Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Using IGBT/MOSFET in linear power supplies

Status
Not open for further replies.

casemod

Member level 1
Joined
Sep 21, 2013
Messages
34
Helped
1
Reputation
2
Reaction score
1
Trophy points
1,288
Activity points
1,618
Hi guys,

I have a question. I dont normally see IGBT's or mosfets being used in linear power supplies. Many of the components on these are only rated for their output load when TJ = 25C and the transistor quickly burns under short circuit conditions.

For a 0-30V 5A (or 0-60V 2.5A) power supply to be fool proof, even if permanently short circuited, I would need about 24*1.41*5A = 170W of power dissipation, lets say a worst case scenario of 200W with a TJ = 125C and a heatsink = 60C.

Its not too hard to find mosfets or IGBT's with such specs, so my question is can these be used in place of the bipolar with suitable modifications to the driving circuit? If not what are the reasons?
 
Last edited:

Yes, full rated power is given at an impossible 25C junction temperature and zero power usually at 150C junction temperature.
That is the standard way of rating things.
Both ends of that range are pure fantasy.

You then draw a straight line between the two imaginary extremes and that will give you the maximum theoretical power limit for the device under the conditions you decide to run it at.

lets say a worst case scenario of 200W with a TJ = 125C and a heatsink = 60C.

125C junction temperature is getting too hot, its at 83% of maximum.
You will therefore need to limit any further temperature rise by limiting the safe power to about only 7% of its power rating at 25C.
That is just not practical.

A much better approach would be to aim for the middle, say about 85C junction temperature, and run the devices at half their rated dissipation at 25C.

For 200 watts of real heat, you would then need 400 watts worth of transistors.
That will run them right up to their theoretical limit.

From that you calculate the total thermal resistance from junction to air.
At 40C ambient that would be a 45C temperature differential junction/air.
At 200 watts that works out to 45/200 = 0.225 degrees C per watt.

That will give you the absolute design limit.
I would be adding at least one extra transistor to that to feel safe.
 

To minimize the heat-sink requirements under short-circuit conditions you can add an electronic fuse limit such as this, which reduces the dissipation to essentially zero when the output is shorted.
 

You could also bolt a thermal switch to the heat sink that either starts a fan, or disconnects the power.
clixon.jpeg
 

Thanks guys.

My concern is, as suggested, the use of an IGBT rather than a BJT and what issues I may encounter while doing so, such as oscillations and so on. I've read some papers and hotspots on the die were mentioned. I am aware of the thermal limitations, hence why I am aiming at the absolute worst case scenario.

I'm thinking in adding a buffer to the output of the last opamp to supply the gate capacitance rather than the current arrangement that drove the BJT with a pull down.

Question : Has anyone thought or seen a PS using a controlled triac rectifier? I'm pondering this to reduce overshoot after the load is removed and reduction in power dissipation.
 
Last edited:

Hotspots on the die and local thermal runaway are definitely a problem with BJTs, particularly at higher voltages, known better as second breakdown. this is mostly a problem at higher continuous current above about 15v across the transistor.
There is a time element involved with this, as its a localised thermal issue.

An IGBT is really a high voltage PNP BJT in disguise, and all the above issues occur with these as well. They make superb switches, but have serious limitations for continuous high power operation in the linear region.
Its why you never see IGBTs normally used as linear regulators.

Mosfets look much more promising, they suffer none of these evils. you can definitely run them continuously in the linear region at high power with safety, and they will even current share reasonably well.

But they have considerable high frequency current gain, and can be difficult to tame if run in common source configuration.
Used as source followers, there can be enough degeneration to calm things down, especially if you can damp the gate circuit with some resistance (or ferrite beads) to kill any high Q resonances with parasitic inductance and the gate capacitance.

None of that is likely to show up in circuit simulation, its more a case of build it and then try to figure out why it is unstable. Layout is critical, another thing circuit simulations will not tell you.
With a bit of patience and some serious thought it should be possible to get something working.

Try running it open loop first.
That will tell you if its basically stable without any deliberate feedback.
Only then will you have any hope of successfully closing the loop.

SCRs and triacs are more ideal for battery chargers and similar applications where fast response to load changes are not required.

They can be used for a bench power supply, but the transient regulation will be very poor and well below most requirements as a general purpose supply.

They could be used as a pre regulator for a linear supply, but it gets pretty complicated to set it all up. Tap switching on the transformer is a lot simpler and works almost as well.
 

Hotspots on the die and local thermal runaway are definitely a problem with BJTs, particularly at higher voltages, known better as second breakdown. this is mostly a problem at higher continuous current above about 15v across the transistor.
There is a time element involved with this, as its a localised thermal issue.

An IGBT is really a high voltage PNP BJT in disguise, and all the above issues occur with these as well. They make superb switches, but have serious limitations for continuous high power operation in the linear region.
Its why you never see IGBTs normally used as linear regulators.

Mosfets look much more promising, they suffer none of these evils. you can definitely run them continuously in the linear region at high power with safety, and they will even current share reasonably well.

But they have considerable high frequency current gain, and can be difficult to tame if run in common source configuration.
Used as source followers, there can be enough degeneration to calm things down, especially if you can damp the gate circuit with some resistance (or ferrite beads) to kill any high Q resonances with parasitic inductance and the gate capacitance.

None of that is likely to show up in circuit simulation, its more a case of build it and then try to figure out why it is unstable. Layout is critical, another thing circuit simulations will not tell you.
With a bit of patience and some serious thought it should be possible to get something working.

Try running it open loop first.
That will tell you if its basically stable without any deliberate feedback.
Only then will you have any hope of successfully closing the loop.

SCRs and triacs are more ideal for battery chargers and similar applications where fast response to load changes are not required.

They can be used for a bench power supply, but the transient regulation will be very poor and well below most requirements as a general purpose supply.

They could be used as a pre regulator for a linear supply, but it gets pretty complicated to set it all up. Tap switching on the transformer is a lot simpler and works almost as well.


Thanks

So are you saying a Mosfet is a better choice then? I actually used a Mosfet for the tests. I didn't check overshoot, but seemed to hold on OK otherwise, despite being operated right at the limits. I'm aware these share the current better than an IGBT as well, although the idea with the IGBT was to use a single device with the required power dissipation. Transistors are currently limited to about 400W at 25C. IGBT's can go way above this.

Yes the aim is to use the controlled rectifier as a pre regulator, to keep the voltage across the transistor minimized to say 5-10V. As long as the supply operated at a constant voltage the output should be fairly stable.

Say if the output were 5V the DC link would be regulated at 10 or 15V rather than 24*1.41 = 35V.
 

Yes the aim is to use the controlled rectifier as a pre regulator, to keep the voltage across the transistor minimized to say 5-10V. As long as the supply operated at a constant voltage the output should be fairly stable.
SCRs will be much better for this than a triac, because the commutation is forced each half cycle.

Mosfets are good because the bit that gets hot increases in resistance.
BJTs and IGBTs are poor because the bit that gets hottest reduces in resistance and hogs the current. Its an accelerating self destruct mechanism.

Mosfets tend to be thermal equalisers, both between devices, and across a single die, so the whole thing is a lot more rugged.
 

SCRs will be much better for this than a triac, because the commutation is forced each half cycle.

Mosfets are good because the bit that gets hot increases in resistance.
BJTs and IGBTs are poor because the bit that gets hottest reduces in resistance and hogs the current. Its an accelerating self destruct mechanism.

Mosfets tend to be thermal equalisers, both between devices, and across a single die, so the whole thing is a lot more rugged.

SCRs will be much better for this than a triac, because the commutation is forced each half cycle.

Mosfets are good because the bit that gets hot increases in resistance.
BJTs and IGBTs are poor because the bit that gets hottest reduces in resistance and hogs the current. Its an accelerating self destruct mechanism.

Mosfets tend to be thermal equalisers, both between devices, and across a single die, so the whole thing is a lot more rugged.

That works pretty well in saturation for a switching regulator, my concern is that as an analog amplifier working on the linear region, for the same gate voltages one device may inherently conduct more current than the other - the characteristics while operating on such region are not so well defined. Would you think my concern is not justifiable and the change in internal die temperature would equalize naturally?

I'm guessing SCR or triac, it will pop out of conduction as soon as the line voltage drops below the capacitor voltage, due to lack of self holding current. Anything I'm perhaps forgetting?
 
Last edited:

There will always be a need for some ballast resistance in the sources to balance the current between individual mosfets, just as emitter ballast resistors are always required for individual BJTs when run in parallel.

At very low output current, balance is not important. But flat out it becomes of far more significance.

You cannot run your scr or triac straight into a reservoir capacitor, the current spike at turn on would be horrendous. There absolutely must be a choke after the rectifier to even out the current flow over the conduction cycle.

The problem then is that the current keeps flowing during and throughout the voltage zero crossing, and triacs in the transformer primary tend not to always reliably turn off.
Back to back scrs in the transformer primary, or used as rectifiers in the secondary are a far more robust and reliable solution to follow through conduction problem.
 

There will always be a need for some ballast resistance in the sources to balance the current between individual mosfets, just as emitter ballast resistors are always required for individual BJTs when run in parallel.

At very low output current, balance is not important. But flat out it becomes of far more significance.

You cannot run your scr or triac straight into a reservoir capacitor, the current spike at turn on would be horrendous. There absolutely must be a choke after the rectifier to even out the current flow over the conduction cycle.

The problem then is that the current keeps flowing during and throughout the voltage zero crossing, and triacs in the transformer primary tend not to always reliably turn off.
Back to back scrs in the transformer primary, or used as rectifiers in the secondary are a far more robust and reliable solution to follow through conduction problem.

Yes, I will use a choke rated for the current I need at the output to smooth things out. The SCR/TRIAC would be placed on the secondary arranged in a way that the freewheeling currents from the choke do not interfere with its operation.

Thank you for your thoughts. I will, on due course, make some tests with this and report back.

One last question:

Are linear regulators capable of cleaning the noise from a switcher placed before them? I'm thinking either a buck converter or even an active PFC stage would be good as pre-regulators but I'm concerned with noise that may propagate to the output. What I have in mind works at no more than 100KHz
 

A choke input filter will remove very high di/dt that the scr/triac would otherwise create. The reservoir capacitor should create a fairly smooth dc with only slight ripple at twice mains frequency.

If your linear regulator feedback loop is worth half a damn, it should very easily cope with that.

I would very strongly recommend you stick with SCRs rather than a triac.
The second gate is a nuisance, but a pulse transformer with two secondaries takes care of that problem.

All these words of caution come from several years of designing commercial high power phase controlled battery charging equipment.

Triacs are an instrument of the devil..........!
 

A choke input filter will remove very high di/dt that the scr/triac would otherwise create. The reservoir capacitor should create a fairly smooth dc with only slight ripple at twice mains frequency.

If your linear regulator feedback loop is worth half a damn, it should very easily cope with that.

I would very strongly recommend you stick with SCRs rather than a triac.
The second gate is a nuisance, but a pulse transformer with two secondaries takes care of that problem.

All these words of caution come from several years of designing commercial high power phase controlled battery charging equipment.

Triacs are an instrument of the devil..........!

Thank you, will do so.

I'm not too worried with the SCR switching at 100Hz, the concern I was asking was if I were to replace the SCR with a switching regulator of some sort as a pre regulator before the linear stage (just pondering options).

When you say SRC's are you suggesting me to use two as to replace the rectifier diodes? I was thinking in using one after the bridge before the choke, capacitor and freewheeling diode.
 

Thanks guys.


Question : Has anyone thought or seen a PS using a controlled triac rectifier? I'm pondering this to reduce overshoot after the load is removed and reduction in power dissipation.

Back in the 1970s and early 1980s, when the only real power semiconductors were either SCR or Triacs, yes they were commonly used as the regulating elements. I actually built a 1 Kw battery charger using these.
Powerline harmonics were terrible, though.
Later (and perhaps even today) they were used to bypass the series inrush-current-limiting resistor.
 

The best lab supplies in the 70's came from Lambda and they used SCR's with phase control to regulate the minimum dropout voltage required (aka pre-regulator) for the stage 2 Power Linear regulator.

This way the tradeoff between peak/avg pulse currents and related ripple voltage could be optimized for least power loss and lowest ESR in the power supply, making this lab supply almost ideal.

CC adjustments did not include the surge discharge of the output capacitor but there was active internal loading which could change the voltage faster in CC mode with step loads, but you could still create a good arc with short circuits being power limited.

You could also remote program these PSU's with 200 -2k Ohms per volt (I forget) as well as remote sense. I used an HP instrument board off HP-IB Resistor ladder as a DAC to control the voltage in xx mV steps from a remote desktop computer in 1977 { HP9825}, then setup a long distance RS.485'data network and used the,numeric keypad to either program the remote power supply or adjust it in small increments using (+\-) keypad then enter wouldmcapture the remote voltage on a long umbilical cable from payload on launchpad and display it instantly on the computer.'s 80'character LED display. Then I could select up to 10 PSU's and control each for different battery packs for each experiment had its own., It was then automated with upper and lower control limits and flagged on a CRT field and every key was mapped into a user programmable power relay control function with feedback to the remote control between two .HP9825's a mile apart..


Much later I realized this what we now call SCADA.
 
Last edited:

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top