Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Mosfet series regulator

Status
Not open for further replies.

M1EUF

New Member
Im thinking of making up a13.8v 25a power supply but dont want to use a whole bunch of those ancient 2N3055s. I have in my junk box 10 of SNP60N03s and an thinking of using, say, 6 of these in parallel, on a decent heatsink of course. I will be fitting a resistor in each source lead of say 0.1ohms in an attempt to keep the current through each device roughly equal. Im hoping that, with the high input impedance of the gate circuit, I can drive them all on a single op amp. Will I have any problems with a single device hogging the current? your suggestions please, but must not be an SMPS.

Regards, John M1EUF
 
Can't find a datasheet on the mosfets, but you won't be able to drive them with a single opamp.

Mosfet gates are caps. They need to be charged quickly for the mosfet to dissipate the least amount of power. Typical power mosfets usually need 1A or more to switch quickly enough. Large ones can take up to 10A or more. Op amps are not power devices and just cannot supply this kind of current. You will need drivers for the mosfets.
 
If they are in a linear mode the gate capacitance issue is not relevant. Until the switching frequency gets into the 10's of KHz the combined capacitance may still not be much of an issue.
Being voltage controlled devices you may run into control circuit issues though. That will depend upon what type of control and feedback system you are using to maintain the voltage regulation.
 
Can't find anything about the particular characteristics of the SNP60N03 but, in general, MOSFETs can have significant differences in their gate turn-on voltage (Vgs) which makes them difficult to match in a parallel configuration in the linear-mode with a common gate drive. The transistor with the lowest turn-on voltage will tend to hog the current. The source resistor would have to be large enough to drop a voltage comparable to the difference in Vgs between the transistors. That could waste a lot of power.

I don't know of any solution to that, short of using a op amp driver for each transistor. Each op amp would drive the transistor gate to have a similar source voltage with a small source resistor to provide the working voltage (thus balancing the currents) for each transistor.
 
If you do not short the output of your supply then only one Mosfet is needed.
 
Thanks for the replies so far. The MOSFETs in question are rated at 60A Id and 30V. In theory i suppose one device would be adequate, but I cant see me having access to cryogenic cooling equipment so I thought it would be better to use several devices to share the current. They are described as logic level devices and seem to turn on at about 3v. I think im going to do some experimenting here and just 'suck it and see' with a decent dummy load, perhaps I can select several from the batch which have the most similar turn on voltages. If it means using a separate driver for each transistor then I will shelve the idea and use bipolar devices.

Regards, John
 
It doesn't matter how many devices you use in parallel if you don't have enough heat sinking.

The power dissipated is the drop across the device times current through it.

If you have 3v drop across MOSFET at 10 amps that is 30 watts. A large heat sink would be needed.
 
I once did a similar project and:
1) Matched the Vgs threshold voltage, and
2) Used a source resistor to balance out the remaining difference.

I also used a transistor emitter follower to boost the opamp's output, but to tell you the truth, I don't know if it was really required or not. My concern was that the total gate capacitance -which can become quite large- could cause unstability on the opamp output.
 
A 100Ω to 1KΩ resistor between the op-amp output and the gate(s) will keep the op-amp happy. The RC time-constant created by the series resistor and the gate capacitance shouldn't matter....
 
I once did a similar project and:
1) Matched the Vgs threshold voltage.
You looked at the wrong spec.
The threshold voltage is when a Mosfet barely turns on (or is almost completely turned off) with a current of only 0.25mA.

If you want the Mosfet to conduct a lot of current then you should look at the "on-resistance" spec. The gate voltage is 10V for most Mosfets. But their theshold voltage for only 0.25mA is spec'd at 2V to 4V.
 
You looked at the wrong spec.
The threshold voltage is when a Mosfet barely turns on (or is almost completely turned off) with a current of only 0.25mA.

If you want the Mosfet to conduct a lot of current then you should look at the "on-resistance" spec. The gate voltage is 10V for most Mosfets. But their theshold voltage for only 0.25mA is spec'd at 2V to 4V.
Sorry, wrong. You looked at the wrong spec. You are confusing this applicaton with switching on and off.

In a linear voltage regulator, the ON resistance is irrelevant. The transistor is operating in the region where the drain current is controlled by the gate voltage. Look at the curves in the area where Drain to Source voltage is greater than few volts. The threshold voltage indicates the voltage at which the conduction curves begin, and are the most reliable predictors of current sharing in paralleled MOSFETs.
 
Sorry, wrong. You looked at the wrong spec. You are confusing this applicaton with switching on and off.

In a linear voltage regulator, the ON resistance is irrelevant. The transistor is operating in the region where the drain current is controlled by the gate voltage. Look at the curves in the area where Drain to Source voltage is greater than few volts. The threshold voltage indicates the voltage at which the conduction curves begin, and are the most reliable predictors of current sharing in paralleled MOSFETs.
No.
The curves in the datasheet are for "typical" Mosfets. There are strong ones and there are weak ones that have the same part number and the same manufacturer. Their spec's are not typical.
The threshold voltage is a wide range of voltage. You don't know how sensitive and strong is the Mosfet you bought unless you measure its spec's.
 
If you measure the threshold as schmitt trigger said, you can match transistors adequately for service in an amplifying application such as the one in this thread. [edit: incorrectly attributed measurement to MikeML.]
 
Last edited:
Mosfets have different gains. Some have high gain and others have low gain.
The theshold voltage is at the bottom when the Mosfet barely turns on.

If you want to match Mosfets then they should be turned on near their operating current, not just barely turned on.
 
One other thought. I have heard that mosfets have a negative temperature co efficient and will tend to conduct less as their temperature rises. How about mounting each device on a smaller, separate heatsink? could I rely on this to equalise the currents?
Also, I might not yet scrap the idea of using a separate op amp for each device as quad op amps are available...could make up say 4 separate regulators and combine the outputs with diodes...
 
Well, everyone is entitled to an opinion, but the facts were:

1) Before matching the Mosfets for Vgs(th), I had included the source resistors only as a means of equalizing currents, and I was measuring something like 50% current difference between them. (If one was conducting let's say, 1 amp, the other would conduct 1.5 amps).

2) Although the circuit was working apparently well, I decided to reduce the conduction difference further. When I performed the Vgs(th) matching the difference was reduced to about 10%. So matching did work.

3) Now Audioguru does have a point...Vgs(th) is usually measured at low drain currents, in the milliamp range. If this measurement were to be taken at drain currents closer to the actual operating currents, I believe that the difference could be brought further down.
 
I agree, schmitt_trigger. Vgs(th) matching will be helpful for this application.

The heat sink still has to be quite large, since in this circuit the only improvement is reducing the aggregate junction to heat sink thermal resistance. The heat is still there.
 
I have heard that mosfets have a negative temperature co efficient and will tend to conduct less as their temperature rises.

Very true… with parallel devices, one will turn on before the others, but as they “warm-up”, the ones that are conducting more current will get hotter and the Rds on will go up. Now, since current takes the path of least resistance the “cooler” devices will now start to conduct more. While the current will not be exactly the same in each device, when thermal equilibrium is reached, all devices will be “sharing” the load current (for a closed loop system).
 
I wonder if thermal equilibrium can be relied upon then to give me adequate current sharing. Im not bothered about getting bogged down too much in the theory, just something that works adequately. "source resistors only as a means of equalizing currents, and I was measuring something like 50% current difference between them"
50% difference in current sounds pretty closely matched to me so it might not be worth going to the trouble to get it any better.
On another point, how much power does an old 486 type processor dissipate? im wondering if i could use a bank of old computer processor heatsinks and fans to assist the cooling, and how many watts each heatsink can dissipate. I want the project to be cheap and use junk box parts rather than innovative.
 
My 486 pc went through two processors.
The first one was a 486/66 (66MHz) that was in a normal DIP case and did not have a heatsink.
The second one was a super high speed one built like a Pentium that operated at 100MHz and was in a surface-mount case and worked from 3.3V. It was on a small "daughter board" that was the same size as the first one and also did not have a heatsink.
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top