high-side current sensing chips

This is just a suggestion, but consider the circuit below. If the transistors are well matched Vchg should come pretty close to Vgen/10 +

Try this out with SPICE before you run with it: I just threw it down off the top of my head, it's not exactly like anything I've done before nor is it tested at all.

Vgen ___ Vbatt o-----o---|___|---o-----o | | ----->

| 0.02 | Ibatt .-. .-. | |100 | |100 | | | | '-' '-' | | | | >| |< |---o---| /| | |\\ | | | | | | Vchg | '-----o o-----o | | | .-. .-. | |4.7K | |47K | | | | '-' '-' | | | | === === GND GND (created by AACircuit v1.28.6 beta 04/19/05

"gearhead" schreef in bericht news: snipped-for-privacy@m36g2000hse.googlegroups.com...

That Micrel chip is not a current sensor, it is a mosfet driver. Looking for a highside current sensor, try the MAX4374. A ZXCT1008 or ZXCT1009 may be even better. But I doubt you to need that special (and expensive) chips.

Can't be sure what current you want to limit, but usually the field current is limited to limit the output voltage of the generator. Which in turn prevents the battery from being overloaded. This is exactly what the circuit of your schematic is supposed to do. Just use R1 to set the correct voltage. For a 6V lead-accid battery this will be about 6,9V. So even for fully loaded battery the voltage should never exceed that value. You may have to take some voltage loss accross D6 into account which is about 0.2V for this particular diode.

petrus bitbyter

I've done plenty of scheming about how to do it with discretes. But like you said, I'd still want to use a chip in the end, to get matched transistors. Pairs and arrays, I see ony surface mount, including zetex. If I have to go that way, might as well use the ZXCT1009F in SOT-23 with pins 1.9 mm apart according to the diagram. Almost a tenth of an inch, could probably solder it right onto the pads of a perfboard, eh?

- snipped-for-privacy@m36g2000hse.googlegroups.com...

or

it

e.

Current limiting protects the generator, not the battery.

- snipped-for-privacy@m36g2000hse.googlegroups.com...

or

it

e.

Don't say the micrel chip is not a current sensor. It does current sensing. And current limiting is for the generator, not the battery.

Hello,

When you buy about 10 general purpose PNP transistors from one batch, you will probably find several transistors that are within some mV with same Ic. When you design the 2 transistor current sensor at low bias current, self heating can be neglected and it saves you from soldering SMD devices like NXP's BCM857. When matching is impossible, you might add a trimmer potentiometer for nulling.

best regards,

Wim PA3DJS

of

ent

ds

et

and

with

s,

ff

Doesn't gain determine the balance in a current mirror? If so, one would have to match for gain. I did match for gain last time I experimented with current mirrors. I'd like to hear from the experts about the need for this.

About temperature matching. I calculate that one tenth of a degree centigrade difference between the transistors in a mirror would add about one percent error to the current mirror. I wouldn't want temperature vagaries causing any more error than that. I'm wondering if you could count on discrete transistors on a board to stay within a tenth of a degree to each other. More fodder for the experts.

Hello,

For such circuits matching on Vbe with constant collector current is sufficient. I normally do it by connecting B to C, bias at required current and measure the voltage across the diode. Of course for high volume applications I use a matched pair.

When 1% additional tolerance is unacceptable, you also should evaluate all other components. I think of current sensing resistor, other resistors, voltage variatons, offsets in control loop etc. Regarding temperature, put them close together, take some epoxy resin and some piece of coper/alu, and temperature difference due to environment will be negligible.

I can hardly imagine that your original application requires that high accuracy. Last year I designed several current limiting circuits for hot swap applications and inrush current limiting and less then 5% accuracy without adjustment and cheap components will be very difficult.

Best regards,

Wim PA3DJS

On Tue, 8 Apr 2008 09:40:29 -0700 (PDT), gearhead put finger to keyboard and composed:

AFAIK, the LM723 requires a minimum supply voltage of 9.5V. The internal reference zener is 7.15V, typical.

- Franc Zabkar

berichtnews: snipped-for-privacy@m36g2000hse.googlegroups.com...

| Don't say the micrel chip is not a current sensor. It does current | sensing. | And current limiting is for the generator, not the battery.

Hmm... Did you check the high-side current sensors I mentioned? As you know the datasheet of the mic5021 can you explain how you get a signal from it that is proportional to the measured current? I don't see it, but I like to learn.

I am not aware of an automotive generator with a current limiter other then the fields. Normaly the generator is big enough to provide the required power for the vehicles appliances with a wide margin but yes, you can damage it by overloading it. Most of the times however you will blow one or more fuses when you try.

petrus bitbyter

- snipped-for-privacy@k13g2000hse.googlegroups.com...

s.com...

to

The MIC5021 does not put out a proportional signal, it trips at 50mV. Yes, maybe it would make the lights flicker if it were to toggle on and off too slowly. And I looked at the sensors you mentioned, okay?

ge

You need to know, this isn't an alernator. Are you familiar with the old-fashioned generators with the long, skinny armatures with segmented commutators? They are different from modern alternators in that they will burn themselves up if you overload them. Is English a second language for you?

le

ple of

current

t.

in

rd

m-

too

bly

but

rel

he

needs

chip

asheet

of

he

/10 +

nd to

OK with

parts

wn off

e nor

ut

ed

f an

h?

e,

My first dog was named Wimpie. And there was a Wimpie character in the old Popeye comic that used to say, "I will gladly pay you Thursday for a hamburger today!" Ah, Saturday mornings in front of the TV.

Thanks for the tip about testing Vbe, it makes sense. I guess putting a bunch of diode-connected transistors in series would really be the best way, you could go down the line testing voltage and sort them out right quick.

gearhead wrote in news:e66fa9e8-d621-4ba3-8c47- snipped-for-privacy@m36g2000hse.googlegroups.com:

I'd pick up a surfboard and give up on the thruhole idea.

berichtnews: snipped-for-privacy@k13g2000hse.googlegroups.com...

berichtnews: snipped-for-privacy@m36g2000hse.googlegroups.com...

| | The MIC5021 does not put out a proportional signal, it trips at 50mV. | Yes, maybe it would make the lights flicker if it were to toggle on | and off too slowly. And I looked at the sensors you mentioned, okay?

| | You need to know, this isn't an alernator. Are you familiar with the | old-fashioned generators with the long, skinny armatures with | segmented commutators? They are different from modern alternators in | that they will burn themselves up if you overload them. | Is English a second language for you?

I remember those old things and the electromecanical circuits that controlled them, though I never got into their inner workings. So I do not know about current control and how it was done.

English is not my native tongue.

Consider a ZDX1009. It is SMD but large enough to be mounted on perf board. The datasheet shows the internal schematic which is very simple so you can also build it using discretes. You have to use matched transistors as others stated already.

petrus bitbyter

(snip)

You mean the ZDS1009, petrus. Check out that circuit in the datasheet -- if it doesn't leak, it won't turn on. The circuit below shouldn't have any problem starting up. Below Ilimit, the collector output will not source any current. Just as load current passes Ilimit the output collector should start sourcing a small current. I can use this collector current in a feedback loop with the voltage regulator and make the load current settle at Ilimit.

V+----+-----Rs-------+-----+---load | | | Re | | | | | | | | \\ PN2907 / | >| x2 |< | |---+----| | /| | |\\ | / | \\ | | | | | +------' | / | | |< | ,-------+--| | | | | \\ | / \\ \\| | |/ \\ |---+----| '---out /| |\\ < PN2222x2 >

| | '------+-------' | current sink I = 2 Ilimit Rs / Re | | gnd

For example, Isink = 200 uA, Re = 1k, Rs = .01, Ilimit = 10 A The current mirrors reach an equilibrium at Ilimit, with 100 uA on each side. Below Ilimit, the right side of the current mirror pulls up and the left side pulls down. The output transistor stays turned off. Above Ilimit, the left side pulls up and the ride side pulls down, drawing some current from the base of the output transistor.

Whaddya think, it might work?

nt

ot

rd.

an

hers

Correction: instead of a mirror it needs separate current sinks on the ground in order to work.

And another transistor on the output for more gain, if necessary.

Correction: instead of a mirror it needs separate current sinks on the ground in order to work.

And another transistor on the output for more gain, if necessary.

Yes, I meant the ZDS1009.

I gave a quick glance and I doubt this circuit will work. The way the outputtransistor is connected will ruin whatever the others might do. For instance, most of the current through the right current source will come through the BE-junction of that transistor. So the balancing for the PN2907 will be gone. Besides, you will have to match both current sources which brings you directly to the original schematic. Sure, it will not start without leakage. As a matter of fact, that leakage is specified in the datasheet. All bipolar transistors I met so far, had some leakage. It's inherent to to physics of the transistor. Once amplified, it will become serious current which brings the circuit to live. I ever build a current sensor like that and it wordked though not very accurate. I used no matched or twins transistors and did not bother to make them thermal connected. At the time, it was good enough for my application.

petrus bitbyter

Doesn't gain determine the balance in a current mirror? If so, one would have to match for gain. I did match for gain last time I experimented with current mirrors. I'd like to hear from the experts about the need for this.

About temperature matching. I calculate that one tenth of a degree centigrade difference between the transistors in a mirror would add about one percent error to the current mirror. I wouldn't want temperature vagaries causing any more error than that.

For C> sakes! You are talking about a motorcycle here. 1% up or down will not make an ioata of difference to the battery charging on the bike.

I'm wondering if you could count on discrete transistors on a board to stay within a tenth of a degree to each other. More fodder for the experts.