voltage regulation question

[MrAl]

Hello there,

Sounds to me like something isnt hooked up
correctly. I've included an attempt at
making a text schematic of the circuit.
Check it over carefully and let us know what
you find out.

</font><blockquote><font size="1" face="Verdana, Arial">code:</font><hr /><pre style="font-size:x-small; font-family: monospace;">_____________________
| |
| e
| -------b[ Q1
+ | c
BATT | |
- ---C--R--|
| | |
| | +
| | LS
| | -
| + |
| LM334 [ ]s------|
| - |
| | |
| | Rset
| | |
| | |
----------------------</pre><hr /></blockquote><font size="2" face="Verdana, Arial">Good luck with your LED circuits,
Al

 

[star882]

What about the classic 2 transistor circuit?
</font><blockquote><font size="1" face="Verdana, Arial">code:</font><hr /><pre style="font-size:x-small; font-family: monospace;">
+ O--- +LS- +-+
| |
> |
470 < |
> |
+< TIP31
| |e
2n2222 >-+
e| |
| >
| < 2(1w), 1(5w)
| <
- O---------+-+</pre><hr /></blockquote><font size="2" face="Verdana, Arial">

 

[MrAl]

Hello there,

starr882:
That's a nice little circuit there.
I found that using a TIP31 required too
much battery voltage to operate at upwards of
300ma or so, but that using a zetex transistor
like the 617 line worked much better. The
output current at low batt (like 3.7v) was
still a bit low though, so i disconnected one
arm of the 470 ohm resistor and connected that
directly to the batt + supply. This gets you
down to about 3.8v with still decent output
current.

I dont know if i would want to use this
circuit with the three battery power supply
(cant really get away with it when you use
three nicd's), but with higher voltage supplies
this is a really nice and low cost circuit.
It's very comfortable with a 6v power supply.

The LM334 designs were aiming at very low
voltage drop when the battery voltage starts to
sink down. Under these conditions, the LM334
design will work down to about 0.1 volt above
the LS drop, which ends up in the neighborhood
of about 3.4 volts.

Here's the circuit the way i modified it:

</font><blockquote><font size="1" face="Verdana, Arial">code:</font><hr /><pre style="font-size:x-small; font-family: monospace;">+ O-+- +LS- ----+
| |
|------- |
> |
470 < |
> |
+--< Zetex 617 line
| |e
2n2222 >---+
e| |
| >
| < 2(1w LS)
| <
- O---------+---+</pre><hr /></blockquote><font size="2" face="Verdana, Arial">Note i didnt test it for use with the 5w LS.
I also didnt test the highest voltage that
could be used before the 617 transistor
overheats.

Good luck with your LED circuits,
Al

 

[tall paul]

So far no dice. i've checked the setup a zillion times now and have etched the circuit board a couple times just for fun.
changing the cap changes the output current significantly, but i still don't get what i want, the higher i make the capacitance, the higher the output. doesn't seem right.
i was planning on making a spice model, both as a check and because it sounds interesting. i downloaded switchcad 3 and added MrAl's luxeon spice models. i was wondering if anyone had the spice info for the 717 transistor and the lm334.
again, all thoughts and comments are welcome.

 

[Doug S]

Originally posted by tall paul:
So far no dice. i've checked the setup a zillion times now and have etched the circuit board a couple times just for fun.
changing the cap changes the output current significantly, but i still don't get what i want, the higher i make the capacitance, the higher the output. doesn't seem right.
i was planning on making a spice model, both as a check and because it sounds interesting. i downloaded switchcad 3 and added MrAl's luxeon spice models. i was wondering if anyone had the spice info for the 717 transistor and the lm334.
again, all thoughts and comments are welcome.

<font size="2" face="Verdana, Arial">Paul, are you using the version of this circuit as diagrammed by Mr Al four posts above? If so, what R and C values are you using? Also, did you add the additional resistor to the LM334 +terminal as I suggested in my 12/13 post and if so, what value did you use?

 

[tall paul]

Doug,
Yes, I'm using the circuit exactly as MrAl described. i started out with R and C values of 100ohms and .01uF but have changed both values up and down. I haven't installed a resistor to the 334 + terminal yet, i'll give that a try tonight and report back.
Thanks,
Paul

 

[Doug S]

Originally posted by tall paul:
Doug,
Yes, I'm using the circuit exactly as MrAl described. i started out with R and C values of 100ohms and .01uF but have changed both values up and down. I haven't installed a resistor to the 334 + terminal yet, i'll give that a try tonight and report back.
Thanks,
Paul

<font size="2" face="Verdana, Arial">I haven't had success with C<0.1uF. Other possible sources of problems: If using the SO-8 package, NSC makes it with two different pin assignments. If using the SGS Thomsen SO-8, it has NC [no connection] to some of the pins that NSC connects to. You are using 3 cells for a power source. Any chance of a bad cell? It would be informative to check the circuit while operating by measuring the battery voltage, and drops across the transistor, sense resistor, and LED.

 

[tall paul]

i know the all the cells are good. I've been monitoring the voltage across the led and Rsense, that's how i've been measuring the success or failure of the ciruit. I've had the most success when i've increased C, currently i have .2uF and the voltage across Rsense is .067, so it's a little high now, but that could be just an issue of my volt meter not be 100% correct.

 

[Doug S]

Originally posted by tall paul:
i know the all the cells are good. I've been monitoring the voltage across the led and Rsense, that's how i've been measuring the success or failure of the ciruit. I've had the most success when i've increased C, currently i have .2uF and the voltage across Rsense is .067, so it's a little high now, but that could be just an issue of my volt meter not be 100% correct.

<font size="2" face="Verdana, Arial">Paul, it sounds like you may have it working OK. That 0.067V across the sense resistor is well within the datasheet tolerance on the 0.064V sense voltage as the tolerance is +/-8%. If you have a variable power supply, you could get added assurance that its working OK by varying the input voltage and confirming that the current remains relatively constant. I do encourage you to add the other resistor to prevent the LM334 from being damaged when the batteries run down. It also will improve the efficiency when operating at low battery voltage as it prevents excessive current from bypassing the LED.

 

[tall paul]

thanks for the advice doug, i'll let you know what i find out tonight.

 

[tall paul]

ok, all is good. i switched to a set Nimh AA's and the voltage across the sense resistor remained at .067 volts. the final setup was c1=.2uF, R1=100ohms, and Rsense=.14ohms. i plan on putting a resistor on the 334 + terminal tonight.
thanks for all the help.

 

[Doug S]

Originally posted by tall paul:
ok, all is good. i switched to a set Nimh AA's and the voltage across the sense resistor remained at .067 volts. the final setup was c1=.2uF, R1=100ohms, and Rsense=.14ohms. i plan on putting a resistor on the 334 + terminal tonight.
thanks for all the help.

<font size="2" face="Verdana, Arial">Pleased to help. This is a great circuit. I am surprised that it is not more widely used.

 

[MrAl]

Hello again tall paul, Doug S,

Looking at the circuit again, im in favor
of trying a resistor at the v+ terminal of the
LM334 also. Something like 200 to 500 ohms.
The smaller the cap the better too, as this
helps reduce input surge during turn on.

What may be happening is that since the
be diode is in series with the LM334, the
LM334 gets the full supply voltage during
turnon, which isnt good at all. This might
force it out of regulation and 'latch' the
v+ terminal high. With an increasing cap
value, the cap momentarily 'shorts' out the
LM334, allowing it time to start regulating
normally.

BTW, zetex publishes the spice data on the
717 transistors, but since i couldnt find
a model for the LM334 i had to create one,
which i didnt match up the gain for (as published
in the data sheet). Even so, it does work
better then nothing so if you still want to
try the simulation ill post it.

Good luck with it, and let us know how it
works out ok?

Take care,
Al

 

[Doug S]

Al, I don't think that any brief startup transient is an issue considering that he is using 3 alkaline cells as a power source. Many modders direct drive Luxeons with 3 cells [not to imply that I endorse this]. I've worked with the LM334 quite a bit and have never seen the latch-up behavior that you speculate on. Rereading what you've written, I suspect that it doesn't really say what you intended: the only diode in series with the LM334 is the base-emitter junction of the transistor and so what if the LM334 sees full supply voltage, its rated to 40V.

 

[MrAl]

Hello again tall paul and Doug,

paul:
I agree that there must be something more basic wrong in your setup.

To find out what is going on, you'll have to perform a few basic
tests which shouldnt take more then about 10 minutes. What we have
to do is connect a circuit up and make a few measurements. This
will tell us what is going on and why it's not working right.

First, we will connect the most basic circuit to make sure
your LM334 is operating properly to begin with, then we
will proceed to the same kind of circuit we wish to use
in the end.

Referring to page 8 on the data sheet:

at the bottom of the page, find the circuit labeled
"Basic 2 terminal current source".

using a +5v power source...(which can be three fresh alkies in series)

1. connect this circuit using Rset=10 ohms.
2. connect a 100 ohm resistor (we'll call this Rx) between +Vin and +5v.
3. connect -Vin to ground.
Now, apply power and measure the voltage across Rset and the
voltage ACROSS Rx and record the results and post them here.

Next, connect the "Higher output current" circuit
(on page 8, the circuit we have been looking at here)
using a 2N4003 or 2N2905 or similar transistor, then:

1. use the same Rset (10 ohms)
2. connect the 100 ohm resistor (Rx) between Rset and the collector
(this goes in place of where the LS was intended to be connected).
3. connect -Vin to ground, +Vin to +5v.
4. dont connect R1 or C1.
Now, apply power and measure the voltage across Rset and the voltage across Rx
and record the results and post them here.

Once we get these numbers in, we can tell what's wrong.

Take care for now,
Al

 

[tall paul]

Hi Doug and MrAl,

I had set aside the project for a couple days mostly due to lack of time and a little frustration. This weekend i made some progress though. i hooked up my o-scope and went to town. i've built a couple ciruits so far, one meant to deliver 325ma for a flashlight and one at 450ma for a bike light. the 425ma was working pretty good, meaning that the voltage across the Rsense was within spec (the .067V that i reported earlier). the 325 ma circuit wasn't fairing so well, i increased the cap like doug suggested but it still wouldn't pan out. after hooking up the o-scope, i could tell that the circuit was oscillating. if i would short the cap, the oscilallation would stop and the circuit would regulate as designed. after the short was removed, the circuit would continue to act fine (regulate with no oscillations). if the switch was turned off then back on, the oscillations would return. so, i figured that initial turn on condition was the problem, sending the ciruit into oscillation. i decided to put a cap across the battery terminals and everything worked great after that. can you guys see anything wrong with this solution?
i'll try MrAl's suggestion to put a resistor on the V+ terminal, right now i have a smaller one there just to make sure that too much current isn't allowed through after the battery voltage drops too low and the circuit quits regulating. I could also run the test MrAl suggested if this post doesn't answer all the questions.
thanks for all the help guys.

 

[MrAl]

Hello again tall paul,

The reason for the above tests was to find out
if everything was connected properly in your
circuit, and that the LM334 was working ok.

I changed my mind on the added resistor, because
the circuit should work as is on the data sheet.
If the tests i gave you all work out ok, then
your higher current circuit should work too.

With the test circuits, you dont need any R
and C for compensation either. This helps
isolate the problem too.

Take care for now,
Al

 

[Doug S]

Originally posted by tall paul:
Hi Doug and MrAl,

I had set aside the project for a couple days mostly due to lack of time and a little frustration. This weekend i made some progress though. i hooked up my o-scope and went to town. i've built a couple ciruits so far, one meant to deliver 325ma for a flashlight and one at 450ma for a bike light. the 425ma was working pretty good, meaning that the voltage across the Rsense was within spec (the .067V that i reported earlier). the 325 ma circuit wasn't fairing so well, i increased the cap like doug suggested but it still wouldn't pan out. after hooking up the o-scope, i could tell that the circuit was oscillating. if i would short the cap, the oscilallation would stop and the circuit would regulate as designed. after the short was removed, the circuit would continue to act fine (regulate with no oscillations). if the switch was turned off then back on, the oscillations would return. so, i figured that initial turn on condition was the problem, sending the ciruit into oscillation. i decided to put a cap across the battery terminals and everything worked great after that. can you guys see anything wrong with this solution?
i'll try MrAl's suggestion to put a resistor on the V+ terminal, right now i have a smaller one there just to make sure that too much current isn't allowed through after the battery voltage drops too low and the circuit quits regulating. I could also run the test MrAl suggested if this post doesn't answer all the questions.
thanks for all the help guys.

<font size="2" face="Verdana, Arial">Paul, when I read that you had success with the C shorted out, my first reaction was "no way". Once I analyzed it, however, I see that with a fortuitous mix of transistor Hfe, R, and a narrow range of battery voltage, the R could in fact provide a form of compensation to the circuit. Regard it as a curiosity only, it does not lend itself to a real solution. Pondering why you are not achieving success with R and C similar to what has always worked for me [I typically use R=470 and 0.1<C<1.0uF] it occurs to me that I always used the added R at the V+ terminal [let's call it R2 for further discussion] and I can see that it would add to the stability. For your particular application you can go as high as 1K for this resistor. I would suggest one in the 500 to 1000 ohm range. Regarding your fix of adding the C across the battery, I see no problem with this. Hey, it works for you. It should not be necessary, however, once you find the right values for R, C, and R2.

 

[tall paul]

Originally posted by Doug S:
Once I analyzed it, however, I see that with a fortuitous mix of transistor Hfe, R, and a narrow range of battery voltage, the R could in fact provide a form of compensation to the circuit

<font size="2" face="Verdana, Arial">I guess I'm just lucky...
Actually I'm kinda glad things didn't go real smoothly, it ended up being a pretty good learning experience.

 

[Doug S]

Originally posted by tall paul:
I'm kinda glad things didn't go real smoothly, it ended up being a pretty good learning experience.

<font size="2" face="Verdana, Arial">Paul, when you are done playing with it, let us know what your final "as built" configuration is.