How Can I Design a Digitally Controllable Constant Current Source?

[Lexilighty]

Okay, it looks like opamp is trying to turn OFF the mosfet when we call for full current,
Does he try any harder when we call for zero current?
What is voltage at opamp output with RIO output at zero?
pin1 =1.492V
What are voltages at opamp inputs then?
+in pin 3= 0.048V
-in pin 2=1.7632

volts ACROSS solenoid, and we'll check for AC in case we'e oscillating
volts DC =
volts AC = ,, if there's enough to measure get frequency if your meter is so equipped.

Jumper drain(or opamp pin 1) to AGND
and read again
out pin 1 = 0.010V
+in pin 3 =0.036V
-in pin 2 =0.003V

volts ACROSS solenoid, and we'll check for AC in case we'e oscillatng
volts DC = 23.34V
volts AC = 0.6mV,, if there's enough to measure get frequency if your meter is so equipped.

Is there measurable voltage between physical bottom of R1 and GND ? 0.001V

R1 and pin 2 are joined by a wire ?

Directly beside the pin2 on the bread board.

BTW, pin 1 and GND are now 1.391V, piin 2 and GND now 1.594. Pin 3 is 0V with Vref of RIO at 0V and it's 1.654V with Vref of RIO at 1.654.

 

[Lexilighty]

I noticed something strange just now. The carbon-film resistor I have been using is 1/4W. Currently, at max voltage from RIO, we are passing 0.27W through it. Does this matter?

 

[Lexilighty]

Our Circuit is finally working. I realized I was not meeting the wattage requirement of the shunt resistor. I was passing a max voltage of 1.65V through a 10 ohms carbon-film resistor. That makes 165mA across the resistor at full throttle from Mr RIO. Power dissipated across R1 is 0.27W. Carbon-film resistor I was using was 1/4W.

It dawned on me to calculate this few minutes ago. So I replaced the 1/4W 10 ohms with a 7 ohms 15W and bingo! The circuit now works like a charm. There’s still a lot of oscillations in the circuit though. I figure a few caps would do to keep gentle transistor drain current stable. Cheers to perseverance. Mother nature always has a hard way of making us learn. Beats simulation, hehe!

 

[jim hardy]

What is voltage at opamp output with RIO output at zero?
pin1 =1.492V
What are voltages at opamp inputs then?
+in pin 3= 0.048V
-in pin 2=1.7632

Great Scott !
Pin 3 went down meaning Mr RIO requested Mr Opamp to lower current.
Pin 2 didn't decrease meaning we still had current ?.
Neither did pin 1.
Looks like Opamp didn't try any harder to shut off current.

How about when w gave him some help ?

Jumper drain(or opamp pin 1) to AGND
and read again
out pin 1 = 0.010V
+in pin 3 =0.036V
-in pin 2 =0.003V

Aha, the plot thickens
Pin 2 went down, inferring we successfully turned off current through R1.

So the mosfet will turn off current if directed by opamp?
Are you thinking what I'm thinking - Mr Opamp isn't changing his output at all ?
(by the way - for now only run current zero or max because Mr Mosfet will overheat at mid-range currents)

meBigGuy suggested a long time ago a 10K resistor from pin 1 to ground.
That is called a "pulldown" because it helps pull the output voltage toward zero.
We went to the extreme and used a zero ohm pulldown resistor. It looks like that worked.

But my test wasn't bulletproof - we might have bypassed current around R1 again.

If you're not out of patience
would you lift drain wire from opamp pin 1 and briefly hook it to source, top of R1?
Does that drive R1 to zero volts?
Does it stop air out of solenoid?

Then take the gate wire off source.
Select on your DMM 200 milliamp scale
connect meter between that gate wire and AGND

If you get current out the gate wire, the mosfet is shot again

if you get no current out gate and solenoid air stops
then opamp is not living up to its promise of "output voltage can include the negative rail".

We'll try a 1K pulldown, or replace opamp.
I never used a 3403 and its specsheet is less than crystal clear
says in first page text that Vout can include negative supply
but page 4 note 5 says it needs a 10K pulldown resistor to do it.
Page 3 section "Output Voltage range " suggests Vout might go to only ~1.5V shy of negative supply and that's just what we are observing.

Fix may be as simple as a resistor from opamp pin 1 to gnd.
If it needs 10K i'd use 3K.

Does this make sense to you, Lex ?
What do you think, meBigGuy?
I've been known to criss-cross my logic - part of that Asperger i think.
Eternal vigilance is the price of accuracy !

Pin 3 is 0V with Vref of RIO at 0V and it's 1.654V with Vref of RIO at 1.654.

Good - the expensive RIO and divider are okay

BTW, pin 1 and GND are now 1.391V, piin 2 and GND now 1.594.

Just can't seem to get that pin 1 below 1.4volts , can we?

If you want, just try 1k from pin 1 to gnd. Above tests though should remove any lingering doubts about mosfet.

old jim, trying to be thorough
thanks for your patience

 

[jim hardy]

I noticed something strange just now. The carbon-film resistor I have been using is 1/4W. Currently, at max voltage from RIO, we are passing 0.27W through it. Does this matter?

It will not last long. If it smells bad it's probably failing.

Great news on your success - open resistor all the while?

 

[jim hardy]

i'm having difficulty with system it seems to ignore "post reply"

oh well

congratulations on your success
can you describe the "oscillations" ?

Feel your Mosfet the VN2035 will overheat at currents around 100 ma
volts X amps must stay well below 0,56

 

[jim hardy]

Cheers to perseverance. Mother nature always has a hard way of making us learn. Beats simulation, hehe!

You made my day ---- Thank You !

if you've not discarded that carbon 10 ohm resistor,

how many ohms does it read now?

Shutoff current of your mosfet should be a milliamp or less
so your resistor should be 1500 ohms or more?
a pulldown resistor is probably still a good idea
as would be a stout mosfet on a heatsink

 

[Lexilighty]

how many ohms does it read now?

Shutoff current of your mosfet should be a milliamp or less
so your resistor should be 1500 ohms or more?

Actually, it's discarded :(

Right now, I am using a 7 ohms 15W wirewound resistor. Not entirely a bad choice!

And btw, I am using my last IRF510 after I have burned the remaining two I got yesterday. It's glued like a bad habit to a heat sink.

I will try with a 3k resistor to circuit ground as you have suggested.

 

[jim hardy]

All that for a resistor... I'm a little embarrassed--- seems you mentioned someplace early on that it was 1/4 watt.
but we were moving in right direction. And it was fairly methodical.

One should use resistors at half their rating.

Mother nature always has a hard way of making us learn

I've been learning from my mistakes for almost 70 years now and have built up a great repertoire of them.
How're you coming along ?

old jim

 

[Lexilighty]

Something stranger keeps happening though. I returned from an outing and tried to test once again and it seems the transistor, IRF510 is gone again. So I replaced it and I can vary flow across my solenoid valve from 1.17V to 1.65V using the 7 ohms. Bear in mind I am using the IRF510 and it has a heat sink.
I have a 2.9k resistor between pin 1 and GND, a diode is sitting on the +ve terminal of the valve to prevent inductive kick, and 0.1uF metallic film caps between pin 1 and GND/pin 4 and GND to shut off ripples.
Voltage across pin 1 and GND now 10V, pin 2/3 and GND now 1.298V and RIO supply is 1.295V. I have 1.273V across R1. So I am somewhere good on the opamp side.
But on the question of varying the voltage in the solenoid valve, I am doing more than the manufacturer required specs of 100mA and 165mA at 24V but the solenoid works. Maybe there is something I am missing.
At 7 ohms, my volt from RIO to drive 100mA to 165mA for the valve should be 1.15V to 0.665V but it seems this valve is requiring more volts. Hopefully I have not damaged the valve.
Again, your expertise will be appreciated, big Jim.
Thank you for your good insight all the while.

 

[jim hardy]

metallic film caps between pin 1 and GND/pin 4

that one should come out. Opamps don't like capacitive loading very much
try it between 3 (+in) and gnd instead. That'll take noise out of input signal.

Where does your 24 V supply negative wire connect to gnd ?
Physically run your finger along the wires that valve current must take to get from R1's bottom to -24.
It is important for that current to NOT flow right adjacent opamp's V- pin.
Run one wire from bottom of R1 to AGND
and another wire from bottom of R1 to 24V negative. That one carries load current back to supply , separated from opamp,
so that load current cannot inject noise into opamp.

1.273V across R1

=185 ma ? 185/165 = 12% high..
How close to 7 ohms is that resistor? Can you measure it? 112% of 7 ohms is 7.84 ohms... probably it's closer than that.

Voltage across pin 1 and GND now 10V,

That one i don't understand at all yet. 4 or 5 volts should have him fully on.
Is there any AC in it ?
Ahhhh never mind - i'l bet he IS fully on.
Measure volts across MOSFET S-D when he's got ten volts on his gate: should be less than a volt, probably less than 0.1 volt
In other words, your mosfet puts full voltage across valve and there's no more available.
But Mr Opamp thinks you need just a little more so increases gate drive.

Go to about half current and repeat
Pin 1 (opamp out) =
Pin 3 (RIO requested current) =
Pin 2 (Measured current) =
i bet you find everyone happy
and it's good that you don't have enough supply voltage to hurt valve. Adjust it to just a little more than you need for operation.

Sanity check
Is mosfet hot? Is opamp hot ?
Valve should get warm, 24 volts X.18 amps = 4.3 watts
Observe he is specified for 10% hysteresis. So you'll get 10% variation in current depending on direction of approach.

a diode is sitting on the +ve terminal of the valve to prevent inductive kick,

I trust it's connected between the valve's electrical ends ! _(clarity)

This thing is going to work great. Keep refining it in teeny steps.
Measure valve's resistance and write it on side. If valve gets hurt by overheating its resistance will go down as insulation fails. Five years from now it'll be nice to know what it was new. For extra precision, write temperature too.

 

[jim hardy]

Wow didnt think that one was going to post.
had to log out and back in, still took three tries.
Something is haywire with system. It just doesn't obey the "post reply" button.

I'll be back tomorrow.

 

[Lexilighty]

Where does your 24 V supply negative wire connect to gnd ?

It's a 0V and not -24V. The 0V is grounded to same place the GND of the RIO is.

Can you measure it? 112% of 7 ohms is 7.84 ohms... probably it's closer than that.

It's 7.4 ohms actually.

Valve should get warm, 24 volts X.18 amps = 4.3 watts
Observe he is specified for 10% hysteresis. So you'll get 10% variation in current depending on direction of approach.

It sure does feel really warm after a few minutes of operation.

I trust it's connected between the valve's electrical ends ! (clarity)

The cathode is on the positive end of the inductive load while the anode connects to OUT. Is that wrong?

Measure valve's resistance and write it on side. If valve gets hurt by overheating its resistance will go down as insulation fails. Five years from now it'll be nice to know what it was new. For extra precision, write temperature too.

Will do this tomorrow. I'm home now. Again, thanks for yor help, most excellent Jim. You are God-sent!

 

[jim hardy]

The cathode is on the positive end of the inductive load while the anode connects to OUT. Is that wrong?

That's just right ...

from earlier post

At 7 ohms, my volt from RIO to drive 100mA to 165mA for the valve should be 1.15V to 0.665V but it seems this valve is requiring more volts.

Why do you say the valve is "requiring" more volts ? Do you mean volts across R1 or across valve?
Might it just be that your machine sends it more current than it needs? You said at one point your supply was adjusted up to 30 volts.

What happens when you make your RIO send whatever current is required to make 165 milliamps? You could put your DMM in series with valve on current scale and measure directly.
What is voltage across R1 at measured 165 ma?

That'd be a good check to do anyway because your device depends on R1 having stable resistance at its operating temperature. So it's good that you have a 15 watt resistor now, it shouldn't heat up much at all.

Have fun exploring this thing you've built . I really enjoyed the project. Thanks for letting me play !

 

[Lexilighty]

You said at one point your supply was adjusted up to 30 volts.

It's been at 24.53V since I got it working.to ground

I think the problem is with the opamp. It keeps generating unequal voltages across the inputs when I run the circuit for a sufficiently long time (for example, right not pin 2 to GND is 0.288V when pin 3 to GND is 1.652V (RIO is also 1.652V). I have probably lost the opamp. I also get a very hot valve afterwards. But I do have the 10k resistor between opamp V+ and ground. I had to change this opamp yesterday after I got it working and burnt and it looks like something is about not right with this circuit.

 

[jim hardy]

right not pin 2 to GND is 0.288V when pin 3 to GND is 1.652V (RIO is also 1.652V). I have probably lost the opamp. I also get a very hot valve afterwards.

How hot is the mosfet to touch? To be technical, does it "fry spit" ?

0.288 volts/7.4ohms = only 0.038 amps , so if excessive current is going through valve it isn't making it to R1.
Might you put DMM set for ammeter in series with valve ?
and with R1 of only 7.4 ohms
wouldn't 165 ma give voltage of 1.221 ?

Indulge my curiosity
RIO calling for 1.221
meter negative on your AGND where they all tie together (at RIO?)
read
Volts at OUT/bottom of solenoid valve =
Volts at top of R1 =
Volts at opamp pin2 (-in)=
millivolts at physical bottom of R1 = not at RIOgnd, am looking to measure millivolts along the wire from R1 to AGND
millivolts at 0V side of 24 volt supply = right at the supply not at RIO gnd same reason looking for millivolts along that wire
millivolts at opamp V- terminal pin11 = right at opamp same reason

All readings wrt your circuit common tie point.

 

[jim hardy]

Also have you scanned that current schematic ?

Maybe a photo of your breadboard?

Hang in there - this is typical of prototyping.

 

[Lexilighty]

How hot is the mosfet to touch? To be technical, does it "fry spit" ?

Brutally hot.

 

[Lexilighty]

RIO volts now between 1.189V (=165mA) and 0.724V (=100mA) solenoid requirement as set resistor is now 7 ohms.

meter negative on your AGND where they all tie together (at RIO?)
read
Volts at OUT/bottom of solenoid valve =

15.17V

Volts at top of R1 =

15.16V

Volts at opamp pin2 (-in)=

15.17V

millivolts at physical bottom of R1 = not at RIOgnd, am looking to measure millivolts along the wire from R1 to AGND

15.15V ( I smell the rat)

millivolts at 0V side of 24 volt supply = right at the supply not at RIO gnd same reason looking for millivolts along that wire

15.15V (again, a bad sign)

millivolts at opamp V- terminal pin 11 = right at opamp same reason

15.15V

Something weird is definitely going on. Again, thanks for your probing intellect.
I tried measuring MOSFET gate-source resistance, I have 71.9 ohms, drain - source = 3.86K, gate-drain = 4K.

Opamp resistance with all power off:
Pin1 to 11 =159K
Pin2 to 11 7.4 =ohms
Pin 3 to 11 1=58K
Pin 4 to 11 = 9.82K

 

[jim hardy]

Hmm why do we keep killing mosfets and amps ?

i hope your schematic is wrong..

1. Diode goes in parallel with valve not in series
cathode to +24 side
anode to OUT side
you can solder it to the wires or put it on breadboard your choice
why does diode go that way??
'cause when mosfet tries to stop current, inductor tries his best to keep it flowing.
As shown, he tries to force it through Mr Mosfet and a modest inductor can make kilovolts.

 

[jim hardy]

had to close and reopen window to post above
and system only got half of it
very strange

here it is again
----------------------------------------------------------------------------------------------------

Hmm why do we keep killing mosfets and amps ?

i hope your schematic is wrong..

1. Diode goes in parallel with valve not in series
cathode to +24 side
anode to OUT side
you can solder it to the wires or put it on breadboard your choice
why does diode go that way??
'cause when mosfet tries to stop current, inductor tries his best to keep it flowing. He makes his bottom terminal more positive than his top one (Lenz's Law)
As shown, that'd force current down through Mr Mosfet, and a modest inductor can make kilovolts in order to do that.
Moving the anode to bottom of solenoid let's solenoid instead push current around himself thru diode and it flows in that circle until the inductive energy is dissipated. He only has to make 0.6 volts to do that.

2. I trust the capacitor shown in series with pin 3 is mis-drawn. Else it could never have worked.
Pin 3 goes of course to your RIO's voltage divider. Capacitor from junction of pin3-RIOVref to gnd.

Will post this now, don't want an edit timeout.
looks like i got one any way this is extremelyy frustrating
hope next one goes smoother

 

[jim hardy]

Good job on those voltages ! Clearly R1s bottom isn't reliably connected to gnd and neither is amplifier common. That happens sometimes with those push in terminals... intermittents are hideous to troubleshoot.

looks like current could be returning to your RIO's +15 supply through opamp.
might put1k between top of R1 and pin 2, and 1k between pin1 and mosfet gate... to protect RIO.

mosfet gets really hot? heatsink needs to be about 1 square inch per watt... it'll see a watt at 100 ma...

it worked for a while , we'll spiral in on what changed.

 

[Lexilighty]

cathode to +24 side
anode to OUT side

I clearly played that wrong, didn't I?

Moving the anode to bottom of solenoid let's solenoid instead push current around himself thru diode and it flows in that circle until the inductive energy is dissipated. He only has to make 0.6 volts to do that.

This is so beautifully explained. I read this in high school physics some 17 years ago but it has clearly slipped. You have such a sharp edge!

2. I trust the capacitor shown in series with pin 3 is mis-drawn. Else it could never have worked.
Pin 3 goes of course to your RIO's voltage divider. Yes! I got a 10K resistor in there. Sorry I somehow omitted this. Will redraw the circuit morrow morning and have it all together. Capacitor from junction of pin3-RIOVref to gnd. Yes, you are right. The circuit was hastily put together. The other leg of the cap is grounded.

might put 1k between top of R1 and pin 2, and 1k between pin1 and mosfet gate... to protect RIO.

Will implement this tomorrow. Again, I am so grateful for your valued expertise. I am learning great from you!

 

[jim hardy]

I am learning great from you!

no, you're learning by doing.

i feel bad about all the trouble. But it demonstrates the difficulty of conveying ideas accurately by just words..

When we're done I'm going to point you toward Lavoisier's "Introduction to Treatise on Chemistry"
which speaks at some length to that. (i use him a lot) "Science is but language well arranged"..

old jim

 

[Lexilighty]

I have everything together on a prototyping board now and I have equal voltages at pins 2 and 3 wrt GND = 1.452V when RIO max is 1.652V. 1.2K resistor sits between RIO input and pin 3 (discovered the previous 10K I had there before failed when I measured its current this morning: was reading 16K!). V+ supply of opamp is 15V from RIO and +24V supply is interfacing with inductive load and OUT of drain. But my load is not turning on. I kept hearing buzzing sounds but no opening of valve. SO I put DMM on OUT and V+ of power supply and it reads 15V! I take this to mean the voltage of the V+ on opamp since gate of MOSFET and Pin 1 of opamp both read same volts of 5.9V. Mosfet seems ok. Voltage drop across R1 is also 1.452V and circuit seems perfect.
I am operating from a +24V supply so I turned up voltage to +30V to see if I'll get any changes but nope!
So I went back down to 24V, turned up inductor and I discovered another 15.72V drop across the load. When I disconnect the load however, I get 24V through it. Wondering what could be wrong!

 

[Lexilighty]

Okay. I tested the circuit with my other spare valve and I realize that other one now responds to proportionality with same circuit. But the first one I've been using along would not any longer. It still opens up at full voltage across its coils though. Could it be I have lost the first valve?

 

[jim hardy]

But my load is not turning on.

I guess by "load" you mean the air ?

+24V supply is interfacing with inductive load and OUT of drain

Interfacing ? not sure what you mean.

But my load is not turning on. I kept hearing buzzing sounds but no opening of valve.

hmmm remember that line in valve datasheet about voltage may give fluctuations...
buzzing could be the valve oscillating.

See if does the same when you call for midrange air, 130 ma or so

SO I put DMM on OUT and V+ of power supply and it reads 15V!

that says there's about 15 volts across valve, meaning you have about 100 ma through valve?
When it is doing that and buzzing switch your meter to ACvolts , if valve is oscillating you'll read substantial AC.

Okay. I tested the circuit with my other spare valve and I realize that other one now responds to proportionality with same circuit. But the first one I've been using along would not any longer. It still opens up at full voltage across its coils though. Could it be I have lost the first valve?

Could be it's got overheated and damaged. But that it opens at full current says it's worth further investigation.
Can your RIO be programmed easily to step current from 100 ma to max in 10 ma steps, then back down?
If so try that . Valve might have a piece of dirt in it or something that's making it sticky.
Repeat that test several times being observant for clues.
Does it suddenly leap open at full current?
Is there some range of current that makes it buzz?
Is 24 volt supply staying constant?
How's heat on that Mosfet ?

posting this so as to avoid timeout

stability problems can be tricky.

 

[jim hardy]

gate of MOSFET and Pin 1 of opamp both read same volts of 5.9V.

good deal .
Observe from mosfet datasheet http://www.vishay.com/docs/91015/sihf510.pdf
line "Gate-Source threshold voltage" he's allowed 4 volts to start conducting
and from curve 3, typical transfer characteristics, gate voltage goes up with temperature. That's why i asked how hot he's running.

your 1.452 at top of R1 (which is mosfet source terminal) plus 4 volts makes ~5.5 volts so we're passing sanity checks now.

 

[jim hardy]

How much air pressure do you have to this thing?
I notice from curve on page 687 of valve datasheet " PQ30 1.6mm orifice flow vs dp " that at 30 psi flow won't start until about 140 ma

 

[Lexilighty]

I guess by "load" you mean the air ?

I meant solenoid valve.

Interfacing ? not sure what you mean.

Connected. Eternal vigilance, eh?

that says there's about 15 volts across valve, meaning you have about 100 ma through valve?

Nope, valve is operating at full voltage (1.65V) and resistor is now 10 ohms. I am attaching schematic and circuit layout down here.

Does it suddenly leap open at full current?

Yep!

Is there some range of current that makes it buzz?

Full current, I have observed.

Is 24 volt supply staying constant?

If you mean the 24V of power supply,. I bet he is.

How's heat on that Mosfet ?

Fairly okay for now. I've not noticed any hotness on the heatsink plate.

and from curve 3, typical transfer characteristics, gate voltage goes up with temperature. That's why i asked how hot he's running.

The temperature is fairly ambient or below ambient based on what I observe when I touch the heatsink.

How much air pressure do you have to this thing?

I am doing between 0 to 24 psi for moving air through the valve.