[SOLVED] constant current circuit (4A) with unstable Load

[z_mx]

hi everyone
i need to develop constant current circuit (4A +-10%) in temperature range (-45 to 85 C). i mean constant current in temperature range, circuit will work 1 sec. supply voltage 28 V. main problem that my load can be from 1 to 7 ohm (random). i don't know if anybody here had experience this problem and should please guide me through..
thanx

 

[goldsmith]

hi everyone
i need to develop constant current circuit (4A +-10%) in temperature range (-45 to 85 C). i mean constant current in temperature range, circuit will work 1 sec. supply voltage 28 V. main problem that my load can be from 1 to 7 ohm (random). i don't know if anybody here had experience this problem and should please guide me through..
thanx

Hi Z_mx
The whole idea is using a simple hiccup current limiter . it will handle your aim as well . but you may want to change some of the things in a basic hiccup current limiter , for instance something about temperature feedback or perhaps implying it as a switching hiccup system which can yield better efficiency in compare with a linear one .
Best Wishes
Goldsmith

 

[godfreyl]

temperature range (-45 to 85 C)

That's the hard part.
The main problem with the circuit you showed earlier is that it compares the voltage across the current sensing resistor with the Vbe of a transistor, and the Vbe changes with temperature.

The trick is to use a precision voltage reference that doesn't vary with temperature, and compare the voltage across the current sensing resistor with that. You could use something like an LM285, LM385 or MP5010.

Here's a circuit to show the idea. It's not a finished design, just something to start with. In particular, the current mirror should probably be a lot better.



Q1 and Q2 form a differential pair that compares the voltage across R5 with the voltage across D1. The Vbe of each transistor still varies with temperature, but that doesn't matter because the two temperature coefficients cancel out.

The total current through R1 and the two transistors will change with temperature too but that also doesn't matter because Q5 only sees the difference between their collector currents.

High frequency compensation of this circuit may be a bit tricky. My gut feel is that Miller compensation of either Q5 or Q6 will likely be a train-smash, and that a shunt capacitor as shown is probably the most stupid-proof option.

There are some other practical considerations to think about. e.g. R5 has to dissipate about 5W of heat at up to 85°C, and you need to worry about it's temperature coefficient as well. Similarly, the MOSFET has to dissipate up to 100W at up to 85°C. I suspect you will have to use multiple MOSFETS and a sizable heatsink and/or forced air cooling,

If you want to reduce the power wasted in the current sensing resistor and reduce the voltage drop across it, you could use a lower value resistor and compare the voltage across it to a fraction of the reference voltage. Here's one way to do that:



Just for fun, I used a CA3046. Hence the diode connected transistors and Q5 doing nothing useful. The individual transistors in the chip aren't very good but they are well matched, which is what's important here. Also, it's cheap.:smile: Note that the currents through Q3 and Q4 are equal, and R7's value is equal to the parallel combination of R1 and R3.

hiccup current limiter

He wants a constant current source, not a protection circuit.

 

[goldsmith]

He wants a constant current source, not a protection circuit.

Hi Godfrey

Of course i know he wants a constant current source . as i've saw his another thread which was exactly like this one , ( it has a mosfet and then load in series with the mosfet ) i've understood what he wants exactly . and certainly a simple hiccup current limiter can work as a perfect current source . i didn't show hime any circuit at first because i thought thus he will try to think about how to do that ( teaching how to take fish instead of giving fish ! ha ha )

the whole circuit will be simpler and it will have only two or three transistors . ( just ! )
Best Wishes and regards to you
Goldsmith

 

[godfreyl]

@Goldsmith:

OK, I understand your problem now. In another thread you showed this circuit and you called it a hiccup current limiter:

Here a simple hiccup current limiter comes :

But that circuit is not a hiccup current limiter.

Here is a description of how a real hiccup current limiter works.

The three major types of current-limit-protection mechanisms are constant, foldback, and hiccup. Hiccup current limit performs the best of the three types; however, the implementation is rather complex. In this scheme, upon detection of an overcurrent event, the whole power supply shuts down for an interval before it tries to power itself up again. The cycle repeats until the overcurrent fault disappears.

I do agree the constant current circuit can be made with an opamp and less transistors. That would be simpler.

 

[goldsmith]

OK, I understand your problem now. In another thread you showed this circuit and you called it a hiccup current limiter:

Quote Originally Posted by goldsmith View Post
Here a simple hiccup current limiter comes :
But that circuit is not a hiccup current limiter.

Here is a description of how a real hiccup current limiter works.

Do you know why a circuit can be called as hiccup ? or what the word hiccup does mean ?! means on/off state , as fast ! to keep the current in the desired value . the word hiccup has been used in many of the text books for such a circuit that i've used . every circuit that can deliver fast on/ off state to keep the current constant , called hiccup . so a hiccup circuit can work as a protection or current source or everything else that depends on the designer .

I do agree the constant current circuit can be made with an opamp and less transistors. That would be simpler.

Not just with an op amp , you can design a constant current circuit just with three transistors and no op amp . again it depends on the designer to make what kind of choice in design process .

 

[godfreyl]

Do you know why a circuit can be called as hiccup ?

Yes.

or what the word hiccup does mean ?!

Yes. English is my home language.

means on/off state , as fast !
[snip]
every circuit that can deliver fast on/ off state to keep the current constant , called hiccup .

The circuit you showed does not "deliver fast on/ off state". It gives smooth, continuous current. No hiccups.

the word hiccup has been used in many of the text books for such a circuit that i've used .

No it hasn't. Compare your circuit to the ones in the books.
Your circuit is called a constant current source, not a hiccup current limiter.

you can design a constant current circuit just with three transistors and no op amp

Not if you want stable output current over a wide temperature range.

 

[goldsmith]

The circuit you showed does not "deliver fast on/ off state". It gives smooth, continuous current. No hiccups.

Your answer is surprising me ! how you say that ? you can easily decrease speed of loop by a relaxed network and then see it is exactly what i said and it is hiccup . it's speed is pretty fast and if you want see it's right form you need to decrease speed of loop . these are things that i'm using since many years . and i'm sure about their performance . not just me , many of the text books and professors around the world .
Ok , now what you need is simulating the circuit with decreased loop speed . try to do it .

No it hasn't. Compare your circuit to the ones in the books.
Your circuit is called a constant current source, not a hiccup current limiter.

Again you need to do what i told about decreasing speed of the loop

Not if you want stable output current over a wide temperature range.

I agree . but why i told that to you ? you didn't mention anything about temperature and i told it is possible to do that . of course i know the original poster mentioned about the wide range of temperature .
Certainly if you need to guarantee it's performance instead of variations in temperature , you need to add just two diodes . if you need very high precision you must take a feed back from the temperature of ambient .
That's all !

 

[godfreyl]

it is exactly what i said and it is hiccup .

The MOSFET does not switch and off, it conducts continuously. Thus it is not a hiccup circuit.

Maybe your circuit oscillates because there is no frequency compensation. In that case the output is not a constant current.

Ok , now what you need is simulating the circuit

No thanks. I've got better things to do. Maybe you should simulate your circuit. I think you will find the MOSFET does not switch on and off. Thus you might learn something new!

 

[goldsmith]

The MOSFET does not switch and off, it conducts continuously. Thus it is not a hiccup circuit.

Maybe your circuit oscillates because there is no frequency compensation. In that case the output is not a constant current.

Quote Originally Posted by goldsmith View Post
Ok , now what you need is simulating the circuit
No thanks. I've got better things to do. Maybe you should simulate your circuit. I think you will find the MOSFET does not switch on and off. Thus you might learn something new!

I've learned all of the things that i needed before . and i don't need to learn how a simple circuit does work . and as you aid you don't have time to learn because you have best things to do . so i think we need to don't talk about things that we both don't need . ok ? it will be better for both of us . and furthermore i don't need to demonstrate things to the other people .
Here we are to help other people and sometimes learning . and i'm trying to help someone . if you think i can't help , then i think you can tolerate and see the result and if i couldn't help then you contribute the discussion . ok ? so be patience my friend .

 

[D.A.(Tony)Stewart]

I had never heard of a "hickup current limiter" until recently in my last 40 yrs of Electronics. SO I decided to learn.

The three major types of current-limit-protection mechanisms are constant, foldback, and hiccup. Hiccup current limit performs the best of the three types; however, the implementation is rather complex. In this scheme, upon detection of an overcurrent event, the whole power supply shuts down for an interval before it tries to power itself up again. The cycle repeats until the overcurrent fault disappears. With such operation, the dissipation in the power supply itself is minimal.

I would expect the hick-up limiter to have the elements of a linear current sensing limiter but used in a different way.

When I imagine the short circuits that can be caused by CMOS with SCR latchup or bridge shoot-thru with thermal runaway, a hick-up restart makes sense. That's why I never heard of one. I learnt from the beginning to try to avoid/protect/filter/clamp to prevent these issues. But it makes sense, for some apps. A hickup counter might be useful too.

more **broken link removed**

Now getting back to the original question...

4A for 1 second with 1~7 Ohms from 28V was it?

This should be doable with any linear design that uses low drop out or LDO switching since 28V/7=4 is the limit.
But as Goldsmith pointed out I think, this can heat up your components pretty fast at 1 Ohm. So regulating it can be easy with low Ron Mosfets but getting rid of the heat will be a major issue with 24V drop x 4A ~ 100 Watts being dissipated. (wasted)

The more efficient way to achieve this in this case is to use a SMPS buck converter to provide 28 to 4V range with 4A CC mode and an enable function.

The demand for Buck CC drivers has increased now as LED drivers, so should be lots of ready to go designs. eg

For me, I would buy a laptop charger rated for >100W with universal capability and a 6pin header. WHy it has remote sensing here which you can use for changing it to CC mode. That's how I powered the LEDs in my photo

 

[crutschow]

......................... you can easily decrease speed of loop by a relaxed network and then see it is exactly what i said and it is hiccup . it's speed is pretty fast and if you want see it's right form you need to decrease speed of loop . these are things that i'm using since many years . and i'm sure about their performance . not just me , many of the text books and professors around the world .
.......................

The constant current circuit you show may oscillate due to a zero in the feedback loop from the added again of the NPN transistor but I would not classify that as a hiccup limiter. It is simply a limiter with uncontrolled oscillation due to feedback instability. A true hiccup limiter has some sort of timers to turn the current on and off for fixed period of times under short circuit conditions to limit the power dissipation in the series current limiter transistor or power supply. Typically the ON period is much shorter than the OFF period for a true hiccup limiter.

 

[D.A.(Tony)Stewart]

yes zapper the oscillation is from additional phase shift outside the Op Amp.

 

[godfreyl]

I had never heard of a "hickup current limiter" until recently

Likewise. It's surprising how little mention it gets on the internet.

When I searched for "hiccup current limiter" on Google, it only found about 3310 results, and only bothered to show the first eleven, nine of which were quotes of Goldsmith on this forum.

Narrowing the search down a bit, it turns out Google can only find four (not a typo) pages not related to something Goldsmith posted here.

- - - Updated - - -

Correction: There's plenty of good info out there, it's just that almost nobody calls it a "hiccup current limiter".

Searching for "hiccup overcurrent protection" or "hiccup short circuit protection" gives lots of good results.

 

[D.A.(Tony)Stewart]

yes over current protection is the standard term as you know. as in OCP, OVP, and CV mode or CC mode for regulation.
Hiccup is what you need when a fold-back still latches the SCR fault condition or there is still a screwdriver across the rails.

However not all CC mode regulators are Short circuit protected from inadequate protection and not all CC regulators prevent surge currents well above the set limit in case of an arc. Case in point.. a Slaughter Hipot test used to test every DUT in production on the AC mains for leakage and primary surge withstanding voltages such as 2 to 3kV. Leakage is often in a PSU line filter to ground in the order of 100 uA, but I discovered on one OEM when the secondary was also grounded , isolation diminished to 1kV, so I had to add a large series resistor in the line test so that the PSU would not get fried from the surge.

I have assisted in diffusion bonding current & VI power instrument regulation in order to weld steel tubes together with a Zirconium shim. The constant current desired was variable from 10kA to 100kA as the tube ESR reduced by the welding process around the tube. Essentially resistance welding with DC but with a variable short on the object being welded. I just used the 8"diameter copper/steel arm on the welder to tap in 2 screws for a current shunt monitor to get 50mV. I digress ..memoires circa '79. EMI was also a challenge.

 

[goldsmith]

Correction: There's plenty of good info out there, it's just that almost nobody calls it a "hiccup current limiter".

Searching for "hiccup overcurrent protection" or "hiccup short circuit protection" gives lots of good results.

Hi again
I think this article deals with hiccup mode current limiter :
https://www.google.com/url?sa=t&rct...XBrSxSMYmkRBsOc2IyNS2qQ&bvm=bv.45175338,d.Yms
is the problem between us just about a name ? is that makes any sense ?
I saw in EDN too that hiccup mode current limiters are used . and in some other books too . if you need more sources i can bring them here .
and about oscillations in the circuit , sometimes unstable and uncompensated circuits can do something that we want . it is why i didn't use any compensation in that circuit to make it unstable to force it to do what i mean and what i've looked from a hiccup mode current limiter .
But i still couldn't understand what was the problem that you stuck on it . the question could be easily answered by what i've told at my first post . and there were not needed to such a long discussion between you and i here in this thread .
By the way please go to the page three of the article that i've attached .
Best Wishes
Goldsmith

 

[z_mx]

hi!
thanks for everyone for the replies
i solved this problem! here is a circuit:

Vzener=4.7V
here is a simulation:

as you can see: current on R3 from 3.94A to 4.1A. tolerance less than 10%.

 

[godfreyl]

i solved this problem! here is a circuit:

Hi
There's a couple of problems with that circuit.

Firstly, you've got 4A flowing through M6 and the load but you've also got 4A flowing through M3,M1, R2 and R4, wasting an extra 112W as heat. That's going to make your cooling problems much worse than they need to be.

Second, and more important, the current mirror made from M3 and M6 isn't going to work as you expect, even if you have perfectly matched MOSFETs. The problem is that M6 is going to get hotter than M3 so their Vds(thresh) is going to be different, causing the currents to be different.

You've got the right idea driving the load from the top with a P-channel MOSFET but it would be better to drive that MOSFET directly with the opamp, and sense the current with a resistor between the MOSFET's source and V+.

 

[goldsmith]

Hi
I agree with Godfrey about that issue that your circuit is pretty dissipative . so you can go through this circuit :


If you change V3 , you will be able to control current through the load in your desired range . it will given by : V3/0.1

Good Luck
Goldsmith

 

[z_mx]

hi my friends!
i agree! but don't forget that this circuit will work 20ms. i don't need constant current source for the long time (this will be one puls).
thanks))