[Eliminateur]

hahah i love the steampunk tone that vintage ammeter provides!

yeah, my circuit will be working all the time above 15V, so i'm not too worried about the low end.
Again, i'm looking forward to the completed "report", even thought you're using a BJT the principles should stay the same.

I'm still bashing my head against the local providers because they don't seem to even heard of a precision resistor smaller than 1R T_T

[johnmx]

If you don’t need high accuracy you can always use the shunt resistor of any cheap multimeter. You can buy multimeters starting at 5€ in local Chinese stores, at least here in Portugal you can.

[Eliminateur]

@Johnmx:
i need 1.953mV accuracy (0.5v/256 A/D-D/A steps), and the cheap DMM shunts aren't rated AFAIK, so i'd have no way of knowing what the value is(and measuring sub 1R values is.... complicated)

[DJPhil]

With a mulitmeter adapter for milliohm resistance measurement (like this) you might be able to make your own shunt out of thick copper wire by trimming it to length.

Do you think that would help?

[Eliminateur]

Very useful Phil, i need to get my paws on a LM317(i don't think i have any lying around...) and a 12R precision resistor....(actually it would be advisable to get a precision Vreg as well...).

i'd need to substract the parasitic resistance of the wires(as i don't have a relative value DMM) and that's it!.

i have tons of shunts from defunct computer PSUs, now i can test them!

[DJPhil]

Glad I could help.

If you want to go completely wild you could use a precision voltage reference and an opamp to make a precision current source. There's lots of ways to get a precise current, but this one always comes to mind as I found it in an old Burr Brown appnote. It'd essentially replace the LM317 and set resistor in the previous project. It's a bit of overkill, and expensive in comparison, but I thought I'd mention it in case.

[Psi]

If you ever need a high wattage dummy load and you just dont have one. A glass or bucket of water will act as a very good heatsink for a large resistor if you have the resistor underwater.
You can overrate a 10W resistor to handle 100W or more quite easy doing this, especially if your only testing and dont need it to be on continuously.

Obviously it isn't a good idea with high voltage though.

[alm]

i need 1.953mV accuracy (0.5v/256 A/D-D/A steps), and the cheap DMM shunts aren't rated AFAIK, so i'd have no way of knowing what the value is(and measuring sub 1R values is.... complicated)

Just shove a lot of current through the shunt (10A or so, whatever your DMM and shunt can handle). Put a DMM in current mode in series with the shunt. Read current from the first DMM, measure voltage over the shunt with the second DMM, use Ohm's law to calculate resistance. Note that resistance may depend on current, so for max. accuracy, you may want to measure at various currents.

[allanw]

If you ever need a high wattage dummy load and you just dont have one. A glass or bucket of water will act as a very good heatsink for a large resistor if you have the resistor underwater.
You can overrate a 10W resistor to handle 100W or more quite easy doing this, especially if your only testing and dont need it to be on continuously.

Obviously it isn't a good idea with high voltage though.

Tap water is pretty conductive though, right? Pure water should be fine though.

[Eliminateur]

@DJphil: excellent, i might just use that, i need the utmost precision in measuring the shunt as it's a very low value(or having a 0.01R precision DMM which i don't have....)
@psi: not very useful for a "production" equipment sealed/repeatable/easy-to-use/automated device
"prior to use, sink resistor in a buckets of distilled water", no, i don't see that happening
@alm: i could try that to characteriuze, but again, using a dmm shunt is not very useful for a production run

All in all, what i'm worried about when using such a low value shunt is that the parasitic effects of traces and SNR suddenly become pronounced for low currents, opamp selection also becomes crucial with -Vcm having to be rail in or <Vss with ultra low offset.
But if i go to higher shunt values, resistor power skyrockets( with 0.1 i'm under 2.5W, but too close to use a 3W resistor for my taste, with 0.2 i already need a >5W resistor)

[johnmx]

Pay attention to the resistivity temperature dependence. If you want precise measures you need to keep the temperature of the shunt resistor as constant as possible. A higher resistor value dissipates more energy and will affect more.

[alm]

All in all, what i'm worried about when using such a low value shunt is that the parasitic effects of traces

This can be reduced by using kelvin connections.

and SNR suddenly become pronounced for low currents, opamp selection also becomes crucial with -Vcm having to be rail in or <Vss with ultra low offset.
But if i go to higher shunt values, resistor power skyrockets( with 0.1 i'm under 2.5W, but too close to use a 3W resistor for my taste, with 0.2 i already need a >5W resistor)

This is why most DMM's use several shunts, depending on the range. You could also look at Dave's µcurrent, it also handles pretty low voltage drops over the low-resistance shunts. But it's a trade-off between resolution/accuracy and voltage drop/dissipation.

[Eliminateur]

Dave ucurrent is nice, i took a look and found interesting things for my project.
Yeah.. i'm still pondering about resolution/accuracy vs power, i don't really care much about dissipation, as the power will go either to the MOSFET or the Shunt so it's all the same to me having it in one or the other part(albeit having it in the mosfet makes it easier to dissipate)

The other issue is that -at least for the very few stores that i've found that don't charge 100$ shipping to me, well, 40.... the best and has pretty limited supply- a 0.2R 10W 1% resistor costs eight times as much and comes in Axial config(so that adding wires for the board would add even more parasitics) vs through-hole OAR-style ones that cost less than a buck for 0.1R 3W.
Even Vishay LVR 5W is 10c more than the OAR.... ( i think i'll go with that one if i buy one).

Hadn't thought of multiple shunts, the idea is nice but adds problems with switching the banks as i'd need at least one extra power resistor, 2 more mosfets to switch the shunt banks(i could use a TO-220 logic level cheap mosfet for this as it shouldn't dissipate much), some kind of driver for those mosfets, another precision opamp for the second shunt and that would end up rising the complexity, board layout and costs through the roof

ofc i'm thinking about kelvin connections(even though the kelvin leads could add to parasitics also)

[Eliminateur]

hmmm i have a couple design "problems" with changing to 0.1R which i hope you can give me a hand:
Ifi choose a 0.1R resistor and want to keep the adjustment from 0-5V on NI then i need to add a gain of 10 from the sense to the NI input of main OP, i though about adding a second opamp with gain of 10 to feed the 1st one AND i can also use it as my ADC input.
BUt i'm having a stupid problem, how do i get a gain of 10 with normal commercial resistors?
Also, is my choice a sound one?, because i'd essentially be killing two birds with one stone as i'd still need to map the Vsense to 0-5V...

Also 2, to get rid of the offset in Vout from that amplifying amp, what can i do?, virtual ground?, rail to rail out?,m but there don't seem to be many R-t-r out low offset ones...
here's the circuit...


i'm at a complete loss, i don't know how to put a gain of 10 in the same loop or how to deal with the output offset of ~1V, another option is to completely use split power supply, then the 0 would be possible....(also, i'm trying to stay away from SMD, it would be a problem later to mount)
ideas?

[jahonen]

You can get gain of 10 for example with E12 resistors of 2.7k from output to inverting input and 330 ohms in parallel with 3.3k from inverting input to ground.

Be careful if you use an opamp in feedback loop, you can create all kinds of oscillators if the feedback amplifier is not fast enough. That can be solved by limiting the error amplifier gain at higher frequencies like I described above.

If you want to be absolutely sure that you can go to 0 volts, you must use a split supply. If your VDD is +5, then you can create the negative voltage using ICL7660 (or MAX660 for more current), or even with ubiquitous 555 with bit more components and less efficiency.

Regards,
Janne