Amplified(?) voltage limiter design (behavior at sagging input voltages)

Thread Starter

Saluki

Joined Feb 23, 2019
80
I'd like to be able to use a moderately varying source for 13 volts at well under 5 amps*, tapping from an automobile electrical system, and limit the output voltage to various LED modules to ~13.6 volts maximum. The load consists of LED modules I purchased which are evidently 3x 100 mA LEDS in series with a ~3 ohm resistor to somewhat limit current. However, powered off a variable DC power supply the current increases rather quickly at over 14 volts, and at 14.5 volts they get hot quite quickly -- In testing I removed the 14.5 volts quickly fearing the LEDs would "burn out". At 13 volts the LED modules are sufficiently bright and only a little warm.

*Note that "0" through at least 8 modules may be switched on at any given time. So, a constant current / LED driver doesn't work unless I want to feed each LED module with its own driver, after its switch or relay. :p

Note that below 13v, it is desirable to have the input to the modules follow the source (battery) voltage, as the weakening light output is a good indicator one is draining the battery too far. (I might even want to enhance that effect, but that's another circuit for a different discussion, I think. I could simply add a voltage display of the source, also, although that works against conserving battery power when the engine is off. In any event, a very sudden shutdown is undesirable.)

A simple zener-resistor "regulator" would "work", but is rather wasteful of the source power. I'm thinking a simple zener* + pnp-transistor or Darlington transistor (and a couple resistors) regulator would do the trick (circuits available online) but I am not sure of behavior as the source voltage falls UNDER the voltage needed to trigger limiting in the first place.

https://www.bristolwatch.com/ele/zener_power_supply.htm

I do NOT need "power supply for audio op amp" quality regulation. I just want to lop off the voltage to the modules when the "battery voltage" runs a little "hot", and follow the battery voltage otherwise.

I do not know how a PWM buck regulator will behave with sagging input voltage. The voltage differential is probably a "killer" anyway. A buck-boost regulator will draw more current from the battery at the worst possible time.

I'll add that one sees a great number of reviews on Amazon, etc., of LED lights and modules "burning out" in automotive and related use. I've tested a number and most* of the same lights and modules run fine (again, often just barely warm) at their nominal 12v rating, even up to 13v or so. I've never had one fail due to overheating if powered off a "12v" source that wasn't being also fed by a charger / alternator that could take the battery above 13.5 volts or so.

Thoughts? Will the zener+transistor regulator do the trick?

(Worst case, I can order parts, breadboard the thing, and see how it runs when I vary the voltage input. But discussions add to my and others knowledge base.) :)

Thanks!
 

crutschow

Joined Mar 14, 2008
34,428
A simple zener-resistor "regulator" would "work", but is rather wasteful of the source power. I'm thinking a simple zener* + pnp-transistor or Darlington transistor (and a couple resistors) regulator would do the trick
Any linear method to limit voltage will waste energy, but I don't see that as being significant in a vehicle application.
For 8 modules at 100mA each, the dissipated power would be about (14.5V-13V) * 800mA = 1.2W.

Below is the LTspice sim of an emitter follower regulator with a cheap TL431 voltage reference that may work for you.
The maximum output is about 12V, which can be varied by changing the value of R1, with the output being regulated at a Ref voltage of 2.5V.
The follower is a Sziklai pair, which has a lower in-out voltage drop (about 1V here) than a Darlington stage at low voltages, so it starts to come out of regulation at a battery voltage of about 13V.

You may prefer to avoid that 1V drop and have it regulate down to 13V input and output.
For that a slightly more complex LDO circuit should work, (bottom sim).
Note that the input and output voltages are now essentially equal below the regulation point, as determined by the MOSFETs on-resistance times the output current.

The P-MOSFET should be one in a TO-220 case to dissipate the power.

1707512841214.png

============================

1707515691501.png
 
Last edited:

MisterBill2

Joined Jan 23, 2018
18,491
It seems like a simple BUCK voltage regulator circuit will do best, to run the lights on 12.0 volts of maybe even 11.8 volts. And there should be such a DC input module available someplace. That scheme will allow switching individual LEDs as needed.
 

LowQCab

Joined Nov 6, 2012
4,071
The best arrangement is to get rid of the Resistors and install a Current-Regulator on each LED-Strip.
Otherwise, You can probably get away with Regulating the Voltage with a simple Circuit.
This Circuit needs to be mounted in an Aluminum-Box for Heat-Sinking.
The unmarked Resistors can be replaced by a 10K Trim-Pot.
Ask if You want FET recommendations.
.
.
.
Heavy Duty LDO Regulator .png
.
 

MisterBill2

Joined Jan 23, 2018
18,491
Quite probably the TS will not be needing to operate all of those LED light sources at maximum rated power. And certainly the 100 mA current listed is the maximum power for rated LED lifetime. The reduction in brightness at a reduced power level will be difficult to notice and not a problem. And my observation of automotive supply voltage is that with the engine running it is mostly closer to 13 volts than 12 volts, which will give a decent buck regulator plenty of headroom.
 

crutschow

Joined Mar 14, 2008
34,428
Otherwise, You can probably get away with Regulating the Voltage with a simple Circuit.
Your posted circuit generates limit-cycle oscillations in my simulation under certain transient conditions.
I was not able to eliminate them by adding compensating capacitors anywhere in the loop.
 

LowQCab

Joined Nov 6, 2012
4,071
I'm interested to know more .......

I've never heard the term "Limit-Cycle-Oscillations" used before.

It would seem that some overshoot, and similar anomalies,
would not cause any problems in this type of application.

Did your testing include removing the optional-Output-Capacitor ?
.
.
.
 

Thread Starter

Saluki

Joined Feb 23, 2019
80
Quite probably the TS will not be needing to operate all of those LED light sources at maximum rated power. And certainly the 100 mA current listed is the maximum power for rated LED lifetime. The reduction in brightness at a reduced power level will be difficult to notice and not a problem. And my observation of automotive supply voltage is that with the engine running it is mostly closer to 13 volts than 12 volts, which will give a decent buck regulator plenty of headroom.
Indeed, these modules appear as though they were really designed for ~13 volts. That rather makes sense, as if they were designed for 12v, the module would be running quite hot at the 13 plus-a-bit volts the car charging system typically charges the "12v" battery to. In fact, on occasion the car measurably gives me a rail slightly above 14.5v. (Subarus have a bit of a reputation for burning out even incandescent lamps early in their rated lifetimes, so, there may be a "connection" here. Yes, I know, bad pun! But our Chevy Tahoe runs nearly as high a voltage sometimes, and I've monitored it far less.)

There's actually a pretty good chance all modules may be on simultaneously at times, so, as a safety margin I really want to go with 2x, which would put me around 1.6 A as a design target.

The modules dim just enough to notice in most situations, between 13v and 12v, but not drastically*, (which is good in my case) but are still visibly emitting weak light at ~ 8-9v.

*Sort of like, when using a flashlight, about the time you think "Hmm, almost time for fresh batteries..."

Again, the question is what the buck regulator will do as source voltage decreases. You may misunderstand - I WANT to see modest or worse dimming if the source voltage decreases significantly below ~13v, but I don't want it to chop off abruptly, as it'd be too easy to assume a problem or reason other than the battery getting a bit low with the engine off for a while. I may not be the only user, either. I hasten to add that in 2024 "standard" 12v car batteries may not survive being deep cycled more than a few times, and even deep cycle marine batteries do NOT like being fully discharged very many times. I found that out the expensive and rather inconvenient way... :oops:
 
Last edited:

Thread Starter

Saluki

Joined Feb 23, 2019
80
Any linear method to limit voltage will waste energy, but I don't see that as being significant in a vehicle application.
For 8 modules at 100mA each, the dissipated power would be about (14.5V-13V) * 800mA = 1.2W.

Below is the LTspice sim of an emitter follower regulator with a cheap TL431 voltage reference that may work for you.
The maximum output is about 12V, which can be varied by changing the value of R1, with the output being regulated at a Ref voltage of 2.5V.
The follower is a Sziklai pair, which has a lower in-out voltage drop (about 1V here) than a Darlington stage at low voltages, so it starts to come out of regulation at a battery voltage of about 13V.

You may prefer to avoid that 1V drop and have it regulate down to 13V input and output.
For that a slightly more complex LDO circuit should work, (bottom sim).
Note that the input and output voltages are now essentially equal below the regulation point, as determined by the MOSFETs on-resistance times the output current.

The P-MOSFET should be one in a TO-220 case to dissipate the power.

View attachment 314857

============================

View attachment 314863
Those look promising. Does output voltage continue to track the input voltage below an input of 12v?
 

crutschow

Joined Mar 14, 2008
34,428
I've never heard the term "Limit-Cycle-Oscillations" used before.
It's generally non-sinusoidal oscillations, similar to the non-linear waveform from a relaxation oscillator.
A system can sometimes exhibit either one, depending upon the starting and operating conditions.
Some systems will look stable from a Bode plot but still exhibit limit-cycle oscillations under transient conditions.
This regulator seems to be one of those.
From Google--
Limit cycle oscillations at resonances
The stable limit cycles are of importance as they model systems with self-excited vibrations or oscillations. Systems with a stable limit cycle will, independent of the starting conditions, settle into a steady trajectory in the phase plane, where there is a balance between generation and dissipation of energy.

Did your testing include removing the optional-Output-Capacitor ?
Sim below:

1707586062235.png
 
Last edited:

Thread Starter

Saluki

Joined Feb 23, 2019
80
Any linear method to limit voltage will waste energy, but I don't see that as being significant in a vehicle application.
For 8 modules at 100mA each, the dissipated power would be about (14.5V-13V) * 800mA = 1.2W.

Below is the LTspice sim of an emitter follower regulator with a cheap TL431 voltage reference that may work for you.
The maximum output is about 12V, which can be varied by changing the value of R1, with the output being regulated at a Ref voltage of 2.5V.
The follower is a Sziklai pair, which has a lower in-out voltage drop (about 1V here) than a Darlington stage at low voltages, so it starts to come out of regulation at a battery voltage of about 13V.

You may prefer to avoid that 1V drop and have it regulate down to 13V input and output.
For that a slightly more complex LDO circuit should work, (bottom sim).
Note that the input and output voltages are now essentially equal below the regulation point, as determined by the MOSFETs on-resistance times the output current.

The P-MOSFET should be one in a TO-220 case to dissipate the power.

View attachment 314857

============================

View attachment 314863

Any linear method to limit voltage will waste energy, but I don't see that as being significant in a vehicle application.
For 8 modules at 100mA each, the dissipated power would be about (14.5V-13V) * 800mA = 1.2W.

Below is the LTspice sim of an emitter follower regulator with a cheap TL431 voltage reference that may work for you.
The maximum output is about 12V, which can be varied by changing the value of R1, with the output being regulated at a Ref voltage of 2.5V.
The follower is a Sziklai pair, which has a lower in-out voltage drop (about 1V here) than a Darlington stage at low voltages, so it starts to come out of regulation at a battery voltage of about 13V.

You may prefer to avoid that 1V drop and have it regulate down to 13V input and output.
For that a slightly more complex LDO circuit should work, (bottom sim).
Note that the input and output voltages are now essentially equal below the regulation point, as determined by the MOSFETs on-resistance times the output current.

The P-MOSFET should be one in a TO-220 case to dissipate the power.

View attachment 314857

============================

View attachment 314863

Out of curiosity, what do you see as the biggest disadvantages of the simple zener-Darlington transistor circuit, in my application?

1707586051386.pngNot that my input will be 18v!
 
Last edited:

crutschow

Joined Mar 14, 2008
34,428
Out of curiosity, what do you see as the biggest disadvantages of the simple zener-Darlington transistor circuit, in my application?
  1. The output voltage value depends upon the Zener voltage tolerance, which is typically ±5%.
  2. Due to drop of the emitter-follower Darlington stage, the output voltage can be no closer than 2V below the input voltage.
Out of curiosity, what don't you like about the circuits I posted?

Small note: When you reply post has a quote, do your reply outside of the quote so it's separate.
 

Thread Starter

Saluki

Joined Feb 23, 2019
80
  1. The output voltage value depends upon the Zener voltage tolerance, which is typically ±5%.
  2. Due to drop of the emitter-follower Darlington stage, the output voltage can be no closer than 2V below the input voltage.
Out of curiosity, what don't you like about the circuits I posted?

Small note: When you reply post has a quote, do your reply outside of the quote so it's separate.
And I just now went back into "edit" but nothing I do seems to fix it. I guess I could delete and repost.

In answer to "Out of curiosity, what don't you like about the circuits I posted?"

They look good to me! Particularly the 2nd circuit (at a bit higher parts count.) I'm just trying to better understand how each approach compares, what the limitations are, etc.

Oh, BTW, I'm still curious what your circuits do when the input falls under 12v. An approximate answer is probably fine. :)

Thanks!
 
Last edited:

LowQCab

Joined Nov 6, 2012
4,071
It's generally non-sinusoidal oscillations, similar to the non-linear waveform from a relaxation oscillator.
A system can sometimes exhibit either one, depending upon the starting and operating conditions.
Some systems will look stable from a Bode plot but still exhibit limit-cycle oscillations under transient conditions.
This regulator seems to be one of those.
From Google--
Limit cycle oscillations at resonances
The stable limit cycles are of importance as they model systems with self-excited vibrations or oscillations. Systems with a stable limit cycle will, independent of the starting conditions, settle into a steady trajectory in the phase plane, where there is a balance between generation and dissipation of energy.

Sim below:

View attachment 314925
.
Thank You for taking the time to explain your point,
but I'm reasonably sure that even a small Output-Capacitor will
completely smooth out a roughly ~200kHz oscillation,
which is also definitely dependent on the accuracy of the modeled Components
and could easily change based on various outside influences, varying-loads, temperature, etc.

There's certainly more than one way to skin a cat !!!

Up in the ~$30.oo Project-Price-Range,
I would personally be leaning towards a
Switch-Mode-Regulator if the project must be "All-In-One".

I'm still thinking that the best solution is individual Current-Regulators on each Load,
this spreads-out the already much lower heat that must be dissipated to almost nothing,
and, Voltage would "sag" a little bit earlier, which is something that he Thread-Starter requested.
.
.
.
 

Thread Starter

Saluki

Joined Feb 23, 2019
80
.
Thank You for taking the time to explain your point,
but I'm reasonably sure that even a small Output-Capacitor will
completely smooth out a roughly ~200kHz oscillation,
which is also definitely dependent on the accuracy of the modeled Components
and could easily change based on various outside influences, varying-loads, temperature, etc.

There's certainly more than one way to skin a cat !!!

Up in the ~$30.oo Project-Price-Range,
I would personally be leaning towards a
Switch-Mode-Regulator if the project must be "All-In-One".

I'm still thinking that the best solution is individual Current-Regulators on each Load,
this spreads-out the already much lower heat that must be dissipated to almost nothing,
and, Voltage would "sag" a little bit earlier, which is something that he Thread-Starter requested.
.
.
.
I might pick up a single current regulator just to experiment with the behavior with sagging input voltage.. Small ones seem to be pretty cheap, and, if not here, an application for the little beastie is sure to turn up...

Thanks for the input! (Another unintended pun - hahaha!)
 

Thread Starter

Saluki

Joined Feb 23, 2019
80
The outputs will continue to follow the input down to a low voltage.
Sims below:

View attachment 314932

View attachment 314933
Ah, thanks. VERY interesting / useful. Even the circuit that tracks the input with a roughly -1v difference in the output, up to the point where limiting kicks in, could be handy either:

1) In a situation with high ambient heat, to protect the LED module at the expense of a little less light, or;

2) With a different module / LED light that already runs rather hot at 13 volts. (In fact I have some uninstalled fog lamps that do just that. Granted they also came with just a little streak of heat sink grease between the Aluminum housing / heat sink and the (aluminum?) backing of the LED board.) So many Chinese products seem to be this way - not a bad design, but horrible / slap-dash execution in production...

Off to the web to roust up parts. And I'd just ordered some TIP31c's... :rolleyes:
 
Top