Linear voltage regulator design

[Simon]

I'm going to design a linear voltage regulator for a project I am working on. I have loads in the system that have 40V mosfets in them and this is a "nominal" 24V automotive system so basically *I have to deal with surges of 40V. So I plan to use a linear voltage regulator that will try to regulate to 30V, I expect usually to see no more than 28V so most of the time my regulator should just be on full throttle passing all of the current through and only actually regulate should the voltage exceed 30V, so by doing this i hope to reduce any heat dissipation (from 2x7.5A) and only have to dissipate the least possible with a standard auto supply spec expecting me to be able to live indefinitely at 33V and survive 500mS  40V surges every 5S (test case) for which I hope heat sink thermal inertia will do nicely.

I plan to use a PNP darlington as the main pass transistor as using an NPN would mean i contently have to deal with a dropout voltage: http://docs-europe.electrocomponents.com/webdocs/11a6/0900766b811a658c.pdf I'd pretty much use the text book design of an op-amp comparing the output to a reference and driving a small NPN transistor that drives the base of the power PNP pass transistor.

http://www.simplecircuitdiagram.com/2010/05/19/the-low-dropout-ldo-regulator/
As most op-amp come 2 in a package I can build two regulators so halving the power dissipation. Any gotcha's ? Any particular op amp of choice or is a run of the mill one good?

[T3sl4co1l]

So.... wait, what output voltage and current?  And what input?

You want a 30V LDO to provide ride-through for... all loads (2 x 7.5A)?  Not just for the sensitive little control stuff?

What's wrong with running everything at (transient peak up to) 40V?  Tons of stuff is available for that.  As you say, your transistors already handle 40V.

Are the loads 7.5A continuous all the time, or intermittent?  A PNP transistor would be rather inefficient for either case.

A complete off the shelf solution would be LTC4380, https://cds.linear.com/docs/en/datasheet/4380fb.pdf slam dunk no need to think about it, application circuit provided.  Well worth the price of an LT chip!

Tim

[jaycee]

Automotive is generally hell.. as you correctly say it's nominal 24V but load dumps and alternator spikes can produce pulses of many times that.
There are specialist regulators designed to deal with all this horror which you might be best using to ensure a steady 24V supply.

Of course there's other techniques such as using TVS diodes to protect the input, input chokes, and such. I read a really good application note on all this stuff once, but unfortunately I can't remember what it was called!!

[Simon]

Of course there's other techniques such as using TVS diodes to protect the input, input chokes, and such. I read a really good application note on all this stuff once, but unfortunately I can't remember what it was called!!

This not for your typical low energy 250V spike, this is for consistent supply voltages within the norm that the load cannot handle so i have to be prepared to dump 45W continuous. The pass transistor is rated for 120V so that a TVS will easily protect it.

A 36V TVS will have a reverse standoff voltage of 33V (normal operating condition), the actual cut in voltage will be anywhere between 36 and 39V and maximum clamping voltage could be as high as 48V so nope TVS's are goo but not that good. I dare not take my reverse stand off voltage anywhere near the normal operating voltage or I would just crowbar the supply, blow fusees and generally not make people very unhappy.

the chip Tim suggests might work but then it's pretty much just taking me back to rolling a power regulator. I'm not just looking at an "active TVS" here, I will potentially have to drop 3V non stop at 15A (split two ways)

[Simon]

A complete off the shelf solution would be LTC4380, https://cds.linear.com/docs/en/datasheet/4380fb.pdf slam dunk no need to think about it, application circuit provided.  Well worth the price of an LT chip!

Tim

Unfortunately it appears to have a 250KHz charge pump, this may cause EMC issues as this one is pretty sensitive,

[Scrts]

What is your sensitive area in terms of EMC? Maybe you can use 2MHz DC/DCs?

[Simon]

Vary sensitive, No switch modes can be used at all. As it is i am hoping that this linear reg will help take out some of the noise the load makes and I have a further linear reg after it that supplies a load known to be a bit noisy due to it's DC/DC but if I supply voltage just under it's output it will just act as a pass through and "go quiet"

[T3sl4co1l]

No, how sensitive?

SMPS are used in the world's most sensitive instruments.  It's not an exclusive problem, it's a "git gud" problem.

I'm working on such a project right now; would you like to do an NDA and a few hours consultation to work it out?

Tim

[Simon]

No, how sensitive?

SMPS are used in the world's most sensitive instruments.  It's not an exclusive problem, it's a "git gud" problem.

I'm working on such a project right now; would you like to do an NDA and a few hours consultation to work it out?

Tim

Tim

Thank you for the offer but this is for my employer so we are just keeping it simple (or it needs keeping simple for them before we all freaqk out ). We have had problems with anything that wiggles a current so are playing this safe. Things would also quickly get tricky needing all parts of the SMPS to work at up to 120V. As things are currently planned I can use a zener regulator for my control circuitry that can be easily design to withstand any spikes thrown at it and everything else will sit behind the darlingtons at no more than 30V.

[T3sl4co1l]

You're switching motors, right?  Is it extrememly bad for them to start turning during such a rare event as this?

In other words, turn the switches on so all the voltage drops across the motor.

Tim

[Simon]

No I'm not switching them on and off I'm looking to do a linear regulator to hold the supply at 30V so that they can happily work away whilst the regulator takes up any extra. So my concern is a surge, that is 40V for 500mS every 5S. The regulator should take care of this, as the transistors are rated for 120V a suitable TVS can clamp the low power 250V spikes and if in clamping it gets to 50V that is not a problem whereas the motors I am told have 40V mosfets in them.

[T3sl4co1l]

Every 5s?  (S is siemens, BTW)

What the hell would do that?

Load dump is a one-time event.  It's so uncommon that it hardly seems worth testing at all (and, probably, a lot of aftermarket stuff isn't), but it's always done for proper OEM hardware, just in case.

Now I'm even more confused what you're doing here...

Tim

[Simon]

Right, the motors (brush-less with builtin drive electronics) have mosfets with a maximum Vds of 40V, the manufacturer would really like to see them run on 20V (50% derating), they don't mind 24V but I'm happy with 30V, realistically in a vehicle 27V is quite likely.

But there can be surges of 40V for 500ms every 5s from a source impedence of 0.5ohms. This will break the motors and is not the sort of thing a TVS or other crobaring device would have to absorb too much power. So the only way out I see is pre regulate the power supply. So normally I will just pass all the power through with minimal drop but should the voltage exceed 30V the regulation will kick in and hold down the supply to the motors at 30V. It is perfectly acceptable by the power supply standard that the "battery" voltage can be 33V permanently so i could see myself having to drop a constant 3V as well as absorbing the extra excursion to 40V, yes this is an occasional occurance but it only takes one to blow the motor drivers.

[T3sl4co1l]

But there can be surges of 40V for 500ms every 5s
...
yes this is an occasional occurance but it only takes one to blow the motor drivers.

0.2Hz is not what I would call "occasional".  It sounds more like you need a switching [pre-]regulator...

Tim

[Simon]

Like i say we can't use a switchmode. The test is 0.2Hz (10% duty) like you say realistically it won't happen very often but for test purposes it has to satisfy that. At the end of the day it equates to 1 extra continuous volt which is fine.

[Zero999]

I'm going to design a linear voltage regulator for a project I am working on. I have loads in the system that have 40V mosfets in them and this is a "nominal" 24V automotive system so basically *I have to deal with surges of 40V.

Right, the motors (brush-less with builtin drive electronics) have mosfets with a maximum Vds of 40V, the manufacturer would really like to see them run on 20V (50% derating), they don't mind 24V but I'm happy with 30V, realistically in a vehicle 27V is quite likely.

I don't see the need for a regulator or any surge suppression.

The MOSFETs are rated to 40V, therefore they will be able to withstand 40V spikes, possibly even a little more, for very short periods of time. I agree, that de-rating is desirable for reliability reasons and that it's bad practise to continuously subject components to their absolute maximum ratings, but this is a stress test, rather than normal operation.

I think adding more components, such as a low drop-out regulator, would decrease the reliability of the system, as a whole, as there are more components to fail. Low drop-out regulators can also be a pain to design and can sometimes oscillate, under certain load conditions.

But there can be surges of 40V for 500ms every 5s from a source impedence of 0.5ohms. This will break the motors and is not the sort of thing a TVS or other crobaring device would have to absorb too much power.

I don't see how that can break the motors, which will be able to withstand much higher voltages, for short periods, without damage.

As far as clamping is concerned, it's doable, with a zener diode and a big enough transistor. I've done a similar thing before to absorb the excess voltage generated by a motor breaking. I used a Darlington pair and a 30V zener diode. I doubt it's worth it in this case though.

[Simon]

The problem here is that I need to work a continuous 33V and i really don't want to subject the motors to more than 30V and crowbaring is not an option, but I did want to avoid any extra regulation at 27V so was thinking of a low drop out. I could use an NPN and zener, that was my original intention, but there will need to be some voltage between the base and collector in order to drive current though the base. Tests done for another project show that this could be a little less than 4V so it's sort of doable but i hoped to only have to make a load of heat if actually necessary. Dropping 4V at 27V will take the motors down to 23V, we get the performance we want with 24V and any further drop due to a drop in vehicle supply will cause even lower supply voltages. This would not be a problem for anything other than not meeting the nominal specs which in this case is not really a problem if there are overriding reason not to. with increased supply voltage the motors will need less current for the same speed. If I always dump heat and have a lower supply voltage with increase current demand then my heat-sinks need to be twice what they already are (88x50mm base 38mm tall), If I make this too unwieldy someone will just throw the baby out with the bath water.....

The reason I can't just expect these to work at 40V is that I still need to protect for spikes, if I use a TVS this will either allow over 40V across the motors or just short the power supply as the reverse standoff will be too low, I have but 7V to play with.

Any particular condition LDO regulators are unstable? small loads maybe where the control loop over compensates ?

[macboy]

Here is a regulator for your consideration. It will limit the output voltage to 30 V, has less than 1 V of dropout with a 15 A load, and has an automatic shut-off if the input voltage exceeds a safe threshold, to protect the pass transistors from cooking.

Q2 drives the output stage by pulling current through the PNP darlington transistor bases. Two darlington transistors (simulated with two discrete devices each) are used to handle the high power and to reduce the dropout voltage. R4 provides base current (into node Vdrive) to turn on Q2. C3 and R5 (from node Vdrive to ground) are necessary for stability. Both values can and should be adjusted based on actual operation. R5 can be increased quite a lot, 1k works well in simulation. (edit: or does it?)

Zener D1 drives Q1 for voltage regulation. Basically if the output voltage exceeds the D1 zener voltage (plus Vbe of Q1) then Q1 starts to conduct, drawing current away from Q2 (at node Vdrive) which reduces the output. Line and load regulation are surprisingly good for such a simple circuit. 

Similarly, D2+D3 drive Q5 for over-voltage protection. It activates around 44 V with the values shown, turning off the output to protect the pass transistors from excessive dissipation. C1 provides a filter/delay which prevents a short spike from turning off the output.

Either or both Zeners and their transistors could be replaced with a TL431A for better precision and adjustability. Transistors were chosen from LTSpice's library and are not considered optimal or even appropriate for real implementation. Use output capacitor of > 100 uF for good stability (fast transients on the input can cause instability with the 10 uF value used in the asc file attached).

The sim output graph shows:
- nominal input voltage of 27 V giving <1 V dropout at >15 A output, then
- a 40 V surge for 500 ms, output limited to 30 V, then
- a long 55 V spike which is limited to 30 V then shuts down the output, then
- a short 55 V which is limited to 30 V but does not shut down the output, and
- low input voltage still resulting in <1 V dropout.

No warranty expressed or implied. This is just FYI. LTSpice asc file in the zip.

[Simon]

Thank you, pretty much what I was looking at with an op amp in the feedback instead of the transistors. Would a fully descrete version be more stable? Given the low amount of components required I was just going to make two parallel regulators to make then more fail safe (1 regulator per motor) which is completely unnecessary in this case...

I'm not planning an over voltage cut out, I expect the regulator to ride it out which is the whole point of having it. with a new input of 120V instead of 40V I can put in a TVS with a reverse stand off voltage in excess of 40V.

[macboy]

My discrete version is quite simple, so fewer things to cause instability. A drawback to the simplicity is its low precision: dependent on the zener voltage with no way to adjust it, etc.  Using the TL431A could improve that but would also change the loop gain and phase, and stability would need to be reassessed. The over-voltage protection was an afterthought and is entirely optional. Just remove everything in that box if you don't want it.

[Simon]

well I was going to use a zener as a reference anyway but this seems to rely on the amount of current through the zener altering the voltage across the resistors to control it. More volts out, more volts across the resistors, more current through q1 that tries to turn q2 off.