LEDs in Series with LM334 Current Source
Mike H
parking-light assembly on an automobile. The manufacturers design used
some HP designed circuit that simply used current limiting resistors to
drive the leds. For whatever reason something shorted out and things
became warm enough to take the isulation off the power connection at the
circut board, melting the board itself.
Time to build a new circut with new white LED's.
Considering the voltage variability of a automobile alternator I've
decided to build into the new circut a current source rather than a using
a current limiting resistor.
Digging around it appears the LM334 is a good solution. I have 20 LED's
to light in this array.
12 volts with an estimated 3.2 volt forward voltage on the LED's allows me
3 leds per series.
3 leds per series means 7 series assemblies.
As I'm completely new to this, the question that remains for me is what is
the value of the resistor I choose to set the current output of the LM334?
The LM334 has an upper limit of 10ma, but people in the model trains world
have been able to run it at 12ma. That is the goal for me as well, get an
output from the LM334 of 12ma.
Any comments or suggestions are welcomed.
Here is a link to the circut.
http://www.frontiernet.net/~miketoni/circut.jpg
The ideas for this circuit came from the circut on this website:
http://www.pollensoftware.com/railroad/circuit.html
Thanks.
CFoley1064
>From: Mike H miketoni.arggg.@frontiernet.net
>Date: 4/30/2004 10:55 AM Central Standard Time
>Message-id: <Xns94DB6EC10...@66.133.130.30>
Hi, Mike. As long as the car's voltage regulator is working properly, you
shouldn't see that much variability in your power supply. You'll see around
12VDC with the car motor off, and 13.6 to 14.3V with the motor on.
An LM334 is good for 10 mA, and possibly 12 mA, but I think you're going to
need more than 12 mA to get the type of brightness you want with the white
LEDs.
If you really want to use a current source for the LEDs, you might want to try
something like this (view in fixed font or M$ Notepad):
Multiple Current Sources
VCC
o----------o-----------o-----------.
| | | |
| | | |
V .-. .-. .-.
- | | | | | |
| | | | | | |
| '-' '-' '-'
| | | |
V | | |
- |< |< |<
| .---| .---| .--|
o-. | |\ | |\ | |\
| | | | | | | |
| | | | | | | | And so on...
| '--o-----------o------------o--------->
| | | |
| | | |
| | | |
.-. V ~ V ~ V ~
| | - ~ - ~ - ~
| | | | | ~
'-' V ~ V ~ V ~
| - ~ - ~ - ~
| | | |
| V ~ V ~ V ~
| - ~ - ~ - ~
| | | |
| | | |
o----------o-----------o-----------o
|
===
GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de
The diodes can be 1N4001s, the resistor in series with the diodes can be 1K,
and the current sense resistors are chosen as 0.7V/I(LED). For 20 mA, that
would be 0.7V/.02A = 33 ohms. The transistors can be any TO-92 PNP with V(ceo)
of 60V or greater, your choice.
Good luck
Chris
Mike H
news:20040430134401...@mb-m15.aol.com:
>>Subject: LEDs in Series with LM334 Current Source
>>From: Mike H miketoni.arggg.@frontiernet.net
...
>>
>>Here is a link to the circut.
>>http://www.frontiernet.net/~miketoni/circut.jpg
>>
>>The ideas for this circuit came from the circut on this website:
>>http://www.pollensoftware.com/railroad/circuit.html
>>
>>Thanks.
>
> Hi, Mike. As long as the car's voltage regulator is working properly,
> you shouldn't see that much variability in your power supply. You'll
> see around 12VDC with the car motor off, and 13.6 to 14.3V with the
> motor on.
>
In general, the concern is what happens when the voltage regulator dies.
The existing circuit, that failed, relies on resistors to limit current
flow. In my case something catastrophic happened that damaged the PCB.
But more commonly, the LED's in the original circuit simply die to to
excess current.
> An LM334 is good for 10 mA, and possibly 12 mA, but I think you're
> going to need more than 12 mA to get the type of brightness you want
> with the white LEDs.
I haven't received any LED's yet to verify brightness. My goal was to run
them below their rated current as what I have read thus far leads me to
believe that their rated current is generally a maximum and to be avoided.
This application uses the LED's with a reflector and a lenses that
disburses the light in a wider range than a LED normally would. It is
this reflection of light that is visible in the marker lens. Since there
is a right and a left, and I only need to replace a right one. The goal
would be to adjust the current so that the right side would match the left
side.
The stock circuit (on the left side) was developed in 1993, so I'm
guessing the stock design is not all that "bright" in comparison to the
LED's available today. But I may be wrong.
Regarding the circuit you proposed.
If've done some further research and am wondering if all of the components
are necessary. Here is a second circuit diagram I drew where the LED's
are placed into an array, and then the current into that Array is
controled by a single device.
http://www.frontiernet.net/~miketoni/circut2.jpg
Would that work?
Could that be translated to what you proposed?
I.E. reducing the number of components in the circuit.
And can you clarify the calculations for me in your diagram. Where did
the value for .7 come from for the LED?
Thank you for your assistance.
John G
"Mike H" <miketoni.arggg.@frontiernet.net> wrote in message
news:Xns94DBA341...@66.133.130.30...
Unless I missed something your new circuit requires 70
milliamps thru the LM334 and its maximum allowed is only
10ma.
Surely there are larger constant current regulators that
would work here but I have not got time to research this
morning.
The LM334 was most likely used in the model train because he
needed to get the ONE led and regulator in a small space.
You should not have this restriction.
--
John G
Wot's Your Real Problem?
CFoley1064
>From: Mike H miketoni.arggg.@frontiernet.net
>Date: 4/30/2004 4:04 PM Central Standard Time
>Message-id: <Xns94DBA341...@66.133.130.30>
Hi again, Mike. The second circuit shows several strings of three LEDs in
parallel, driven by a current source. That won't work too well, because minor
changes in the Vf of the LEDs will mean that the current will get hogged up by
the strings of LEDs that have the lowest combined forward voltage. That means
major differences in brightness between LED strings. If you match forward
voltages, it might work pretty well, but as a paractical circuit, no. (I guess
the business with having two LEDs in one string is a typo -- that will limit
forward voltage to 6.4V, and only those 2 LEDs will light. Not to mention that
the LM334 has no chance of pumping enough current for all those LED strings.).
Let's take a look at the original "HP" circuit for best and worst cases (view
in fixed font or M$ Notepad):
+12V from regulator ~ ~ ~ GND
___ ~ ~ ~
o---|___|--->|---->|---->|------o
Let's assume you've got a series resistor of 200 ohms +/- 5%, an LED forward
voltage between 3.0V and 3.4V with a 3.2V nominal, and a regulator voltage that
varies between 11.7V (low battery, motor off) and 14.5V (high regulator, motor
on).
Nominal LED current: (13.8V - (3 * 3.2V)) / 200 ohms = 21 mA
High LED Current: (14.5V - (3 * 3.0V)) / 190 ohms = 28 mA
Low LED Current: (11.7V - (3 * 3.4V)) / 210 ohms = 7.14 mA
As you can see, high case to low case is almost a 4:1 ratio. Assuming the
motor is on, you can assume a minimum of 13.5V, which would give a low case of
Low LED Current: (13.5V - (3 * 3.4V))/210 ohms = 15.7 mA
Which gives you less than 2:1 when the motor is on. If you can get better
control over the LED Vf (after all, HP has been making LEDs since the 70s --
they know what they're doing), that means a better ratio.
Now let's assume your voltage regulator has gone south. In that case, your car
voltage may go as high as 16V or so. When that happens, you're killing your
battery, and bad things would happen to incandescent lights too because of the
higher power. In that event your nominal current would be:
Nominal LED current = (16V - (3 * 3.2v)) / 200 ohms = 32 mA
That is 50% over nominal, which isn't a problem if you're at 20 mA. If you're
at or over 40 mA, it might be fatal, depending on the LED.
That gets us back to the current source. The idea is a good one -- keep the
same current going through the LEDs despite variations in voltage. Here's a
segment of the circuit:
12V
.-o-.
| |
| |
D1 V .-.
- | |33 Ohm
| | |
| '-'
D2 V |
- | Ve
| |<
Vb o-|
| |\
.-. |Vc
1K| | |
| | V ~
'-' - ~
| |
| V ~
| - ~
| |
| V ~
| - ~
| |
'-o-'
GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de
You're probably familiar with the assumption that a forward-biased diode will
have about 0.7V across it. Two diodes in series, then, will have about 1.4V
across them. So, Vb is 1.4V less than the 12V. By the same token, the forward
biased PNP transistor will act so there is 0.7V from the emitter to the base.
That will mean that the transistor will tend to act so that Ve is 0.7V less
than the 12V. Your job is to choose the emitter resistor so that 0.7V across
it will give you the current you need. That would be (0.7V)/0.02A = 35 ohms,
or 33 ohms to the nearest standard value.
Of course this is a first cut approximation. Collector current is emitter
current plus base current (which should be less than 1/100th of emitter current
with a stiff base voltage source). More importantly, Vf of the diode and Vbe
of the PNP transistor is somewhat dependent on junction temperature. But under
normal circumstances in an automotive environment, Vbe will only drift by about
0.2V, and more importantly, the Vf of the diode will tend to track Vbe if
they're the same temp. Without going through the math, I can tell you that the
current source will give far better results than the series resistor alone,
like the example above. Also, a jellybean SOT-23 PNP transistor and a 1/8W SMT
resistor will cost a lot less than an LM334. The voltage source of the two
diodes and the 1/4W resistor is used by all the diode strings.
The main issue here is the voltage burden of the current source. However, even
if you assume 3.4V per diode, the current source can drive a 10.2V burden as
long as the voltage source is above 11.7V or so. Below that, the transistor
will begin to saturate, and current will go down. That should be acceptable --
people are used to having lights dim a little when the car is starting. The
main reason automotive manufacturers didn't use current sources is cost and
reduced reliability due to circuit complexity. Also, they specify LED Vf, and
generally will not accept delivery of LEDs with excessive Vf (which can be
controlled, to an extent, in manufacture).
Good luck
Chris
Mike H
news:20040430201738...@mb-m10.aol.com:
...
>
> The main issue here is the voltage burden of the current source.
> However, even if you assume 3.4V per diode, the current source can
> drive a 10.2V burden as long as the voltage source is above 11.7V or
> so. Below that, the transistor will begin to saturate, and current
> will go down. That should be acceptable -- people are used to having
> lights dim a little when the car is starting. The main reason
> automotive manufacturers didn't use current sources is cost and
> reduced reliability due to circuit complexity. Also, they specify LED
> Vf, and generally will not accept delivery of LEDs with excessive Vf
> (which can be controlled, to an extent, in manufacture).
>
Thanks for the feedback and information. This should give me plenty of
information to work with.
Thanks
Robert C Monsen
One other thing about LEDs in parallel is that they will hog more current as
they get hot, because their TC is negative. Thus, as they get hotter, their
Vf gets smaller, causing even more current to get put through the hot one,
leading to even more current, heat, etc, etc.
To overcome this problem, I'm a fan of the multistage current mirror. You
can regulate the current for all strings quite easily by using your current
source like this:
VCC
+
|
+----+-------+---------+--------+------+---------------+
| | | | | |
| | | | | |
|< |< |< |< |< |<
+-----------------------------------------------------|
| |\ |\ |\ |\ |\ |\
| | | | | | ... |
| | | | | | |
+-----+ V V V V V
| - - - - -
.-----. | | | | |
| | | | | | | |
| | | V V V V V
| \|/ | - - - - -
'-----' | | | | |
+----+-------+---------+--------+------+---------------+
|
===
GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de
The box with the arrow is your LM334 current source. If you use it like
this, all the strings in the system will have the same current sourced by
the current source, minus a tiny bit for the biasing.
Its easy to see why this is. Notice that the voltage from emitter to base is
equal on all of the transistors. That means that, according to transistor
theory, the current through them will be equal if they have the same temp
(and Vce, which is a far less important issue.) If you mount them all on the
same chunk of steel, they will stay pretty much the same temperature, so the
current will be pretty much the same.
The nice thing about this scheme is that its independent of Vcc. If Vcc
fluctuates, it still regulates the current properly, since the current sink
isn't dependent on Vcc.
You can also use a resistor as your current source, but you lose the Vcc
independence if you do that. That can apparently be an issue in an
automotive environment, because the voltage can spike quite a bit.
Regards,
Bob Monsen
Mike H
news:Ti_kc.11312$0H1.1209309@attbi_s54:
> One other thing about LEDs in parallel is that they will hog more
> current as they get hot, because their TC is negative. Thus, as they
> get hotter, their Vf gets smaller, causing even more current to get
> put through the hot one, leading to even more current, heat, etc, etc.
>
> To overcome this problem, I'm a fan of the multistage current mirror.
> You can regulate the current for all strings quite easily by using
> your current source like this:
>
Lots of information here and I greatly appreciate it. It gives me more to
think about than I can currently grasp.
I believe that Chris' comment regarding the current limitations of the
LM334 are valid. I did some initial search for another current source
that is as simple as the LM334 but with higher capacity and didn't have
any luck. I'm not sure there would be any advantage with finding one that
is more complex as Chris showed me a design that works with simple
transistors.
In the mean time I've located a LED that has a lower Forward Voltage of
2.2v at 20ma. Thus I should be able to support 5 LEDs in each series.
Question 1 - Is this not a good idea to grow series?
Question 2 - Here is how his circuit looks extended out, keeping the 1k
resistors and 33k resistors. Am I extending this properly?
o---------------------------------------
+12v | | | | | | | |
V .-. V .-. V .-. V .-.
- | | - | | - | | - | |
| | | | | | | | | | | |
| '-' | '-' | '-' | '-'
V | V | V | V |
- | - | - | - |
| |< | |< | |< | |<
-| -| -| -|
| |\ | |\ | |\ | |\
.-. | .-. | .-. | .-. |
| | | | | | | | | | | |
| | V | | V | | V | | V
'-' -\\ '-' -\\ '-' -\\ '-' -\\
| | | | | | | |
| V | V | V | V
| -\\ | -\\ | -\\ | -\\
| | | | | | | |
| V | V | V | V
| -\\ | -\\ | -\\ | -\\
| | | | | | | |
| V | V | V | V
| -\\ | -\\ | -\\ | -\\
| | | | | | | |
| V | V | V | V
| -\\ | -\\ | -\\ | -\\
| | | | | | | |
o---------------------------------------
-12v
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de
Question 3: Are your (Bob) comments regarding the current source related
to the above design in that the current flow through each parallel section
above may vary as each one is being managed within each grouping of 5?
Oh... and re-reading the information from Chris, I wonder if my circuit
above should have looked like this: Rather than that^
o---------------------------------------
+12v | | | | |
V .-. .-. .-. .-.
- | | | | | | | |
| | | | | | | | |
| '-' '-' '-' '-'
V | | | |
- | | | |
| |< |< |< |<
-|---------|---------|---------|
| |\ |\ |\ |\
.-. | | | |
| | | | | |
| | V V V V
'-' -\\ -\\ -\\ -
| | | | |
| V V V V
| -\\ -\\ -\\ -\\
| | | | |
| V V V V
| -\\ -\\ -\\ -\\
| | | | |
| V V V V
| -\\ -\\ -\\ -\\
| | | | |
| V V V V
| -\\ -\\ -\\ -\\
| | | | |
o---------------------------------------
-12v
Robert C Monsen
No, the one above is not a current mirror. The idea behind a current mirror
is to use the current through one transistor to control the current through
the rest of the transistors. The thing above is simply 4 separate circuits
which don't affect each other (assuming the 12V doesn't sag, which it won't
in a car.)
The circuit below is a form of current mirror. However, I think you'll get
better results with the one I posted, which also uses fewer parts, and has a
larger voltage compliance (meaning you can possibly use more LEDs per
string.) The one below also has problems with voltage spikes, which my
posted circuit does not have.
Mike H
>
> "Mike H" <miketoni.arggg.@frontiernet.net> wrote in message
> news:Xns94DD79E20...@66.133.130.30...
...
>
> The circuit below is a form of current mirror. However, I think you'll
> get better results with the one I posted, which also uses fewer parts,
> and has a larger voltage compliance (meaning you can possibly use more
> LEDs per string.) The one below also has problems with voltage spikes,
> which my posted circuit does not have.
...
O.k. Sounds like I'm learning some but not quite enough. Taking your
circuit design. Assuming I need 20ma of current through the system. That
eliminates the LM334, Correct? If not, can you demonstrate how it could
be retained?
So if the LM334 is removed, then what goes in it's place to limit the
current?
The circuit below simply puts the number of LEDS in question in place and
adds a blank space for the current control.
The LED's in question have a 2.2v Forward Voltage when run at 20ma.
With this design are we still assuming "any" TO-92 PNP transistor with a
V(ceo) of 60V or greater?
I truly apprecate all the assistance and explinations.
VCC
+
|
+----+-------+---------+--------+------+
| | | | |
| | | | |
|< |< |< |< |<
+-------------------------------------|
| |\ |\ |\ |\ |\
| | | | | |
| | | | | |
+-----+ V V V V
| - \\ - \\ - \\ \\
| | | | |
| | | | |
.---------. V V V V
| | - \\ - \\ - \\ - \\
| | | | | |
| | V V V V
| | - \\ - \\ - \\ - \\
| | | | | |
| | V V V V
| | - \\ - \\ - \\ - \\
| | | | | |
'----+----' V V V V
| - \\ - \\ - \\ - \\
| | | | |
|---+--------+---------+--------+------+
===
GND
Robert C Monsen
Here is a current source that can replace your LM334, made from a few
discretes.
VCC
+
|
|
| current in
| here = 20mA
| |
.-. |
1k ohms | | |
| | |
'-' |
| |
| |
| |
| |/
+-------| NPN (like 2N3904)
| |>
| |
1N914 V |
- .-.
| | | 33 ohms
| | |
V '-'
1N914 - |
| |
| |
+---------+
|
|
===
GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de
You only need one of these to control all of the strings. (it goes in
the box above.) The way it works is that the NPN transistor has a Vbe
of 0.7V. The two diodes have a combined Vf of 1.4V. Thus, there is
0.7V across the 33 ohm resistor, which means 0.7/33 = 21.21mA will go
through the NPN transistor.
You can also use a common voltage regulator, like an LM317. Its
simple:
.------------.
I in In| |Out
---------| LM317 |----+
= 20mA | | |
| | .-.
| | | |
'------------' | | 62 ohms
|Adj '-'
| |
| |
+----------+
GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de
The LM317 is designed to keep the voltage at Out equal to Adj + 1.25.
It'll put out whatever current it needs to to do this. Since our 68
ohm resistor is between Out and GND, then we have
Iin = 1.25/62 = 20mA.
For this circuit, you need to ensure that the Iin is greater than
about 4V. That isn't really an issue with your circuit, though.
You can get LM317s in TO-92 cases, just like a small transistor.
Note that your voltage compliance (Vc) is going to be less than 12V
(probably more like 11V) so if the forward voltage of your LEDs is Vf
at 20mA, then each string will only support Vc/Vf LEDs. If you put in
more than that, the strings won't light up.
Regards
Bob Monsen
Mike H
$0H1.1631634@attbi_s54:
...
>
> You only need one of these to control all of the strings. (it goes in
> the box above.) The way it works is that the NPN transistor has a Vbe
> of 0.7V. The two diodes have a combined Vf of 1.4V. Thus, there is
> 0.7V across the 33 ohm resistor, which means 0.7/33 = 21.21mA will go
> through the NPN transistor.
>
> You can also use a common voltage regulator, like an LM317. Its
> simple:
...
Thanks again for the assistance. Busy stamping out software bugs in two
different projects started by others and left to me to fix.
I've placed your example into mine and just want to verify the
connections. Does the below look correct?
You also mention a common voltage regulator. If I were to use such a
device do I then get away from my goal of ensuring a specific current to
the leds regardless of voltage?
Thanks
VCC +
|
+--------+-------+---------+--------+------+
| | | | | |
| | | | | |
| |< |< |< |< |<
| +-------------------------------------|
| | |\ |\ |\ |\ |\
| | | | | | |
| | | | | | |
.-. +-----+ V V V V
| | | - \\ - \\ - \\ \\
| | | | | | |
'-' | | | | |
| |/ V V V V
+------| - \\ - \\ - \\ - \\
| |> | | | |
| | V V V V
V | - \\ - \\ - \\ - \\
- .-. | | | |
| | | V V V V
| | | - \\ - \\ - \\ - \\
V '-' | | | |
- | V V V V
| | - \\ - \\ - \\ - \\
| | | | | |
+--------|-------+---------+--------+------+
Robert C Monsen
> "Robert C Monsen" <rcsu...@comcast.net> wrote in news:3idlc.17440
> $0H1.1631634@attbi_s54:
> ...
>
> >
> > You only need one of these to control all of the strings. (it goes
in
> > the box above.) The way it works is that the NPN transistor has a
Vbe
> > of 0.7V. The two diodes have a combined Vf of 1.4V. Thus, there is
> > 0.7V across the 33 ohm resistor, which means 0.7/33 = 21.21mA will
go
> > through the NPN transistor.
> >
> > You can also use a common voltage regulator, like an LM317. Its
> > simple:
> ...
>
>
> Thanks again for the assistance. Busy stamping out software bugs in
two
> different projects started by others and left to me to fix.
I hate it when that happens...
This looks fine, except there are too many LEDs, I think, unless you
are using the standard 1.7V Vf leds, which are not very bright. I
thought you were using those high-brighness white LEDs, which are more
like 3.5V Vf. Just make sure the forward voltage across all the LEDs
is less than about 11V, and you should be fine (with Vcc >= 12V.) You
can add more strings if you need to. I'd say to keep below about 10
strings with this circuit.
Note that the value of the resistor on the bottom left (parallel to
the two series diodes) controls the current through each of the
strings. Less resistance means more current. The relation is current
times the resistance = the voltage drop across a diode, or
R = 0.7/I
where I is the current you'll get through each parallel string.
The voltage regulator circuit I described was really just a constant
current sink; its another way to do what the left bottom side of the
circuit does, except using a single IC + resistor instead of two
resistors, a transistor, and two diodes. It also wastes less current,
cause it doesn't need the bias provided by the resistor + diode pair.
Regards,
Bob Monsen
Mike H
>
> "Mike H" <miketoni.arggg.@frontiernet.net> wrote in message
> news:Xns94DFC7080...@66.133.130.30...
...
>>
>> You also mention a common voltage regulator. If I were to use such a
>> device do I then get away from my goal of ensuring a specific
> current to
>> the leds regardless of voltage?
...
>
> This looks fine, except there are too many LEDs, I think, unless you
> are using the standard 1.7V Vf leds, which are not very bright. I
> thought you were using those high-brighness white LEDs, which are more
> like 3.5V Vf. Just make sure the forward voltage across all the LEDs
> is less than about 11V, and you should be fine (with Vcc >= 12V.) You
> can add more strings if you need to. I'd say to keep below about 10
> strings with this circuit.
...
>
> The voltage regulator circuit I described was really just a constant
> current sink; its another way to do what the left bottom side of the
> circuit does, except using a single IC + resistor instead of two
> resistors, a transistor, and two diodes. It also wastes less current,
> cause it doesn't need the bias provided by the resistor + diode pair.
Currently I'm looking at a Super-Red LED that has a typical forward
voltage of 2.2v, and a MAX of 2.6v. I thought I wanted white until I
thought about the fact the lens in the assembly is already red. Why waste
energy that will be filtered?
2.6v * 5v = 13v. Woops. I guess I'll be adding another string and
decreasing the count per string.
Regarding the LM317 source you described. Could you describe for me how
it works as a current source as it appears to be a regulator of voltage.
Is the theory that by using such a device, voltage is fixed, thus a
resistor can be again used for current limitation. Thus I would no longer
need to concern myself with the fear of possible voltage issues in an
Automotive electrical system
Would the below be how I would integrate the LM317 source then? (I didn't
change the LED count per string yet assume a reduction to 4)
VCC +
|
+--------+-------+---------+--------+------+
| | | | |
| | | | |
|< |< |< |< |<
+-------------------------------------|
| |\ |\ |\ |\ |\
| | | | | |
| | | | | |
+-----+ V V V V
| - \\ - \\ - \\ \\
| | | | |
| | | | |
+---------+ V V V V
| ____ - \\ - \\ - \\ - \\
| | | | | | |
+--|LM317-+ V V V V
|____| | - \\ - \\ - \\ - \\
| .-. | | | |
| | | V V V V
| | | - \\ - \\ - \\ - \\
| '-' | | | |
| | V V V V
+----+ - \\ - \\ - \\ - \\
| | | | |
+-------+---------+--------+------+
Robert C Monsen
No. Consider a device which, given a range of input voltages, can
produce a given voltage. Now, connect that fixed voltage to ground
through a resistor R. By ohms law, the current through the resistor
will be I = V/R. Since V is fixed at about 1.25V for an LM317 with a
grounded "adj" pin, and since the amount of current through the 'adj'
pin is, essentially, zero, then the total current sinked (sunk?) by
this device is V/R. Voila, its a current source! The LM317 can take
voltage differentials up to 30V, so given any voltage from 30V down to
about 3V, the LM317, hooked up as shown through a resistor, will sink
the same amount of current.
>
>
> Would the below be how I would integrate the LM317 source then? (I
didn't
> change the LED count per string yet assume a reduction to 4)
>
Yes, thats it. the resistor R is going to be 1.25/I, where I is the
current you want through each string. So, for example, if you want to
pass 20mA, you use 1.25/0.020 = 62.5 ohms for R. Anything near that
will be fine.
Watson A.Name "Watt Sun - the Dark Remover"
[snip]
>>VCC +
>> |
>> +--------+-------+---------+--------+------+
>> | | | | |
>> | | | | |
>> |< |< |< |< |<
>> +-------------------------------------|
>> | |\ |\ |\ |\ |\
>> | | | | | |
>> | | | | | |
>> +-----+ V V V V
>> | - \\ - \\ - \\ \\
>> | | | | |
>> | | | | |
>> +---------+ V V V V
>> | ____ - \\ - \\ - \\ - \\
>> | | | | | | |
>> +--|LM317-+ V V V V
>> |____| | - \\ - \\ - \\ - \\
>> | .-. | | | |
>> | | | V V V V
>> | | | - \\ - \\ - \\ - \\
>> | '-' | | | |
>> | | V V V V
>> +----+ - \\ - \\ - \\ - \\
>> | | | | |
>> +-------+---------+--------+------+
>> |
>> ===
>> GND
The above schem assumes that all five PNP transistors have the same Vbe
drop, which would be a valid assumption if they were all on the same
chip in the same package. But they're usually not, and they are not
thermally coupled. So there could be a wide variation in the current of
each string of LEDs.
Robert C Monsen
"Watson A.Name "Watt Sun - the Dark Remover"" <NOS...@dslextreme.com>
wrote in message news:c7qq7k$71hlh$1...@hades.csu.net...
How much do you think? What percentage?
Regards,
Bob Monsen
Rich Grise
news:9nioc.75723$kh4.4359226@attbi_s52...
300%.
But that's a very quick off-the-cuff WAG - but I've never really understood
the theory behind a current mirror, which is what this looks like.
Personally,
I would have done just "ordinary" current sources, like this:
+V
|
----o----
| |
[D] |
| [R]
[D] |
| / E
------B|
| \ C
[R] |
| [LED]
| |
----o----
|
-V
But the other circuit is probably more efficient.
Cheers!
Rich
Robert C Monsen
"Rich Grise" <nu...@example.net> wrote in message
news:LTloc.64129$sK3....@nwrddc03.gnilink.net...
Rich:
Current mirrors just depend on the fact that, for equal transistors,
if the Vbe is equal, the current is equal. Its as simple as that. So,
setting a group of transistors with the same Vbe means they'll allow
the same current through them. Use the first one (the one with the
base-collector connection) to set the current, and everybody else is
following along. Its a neat trick.
I'm guessing that for the OPs application, the diference in current
between transistors from the same batch will be less than about %5. If
its more, and he cares, he can use a small resistor between the
positive rail and the emitter, since he can ensure that resistors are
%5 much more easily. However, that decreases the voltage range the
source will generate, meaning less LEDs he can support. Since the
worst that can happen is that one of the LED strings will be slightly
brighter than the others, I'd say that its not an issue.
Regards,
Bob Monsen
Mike H
> I'm working on a circut design to replace the LED's that came in a
> parking-light assembly on an automobile. The manufacturers design
> used some HP designed circuit that simply used current limiting
> resistors to drive the leds. For whatever reason something shorted out
> and things became warm enough to take the isulation off the power
> connection at the circut board, melting the board itself.
>
Ok. To review.
The goal here is to use a current source. Initially the thought was to
use the LM334 as a current source to drive each string of LED's. Then the
idea became a mirrored Current source. But the LM334 current limits put
it outside the realm of driving LED's that need 20mA to reach their
brightness.
A constant current source is prefered due to the variations that can occur
within the automobile charging system, and the delicate nature of LED's
Information from CFoley1064 and Robert Monsen has been very helpful. So
now it's time to look at the preliminary design and look for issues.
Below is diagram using a LM317 voltage regulator in combination with a
resistor to ensure a fixed current source regardless of changes in
automobile voltage output.
Here are the planned devices to use:
12 Volts (as high as 14 v is possible, as low as 11)
1-EA LM317 (http://www.national.com/pf/LM/LM317.html)
1-EA 62ohm Resistor .1 watt 1% tolerance
6-EA 2N2905A PNP Transistors (Has Vceo of 60)
20-EA RL5-R3545 Super-Red LED Foward Voltage of
2.2 typical, 2.6 max (http://www.superbrightleds.com/leds/r2_specs.htm)
When wired as shown below, the goal is to achieve a current level of 20mA
across all of the LED's in a fairly even manner to achieve near equal
brightness across all of them. The theory being that changing the
resistor value can impact the amount of current flowing as it will impact
the setpoint of the voltage regulator.
Questions:
1) Are the devices listed above appropriate for the goals?
2) What is a good source for the above components. (LED's I know where to
go)
3) Any other comments or changes to the circut below recommended?
Thanks
VCC +12
|
+--------+-------+---------+--------+------+-------+
| | | | | |
| | | | | |
|< |< |< |< |< |<
+-+-------+---------+--------+------+-------+-|
| |\ |\ |\ |\ |\ |\
| | | | | | |
| | | | | | |
+-----+ V V V V V
| - \\ - \\ - \\ \\ - \\
+----------+ | | | | |
| ____ | | | | |
| | | V V V V V
+--|LM317---+ - \\ - \\ - \\ - \\ -\\
|____| | | | | | |
| .-. V V V V V
| | | - \\ - \\ - \\ - \\ -\\
| | | | | | | |
| '-' V V V V V
| | - \\ - \\ - \\ - \\ -\\
+------+ | | | | |
| | | | | |
+---------+------+---------+--------+------+-------+
|
|
===
GND -12
John Popelish
I would eliminate the LM317 and add a resistor to each current source
emitter to make them track each other much better. I would set the
base voltage to one of them to force about .6 volts across its emitter
resistor.
Something like:
VCC +12
|
+--------+-------+---------+--------+------+-------+
| | | | | |
| .-. .-. .-. .-. .-.
| | | | | | | | | | |
| | | 33 | | 33 | |33 | | 33 | |33
>| '-' '-' '-' '-' '-'
|---- + | | | |
/| | | | | |
| |< |< |< |< |<
+-----+---------+--------+------+-------+
| |\ |\ |\ |\ |\
| | | | | |
| | | | | |
| V V V V V
| - \\ - \\ - \\ \\ - \\
| | | | | |
| | | | | |
| V V V V V
| - \\ - \\ - \\ - \\ -\\
| | | | | |
.-. V V V V V
| | - \\ - \\ - \\ - \\ -\\
| |1k | | | | |
'-' V V V V V
| - \\ - \\ - \\ - \\ -\\
| | | | | |
| | | | | |
+--------+-------+---------+--------+------+-------+
|
|
===
GND -12
No need for as spiffy a transistor as the 2N905A. I would probably go
with 2N3905. They have a plastic case that is much cheaper and they
are common as dirt. Even Radio Shack should have them. 2N4403 is
beefier but that is not much of a factor with 20 ma of collector
current.
--
John Popelish
Winfield Hill
>
>> Mike H wrote
>>> I'm working on a circut design to replace the LED's that came in a
>>> parking-light assembly on an automobile. The manufacturers design
>>> used some HP designed circuit that simply used current limiting
>>> resistors to drive the leds. For whatever reason something shorted
>>> out and things became warm enough to take the isulation off the
>>> power connection at the circut board, melting the board itself.
>
> emitter to make them track each other much better. I would set the
> base voltage to one of them to force about .6 volts across its emitter
> resistor.
>
> |
> +-+------+-------+--------+--------+-------+-------+--- etc
> | | | | | | |
> | .-. .-. .-. .-. .-. .-.
> | | | | | | | | | | | | |
> | | | | | 33 | | 33 | |33 | | 33 | |33
> >| | | | | | |
> |--- | ------+ | | | |
> /| | | | | | |
> | | |< |< |< |< |<
> | +-----+--------+--------+-------+-------+----- etc
> | | |\ |\ |\ |\ |\
> | |< | | | | |
> +----| | | | | |
> | |\ V V V V V
> | | - \\ - \\ - \\ \\ - \\
> | | | | | | |
> | | | | | | |
> | | V V V V V
> | | - \\ - \\ - \\ - \\ -\\
> .-. | V V V V V
> | | .-. 1W | | | | |
> '-' | | V V V V V
> | | | - \\ - \\ - \\ - \\ -\\
> | | | | | | |
> +-+------+-------+--------+--------+-------+-------+--- etc
> |
> |
> GND
> No need for as spiffy a transistor as the 2N905A. I would probably go
> with 2N3905. They have a plastic case that is much cheaper and they
> are common as dirt. Even Radio Shack should have them. 2N4403 is
> beefier but that is not much of a factor with 20 ma of collector
> current.
Radio Shack has the 2n3904 (npn) and 2n3906 (pnp). They have the 2n4401
(npn), but not the 2n4403. They also have the TO-92 mps2222 and mps2907.
I enhanced your design just a bit with another pnp trannie, so it can
handle more LED strings if desired.
Thanks,
- Win
(email: use hill_at_rowland-dot-org for now)
Mike H
news:40AA79F9...@rica.net:
> Mike H wrote:
...
>> Ok. To review.
>>
...
>> A constant current source is prefered due to the variations that can
>> occur within the automobile charging system, and the delicate nature
>> of LED's
...
>
> I would eliminate the LM317 and add a resistor to each current source
> emitter to make them track each other much better. I would set the
> base voltage to one of them to force about .6 volts across its emitter
> resistor.
>
> Something like:
>
> VCC +12
> |
> +--------+-------+---------+--------+------+-------+
> | | | | | |
> | .-. .-. .-. .-. .-.
> | | | | | | | | | | |
> | | | 33 | | 33 | |33 | | 33 | |33
> >| '-' '-' '-' '-' '-'
> |---- + | | | |
> /| | | | | |
> | |< |< |< |< |<
> +-----+---------+--------+------+-------+
> | |\ |\ |\ |\ |\
> | | | | | |
> | | | | | |
> | V V V V V
> | - \\ - \\ - \\ \\ - \\
> | | | | | |
> | | | | | |
> | V V V V V
> | - \\ - \\ - \\ - \\ -\\
> | | | | | |
> .-. V V V V V
> | | - \\ - \\ - \\ - \\ -\\
> | |1k | | | | |
> '-' V V V V V
> | - \\ - \\ - \\ - \\ -\\
> | | | | | |
> | | | | | |
> +--------+-------+---------+--------+------+-------+
> |
> |
> == GND -12
In the design you propose, what limits the voltage fluctuation and thus
the current change? Say if the 12v is suddenly 14v for a moment. The
point of the LM317 was it's ability to maintain a set output voltage
regardless of the input.
Thanks for the info on the devices
Mike H
news:c8fit...@drn.newsguy.com:
...
>>
>> VCC +12 [ edited by w.h. ]
>> |
>> +-+------+-------+--------+--------+-------+-------+--- etc
>> | | | | | | |
>> | .-. .-. .-. .-. .-. .-.
>> | | | | | | | | | | | | |
>> | | | | | 33 | | 33 | |33 | | 33 | |33
>> | '-'100 '-' '-' '-' '-' '-'
>> >| | | | | | |
>> |--- | ------+ | | | |
>> /| | | | | | |
>> | | |< |< |< |< |<
>> | +-----+--------+--------+-------+-------+----- etc
>> | | |\ |\ |\ |\ |\
>> | |< | | | | |
>> +----| | | | | |
>> | |\ V V V V V
>> | | - \\ - \\ - \\ \\ - \\
>> | | | | | | |
>> | | | | | | |
>> | | V V V V V
>> | | - \\ - \\ - \\ - \\ -\\
>> | 2.2k | | | | | |
>> .-. | V V V V V
>> | | | 220 - \\ - \\ - \\ - \\ -\\
>> | | .-. 1W | | | | |
>> '-' | | V V V V V
>> | | | - \\ - \\ - \\ - \\ -\\
>> | '-' | | | | |
>> | | | | | | |
>> +-+------+-------+--------+--------+-------+-------+--- etc
>> |
>> |
>> GND
>>
...
> I enhanced your design just a bit with another pnp trannie, so it can
> handle more LED strings if desired.
>
...
Same question for you. In the design above, what ensures that the current
across the LED's is constant and does not change when the 12v moves,
possibly as high as 14v?
John Popelish
The current through the 1k resistor has to be more than enough to
supply all the bases of all the LED current sources. As soon as the
current through the left one causes a .6 volt drop across its 33 ohm
emitter resistor, the extra transistor on the left comes on and sucks
up all the excess current from the 1k. This holds the base voltage
for all the current regulators at that which will cause a .6 volt drop
across each of their emitter resistors. 0.6/33=18.2 ma of emitter
current. The exact current through the 1k resistor is not very
important, since only a fraction of it is expected to involve the
current regulator base currents. If the regulator transistors have a
gain of 50, then 5 of them would need about 5*(0.018/50)=0.0018 amps
(1.8 milliamps). At 14 volts, there will be about
(14-1.2)/1000=0.0128amps (12.8 milliamps) passing through the 1k
resistor, so the assumption that only a fraction of its current goes
to base drive is safe.
This circuit is a little temperature sensitive, because the turn on
voltage of the left transistor goes down as temperature rises, but
this is a good idea for LEDs that also need current derating as
temperature rises.
--
John Popelish
Mike H
...
> The current through the 1k resistor has to be more than enough to
> supply all the bases of all the LED current sources. As soon as the
> current through the left one causes a .6 volt drop across its 33 ohm
> emitter resistor, the extra transistor on the left comes on and sucks
> up all the excess current from the 1k. This holds the base voltage
> for all the current regulators at that which will cause a .6 volt drop
> across each of their emitter resistors. 0.6/33=18.2 ma of emitter
> current. The exact current through the 1k resistor is not very
> important, since only a fraction of it is expected to involve the
> current regulator base currents. If the regulator transistors have a
> gain of 50, then 5 of them would need about 5*(0.018/50)=0.0018 amps
> (1.8 milliamps). At 14 volts, there will be about
> (14-1.2)/1000=0.0128amps (12.8 milliamps) passing through the 1k
> resistor, so the assumption that only a fraction of its current goes
> to base drive is safe.
>
> This circuit is a little temperature sensitive, because the turn on
> voltage of the left transistor goes down as temperature rises, but
> this is a good idea for LEDs that also need current derating as
> temperature rises.
>
I think I understand. I'm trying to model this in Linear Technology's
free LTspice tool but their LED Model doesn't fit with the ones I'm
looking at so the values don't work out quite right.
Am I correct then that to change the current flowing across the LED's, I
need to change the 33ohm resistors?
I'm shooting for 20ma across the LED's. But I may want to shoot for lower
and then have a switch trigger an additional increase in current to give a
moment of brighter light.
John Popelish
> I think I understand. I'm trying to model this in Linear Technology's
> free LTspice tool but their LED Model doesn't fit with the ones I'm
> looking at so the values don't work out quite right.
Just use something like 10 1N4148 diodes, instead. As long as there
is still a volt or two dropped across the regulator transistors, the
current should be pretty independent of the supply voltage.
> Am I correct then that to change the current flowing across the LED's, I
> need to change the 33ohm resistors?
Yes.
> I'm shooting for 20ma across the LED's. But I may want to shoot for lower
> and then have a switch trigger an additional increase in current to give a
> moment of brighter light.
When you have a good idea what the 2 current setpoints are, I might
have some suggestions no how to do the switch.
--
John Popelish
Mike H
> Mike H wrote:
>
>> I think I understand. I'm trying to model this in Linear
>> Technology's free LTspice tool but their LED Model doesn't fit with
>> the ones I'm looking at so the values don't work out quite right.
>
...
>> I'm shooting for 20ma across the LED's. But I may want to shoot for
>> lower and then have a switch trigger an additional increase in
>> current to give a moment of brighter light.
>
> When you have a good idea what the 2 current setpoints are, I might
> have some suggestions no how to do the switch.
>
Thanks. I'll do that. I'm going to head over to Radio Shack here shortly
to get some parts to start trying things out and prototype.
Mike
Rich Grise
news:Xns94F08F78C...@66.133.130.30...
> ...
> >> I'm shooting for 20ma across the LED's. But I may want to shoot for
> >> lower and then have a switch trigger an additional increase in
> >> current to give a moment of brighter light.
> >
> > When you have a good idea what the 2 current setpoints are, I might
> > have some suggestions no how to do the switch.
> >
>
> Thanks. I'll do that. I'm going to head over to Radio Shack here shortly
> to get some parts to start trying things out and prototype.
There was talk once of an LED dimming when a cap is discharged through
it. It does. But this particular one, an ordinary RS red LED, dropped
about 2V or whatever it's supposed to be, and from 20-10 mA, it stayed
just as bright to the eye as at 20 - from 10 it seemed to dim linearly,
smoothly down to imperceptibility, and the current tracked right along.
But I wouldn't depend on a current set point for bright/dim - I'd use
PWM. :-)
Cheers!
Rich
Mike H
> "Mike H" <miketoni.arggg.@frontiernet.net> wrote in message
> news:Xns94F08F78C...@66.133.130.30...
>> John Popelish <jpop...@rica.net> wrote in
...
>> >
>> > When you have a good idea what the 2 current setpoints are, I might
>> > have some suggestions no how to do the switch.
>> >
>>
>> Thanks. I'll do that. I'm going to head over to Radio Shack here
>> shortly to get some parts to start trying things out and prototype.
>
> There was talk once of an LED dimming when a cap is discharged through
> it. It does. But this particular one, an ordinary RS red LED, dropped
> about 2V or whatever it's supposed to be, and from 20-10 mA, it stayed
> just as bright to the eye as at 20 - from 10 it seemed to dim
> linearly, smoothly down to imperceptibility, and the current tracked
> right along.
>
> But I wouldn't depend on a current set point for bright/dim - I'd use
> PWM. :-)
Not having looked into it yet my thought is that there may be some
architectural differences in LED that can impact their behavior when
driving them outside their specifications.
Somehow I feel that pulse wave modulation would be a much more complex
design to create. Not that I woundn't mind giving it a try, but I need to
walk before I run.
Thanks.