manage load with digital pot

[kheston]

I've put together a circuit using an AD5220 digital pot, a 2n3904 and a TIP3055. The hope is I can control up to a 3A load using the circuit you see here:

I understand the output voltage will follow any changes in the battery's voltage. No problem.

Will the circuit work? Any suggestions?

[Robert Reed]

Looks like D1 should be reversed.

[kheston]

Missed that...fixed.

[Robert Reed]

Hi Kheston

Assuming an HFe of 30 for your your 3055 xistor and a 3 amp load, you would need 100 ma of base current supplied by the 3904 follower. Again assuming an HFe of 100 for that 3904, would require 1 ma of base current. With the digipot at its upper extreme, thats leaves 11K in the base path. At a 1 ma current flow,11 volts are dropped across this chain and I only see one volt appearing as the base vloltage here. Add a drop of 0.5 volt at each B-E junction and you have 0 volts across the load, and I am being very optimistic with my figures. Something is definately wrong here. Can you explain what the intended load voltages are planned to be?

[kheston]

Robert,

Thanks for looking. This is what I'm hoping to be a more direct route to exciting the field winding on an automotive alternator (see my other threads) than building a complete regulator circuit. In other words, it is an attempt to completely offload to a µC complete control of power sent to the alternator's field coil. This way, if I want to build a standard 2-stage charger or instead a constant current charger, I'm not limited by the regulator circuitry from doing either.

The max voltage this circuit will see at the load end is 19v, which gives me some headroom in the way my regulation algorithm will actually work (because there's little chance I'll ever let the voltage get that high).

My use of the 3904 is an attempt to limit current flow within the ADC5220...which I guess I did too well. Is there a better way?

[MrAl]

Hi there,

Just to note, you do realize that the R1 and R2 resistors (10k each)
limit the output at the arm of the 'pot' to 1/3 Vcc and 2/3 Vcc right?

As Robert was saying, the circuit as is doesnt have enough gain,
and also it's not as compliant as you probably want either.
A good way to fix this is to add an op amp. A cheapie would do
just fine (LM358) where you connect the arm to the non inverting
input, the output of the circuit to the inverting input, and the output
feeds the 1k driver resistor. This will also allow your circuit to
output from 0v to about 10.5v with the adjustment of the pot,
and it will also be very compliant and fairly linear, with the center
of the pot causing about 6v of output.
Of course you have to get rid of those two 10k resistors unless
you want to limit the output to swing from 4v to 8v only.

Also, just wondering, where are you getting the digital signal to
drive the pot from?

[philba]

Is there a reason you don't want to use a mosfet? It seems like a better way to go.

[dyarker]

Keeping load connected directly to battery negative (common I assume), I've made a couple changes.

Using PNPs the "other way 'round" moves the load from the emitter to the collector. This leaves much more voltage to push current through the transistor bases. That takes care of one problem.

Q1 will need more gain than a TIP3055, because at 100mA a 2N3906 has a minimum gain of 30 (about the same for the NPN 3904), not 100. So with your original, or my changed, circuit; you need more gain like Robert and Al said.

Remove R1 so load current can go to zero.

Or, an inverting stage between Q2 and digital pot. In which case leave R1 and remove R2. With an inverting stage with PNP Q1 and Q2, the direction of current adjustment is back to the way you expect. An op amp or third transistor will also provide the additional gain.

I didn't design a whole working circuit for you. It seems you're trying to figure it out yourself. Just some hints.

C U L -

[Robert Reed]

All good suggestions here Kheston, but if you already have this under construction, the simplest thing to try is merely add a 2N4401 (similar to a 3904 but with added current capability) configured as a follower between the 3904 emitter and the 3055 base.This will probably give the drive required and greater isolation to the digipot. There still may be a drop of a volt in that divider string which you may or may not tolerate? But you may want to delete R1,R2 altogether or reduce their value.which will minimize any 3904/digipot interaction. Also R3 could be eliminated entirely if R1 was still used and had 1K or so resistance.

[kheston]

I'll be using a BASIC Stamp to run this. I went with the digital pot instead of a PWM circuit (despite the billion examples of the latter on Google) because I've only got one µC thread. I think I would need to quit pulsing to read voltage, temp and alternator RPM; would rather not.

There's definitely a consensus that my 10k resistors should change or disappear. I hadn't even considered voltage drop in my design. I guess it goes without stating that my circuit analysis skills could use some improvement...trying to school up, honestly. Resource suggestions welcome.

I really appreciate the suggestions. I'll try a few of them and see if I learn something.

--K

[MrAl]

Hi again,


Here is a voltage controller that will work for your app.

Note that one of the specs is that the digital pot can not have more
than 7v for Vdd, so a protection zener is included in this drawing.
The voltage can be varied from about 1v to 10v with the digital pot.
The output is ratiometric, so that if you have the pot set midway
(5k and 5k) and the supply voltage is 12v you get 6v out (half of 12),
but if the supply is 10v you get 5v out (half of 10).

[rshayes]

I don't see any discussion of power levels. If the load is a resistor of 4 ohms (3 A at 12 V), the maximum power in the 2N3055 will be 9 W (1.5 A and 6 V). This will definitely need a heat sink. If the load has a constant current nature, such as a light bulb, the dissipation could go to 36 W (3 A and 12 V).

With a current gain of 30, the base current of the 2N3055 could be 100 mA. The dissipation in the driver transistor would be around 300 mW for a resistor load, and possibly over 1 W with the constant current load. This is excessive for a TO-92 device. The driver transistor should probably be in a TO-220 or other similar package.

A third stage should probably be added to reduce the input current to a level such that it will no load the 10 K impedance of the digital pot. A diode or two in the ground leg of the digil pot might provide some offset and temperature compensation.

[MrAl]

Hi Rshayes,


How do you think the non inverting input of an op amp like LM358 loads
the pot?

I agree the output transistor will need a heat sink.

[kheston]

Will-called the parts at Jameco today.

I've got plenty of heat sinks and grease around. No problem there.

The load is an automotive alternator field winding. It has about 1k of resistance when the alternator isn't spinning. Not sure if that increases or decreases the number of hypotheticals...

I was thinking that I'd test the circuit with a regular 10k pot to see how it performs before splicing in the AD5220.

Thanks for the great suggestions.

[Robert Reed]

Kheston
"It has about 1k of resistance when the alternator isn't spinning. "
No way. Are you sure you measured this correctly. Should be in the low ohms range.