Linear dropper - 28 to 12V, 1A cont, 5A peak - design choices?

[max_torque]

I have a current system that uses a switch mode converter to drop 24 to 36V (28V nom) to 12V.   Average current draw is just 1A, but peaks of up to 5A can be pulled for up to 10sec (fairly non-repetitive).  I need to replace that switch mode convertor with a linear dropper  (for reasons too long winded and complicated to explain here ..... lol)

So, for a linear drop at 1A that's 16W, at 5A it's a massive 80W  yikes!

I would estimate i need to budget an average thermal solution for dissipating around 20W   (170sec at 1A, 10sec at 5A)

As this looks to require multiple pass devices (difficult to get a 20W average dissipation from a single package unless using a huge old TO-3 style device (possible, but ugly!) are there any advantages between a series or parallel type arrangement of pass devices?  ie, drop a few volts at a time, or drop all the volts in one go, but with the current shared between devices??

[jeroen79]

What about using a switched converter to make just above 12V and then drop the remainder to 12V?

[max_torque]

Sorry, has to be completely linear solution for RFI/EMC reasons.

Multiple pass devices will help by both current (or voltage) sharing and adding extra thermal inertia to the system, allowing for the short term high current drain.

I could also use a purely resistive dropper as a first stage, but sizing that would be critical (obviously it would have to pass the 12v 5A requirement at the lowest input voltage i need to support (~18V) which means it can only be 1ohm max, so it can never take the lions share of the load off the pass devices........)

[max_torque]

It looks like a single TO-247 device would be within it's SOA, something like a TIP145:

http://www.farnell.com/datasheets/1662102.pdf

Probably have to drive the base with another smaller buffer amplifier of some sort or other

[Kleinstein]

Even from 36 to 12 V the voltage is relatively low. So there is usually no real problem with the SOA. Series connection might make load sharing easier and with higher voltage ease on the SOA part.

A single TO247 transistor with a series resistor should be enough. With a Darlington type is might not need an extra driver. The regulator circuit might like to have some of the resistance at the emitter, especially if there is a difficult more capacitive load to it.

[Zero999]

What's wrong with the usual LM7812 or LM317 with a large enough pass transistor?

[DBecker]

I would suggest dropping in two stages.  The first stage should drop to 13V-14V using several devices and load-balancing resistors.  The final regulation should be with a single device on its own heatsink, kept as cool as possible.  Select the final regulator based on the noise profile and regulation accuracy that you need, without being overly concerned about stability or thermal noise.

[max_torque]

Luckily i don't have any tight requirement on the regulation, as the loads will typically have their own internal voltage regulation (as they tend to be automotive type loads that run on just about anything between 10 to 15V for example)

[floobydust]

I don't see any current-limiting, which the SMPS probably has.

If you ever might have a shorted load, then the SOA gets exceeded and the pass transistors(s) fail.
One transistor can handle it in a perfect world, but you need something much stronger IMHO.

[edavid]

I could also use a purely resistive dropper as a first stage, but sizing that would be critical (obviously it would have to pass the 12v 5A requirement at the lowest input voltage i need to support (~18V) which means it can only be 1ohm max, so it can never take the lions share of the load off the pass devices........)

If you get advance warning of the 5A peaks, you could use a resistor dropper that is shorted by a relay or transistor.

[rob77]

don't forget that your source has to provide the full 5Amps with linear regulator ! the buck converter did a conversion so your 5Amps @12V output was more likely something around 3Amps @ 28V at the input... with linear regulator you have to have a 5A capable source on the input.

[Ian.M]

It may be best to bite the bullet and go to a full LM723 based regulator PSU  with multiple external pass transistors, to get current limiting so you can protect the pass transistors.   Add a thermistor on the heatsink and a circuit to pull down the compensation node if it overheats and you've got thermal protection as well.

[Zero999]

There's the LM338 which is rated to 5A but it won't be able to do so with such a high a voltage differential. As mentioned above, you could drop the voltage before it, to just over 3V above the nominal output voltage, with a series resistor or power zener diode.

[Pitrsek]

linear solution for RFI/EMC reasons.

Can you elaborate on RFI/EMC reasons? Once you close your stuff into a properly designed metal chassis, and have decent input/output filter, you usually do not have RFI/EMC issues.

What is your allowable voltage output noise?  Do you have problem with magnetic coupling?
Actually I think if you elaborate on the long winded reasons, you might find out that there is some completely different solution.
 

[SeanB]

Typical is a long ( 300mm long) aluminium channel heatsink, with a potted section on the one side, containing the 3 parallel  TIP145 devices, the associated emitter ballast resistors, the small PCB with  the 14V 1W zener diode on it, the thermal cutout ( 70C clixon one), the tiny driver transistor for the 3 pass devices, the reverse polarity protection diodes ( one input, one output and one across the regulator for reverse connection) and some 10uF 50v ceramic capacitors input and output for noise absorbtion, all on a small PCB. Put the 1R 25W metal clad resistor on the heatsink as well, and then put a 40V 5W transorb after the resistor, to handle the load dump voltage surges. Output also put a 20v transorb, for the same reasons, you can buy the 20V device only and stack 2 for the input side, higher power and single part inventory.

You can make the PCB single sided and long, so all the components aside from the thermal cutout are mounted on it, and then use the transistor mounting kit screws to hold it in place. Put the 2 input wires, the 2 output wires and then pot is a soft silicone potting for protection.

Thees are quite common as truck radio adaptors, and designing for 5A output and then using at 1A will be a lot better than hoping the 1A design will survive 5A long period. You have to pot them, as they are going to be exposed to some pretty crappy environments in use, and put some strain relief holes and a tiewrap on input and output sides, along with putting in a provision for an inline 10A fuse for them, in a sealed automotive blade fuse holder.

Bonus points for putting in a relay for switching, that will accept input of 12 to 24V as a turn on signal.

[max_torque]

Output is switched through an Infineon smart high side switch that provides short circuit protection etc, so i don't need anything natively in the pass elements.

Main think is to ensure i have sufficient average heat rejection capability with the necessary short term capability, which is the tricky bit, because a 5 to 10 sec rating isn't really counted as intermittent.

Also, i would like a surface mount solution, which pretty much means multiple pass elements i suspect, as i can't really see any SMC package providing at 20w dissipation?

[Ian.M]

Worst case 36V in.
Highest safe output 14V
Peak current 5A

That gives a voltage drop of 22V and a peak dissipation of 110W

You want to do this all in SMD? <ROTFLMAO>
Either you'll need a massive number of parallel pass devices and acres of PCB area to allow for enough copper to get the heat out of them or you'll have to use an aluminum core PCB closely fitted to a heatsink.

There are times when through hole is more appropriate, and when you need the compliance to clamp multiple power devices to the same headsink is one of them . . . .
By all means make the rest of the board SMD, but IMHO the pass transistors should be a row of paralleled TO-247 or similar package devices.

[mariush]

LM1084 can do 5A guaranteed and most have the over current threshold somewhere around 5.5-6A... the downside is absolute max voltage 29v. 

So you could put 3-4 of these  on a large heatsink, and have every one of these drop some amount of voltage... you'll spread the heat over multiple regulators. If the voltage is lower than the set output voltage, they'll still output  input voltage - 1v

So even with with 18v minimum input voltage, you'd have 17v after the first, 16v after the second , 15v after the third, 12v on the fourth regulator.

These should also be available in to-252 or other smd packages. 

[Zero999]

If a 5A regulator is used, another transistor can be added, as a pre-regulator, see thread linked below:
https://www.eevblog.com/forum/projects/lm317-pre-regulator-for-wide-output-voltage-range/msg1210184/#msg1210184