simple buck switcher for LED drive from 9v?

[Winfield Hill]

Michael wrote...

Winfield Hill wrote ...

R3 is to lift the base of Q1 slightly so you only need to drop less
than .6V accross R1, the circuit would work without it but would be
less effecient. It's a nice circuit, I like it.

Thanks! When Q4 is on, the LED current increases until the sense resistor
voltage is Vhi = Vbe1 - Vd R2/R3, which you can design to be about 250mV.
Then when Q2 turns on and Q4 is off the LED current decreases until the
sense voltage is Vlo = Vbe1 - Vd R2/R3 - (Vd-Vbe) R2/R4 which could be
say 200mV, creating 50mV of hysteresis. With an average voltage of 225mV
and R1 = 22 ohms we get an LED current that ramps from 9 to 11mA and back.

Let's evaluate the inductor voltages. With Q4 on we get Vbatt - Vd - 0.22,
or about +5.7V with a 7.5-volt partially-used battery. With Q4 off we see
-(Vd_LED + 0.22 + Vd1) or about -2.1V. So the inductor charging times will
be about 1/3 as long as the discharging times. Let's design for a 200kHz
switching frequency, or a discharge time of about 3.7us, during which L1's
current drops by 2mA. From dI/dt = V/L we can calculate L = 3.9mH

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Walter Harley]

Seems a common enough task, but it's more rewarding to design a circuit
that's just right, than to copy somebody else's general-purpose design.
First you want a buck regulator in which the bucking inductor is large
enough to operate at a nearly constant current. Most buck-regulator ICs
are meant for voltage regulation and typically provide a 1.25V feedback
pin. They can be easily used for current regulation if the feedback pin
monitors the voltage across a current-sensing resistor, but for a single
red LED load dropping about 1.6V, an additional 1.25V measurement drop
wastes an unacceptable amount of power. Try this efficient LED switcher
circuit I designed that can regulate at the 250mV current-sense level.
_________________________________________________________
| |
| Win's efficient constant-current LED regulator |
| +5 - 9V |
| o--+---+--- E C ------------------+--- L1 --, |
| | | B | | |
| | R6 | Q4 .---+-- | --------+ |
| | | | | | | | |
| | '------+ R4 | | _|_ |
| | | | R3 | D1 _\_/_ |
| R5 R8 Q2 | | ,---, | |
| | | E | /_\ | LED |
| +--------- | -----+--- B | | | |
| | | | C | | | |
| | C C | | | | |
| '- R7 -- B B ---+---+-- | -- R2 --+ |
| E E | | |
| Q3 | | Q1 | | |
| o-------------+------+--------------+--- R1 --' |
| Battery Return |
|________________________________________________________|

My design includes a few subtle concepts, such as Q2's hysteresis effect,
and using the constant LED voltage for biasing. So see if you can figure
out how it works and derive the part values. :>) Note, if Q3 and Q4 are
n- and p-type mosfets (preferred), then R7 and R8 can be eliminated.

I've been trying to get this to simulate, in LTSpice. Following the same
logic that Win stated in another post, I used 22 ohms for R1. I arbitrarily
chose 10k for R2, and thus got 39k for R3. I arbitrarily chose 100k for all
the other resistors, and I used Win's value of 3.9mH for L1. LTSpice only
has two LED models in its default library and neither one is red (bizarrely,
one is a super-bright yellow-orange SMT and the other is a white), but the
difference in behavior was quantitative rather than qualitative.

To my surprise, what I'm seeing is that the oscillation behavior is
determined almost entirely by the values of R5 and R6, apparently as they
charge the gate capacitance of the MOSFET's. (I randomly chose an IRF7406
for Q4 and an IRF9410 for Q3.) Picking different R values, or choosing
MOSFETs with different gate charge, changes the frequency proportionally.

Should I have expected that? Is gate charge a reliable enough
characteristic that I should feel comfortable manufacturing something (in
quantities of a few dozen, anyway) so dependent on it? Or am I doing
something wrong with my simulation, or my choice of components?

 

[Winfield Hill]

Walter Harley wrote...

... I arbitrarily chose 10k for R2, and thus got 39k for R3.
I arbitrarily chose 100k for all the other resistors...
I randomly chose an IRF7406 ... and an IRF9410 ...

To my surprise, what I'm seeing is ... oscillation behavior...

Should I have expected that?

This is such an unacceptable approach to "engineering," I can't
even comment on it, at least not until my stomach settles down.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Mac]

Well, I think I already answered that question, in my initial post. To
answer it more fully: because I'm very unfamiliar with the premises and
ideas behind switch-mode regulators; and I know that there are many dead end
alleys to be explored, and many wrong turns to be taken. It would be fun
and fascinating, and I would learn a lot, eventually. But I have only a few
hours a week to spend on electronics, and I've got a project that needs to
be completed, and this LED circuit is just a small piece of it. So I can't
really afford to invent this circuit from first principles, unfortunately.

A more snippish answer might be "because good engineering consists in part
of knowing what solutions already exist, rather than reinventing them." Of
course, that would be a more convincing answer if I was actually capable of
designing the thing myself in a reasonable time, which I'm not yet.

I deeply appreciate the help that people on this group have given me as I
learn the various technologies I've asked about. I try to return the favor
by answering others' even simpler questions when I can.

Look, if you want this to be simple, easy, cheap and work well, or some
reasonable combination of the above, don't go switch mode. The only
justification for switch mode is a tight power budget. And the non-switch
mode is going to cost you maybe 6 or 7 mA. Only you know whether it is
worth it.

If you just want to keep the current constant over your expected voltage range
of 6 to 9 volts, use an LM117 set up as a current source. The national
semiconductor datasheet shows how to do this (just copy the 50mA constant
current battery charger, but re-engineer it for 10 mA). You can do it with
only one external component. And maybe a decoupling cap.

There are other simple ways, but this one is pretty good.

Mac
--

 

[Walter Harley]

Walter Harley wrote...

This is such an unacceptable approach to "engineering," I can't
even comment on it, at least not until my stomach settles down.

I'm sorry; I didn't mean to upset your stomach.

I think my approach was a bit more thought-out than the posting made it
seem. Here's the back story:

After thinking about the circuit, I decided I'd give it a quick shot on the
simulator to see if it behaved the way that I thought, and to see if I
understood which component values affected what. After all, in most
circuits I've built, there are a lot of values that are noncritical: for
instance, if I'm building a simple noninverting opamp circuit, the ratio
between the two gain-setting resistors is critical, but the absolute values
are less critical, although they do affect noise performance, supply drain,
stability, etc.

Q1 and Q2 were evidently being used in a way that didn't demand any special
characteristics; any small-signal NPN and PNP transistors would do fine, so
I chose on the basis of parts I have in stock. Given your suggestion that
Q3 and Q4 could be either bipolar or MOSFET, I suspected that the precise
characteristics weren't going to be critical, although clearly there were
minimum gate voltage constraints to switch, and benefits to be had from
reasonable Rds(on) for Q4. Now, I spent an hour tonight looking through
catalogs and web sites (the Siliconix/Fairchild Si3861 load switch looks
promising), but this morning I had limited time and a choice: either pick
parts from the very limited LTSpice libraries, or download and modify new
models. In through-hole components, there seems to be a wide gap between
very-low-power RF devices and high-power switching devices. Since I didn't
think these parts were critical, I simply chose power parts with gate charge
and Rds(on) that were nonexceptional (in the context of the LTSpice
libraries). Not finding a red LED in the library was a problem; but I
figured that I'd run the simulation, figure out what voltage was getting
dropped across the LED model I did have access to, and adjust R2/R3
accordingly. Failing that, two ordinary diodes in series would be a decent
approximation, for the sake of simulation.

It was clear that the absolute value of R1 mattered, and I wanted around
250mV drop at the average LED current. The ratio of R2 to R3 mattered, and
for a red LED it wanted to be about 1:4. I "arbitrarily" chose 10k for R2
because it looked like the ratio of R2:R3 was more important than the
absolute value; clearly R2 had to be much higher than R1, but low enough
that the current flow through it (along with that through R3) was enough to
turn on Q1 hard enough to turn off Q3, and that suggested only a ratio
between R5 and R2. 10k seemed a very reasonable choice. I hadn't yet
figured out R4, so that really was just a stake in the ground, based on a
ratio between R3 and R4: if R4 was too low, I'd start draining current away
from the LED, and if it was too high, it wouldn't do anything. My intent
was to come back to that - I had about 30 minutes to spend on it and wanted
to make some progress. Finally, I chose 100k for R5 and R6 because it
hadn't occurred to me that the gate capacitance of these power FETs was so
high that I had a time constant to deal with; as I've said, I'm a newbie
with MOSFETs and with SMPS's. I expected the switching frequency to be a
function of L1, the LED current, and the voltage hysteresis.

When I realized that the simulation was behaving strangely, I plotted
voltages to see what wasn't working as I expected; and noticed that the
voltage at the gates was ramping up, rather than switching abruptly. That
was enough to show me that there was a problem charging the gates, so I
changed R5 and R6 to 10k; lower than that, and the efficiency would be
terrible, since the supply is always draining to ground through one or the
other. And yet it still seemed to be the controlling factor for the
switching frequency.

I apologize again for the upset stomach, and hope I haven't upset it
further. I hope that when you recover you will comment further.

Thanks,
-walter

 

[Walter Harley]

Look, if you want this to be simple, easy, cheap and work well, or some
reasonable combination of the above, don't go switch mode. The only
justification for switch mode is a tight power budget. And the non-switch
mode is going to cost you maybe 6 or 7 mA. Only you know whether it is
worth it.

It is worth it; it's a 9v battery powered device.

Thanks,
-walter

 

[N. Thornton]

Look, if you want this to be simple, easy, cheap and work well, or some
reasonable combination of the above, don't go switch mode. The only
justification for switch mode is a tight power budget. And the non-switch
mode is going to cost you maybe 6 or 7 mA. Only you know whether it is
worth it.

If you just want to keep the current constant over your expected voltage range
of 6 to 9 volts, use an LM117 set up as a current source. The national
semiconductor datasheet shows how to do this (just copy the 50mA constant
current battery charger, but re-engineer it for 10 mA). You can do it with
only one external component. And maybe a decoupling cap.

There are other simple ways, but this one is pretty good.

Mac
--

Well, Win's circuit is pretty, tho I would also say there are easier
ways. An ultrabright LED and a resistor would I think be the easiest
way to get sub-miliiamp drain.

Or if you must go switch mode, one could make a simple oscillator with
less than 4 trs. Accurate regulation is AFAICS not needed here.

Win's circuit strikes me as closer to an IC design than a discrete way
to power a LED. But certainly interesting! And of course it has good
uses.


Regards, NT

 

[Winfield Hill]

Walter Harley wrote...

Winfield wrote ...

Walter wrote...

This is such an unacceptable approach to "engineering," I can't
even comment on it, at least not until my stomach settles down.

I'm sorry; I didn't mean to upset your stomach. [snip stuff]
I hope that when you recover you will comment further.

Sorry for my terse comments, but one simply can't do engineering
by picking parts out of a hat! Clearly the FET capacitances and
the circuit's charge/discharge currents are key to making a high-
speed dc-dc converter (if 200kHz is high speed). Before doing a
Spice model, one should do some simple calculations to determine
the efficacy of the component choices. And availablility in the
model library is certainly NOT an acceptable criteria!

Given that we have a 10mA load, up to 1mA control current wouldn't
be unreasonable, so you could use down to 10k for the gate pullup.

The FETs you chose are 5 to 7A monsters with very high capacitance,
up to 1500pF for the gate-source and 1000pF for the drain (at 5V)!
according to the data-sheet curves. Hey, with 100k resistors this
is a molasses 250us time constant, not at all suitable! Clearly
small mosfets will be required, such as the original Supertex vn01
and vp01 we mention in AoE, or the classic nmos 2n7000 (I suggest
the Fairchild data sheet) with a bs250, etc, for the p-type switch
(Vishay). These are spec'd at about 23pF Ciss and 11pF Coss and
have about 1nC of gate-drain charge. A 10k gate pullup resistor
generates about 2V/10k = 0.2mA of gate discharge current and can
be expected to achieve a turn-off time of about t = Q/i = 5us.

Hmm, acceptable for use in a 25kHz switcher perhaps, but a little
slow for our circuit. We could add a gate-drive circuit with two
more parts, but it appears the smallest common MOSFETs are too big,
and we'd be better off sticking with just bipolar transistors. :>)

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Walter Harley]

Sorry for my terse comments, but one simply can't do engineering
by picking parts out of a hat! Clearly the FET capacitances and
the circuit's charge/discharge currents are key to making a high-
speed dc-dc converter (if 200kHz is high speed).

It's that "if" that I hadn't realized. I don't have many intuitions about
DC-DC converters, and the ones I do have appear to be wrong.

By the way, looking at the datasheet, the Si3861 load switcher I'd thought
had promise, doesn't; it's too slow. Based on an hour or so of hunting
through web sites and datasheets, I wasn't able to come up with MOSFETs that
would work. I'll have another go with bipolars, as you suggest.

Thanks again, Win. Hope your stomach's better.

-walter

 

[Walter Harley]

Well, Win's circuit is pretty, tho I would also say there are easier
ways. An ultrabright LED and a resistor would I think be the easiest
way to get sub-milliamp drain.

That would certainly be an easier fix, but I've not been able to find
ultrabright LED's in the packaging that I desire. The device in question
gets used by musicians, who ironically tend to be hard on physical devices;
so I've been using LED's that have metal panel mounting reflector bezels.
I've found that plastic bezels break too easily: the LED's get shoved back
into the chassis. And simply PC-mounting the LED and poking it through a
hole in the panel barely survives the music store.

I have thought about trying to buy both, and pull the normal LED's out of
the metal bezel and reassemble it with the high-brightness ones. But that's
a lot of labor.

[Hmm. Poking around on Lumex' web site just now, I see they've got an
appropriately-packaged LED that's 100mcd at 20mA, which is better than the
one I'm using now, though far from the 1000+mcd that some ultrabrights
claim. I think I missed it before now because its specs are misprinted in
the Digikey catalog.]

I've also thought about entirely scrapping the "power on" light (which is
what this is) and putting in some sort of clever circuit to turn the power
off automagically after a period of nonuse. It might end up using less of
the battery than the LED does, although I'd have to work pretty hard for it
to be cheaper. Meantime, I figured I'd learn something by exploring the
switching-regulator option.

-walter

 

[Robert Monsen]

I'm playing with a step down switcher design I put together, using a
capacitor as the energy storage device. It uses a micropower uC to control
the energy flow. There is a vp0106 to gate in the charge. Seems pretty
simple (a relative novice like myself whacked it together in about an hour.)
Perhaps too simple. Does this work?

My guess as to why this would be more efficient than, say, a resistor is
that the cap allows the voltage across the vp0106 and 10 ohm resistor to be
lower during the energy transfer phase. The voltage quickly builds up during
the charging phase (something like 2.75V from the nominal charge level in
12uS.) The cap then bleeds the charge back to ground through the LED,
quickly dropping the voltage to a lower value during the long pause between
charges. Thus, the total power required is less cause IV is less.

In order to validate that this does indeed result in better efficiency, I'd
like to measure the power used by the whole circuit, and also by the LED.
Given my tools (multimeter, oscilloscope, signal generator,) is there a
simple way to measure these values? (I know, stone knives and bear skins...)

VCC
+
|
+-------+
| |
.-. .-.
| | | |
10 | | | | 100k
'-' '-'
| |
| |
| |
+-||----+
vp0106 <-|| |
+-|| |
| |
+-------+ |
| | | 1MEG
| | \| ___
3.3uF | _|_ |-----|___|--[0,5] from uC
--- \ / <|
--- --- | 17.5ms low, 12us high
| | | open loop (for now)
| | .-.
+---+---+ | |
| | | 10k
| '-'
+---------+
|
===
GND
created by Andy´s ASCII-Circuit v1.24.140803 Beta www.tech-chat.de

The strange duty cycle comes from the PIC's watchdog timer. I put it into
sleep mode to save power between pulses. The watchdog timer wakes it up
after something like 18ms. Another odd thing is the uPower 5V regulator I'm
using to power the PIC (an S81250SG, which, together with the PIC, is not
shown.) Both of these, together, seem to draw less than 10uA on average, so
I'm neglecting the power consumption of these devices (90uW is small
compared to what the LED uses.)

Depending on whether this makes sense, and if I can measure the efficiency,
I can easily add some feedback using the onboard comparator to stabilize the
voltage by controlling how many uS to power the cap. This will,
unfortunately, up the power requirements of the PIC slightly due to the
energy cost of charging the sample and hold circuit, and powering the
comparator.

Regards,
Bob Monsen

 

[Spehro Pefhany]

I'm playing with a step down switcher design I put together, using a
capacitor as the energy storage device. It uses a micropower uC to control
the energy flow. There is a vp0106 to gate in the charge. Seems pretty
simple (a relative novice like myself whacked it together in about an hour.)
Perhaps too simple. Does this work?

My guess as to why this would be more efficient than, say, a resistor is
that the cap allows the voltage across the vp0106 and 10 ohm resistor to be
lower during the energy transfer phase. The voltage quickly builds up during
the charging phase (something like 2.75V from the nominal charge level in
12uS.) The cap then bleeds the charge back to ground through the LED,
quickly dropping the voltage to a lower value during the long pause between
charges. Thus, the total power required is less cause IV is less.

In order to validate that this does indeed result in better efficiency, I'd
like to measure the power used by the whole circuit, and also by the LED.
Given my tools (multimeter, oscilloscope, signal generator,) is there a
simple way to measure these values? (I know, stone knives and bear skins...)

Here's a simple averaging circuit that you can use to get some idea.
It will drop the Vdd a bit; you can increase it slightly to compensate
so that Vdd to the circuit is whatever it should be (eg. 5.00VDC).

DVM reads 1mV/uA average; power = Vdd * Iavg

+ DVM -
+---------------+--------------o o
| | |
.-. | + |
| | 10K 5% ### 100uF Low Leak |
| | --- +---------+
'-' | |
Vdd | ___ | |
o---+---|___|-------+-----------+--o Vdd to circuit
| + |
1K00 1% ### |
--- |
| 1000uF ---
| --- 100n ceramic
| |
=== ===
GND GND


When you've got a number, try a 15K resistor from the LED anode to Vdd
and compare power consumption and LED brightness..

Best regards,
Spehro Pefhany

 

[Walter Harley]

I'm playing with a step down switcher design I put together, using a
capacitor as the energy storage device. It uses a micropower uC to control
the energy flow. There is a vp0106 to gate in the charge. Seems pretty
simple (a relative novice like myself whacked it together in about an hour.)
Perhaps too simple. Does this work?

My guess as to why this would be more efficient than, say, a resistor is
that the cap allows the voltage across the vp0106 and 10 ohm resistor to be
lower during the energy transfer phase. The voltage quickly builds up during
the charging phase (something like 2.75V from the nominal charge level in
12uS.) The cap then bleeds the charge back to ground through the LED,
quickly dropping the voltage to a lower value during the long pause between
charges. Thus, the total power required is less cause IV is less.

Um, I don't think so, if I understood. You're still dropping the same
amount of voltage across the resistor; you're just doing it in short bursts
of high current, rather than continually. But it's the same average
dissipation. The voltage across the resistor is Vcc minus the capacitor
voltage; and the capacitor voltage must be approximately the same as the
LED's threshold voltage, or else you blow up the LED. You need something
there to limit the current: either a resistor, which is lossy, or an
inductor, which isn't.

I think if you want to do this with capacitors, you need to make a
capacitive voltage divider, i.e., charge a bank of capacitors in series, and
then discharge them in parallel (or one at a time). One capacitor won't do
it.

Also, my recent experience leads me to be suspicious of a 100k resistor
trying to charge a MOSFET gate in the 12uS on-time I don't know what the
gate charge of a vp0106 is, but since it's switching a peak current of
(Vcc - Vled) / 10, which might be more than half an amp, it must be
relatively large.

And, pulling short pulses of very high current from the battery means you're
wasting power in the internal resistance of the battery, not to mention
putting a lot of hash on the power supply wiring. What we're aiming for
here is to pull a steady, low current at relatively high voltage from the
battery; then transform to higher current at lower voltage.

-walter

 

[Winfield Hill]

Walter Harley wrote...

I don't know what the gate charge of a vp0106 is, but since it's
switching a peak current of (Vcc - Vled) / 10, which might be more
than half an amp, it must be relatively large.

Actually it's pretty low. The VP01 is one of the parts I suggested
for a small p-channel mosfet. It's input capacitance is about 45pF
and it only requires about 0.3nC to charge / discharge and switch,
see http://www.supertex.com/pdf/datasheets/VP0106.pdf I've been
a longtime fan of the Supertex VN01 and VP01 FETs, and you'll find
plots of their Id - vs - Vgs relationships in our book, page 123.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Zak]

I've also thought about entirely scrapping the "power on" light (which is
what this is) and putting in some sort of clever circuit to turn the power
off automagically after a period of nonuse.

There exist switches with a really clear 'eye' indicator that shows
whetehr they are engaged or not. They contain a brightly colored part
that is covered by shutters when teh switch is off. The bright
fluorecent green to black transition is quite clear.

Otherwise it might be possible to use an HE LED on the PCB, behind a
clear window of some sort for protection.


Thomas

 

[Robert Monsen]

On Wed, 29 Oct 2003 18:54:02 GMT, the renowned "Robert Monsen"

[snip request on way to measure efficiency, and reason to do this]

DVM reads 1mV/uA average; power = Vdd * Iavg

+ DVM -
+---------------+--------------o o
| | |
.-. | + |
| | 10K 5% ### 100uF Low Leak |
| | --- +---------+
'-' | |
Vdd | ___ | |
o---+---|___|-------+-----------+--o Vdd to circuit
| + |
1K00 1% ### |
--- |
| 1000uF ---
| --- 100n ceramic
| |
=== ===
GND GND


When you've got a number, try a 15K resistor from the LED anode to Vdd
and compare power consumption and LED brightness..

Best regards,
Spehro Pefhany

Thanks. Despite Walter's misgivings about the circuit in general, I am going
to try doing this. However, I'd like to be able to measure the LED 'in
circuit', since I think that getting the brightness equal will be
problematic. I seem to have acquired an AN633 multiplier chip somehow, and
thought I might be able to build a nice power meter using it:

cygwin\emacs-21.2\bin/

VCC
+
|
|
+---------+
| unknown |
+---------+
|
+--------------- X1
|
.-.
| | Load to measure
| |
'-'
|
+--------------- X2 and Y1
|
.-.
| | 0.1
| |
'-'
|
+---------------- Y2
|
===
GND




AD633
X1+----+ (X1-X2)x(Y1-Y2)
X2| |------------------ Z
Y1| |
Y2+----+


C
+--||----+
| |
R | |\| |
Z --o o-/\/\/\---+--|-\ |
\ | >--+---- Out
+---|+/
| |/|
|
===
GND

Its actually an energy meter, which integrates the amount of energy
that has passed a certain point in the circuit. The switch will be some
kind of jfet I think. The timing will be done using a PIC.

3 questions:

1) Will it measure what I think it'll measure?
2) Will the component values matter that much (ie, the cap, resistors, etc?)
I don't have any low-leakage caps, unfortunately.
3) Will the AN633 and whatever opamp be fast enough to be accurate?

I have an MCP62914 opamp, and an LF555. Which might be better for this
application? Will it even matter? (I guess thats 4 questions...)

I think I just want to get a voltage reading for the energy for a given time
across the LED, then another reading for the same time across the entire
circuit, and divide them to get the efficiency...

Regards,
Bob Monsen

 

[Walter Harley]

"Winfield Hill" wrote...

Walter Harley wrote...

Actually it's pretty low. The VP01 is one of the parts I suggested
for a small p-channel mosfet. It's input capacitance is about 45pF
and it only requires about 0.3nC to charge / discharge and switch,
see http://www.supertex.com/pdf/datasheets/VP0106.pdf I've been
a longtime fan of the Supertex VN01 and VP01 FETs, and you'll find
plots of their Id - vs - Vgs relationships in our book, page 123.

(I really shouldn't be doing this at work... oh well.) From that page, I
see that it's rated 0.8A max pulsed current. If his Vcc is coming from a 9v
battery, he's right on the edge, at least if he ignores the internal
impedance of the battery and the on resistance of the FET.

But it specs Rds(on) at 11 ohms typical, for Vgs = 5v. So I guess I
shouldn't worry; with 2 ohms from the battery, 11 ohms from the FET, 10 ohms
from the resistor, and a couple ohms ESR from the capacitor, the peak
current will be only 300mA. No sweat. (No efficiency, either, I believe;
but I already argued that.)

At about US$0.50 in unit quantities from Mouser, that's a nice component,
handy for switching the sort of midrange currents we've been talking about;
I remember you mentioning it in AoE. When hunting for MOSFETs I figured
that one of the few things about electronics that might have changed
substantially since AoE 2 was published was the state of the art in
affordable/available MOSFETs, so I wasn't sure it was still a good reference
for that. Has so little changed? Or is it just that the changes have
focused on very-high-current components?

Thanks,
-walter

 

[Winfield Hill]

Walter Harley wrote...

Meantime, I figured I'd learn something by exploring
the switching-regulator option.

If this is for a real (as opposed to philosophiocal)
application, you may have more interest in a minimum
parts-count solution. I asserted that most adjustable
IC switchers have a 1.25V sense voltage, which if used
directly with the sense resistor would waste 44% of the
power applied to the series LED and resistor.

But Linear Technolgy offers their LTC1779, a low-power
buck regulator with an 800mV Vref, which reduces the
sense-resistor loss to 33%. It costs $2.25, small qty.
http://www.linear.com/prod/datasheet.html?datasheet=666

LTC also offers the LTC3405A with an 800mV reference,
but it doesn't allow for inputs higher than 6V.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Walter Harley]

"Winfield Hill" ...

If this is for a real (as opposed to philosophiocal)
application, you may have more interest in a minimum
parts-count solution.

I have both a real application and a philosophical one. The real
application is a battery-powered headphone amplifier (used by electric
bassists when practicing) that I make and sell. I make 50-100 of them a
year - not exactly the big time but enough that manufacturing efficiency and
parts cost matter. Right now, the power LED is a big factor in the battery
life.

The philosophical application is that since switching amps and switching
power supplies seem to be important technologies these days, I'm trying to
learn more about how they work and how to design them. I figured I'd start
with something small and practical, before moving on to high-powered
designs.

[...] Linear Technolgy offers their LTC1779, a low-power
buck regulator with an 800mV Vref, which reduces the
sense-resistor loss to 33%.

I'm happy to have the solutions to the two goals be different I'll look
more closely at that datasheet later tonight.

Thanks yet again.

-walter

 

[Jim Thompson]

"Winfield Hill" ...

If this is for a real (as opposed to philosophiocal)
application, you may have more interest in a minimum
parts-count solution.

I have both a real application and a philosophical one. The real
application is a battery-powered headphone amplifier (used by electric
bassists when practicing) that I make and sell. I make 50-100 of them a
year - not exactly the big time but enough that manufacturing efficiency and
parts cost matter. Right now, the power LED is a big factor in the battery
life.

The philosophical application is that since switching amps and switching
power supplies seem to be important technologies these days, I'm trying to
learn more about how they work and how to design them. I figured I'd start
with something small and practical, before moving on to high-powered
designs.

[...] Linear Technolgy offers their LTC1779, a low-power
buck regulator with an 800mV Vref, which reduces the
sense-resistor loss to 33%.

I'm happy to have the solutions to the two goals be different I'll look
more closely at that datasheet later tonight.

Thanks yet again.

-walter

If it's just to indicate power is on, why not use a flasher of some
kind? The average power can be made quite low. Numerous designs have
been posted to alt.binaries.schematics.electronic

...Jim Thompson