Searching \ for '[EE:] Son of simple step-up SMPS challenge' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/index.htm?key=son+simple+step
Search entire site for: 'Son of simple step-up SMPS challenge'.

Exact match. Not showing close matches.
PICList Thread
'[EE:] Son of simple step-up SMPS challenge'
2002\10\26@081816 by Russell McMahon

face
flavicon
face
part 1 13824 bytes content-type:text/plain; (decoded 7bit)

Challenge:

   Convert low voltage DC such as battery or 3v or 5v rail
   to "somewhat higher" dc voltage at modest power levels
   for e.g. processor operation, FET gate drive, LED drive,
   Op-Amp supply, Programmer Vpp supply, and more.

Somewhat over a year ago I proposed a design challenge for a simple circuit
to step up low voltages to somewhat higher ones for a range of applications.
I believe there was and is a demand for such capability and that such a
circuit may even provide functionality not readily achieved with available
ICs. While there was some initial response to this challenge it died an
early death, probably due to the events of September 11th.

I'd like to re-propose it and suggest a range of applications and target
parameters. Those interested are welcome top redefine the targets more
broadly or into several categories if felt desirable.

I have a specific prospective application in mind for my own use (4v to 12v
converter for FET gate drive) but this is only one of many possible
applications. I also would like a low cost, compact, efficient white LED
driver that provides essentially constant output and reliable starting over
a full range of single cell voltages (<0.9V - > 1.5V)

Ideally a single core design will allow a variety of input and output
voltages and power levels with suitable simple component changes. Low cost,
(non)complexity, efficiency, low parts count, gee-whiz factor, single
inductor, ...  all score brownie points.

Target power level is modest and will vary somewhat with application. e.g. a
white LED may require 10 to 20 mA at 3v, A FET gate supply a few mA at 12v,
an opamp supply 1 to 10 mA at 12v, a processor 5 to 20 mA at 3v to 5v etc.

Target efficiency is "as high as possible and as low as appropriate". e.g. a
design that rings an inductor and uses a clamp zener to dissipate any
surplus energy would have poor efficiency at less than full power output but
such an approach may well be acceptable in many applications.

A highly desirable aim is use of a single inductor with a single winding
although multi winding inductors or multiple inductors may well be
acceptable. (Single winding inductors at this power level are widely
available and low cost).

Lowest possible starting and operating voltage is a desirable aim although
the importance of this will vary with application (e.g. a 3v processor
supply
to 12v FET drive application only requires that the circuit start and run
properly on the 3v supply.) A "run the 1.5v cell utterly dry" design  will
ideally run on under 0.9v and will hopefully also start on less than a volt.

Number and type of active devices is "free" although obviously the less the
better commensurate with meeting other design aims. I anticipate that most
designs would use one to 3 transistors although more may be appropriate. Use
of an IC may be appropriate, although this may limit lower starting and
operating voltage. That said, a 4069 CMOS inverter IC has a minimum Vdd of 3
volt (but you will almost certainly need at least 1 transistor as well to
allow "flyback" voltage to be used).

Low cost is desirable although the real cost of a given design will vary
with use - e.g. a commercial product will consider PCB area and component
installation costs.

Too complex a design will be beaten in cost and simplicity by commercial
alternatives. Few ICs aim at the very low power / low voltage / low cost
market so there is a niche. The ability to start and run at voltages under 1
volt would place the design in a relatively exclusive club.


_______________


As my initial contribution to this challenge I submit my comments and
circuit
from a year ago for a white (or other) LED driver circuit (led1cel2.gif),
plus a modified version of this designed to produce a positive output
voltage and allow higher input voltages (smps512.gif) While I have named the
diagrams SMPS512 it in fact has NO means of regulation shown and no RC
component values shown. R2 is added to the LED flasher version to increase
Q2 turn on time during the inductor drive pulse and R3 is added to stop Q2
trying to apply full supply to Q1 base :-).

NB - neither of these circuits is optimal.
*** THE CIRCUIT DOES OSCILLATE ***
The circuit has been built and tested in practice.
This is not a paper only dream.
(Hopefully that makes it clear enough :-)
The above added because this circuit not surprisingly attracts detractors
due no doubt to its interesting mechanism of operation.

I imagine (but have not yet tested) that a zener from point g to point c
would provide regulation of sorts. When out exceeded approx Vin + Vzener -
0.6 Q2 would be held off. This has the disadvantage of reducing Vout with
decreasing Vin - less of a problem for Vout >> Vin. In most applications an
output zener shunt regulator would suffice.

Brief description of operation.

NB      R1 is sized large enough to NOT be able to support max drive
required by Q2 to keep Q1 turned on at peak inductor current. !!!!!!!!!!
It's function is to START on drive and possibly provide timing.

R2 small wrt R1

Letters a to h refer to equivalently labelled points on circuit diagram.
Startup. Q1, Q2 off.
c, f at Vin as C1 uncharged.
C1.c charges downwards via R1.
Q2 turned on via R1+R2 when b reaches approx Vin-0.6 (setting lower bound on
starting voltage).
Q2 turning on turns on Q1 via R3.
Q1 turning on pulls f and therefore c low via C1, turning Q2 harder on
causing regenerative turn on.
c now increases as C1 charges via base Q2,
Current increases in L (approx a ramp)
As drive to Q1 DECREASES as C1 charges, drive to Q2 (BetaQ1 x !bq1)
decreases until current in L cannot be supported by Q2. As current in Q2
prevents current ramp up in L field starts to collapse and inductor starts
to 'ring".
f rising increases c further adding to regenerative turn off.
f will ring either to
   - limit set by load
   - limit set by clamp output zener (not shown)
   - limit set by LC in tank cct (as no C across L c is only parasitic and
small)
   - reverse breakdown Q2 be junction (R2 reduces this prospect)
In practice the first two effects are most likely unless there is no load at
all.

With f held high c will now start downwards due to R1 (something like
bricklayer and barrel story :-) )
This will be hastened with increasing load as f will decay sooner.
Voltage at c will sooner or later reach starting point and cycle repeats.

Notes:
- R2 & R3 are refinements from LED driver version but don't alter basic mode
of operation.
- R1 size and Q2 beta are important factors in achieving turnoff.
- IF L saturates it should do so not too far below the R1-Q2beta limit or
there will be large unproductive current spikes in Q1 collector/inductor.
Saturation is not necessary but provides another rmeans of triggering
regenerative turnoff/ring phase.
Efficiency is not liable to be fantastic but Roman can probably tweak it to
get 80% plus :-)


++++++++++++++++++++++++++++++++++++++++++++++

LED flasher / driver

FROM LAST YEAR

           ********   Before we start:   ************

       - This is a REAL operating circuit.
       - This circuit DOES oscillate (& very well) in practice *.
       - It IS possible to describe a formal feedback mechanism and mode of
          operation.

                                   OK

WHAT IT DOES:

                                    1 cell to LED driver

                   or              Low voltage to higher voltage step up

                   or              Flasher

   With minimal parts count.

I suspect this wins the minimum component count with a single winding
inductor as per Roman's spec.

I put this under the above heading as in some ways it's a continuation of
that theme.
It also addresses the LED torch and LED from 1 cell applications.

I expect that Roman & Alice will have lots of fun further developing this
circuit, Jinx will use it in 3 unexpected applications in the next two weeks
and xxx & yyy will have lots of fun "commenting" on it :-).

This is based on a very time honoured flasher circuit which I have used for
other applications.in the past.
This is as stripped down as it seems to be possible to get it with an
inductor added to provide voltage step up.
Adding some components (typically 1 or 2 resistors) alters and may improve
performance.

Other applications might be a voltage step up for e.g. 5 --> 12v for
programming, solenoid drive as per Jinx's recent application, RS232 supply
etc.
I haven't optimised this or measured efficiencies but as shown it has a
remarkably square drive waveform and may even be somewhat efficient.

OUTPUT

As shown the inductor "rings" when Q1 is turned off delivering NEGATIVE
output below ground.
To reverse the circuit to supply positive output above Vin swap Q1 and Q2
types and swap ground and Vin connections.

LOAD

To use as a voltage supply replace the LED with a diode and filter
capacitor.
Usually only a single LED would be used.
LED1 is driven solely by the flyback voltage
LED2 sees" both flyback voltage and input voltage.

Arrangement 2 is more efficient as the input voltage is added to the flyback
and this part of the supplied voltage is essentially "100% efficient"

The LED1 arrangement has the advantage that if Vin exceeds LED normal
forward voltage somewhat (say up to 5 volts) the LED will still operate at
less than destruction current. Efficiency will suffer in this mode but we
now have a LED that will operate from Vin = 0.7 volts to Vin = ???

When operated with only an inductor as load (no LED) my example rings to
about 25 volts limited (probably) by L1 to C1 ratio and scope and stray
capacitive loading.

A LED which will not "glimmer" whatsoever when connected to a single cell
can be run across the whole cell operating range.

COMPONENT VALUES

I have not shown component values (but see example below) as performance is
immensely affected by component choice.
The circuit is reasonably "designable" but the operation is surprisingly
tricky considering the component count.

Rather than play with this further I thought I would release it to the eager
masses (well, Roman and Alice anyway :-) ) as they are much more likely to
extend and optimise it than I am at present.

EXAMPLE:

Single no-name brand le Clanche AA cell (standard penlight battery) half
flat. V = 1.2 volt
L1 = 330 uH miniature choke **
R2 = 1M
C1 = 100 pF
Fosc approx 10 kHz
HP high brightness Red LED
I_Battery approx 4 mA
Surprisingly bright.

** - Dick Smith Electronics R5234
A slightly physically larger 2.5 mH inductor gives somewhat brighter output
at somewhat increased current. I gain the impression that the higher
inductance is more efficient (as would be expected).

CAPACITOR

I have shown C1 as an electrolytic to denote polarity.
When used for a continuous supply (as here) the cap will be so small that a
non electrolytic will invariably be used.
When used as a flasher (see below) an electrolytic may be appropriate.

VARIATIONS.

This circuit can be extended and amended vastly.
A few guidelines:

Placing a resistor "R1" in series with C1 will have a significant effect on
discharge times (as it removes the Q2 Vbe clamp effect on the capacitor).

Placing a resistor between Q2 collector and Q1 base (try 1K to start) will
affect discharge and on cycle times.

This circuit delivers a negative voltage relative to ground.
A mirror image circuit may be built by swapping Q1 with Q2 and ground with
Vin to make a circuit providing voltage ABOVE Vin.

L1 may be altered significantly, varying energy storage for a given
frequency.

Changing frequency will affect energy delivered and therefore LED brightness
(or available output power)

Adding resistors to alter mark space (charge discharge) will affect power
delivered.
.
OPERATING VOLTAGE

Oscillation starts at about 0.65 volts but with components shown above
doesn't give notable LED brightness till about 0.8 to 0.9 volts in.
More is better.
Output waveform squares up nicely by Vin = 0.8 volt or so.

EFFICIENCY

Doesn't look marvellous.
Didn't do formal tests but some rough measurements suggest well under 50%
efficiency !!!!
If so, should be able to be improved substantially, at cost of extra
complexity.
.

REGULATION

No ! :-)

If used as a voltage supply I suspect that a simple zener and series R
connected from output to an appropriate transistor base will allow the
oscillator to be disabled when desired Vout is reached.

Given the power levels involved a simple shunt zener may be better and
easier. If low power maintenance of a voltage is desired then the zener
scheme would allow the oscillator to "burst" as required to maintain
voltage.

FLASHER

When C is made large (say 1 uF range) the frequency of operation will be so
low that individual output pulses will be individually distinguishable. In
this case, provided the energy in the inductor is adequate, the LED
'flashes". The design will need to be arranged to provide requisite energy.

WORTHWHILE?

Possibly not - but it's fun.
A simple 2 transistor cross coupled multivibrator would use  a few more
parts but be rather more designable.
Probably worth a look though.


SO

OK - over to Roman, Alice, Jinx, ... - any other improvers out there ???



regards

               Russell McMahon



_______________________


* For oscillation the current in R2 *Beta Q1 x Beta Q2 MUST be lower than
the load current.
This means that it will stop oscillating when the load resistance is
increased above this level.
The low resistance inductor load generally meets this requirement.







----------------------------------------------------------------------------
----





part 2 1637 bytes content-type:image/gif; (decode)


part 3 2639 bytes content-type:image/gif; (decode)


part 4 131 bytes
--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\26@105900 by Andy Kunz

flavicon
face
>I have a specific prospective application in mind for my own use (4v to 12v
>converter for FET gate drive) but this is only one of many possible
>applications. I also would like a low cost, compact, efficient white LED
>driver that provides essentially constant output and reliable starting over
>a full range of single cell voltages (<0.9V - > 1.5V)

I use an LT1930 for these.  Pretty cheap, whole circuit fits in about 1cm
square, VERY clean output.  2.5V to 15V input, 2.5 to 25V output, 1A input
current limit.

With a little ingenuity (op-amp across a shunt), you can use it as a
constant-current source as well.

Use the circuit from their proto board rather than the data sheet.  It's
cleaner.

Andy

-------------------------------------------------------------------
Race Boats  - spam_OUTandyTakeThisOuTspamRC-Hydros.com      http://www.RC-Hydros.com
Airplanes   - .....andyKILLspamspam@spam@FlyingHobbies.com  http://www.FlyingHobbies.com
Electronics - andyspamKILLspamMontanaDesign.com  http://www.MontanaDesign.com
-------------------------------------------------------------------

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\26@113158 by mark

flavicon
face
On 27 Oct 2002 at 1:16, Russell McMahon wrote:

>
> I have a specific prospective application in mind for my own use (4v to 12v
> converter for FET gate drive) but this is only one of many possible
> applications. I also would like a low cost, compact, efficient white LED
> driver that provides essentially constant output and reliable starting over
> a full range of single cell voltages (<0.9V - > 1.5V)
>

Something like this could be a start:

http://www.emanator.demon.co.uk/bigclive/joule.htm

Marcelo

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\26@121830 by Dave Tweed

face
flavicon
face
Russell McMahon <.....apptechKILLspamspam.....PARADISE.NET.NZ> wrote:

> Challenge:
>
>     Convert low voltage DC such as battery or 3v or 5v rail
>     to "somewhat higher" dc voltage at modest power levels
>     for e.g. processor operation, FET gate drive, LED drive,
>     Op-Amp supply, Programmer Vpp supply, and more.

I have a requirement to take the output of a Ni-MH battery pack that has a
terminal voltage of 3-4 volts and convert this with good efficiency to 5V
at up to 100 mA. I came up with a 2-transistor circuit that I posted on
10/1 (reproduced below). It's remarkably similar to your circuit, with the
addition of output voltage regulation.

I achieved regulation by sourcing the emitter of the PNP from a fixed
reference. DC feeback from the output node keeps the transistors turned off
whenever the output voltage rises too high. In order to keep the reference
current low, I use a "capacitor discharge" circuit to provide the base
current for the NPN. The combination of falling base current and rising
coil current initiates the switch-off, which is then sped up by positive
feeback through the 1nF capacitor.

Efficiency is good, but not great. With no load (i.e., just the feedback
divider, ~0.3 mA), the circuit draws about 4 mA. About half of that is the
reference circuit. With moderate loads (10-30 mA), I'm getting about 55-60%
efficiency, which seems low. I'm using a cheap choke that has about 3 ohms
resistance, and the peak coil current is 200 mA. I figure that accounts for
at least half of my losses.

                                      diode
   3-4 V o----+------------LLLLL---+---AK----+---+----o 5V @ 100 mA
              |            68 uH   |         |   |+
              R                    +------+  R  --- 100uF
         1000 R                    |  .001|  R  ---
              R      PNP           |     --- R   |
              |             100    C     --- |   V
              +--+---E C---RRRRR--B  NPN  |  |  Gnd
              |  |    B            E      +--+
       LM336  Z ---   |            |      |  |
        -2.5  Z ---   |            V      |  R
              Z  |.047|    2200   Gnd     |  R
              |  V    +---RRRRR-----------+  R
              V Gnd                          |
             Gnd                             V
                                            Gnd

I'm going to try the following circuit next. I've added an emitter follower
to reduce the quiescent current of the reference supply, eliminated the
capacitor that discharges into the base of the main switch, and added a
diode that disconnects the output voltage feedback during the switching
cycle. Also, I now have a much better coil: a 47 uH toroid with just 50
milliohms resistance.

                                       diode
   3-4 V o----+---+-------LLLLL---+----AK------+---+----o 5V @ 100 mA
              |   |       47 uH   |            |   |+
              R   |               +-----+      R  --- 100uF
          12K R   |               | 330p|      R  ---
              R   |NPN            |    ---     R   |
              |   C               |    ---     |   V
              +--B    PNP  390    C     |      |  Gnd
              |   E--E C--RRRRR--B  NPN +--AK--+
       LM336  Z       B           E     |diode |
        -2.5  Z       |           |     |      R
              Z       |           V     |      R
              |       |    68K   Gnd    |      R
              V       +---RRRRR---------+      |
             Gnd                               V
                                              Gnd

However, for really high efficiency and low parts count, it's hard to beat
ICs like the Maxim MAX1678 ($1.89 in 100s from Digi-Key), which can
maintain 80% efficiency all the way down to 1 mA of load current in this
application. I got a couple of samples and will be testing it out soon as
well.

-- Dave Tweed

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\26@130305 by Dave Dilatush

picon face
Russell wrote...

>Challenge:
>
>    Convert low voltage DC such as battery or 3v or 5v rail
>    to "somewhat higher" dc voltage at modest power levels
>    for e.g. processor operation, FET gate drive, LED drive,
>    Op-Amp supply, Programmer Vpp supply, and more.

[snip]

>WORTHWHILE?
>
>Possibly not - but it's fun.

Good summation, Russell.

In addition to the LT1930 that Andy Kunz mentioned, Maxim also
has a number of chips that perform this function, such as the MAX
608 and the MAX1522/3/4 series.  None are terribly expensive, and
all are easy to use and perform well.

Some things just aren't worth "rolling your own" anymore, other
than for the sheer joy of dabbling.

DD

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\26@162033 by Peter L. Peres

picon face
part 1 1362 bytes content-type:TEXT/PLAIN; charset=US-ASCII
I don't know about efficiency but you have to know that you are using a
circuit that exists since times immemorial afaik. You can remove the
resistor in series with the base of Q1 to save one part. Then the base
current of Q1 will be limited by Rl. This circuit has been used for
everything from metronomes to incandescent light flashers to sirens afaik.
I have never seen it used with a coil (so far). The reason is that the
current in the load assumes a triangular waveform for a lot of the
switching time and this is bad for efficiency.

The one-transistor flyback converters that are used in flashes etc should
serve as a better starting point for this kind of circuit imho. The basic
schematic is like that of the led driver with a toroid with 2 wires that
was linked on the piclist before. Flashing (pulsing) output and voltage
regulation are possible.

Since these involve a transformer and this seems to cause ppl to avoid
them here is an (untried) solution to implement the transformer using two
(identical or not, but same sized) ready-made chokes (see attached
drawing):

This is not tried as such but I have sound reasons to believe that it will
work fine. I have posted a similar circuit to Alice C. some time ( more
than a year ?) ago. The circuit using a transformer is well tried
(obviously). Efficiency is not bad and not measured ;-).

Peter


part 2 9603 bytes content-type:APPLICATION/PDF; name="ledflash1.pdf" (decode)

part 3 131 bytes
--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\27@050403 by Roman Black

flavicon
face
Russell McMahon wrote:
>
> Challenge:
>
>     Convert low voltage DC such as battery or 3v or 5v rail
>     to "somewhat higher" dc voltage at modest power levels
>     for e.g. processor operation, FET gate drive, LED drive,
>     Op-Amp supply, Programmer Vpp supply, and more.

> Target efficiency is "as high as possible and as low as appropriate". e.g. a
> design that rings an inductor and uses a clamp zener to dissipate any
> surplus energy would have poor efficiency at less than full power output but
> such an approach may well be acceptable in many applications.

> Lowest possible starting and operating voltage
> A "run the 1.5v cell utterly dry" design  will
> ideally run on under 0.9v and will hopefully also start on less than a volt.


Hi Russell, when I was fiddling with the 2-tran
SMPS circuits I tried a 2-tran boost circuit to
run from a single 1.5v cell.

I got it starting (and running) reliably from around
0.8v up to 1.6v, giving (about) 5v out. I say "about"
5v as Vout:Vin and Vout:Iout regulation were not
great. Efficiency was about 55% to 60% with 1v input.

But it did run a 5v PIC from a cell, and was mega
cheap with 2 cheap transistors and one simple inductor
and 1N4148 boost diode.

At the time I was not really impressed with the
efficiency or regulation, and didn't bother writing
it up properly. Lately I bought one of the Dick Smith
1.5 -> 9v converter kits, and was disappointed to find
that even with specialist chip it only run down to
about 1.2v input, efficiency only 60% and regulation
was quite poor also!

I still have the circuit and some tested figures
if you think it qualifies for your requirements.
-Roman

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email EraseMElistservspam_OUTspamTakeThisOuTmitvma.mit.edu with SET PICList DIGEST in the body


2002\10\27@050819 by Roman Black

flavicon
face
Dave Dilatush wrote:

> In addition to the LT1930 that Andy Kunz mentioned, Maxim also
> has a number of chips that perform this function, such as the MAX
> 608 and the MAX1522/3/4 series.  None are terribly expensive, and
> all are easy to use and perform well.
>
> Some things just aren't worth "rolling your own" anymore, other
> than for the sheer joy of dabbling.


Unless of course you are a manufacturer and
want to replace a $1.50 chip with $0.06 worth
of transistors. ;o)
-Roman

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email listservspamspam_OUTmitvma.mit.edu with SET PICList DIGEST in the body


2002\10\27@142710 by Dave Dilatush

picon face
part 1 4331 bytes content-type:text/plain; charset=us-ascii (decoded 7bit)

Roman wrote...
>Dave Dilatush wrote:
>
>> In addition to the LT1930 that Andy Kunz mentioned, Maxim also
>> has a number of chips that perform this function, such as the MAX
>> 608 and the MAX1522/3/4 series.  None are terribly expensive, and
>> all are easy to use and perform well.
>>
>> Some things just aren't worth "rolling your own" anymore, other
>> than for the sheer joy of dabbling.
>
>Unless of course you are a manufacturer and
>want to replace a $1.50 chip with $0.06 worth
>of transistors. ;o)

Well, sort of: as Russell pointed out, the circuit he posted
(smps512.gif) isn't complete yet since it has no means of
regulating its output voltage.  For one possible complete design
based on what Russell posted, see the attached.  In this case,
I've (more or less) optimized the circuit to generate a 5 volt
output from a single alkaline cell.  Zener diode D2 sets the
output voltage by providing feedback through transistor Q3.
Diode D3 has been added to prevent B-E reverse breakdown of Q2
during flyback.  I haven't built it but simulation suggests it
ought to work, with roughly 60% efficiency at 5 mA load current.


So we've got 3 transistors, 2 diodes, 1 Zener diode, four
resistors and 2 capacitors.  That's a dozen components, versus
whatever the IC solution takes.

This discrete solution is probably still cheaper than what the IC
and its related components would cost, but a manufacturer (at
least, any manufacturer who's smart enough to stay in business
very long) will have other considerations besides the purchase
price of the components when comparing alternative designs.

Component count and circuit board area may be an issue, as well
as reliability- the more parts a design has, the more likely
something will fail.  If one design has a lot more components
than another, will the extra components fit in the available
space?

Production volume is a consideration, too: is it worth spending
several weeks (maybe even more) of engineering time to come up
with a robust switching regulator design using discrete
components, when a single IC can do the job?  If volume is high
enough, the answer may be "Hell, yeah!"; if volume is low, it
might be "Hell, no!"; and in between, the question bears careful
examination.

Time to market may be another consideration.  Is it really
worthwhile tying up engineering manpower working on a circuit
function that could just as easily be purchased as a chip,
instead of attending to other parts of the design that may be
much more crucial to the product's success?  Sometimes the extra
delay in getting a product out on the market is acceptable; other
times it's not.

Circuit performance may be an issue- indeed, it may be the
biggest issue in something like this: most of the IC switching
regulators available today will outperform one of these
"3-transistor wonder" circuits by a wide, WIDE margin in terms of
line and load regulation, output voltage accuracy, temperature
stability, efficiency, input voltage range and fault tolerance.
For some applications, the 3-transistor wonder may be adequate;
for others, it may fall pathetically short of being good enough.

What I like to call "quality of design" comprises a set of issues
that affect production yield and field return rate- things like
design margins, component de-rating, sensitivity to component
value variations (both unit-to-unit and over temperature),
typical characteristics vs. worst-case, adherence to design rules
and the like.  One design rule this 3-transistor wonder circuit
runs afoul of (specifically with regard to Q1) is pretty
common-sense and widely observed: never rely on transistor beta
as a means of limiting current flow, because it varies too much
from unit to unit, from lot to lot, and over temperature.

Experience has taught me over the years that it frequently pays
to overdesign voltage regulators- that is, to intentionally make
them better than apparently required.  Operating voltage tends to
affect many aspects of circuit performance in subtle,
hard-to-troubleshoot ways and a good, tightly regulated supply
voltage in a product can make for a lot fewer production
headaches.  Just my observation; YMMV, of course.

DD


part 2 3469 bytes content-type:image/gif; name=stepup1.gif (decode)


part 3 136 bytes
--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email @spam@listservKILLspamspammitvma.mit.edu with SET PICList DIGEST in the body


2002\10\27@171533 by Dave King

flavicon
face
At 08:01 PM 27/10/02 +1100, you wrote:
>Dave Dilatush wrote:
>
> > In addition to the LT1930 that Andy Kunz mentioned, Maxim also
> > has a number of chips that perform this function, such as the MAX
> > 608 and the MAX1522/3/4 series.  None are terribly expensive, and
> > all are easy to use and perform well.
> >
> > Some things just aren't worth "rolling your own" anymore, other
> > than for the sheer joy of dabbling.
>
>
>Unless of course you are a manufacturer and
>want to replace a $1.50 chip with $0.06 worth
>of transistors. ;o)
>-Roman

Uh try $4.50 +MBR750 zener cost +

6 cents seems like a hell of a deal ;-]

Dave

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
email KILLspamlistservKILLspamspammitvma.mit.edu with SET PICList DIGEST in the body


2002\10\28@071433 by Russell McMahon

face
flavicon
face
Let's have a look at some of the comments so far and see how well they match
the spec. As noted originally, the spec can be extended if people see fit.

Original spec was for a DC-DC step up from either single cell or a low
voltage rail to a LED, a processor supply, an op amp rail, a FET gate drive
or a programming supply. Current requirement would be from about 1 mA to 20
mA depending on the specific application.

I'll use Digikey prices unless otherwise noted. DK are not the cheapest
source but give a consistent reference point.

Decoupling caps on input and output not counted in parts counts.


               Russell McMahon


==========
Andy
> I use an LT1930 for these.  Pretty cheap, whole circuit fits in about 1cm
> square, VERY clean output.  2.5V to 15V input, 2.5 to 25V output, 1A input
> current limit.

Nice IC
Operation at 1 MHz+ allows small inductors.
Far too dear - $4 in 1's and $1.90/1k
High current compared to spec (1A+ switch)
EXTERNAL Schottky needed.
Vin 2.6v minimum (guaranteed)
36 volt switch so good for over 30v out

Needs 3 cells for operation at dead flat.
Not really suited to LED application.
OK for uP, FET, Vpp, OpAmp
Somewhat overkill powerwise in most apps.

1 IC, 1D, 2R, 1L
_______________
Marcelo
> Something like this could be a start:
> http://www.emanator.demon.co.uk/bigclive/joule.htm

This almost unbelieveable circuit is the "Joule Thief" single cell to White
LED driver also mentioned here a few weeks ago.
Main "problem" with this design is the non-standard two winding inductor. If
this is tolerated then a wide range of multi-winding designs become
possible.

This is about as low a parts count as you could get. (1 Tr, 1R, 1 x 2
winding L !!!)
Efficiency not known but will probably be poor due to hard to optimise
design. Adding parts still will produce a low parts count design.

IF efficiency is tolerable this may be a real winner for most of the
applications.

_____________________
Dave T
> I have a requirement to take the output of a Ni-MH battery pack that has a
> terminal voltage of 3-4 volts and convert this with good efficiency to 5V
> at up to 100 mA. I came up with a 2-transistor circuit that I posted on
> 10/1 (reproduced below). It's remarkably similar to your circuit, with the
> addition of output voltage regulation.

Discrete design with
2 TR, 5R, 1Z, 2C, 1L
Achieves step up and regulation with 2 transistors. Relies on beta
saturation or inductor saturation for startup but apparently starts well.
Efficiency below that of IC designs. Could be improved with extra work (and
probably another transistor).

____________
Dave T
> However, for really high efficiency and low parts count, it's hard to beat
> ICs like the Maxim MAX1678 ($1.89 in 100s from Digi-Key), which can
> maintain 80% efficiency all the way down to 1 mA of load current in this
> application.

Digikey $US1.90 / 1k according to me (may depend on version)

Another nice IC, but also with limitations.
Aimed at 1 cell or 2 cell opertation.
0.87v guaranteed startup (excellent!)
Inbult synchronous rectification and main switch.
Min confign uses 1 IC, 1L !!!
Add 2 R for voltage setting.
Add 2 R for optional power failure / low-battery input sense.
Add 1R for optional power fail out drive.

Vin max is sadly only 5.5V
Vout max is STUPIDLY only 5.5v making it useful for LED and procesor,
marginal for opamp and probably useless for FET and Vpp. A really nice
design made much less useful by an unnecessarily narrow spec.

Efficiencies up to 90% at 50 mA / 2.5v in / 3.3v out (due in large part to
synchronous rectification).

Worth a close look for battery to white LED drivers
eg 80% efficiency at 0.85v in, 3.3v out, 20 mA !!!!!!!!!!!!!!

____________________________
Peter
> I don't know about efficiency but you have to know that you are using a
> circuit that exists since times immemorial afaik.

"Very time honoured" was how I put it part way through the post. I 1st saw
it used as a mose code oscillator several decades ago. I've never seen it
with a "rung" inductor in the output.

> The one-transistor flyback converters that are used in flashes etc should
> serve as a better starting point for this kind of circuit imho..........
> Since these involve a transformer and this seems to cause ppl to avoid
> them here is an (untried) solution to implement the transformer using two
> (identical or not, but same sized) ready-made chokes (see attached
> drawing):

I tried coupling two chokes by placing them in proximity before now without
enough success. removing the platic coating to achieve ferrite to ferrite
coupling may help.
_____________________________________
Roman:
{Quote hidden}

Please supply. The more the merrier.

______________________
Dave D
I'm glad Dave decided to join the party after initial reservations :-)
Nice addition.
{Quote hidden}

The regulation takes 1 Z, 1R, 1 tr.
This could be dropped to 1 Z if you don't mind shunt regulating the output.
This destroys efficiency but may be an entirely viable solution for eg 5v to
Vpp, FET drive or Op amp supply applications.
D3 can almost certainly be omitted in most applications but the reason for
its inclusion is appreciated.

------------------------------------------
Someone (maybe several people?) mentioned MAX1522 / 1523 / 1524
$US1.70 / 1k est by Maxim (AFAIR)

Nice IC but parts count may even exceed discrete designs.
External diode AND main FET !!!
2.5v min Vin or 1.5v with bootstrap supply feedback.
Standard design needs 1 IC, 1L, 1 FET, 1 D, 2 R, 2C.
Bootstrap (1.5v start) version needs 1 IC, 1L, 1 FET, 1 D, 3 R, 3C.

Vin max 5.5v
Vout limited by external pass FET & diode (eg 28v easily achieved)
Power level set by external FET & diode
Low quiescent currents (25 uA typical)
Soft start and short-circuit output shutdown modes

Useful IC. While external parts add to parts count they allow user
selectable Vout and power levels.
____________________________

MC34063A, MC33063A
Mentioned by someone recently.
A very old design and suffers in several respects compared to modern ICs BUT
very very cheap compared and more flexible than many. Can be used for step
up or step down, quite a low voltage internal reference for older IC
allowing low voltage operation.

** $US0.30 ** approx even in small quantities (from the right place) !

Bipolar.
Emitter/Collector of driver available for flexibility.
Current sense in high side supply.
3v to 40v input
External diode, internal 1.5A switch or drives external switch.
8 pin dip/so-8, soeia-j8
100 kHz max so larger inductors than best modern practice.
Quiescent current in 1 to 4 mA range typically (depends on Vin)
The bipolar darlington output and modest saturation and beta specs will not
produce the ultimate efficiency designs.

Can produce eg 28v 150 mA using internal switch.

Unsuited to flat battery operation (3v min)
Suited to 3v or 5v in.
No good for LED drive
Opamp, FET, Vpp drive apps.

CHEAP !!!!!!!!!!!!!!!!!!!!!!!!
(Maybe not as cheap parts wise as bottom end discrete, but close)

Min step up cct    1 IC, 1 L, 3 R, 1 D, 1 C
Current limit add 1 R

Min step down cct    1 IC, 1 L, 2 R, 1 D, 1 C

____________________________

LT1937
Dedicated white LED converter
Nobody mentioned these (yet)

$US2.50/1, $US.130/1k


LT1937 is designed to drive 2, 3 or 4 LEDs in series with constant current
feed.
1.2 MHz for small inductor size.
Targetted at Li-Ion supply.
2.5v to 10 v vin

Could be used for other apps.

__________________________

--
http://www.piclist.com hint: To leave the PICList
RemoveMEpiclist-unsubscribe-requestTakeThisOuTspammitvma.mit.edu


2002\10\28@075618 by Alan B. Pearce

face picon face
>MC34063A, MC33063A
>Mentioned by someone recently.
>A very old design and suffers in several respects compared to modern ICs
BUT

This chip is a cut down version of the 78S40. By using a 78S40 you do get an
additional diode (not schottky though) and op-amp. Do not know about prices
though.

--
http://www.piclist.com hint: To leave the PICList
spamBeGonepiclist-unsubscribe-requestspamBeGonespammitvma.mit.edu


2002\10\28@122047 by Peter L. Peres

picon face
On Mon, 28 Oct 2002, Alan B. Pearce wrote:

*>>MC34063A, MC33063A
*>>Mentioned by someone recently.
*>>A very old design and suffers in several respects compared to modern ICs
*>BUT
*>
*>This chip is a cut down version of the 78S40. By using a 78S40 you do get an
*>additional diode (not schottky though) and op-amp. Do not know about prices
*>though.

No ?! MC34063A is a full featured fixed frequency SMPSU (without the
diode), 78S40 uses the (infamous)  Schmitt principle and is variable
frequency. For low cost, low power a slow diode can be used instead of the
Shottky (eg 1N4004 - I tried this). This will lower efficiency but also
noise (the diode switches slower). The efficiency problem should be lowest
with step-up converters.

Incidentally imho the 78S40 diode is a poor joke even on a 1N4004.

Peter

--
http://www.piclist.com hint: To leave the PICList
TakeThisOuTpiclist-unsubscribe-requestEraseMEspamspam_OUTmitvma.mit.edu


2002\10\28@122047 by Peter L. Peres

picon face
On Tue, 29 Oct 2002, Russell McMahon wrote:

*>Peter
*>> I don't know about efficiency but you have to know that you are using a
*>> circuit that exists since times immemorial afaik.
*>
*>"Very time honoured" was how I put it part way through the post. I 1st saw
*>it used as a mose code oscillator several decades ago. I've never seen it
*>with a "rung" inductor in the output.
*>
*>> The one-transistor flyback converters that are used in flashes etc should
*>> serve as a better starting point for this kind of circuit imho..........
*>> Since these involve a transformer and this seems to cause ppl to avoid
*>> them here is an (untried) solution to implement the transformer using two
*>> (identical or not, but same sized) ready-made chokes (see attached
*>> drawing):
*>
*>I tried coupling two chokes by placing them in proximity before now without
*>enough success. removing the platic coating to achieve ferrite to ferrite
*>coupling may help.

I tried the same and it worked. I used bare ferrite dumbell chokes and
wrapped the ends with soft tinned steel plate strips (about 0.3mm thick).
When measured with a generator and load this seems to show k ~= 0.5. I
have not tried this in an oscillator. The transformer driven oscillator
circuit is one half of the Royer (?) oscillator which drives nearly all
laptop and fl light drivers in the world (except high efficiency Royer
circuits are driven by a specialised IC to forced secondary resonance).

Peter

--
http://www.piclist.com hint: To leave the PICList
RemoveMEpiclist-unsubscribe-requestspamTakeThisOuTmitvma.mit.edu


2002\10\28@123131 by Alan B. Pearce

face picon face
No ?! MC34063A is a full featured fixed frequency SMPSU (without the
diode), 78S40 uses the (infamous)  Schmitt principle and is variable
frequency. For low cost, low power a slow diode can be used instead of the


Well try this link and see what you think then :))
http://www.onsemi.com/pub/Collateral/AN920-D.PDF

They specifically talk of them they way I suggested. On the second page
under "general description it specifically says they are the same with the
differences I noted.

--
http://www.piclist.com hint: To leave the PICList
piclist-unsubscribe-requestEraseMEspam.....mitvma.mit.edu


2002\10\28@125841 by Peter L. Peres

picon face
On Mon, 28 Oct 2002, Alan B. Pearce wrote:

*>Well try this link and see what you think then :))
*>http://www.onsemi.com/pub/Collateral/AN920-D.PDF

Ok, I stand corrected. Figure 26 shows a step-up converter with >90%
efficiency using the MC3... part and a tapped inductor.

Peter

--
http://www.piclist.com hint: To leave the PICList
EraseMEpiclist-unsubscribe-requestspammitvma.mit.edu


2002\10\29@054157 by Russell McMahon

face
flavicon
face
> Ok, I stand corrected. Figure 26 shows a step-up converter with >90%
> efficiency using the MC3... part and a tapped inductor.

The tapped inductor is required for if best possible efficiency is required
due to the dual factors of a NPN high side switch and a Darlington output.
You need to provide a modest high side drive voltage ABOVE Vin effective to
get a saturated switch. The tapped inductor effectively gives you this. If
you don't provide such an arrangement the output switch is not saturated and
at lower input voltages the consequent efficiency loss is significant.

       RM

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads


2002\10\30@110054 by Russell McMahon

face
flavicon
face
White (or other) LED flashlights - 1 - 2 & 2  - 4 cell designs.


http://mikro.e-technik.uni-ulm.de/persons/lares/LED_flashlight.html#V1

Nicely done albeit a little commercial in rendition :-)

University of Ulm.
Two different IC's used.
65% to almost 80% efficiency depending on battery / LED combinations.

.

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


2002\10\31@004100 by Roman Black

flavicon
face
Russell McMahon wrote:

> Original spec was for a DC-DC step up from either single cell or a low
> voltage rail to a LED, a processor supply


Hi, here is an "entry" for Russell's SMPS challenge.
It's not a high performance circuit (yet) but is MEGA
CHEAP and will run a PIC from a single 1.5v cell
(like an AA cell which are the cheapest battery).

The circuit WILL start and run from as little as 0.9v.
It is the SIMPLEST CHEAPEST circuit I could get to run
on 1v or less with a single "off the shelf" inductor.


               470uH
               0.6 ohm            1N4148
In   --*----*------L---*------*-------D--------*----*----
0.9v   |    |          |      |      a k       |    |  +5v
to     |    |          |      |                |    |  out
1.8v   |    |          |      |                |    |
      |    |          |  1.5 |                |    |
      R1   R2 15k     |  nF  C2               |    |
  120 |    |          |      |                |    |
  ohm |    '----------|------*                |    |
      |               |      |            5.1v|    C
      |               |      |              Zener  |10uF
      |          ,----|------|-----,          |    |
      |          |    |      |     |          |    |
      |          |    C      |     C          |    |
      '----------*--B        *---B   Q2       |    |
               Q1     E      |     E NPN      |    |
              NPN     |  1nF |     | BC337    |    |
             BC337    |      C1    |          |    |
                      |      |     |          |    |  Gnd
Gnd ------------------*------*-----*----------*----*----


Oscillation starts as Q1 turns on first, as it
has a low base R1 and no base capacitance. After
Q1 is on, C1 will eventually charge via R2 and
when Q2 turns on it kills Q1.

When Q1 turns off the inductor back emf keeps
Q2 on via feedback through C2 and C1. As the
back emf finishes and the inductor voltage drops
C2/C1 feedback turns Q2 off and the cycle repeats.

Unlike some other boost circuits oscillation is
maintained by a flip/flop type action with each
transistor turning the other off and with an
added R2:C1 time constant, quite similar to my
Black regulator in concept. ;o)

Output power depends on the oscillation which is
mainly determined by input voltage. The circuit
above is more suitable for low power PIC up to
a few mA, MAINLY good for loads that draw a more
"constant" current and are switched on via a
mechanical switch.


----------------------------------------------------------------

The next circuit uses the same parts but now the
zener is used to turn off Q2. This means that the
circuit tailors oscillation to load current and it
is good for more output current than the last
circuit. However voltage regulation is not as good
as the previous circuit.

(Note! A third circuit was tested using an extra
transistor and resistor with the zener diode,
this gave better regulation and performance but
at increased parts cost)



               470uH
               0.6 ohm            1N4148
In   --*----*------L---*------*-------D--------*----*----
0.9v   |    |          |      |                |    |  +5v
to     |    |          |      |                |    |  out
1.8v   |    |          |      |                |    |
      |    |          |  1.5 |              Zener  |
      R    R 15k      |  nF  C           4.6v |    |
  120 |    |          |      |                |    |
  ohm |    '----------|------*----------------'    |
      |               |      |                     C
      |               |      |                     |10uF
      |          ,----|------|-----,               |
      |          |    |      |     |               |
      |          |    C      |     C               |
      '----------*--B        *---B   Q2            |
               Q1     E      |     E NPN           |
              NPN     |  1nF |     | BC337         |
             BC337    |      C     |               |
                      |      |     |               |  Gnd
Gnd ------------------*------*-----*---------------*----



Here are the figures for the 2nd circuit tested
for three input voltages;


Vin     Iin     Vout    Iout
1.6v    11.6mA  4.69v   0mA     (quiescent)
       18.3    4.50    2
       24.8    4.33    4
       31.5    4.19    6
       38.4    4.05    8
       45.0    3.95    10

1.2v    8.4mA   4.60v   0mA     (quiescent)
       17.2    4.34    2
       26.0    4.12    4
       34.7    3.97    6
       44.5    3.78    8
       53.3    3.65    10

0.9v    5.6mA   4.45v   0mA     (quiescent)
       17.5    4.10    2
       29.6    3.68    4
       32.8    2.47    6    (R1/Q1 beta not enough)


Notes; the circuit is quite dependant on the
value of R1 being low enough to provide good turn
on of Q1. The test above was a very quick test
to see if operation was reliable from 0.9v input
to 1.8v input range.

At the higher input voltage R1 is too low and wastes
a lot of quiescent current. Once input voltage drops
R1 needs to be a low value to continue good hard
oscillation.

For a narrower input voltage band (say 1.2v to 1.5v)
it is possible to tune the circuit for much better
quiescent current and regulation improves too.

I used a $0.03 BC337 transistor with measured
low-current beta of 220. With a specialty high-gain
transistor for Q1 most of these problems would
be solved and only cost some cents more.

I did not spend any time trying to get this thing
working well, just a quick test to see if a PIC
could be run from a single cell for a few cents in
parts.

Issues;
* R1 depends on input voltage range
* Q1 needs decent gain
* voltage regulation is not great
* output cap must be small
* no input cap (needs mech switched 1.5v input)

Improvements;
* extra transistor + R fixed regulation
* high gain Q1 will fix quiescent current and
 improve efficiency
-Roman

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\31@070356 by Russell McMahon

face
flavicon
face
part 1 923 bytes content-type:text/plain; (decoded 7bit)

Self oscillating 4 transistor boost converter.

Haven't looked at it in detail but looks promising.
Arguably pushing upper complexity limit in most cases compared with IC
design.

From


http://ourworld.compuserve.com/homepages/Bill_Bowden/page4.htm#ps5.gif

They say

In this small switching power supply, a Schmitt trigger oscillator is used
to drive a switching transistor that supplies current to a small inductor.
Energy is stored in the inductor while the transistor is on, and released
into the load circuit when the transistor switches off. The output voltage
is dependent on the load resistance and is limited by a zener diode that
stops the oscillator when the voltage reaches about 14 volts. Higher or
lower voltages can be obtained by adjusting the voltage divider that feeds
the zener diode. The efficiency is about 80% using a high Q inductor.





       Russell McMahon


part 2 3689 bytes content-type:image/gif; (decode)


part 3 131 bytes
--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\31@210413 by Roman Black

flavicon
face
Russell McMahon wrote:

> Original spec was for a DC-DC step up from either single cell or a low
> voltage rail to a LED, a processor supply


Hi, here is an "entry" for Russell's SMPS challenge.
It's not a high performance circuit (yet) but is MEGA
CHEAP and will run a PIC from a single 1.5v cell
(like an AA cell which are the cheapest battery).

The circuit WILL start and run from as little as 0.9v.
It is the SIMPLEST CHEAPEST circuit I could get to run
on 1v or less with a single "off the shelf" inductor.


               470uH
               0.6 ohm            1N4148
In   --*----*------L---*------*-------D--------*----*----
0.9v   |    |          |      |      a k       |    |  +5v
to     |    |          |      |                |    |  out
1.8v   |    |          |      |                |    |
      |    |          |  1.5 |                |    |
      R1   R2 15k     |  nF  C2               |    |
  120 |    |          |      |                |    |
  ohm |    '----------|------*                |    |
      |               |      |            5.1v|    C
      |               |      |              Zener  |10uF
      |          ,----|------|-----,          |    |
      |          |    |      |     |          |    |
      |          |    C      |     C          |    |
      '----------*--B        *---B   Q2       |    |
               Q1     E      |     E NPN      |    |
              NPN     |  1nF |     | BC337    |    |
             BC337    |      C1    |          |    |
                      |      |     |          |    |  Gnd
Gnd ------------------*------*-----*----------*----*----


Oscillation starts as Q1 turns on first, as it
has a low base R1 and no base capacitance. After
Q1 is on, C1 will eventually charge via R2 and
when Q2 turns on it kills Q1.

When Q1 turns off the inductor back emf keeps
Q2 on via feedback through C2 and C1. As the
back emf finishes and the inductor voltage drops
C2/C1 feedback turns Q2 off and the cycle repeats.

Unlike some other boost circuits oscillation is
maintained by a flip/flop type action with each
transistor turning the other off and with an
added R2:C1 time constant, quite similar to my
Black regulator in concept. ;o)

Output power depends on the oscillation which is
mainly determined by input voltage. The circuit
above is more suitable for low power PIC up to
a few mA, MAINLY good for loads that draw a more
"constant" current and are switched on via a
mechanical switch.

(improved circuit is in the next email)
-Roman

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\31@210416 by Roman Black

flavicon
face
(this is the second half of the post)

The next circuit uses the same parts but now the
zener is used to turn off Q2. This means that the
circuit tailors oscillation to load current and it
is good for more output current than the last
circuit. However voltage regulation is not as good
as the previous circuit.

(Note! A third circuit was tested using an extra
transistor and resistor with the zener diode,
this gave better regulation and performance but
at increased parts cost)



               470uH
               0.6 ohm            1N4148
In   --*----*------L---*------*-------D--------*----*----
0.9v   |    |          |      |                |    |  +5v
to     |    |          |      |                |    |  out
1.8v   |    |          |      |                |    |
      |    |          |  1.5 |              Zener  |
      R    R 15k      |  nF  C           4.6v |    |
  120 |    |          |      |                |    |
  ohm |    '----------|------*----------------'    |
      |               |      |                     C
      |               |      |                     |10uF
      |          ,----|------|-----,               |
      |          |    |      |     |               |
      |          |    C      |     C               |
      '----------*--B        *---B   Q2            |
               Q1     E      |     E NPN           |
              NPN     |  1nF |     | BC337         |
             BC337    |      C     |               |
                      |      |     |               |  Gnd
Gnd ------------------*------*-----*---------------*----



Here are the figures for the 2nd circuit tested
for three input voltages;


Vin     Iin     Vout    Iout
1.6v    11.6mA  4.69v   0mA     (quiescent)
       18.3    4.50    2
       24.8    4.33    4
       31.5    4.19    6
       38.4    4.05    8
       45.0    3.95    10

1.2v    8.4mA   4.60v   0mA     (quiescent)
       17.2    4.34    2
       26.0    4.12    4
       34.7    3.97    6
       44.5    3.78    8
       53.3    3.65    10

0.9v    5.6mA   4.45v   0mA     (quiescent)
       17.5    4.10    2
       29.6    3.68    4
       32.8    2.47    6    (R1/Q1 beta not enough)


Notes; the circuit is quite dependant on the
value of R1 being low enough to provide good turn
on of Q1. The test above was a very quick test
to see if operation was reliable from 0.9v input
to 1.8v input range.

At the higher input voltage R1 is too low and wastes
a lot of quiescent current. Once input voltage drops
R1 needs to be a low value to continue good hard
oscillation.

For a narrower input voltage band (say 1.2v to 1.5v)
it is possible to tune the circuit for much better
quiescent current and regulation improves too.

I used a $0.03 BC337 transistor with measured
low-current beta of 220. With a specialty high-gain
transistor for Q1 most of these problems would
be solved and only cost some cents more.

I did not spend any time trying to get this thing
working well, just a quick test to see if a PIC
could be run from a single cell for a few cents in
parts.

Issues;
* R1 depends on input voltage range
* Q1 needs decent gain
* voltage regulation is not great
* output cap must be small
* no input cap (needs mech switched 1.5v input)

Improvements;
* extra transistor + R fixed regulation
* high gain Q1 will fix quiescent current and
 improve efficiency
-Roman

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.


2002\10\31@234109 by Russell McMahon

face
flavicon
face
> Hi, here is an "entry" for Russell's SMPS challenge.
> It's not a high performance circuit (yet) but is MEGA
> CHEAP and will run a PIC from a single 1.5v cell
> (like an AA cell which are the cheapest battery).

These designs look excellent as a starting point.
This basic 2 transistor design will almost certainly fill the  Vpp, Opamp
rail, & FET gate drive applications. For these fixed input voltage,
relatively fixed output voltage applications, efficiency should be able to
be optimised. Due to the low passive parts count, the addition of a third
transistor is no great disaster.

As both transistors in the basic design share grounded emitters this would
lend itself to a surface mount dual transistor package with commoned
emitters.

I can feel a moderate power version coming on using the ULN2803 / 2003 or
variants. Unusual enough an application to cause some real puzzlement on
first encounter. There is enough power handling capability there to drive
significant loads eg possibly automotive cellphone or notebook chargers that
require more than 12 volts. Note that some of the less seen ULN2xxx variants
use higher value input resistors, allowing lower losses in the front ends.
The darlington configuration is a slight shame but not a fatal flaw. Maybe
even the internal catch diodes can be used as the output diode !!! :-) Using
a 3rd section / transistor as a voltage regulator, using a grounded emitter
resistor driven by a zener to output to clamp the base of Q1, allows a full
regulator using half a ULN2003 / 2004 / .... This means two such ccts could
be implemented in a  single IC. BC337 etc has a cheaper component cost but
more parts.

The design is perhaps less suited for the single cell to LED application in
its present state as its output trails off rather badly as Vin drops.
Probably another transistor to add some dynamic drive variation with Vin
would help. A multi-cell LED version may be OK. The droop of current
capability with voltage out would allow a LED to find its own operating
point rather better than a design with good hard regulation which would then
need current control.

5v to Vpp: Discrete version:

   1 IC (dual SOTxxx  transistor), 2 R, 2 C, 1 D, 1 L. 1 Z.

ULN2003 version (theoretical at this stage)

   1 IC (1/2 actually), 1 R, 2 C, 1 L, 1 Z

  Assumes R1 is internal & R2 added.
  Assumes internal catch diode OK (Add 1 D if no go)

A ULN200x version may allow a low voltage optimised version which starts on
0.9 or less volts to "pump up" / bootstrap  the Vin_local for a second
higher power optimised version which then takes over. This is precisely what
is done in some commercial designs which are intended to operate down to
very low Vin. The inductor ALWAYS draws power from the true Vin but the main
controller runs from a higher voltage bootstrapped Vin thereby allowing far
more efficient operation. Once the main system is running the startup supply
can be disabled. Component count could still be modest and startup to below
commercially realisable voltages may be possible. For the standard dry cell
(Le Clanche / Carbon Zinc / Alkaline etc) the usual endpoint is considered
to be 0.9v but useable energy is still there down to about 0.8V. A converter
that ran WELL at say 1V would be much better for eg NiCd applications where
the bulk of energy is available at lower voltages for std primary dry
batteries. Good low voltage operation allows a minimum number of solar cells
to be boosted to useful voltage levels.



       Russell McMahon

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.



'[EE:] Son of simple step-up SMPS challenge'
2002\11\02@072734 by Russell McMahon
face
flavicon
face
part 1 2140 bytes content-type:text/plain; (decoded 7bit)

*** Yes! This circuit DOES oscillate as shown ***

Here's a simple RC oscillator that MAY show promise for driving very low
voltage boost inverters.

The basic idea came from the 4 transistor boost convertern cicruit I posted
a few days ago but this has been changed so that the oscillator uses two
transistors only. This is essentially a 2 transistor Schmitt trigger with
feedback. Values shown work but are typical only and will need to be adapted
to a specific design.

The oscillator starts well at Vin down to less than 0.9 volt

Adding R5 & Q3 in place of R4 with an inductor to ground from collector of
Q3 gave enthusisatic negative going pulses (as would bee xpected). I played
with, rectification and a LED load. Efficiency was not marvellous but can
probably be made quite respectable. The oscillator proper can be designed to
have low current drain.

While its function may not be obvious to the uninitiated, Q2 is absolutely
vital for oscillation. It serves to provide Schmitt hysteresis by providing
a lower voltage across R1 with Q1 off, Q2 on and a higher voltage with Q1
on, Q2 off.

The only critical design demand for oscillation is that !q1_on is lower than
!q2_on. Fir fully saturated transistors tghis is satisfied by R3 rather
larger than R4 as shown.

The advantage and potential disadvantage of this circuit is that operation
is RC time constant controlled, independent of inductor flyback timing. This
can allow confidence in coil driving and on time etc but care must be taken
with duty cycle allowed and not driving the inductor long enough to allow it
to be severely saturated.

The circuit may be redrawn as a lontailed pair cross coupled oscillator (of
sorts) although the connection of both bases to one collector and no bases
to the other may bend the brain somewhat until it is appreciated that
feedback is via the common emitter coupling.

I tried an inductor in place of R4 and while this did produce flyback action
it was not very satisfactory. More experimentation MAY produce a useable
design. I wonder what Roman can do with this? :-)


       Russell McMahon


part 2 3359 bytes content-type:image/gif; (decode)


part 3 105 bytes
--
http://www.piclist.com hint: To leave the PICList
RemoveMEpiclist-unsubscribe-requestEraseMEspamEraseMEmitvma.mit.edu


2002\11\03@095604 by Roman Black

flavicon
face
Russell McMahon wrote:
>
> *** Yes! This circuit DOES oscillate as shown ***
>
> Here's a simple RC oscillator that MAY show promise for driving very low
> voltage boost inverters.

> The advantage and potential disadvantage of this circuit is that operation
> is RC time constant controlled, independent of inductor flyback timing. This
> can allow confidence in coil driving and on time etc but care must be taken
> with duty cycle allowed and not driving the inductor long enough to allow it
> to be severely saturated.

> I tried an inductor in place of R4 and while this did produce flyback action
> it was not very satisfactory. More experimentation MAY produce a useable
> design. I wonder what Roman can do with this? :-)


I'm guessing from looking at the circuit that
it wouldn't saturate well ie not a great square
wave in the inductor??

Also the emitter resistor is a similar problem to
the 5v -> 13v converter I posted some time back,
good for oscillation but can be a major problem
when you are trying to run the thing from 1v or
less. ie, if replacing R4 with a boost inductor
what voltage appears across R1 when in the ON
state? You are only going to get inductor to
pass uA at best??

The last circuit I posted has no emitter resistor
and the base resistor was only 120 ohms or so
as this enables it to drive Q1 into hard saturation
even with input voltages of 1v or less.

With your circuit at (say) 50mA ON you would need
R1 to be very low, say 4 ohms, giving 0.2v lost on
R1 and Q2 base at 0.6+0.2 =0.8v only giving 0.2v
across R3 or 60uA available to drive Q2, which
obviously won't saturate Q2!

I think it is really going to need the 3rd transistor
to be a viable SMPS power circuit, and still don't
see any real advantages over my (or Dave Tweed's)
2-tran circuit. :o)
-Roman

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads


2002\11\04@043522 by Russell McMahon

face
flavicon
face
> > Here's a simple RC oscillator that MAY show promise for driving very low
> > voltage boost inverters.

> I'm guessing from looking at the circuit that
> it wouldn't saturate well ie not a great square
> wave in the inductor??

Not with the 2 transistor design, as you would expect. I only tried that at
the end of the session just to see how it would go. A 2 transistor version
could be optimised but I don't see that as its target. When used as a driver
for a 3rd transistor base you can get emough drive for saturated operation.

> I think it is really going to need the 3rd transistor
> to be a viable SMPS power circuit, and still don't
> see any real advantages over my (or Dave Tweed's)
> 2-tran circuit. :o)

As I noted, it's a two edge sword. As I noted, it frees up the oscillation
frequency and control from the inductor characteristics or how you drive the
inductor. This means it's hard to get every last skerrick * of performance
from the inductor BUT it would probably make it easier to drive it well
without the drive tailing off at the ends. (This is essentially a Schmitt
trigger oscillator which should allow.reasonably good transitions due to
regenerative action).

If you want absolute minimum component count the 2 transistor design is
probably a better bet.


       Russell





______________________

* skerrick n.
(usu. with neg.) esp. Austral. colloq. the smallest bit (not a skerrick
left). [N. Engl. dial.; orig. uncert.]

The Oxford English Reference Dictionary, ) Oxford University Press 1996

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics


2002\11\04@085302 by Roman Black

flavicon
face
Russell McMahon wrote:
>
> > > Here's a simple RC oscillator that MAY show promise for driving very low
> > > voltage boost inverters.

> As I noted, it's a two edge sword. As I noted, it frees up the oscillation
> frequency and control from the inductor characteristics or how you drive the
> inductor. This means it's hard to get every last skerrick * of performance
> from the inductor BUT it would probably make it easier to drive it well
> without the drive tailing off at the ends


Ok so the main use you see is as an oscillator
to drive the power stage. I'm still not sold, with
low voltage input wouldn't it be best to tailor
oscillation more around the inductor current,
so that when the switch turns OFF you get a nice
packet of energy to the output?

It is an interesting circuit though. :o)
-Roman

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics



'[EE:] Son of simple step-up SMPS challenge'
2003\03\22@080456 by Russell McMahon
face
flavicon
face
Elektor magazine, January 2003 issue, pages 54 - 58 has an excellent article
"Small DC-DC converters" by Prof Dr Ing M Ossman.

He describes a cross section of 2 and 3 transistor circuits that cover the
range of small smps that we have discussed under this and similar threads.
Some remarkable albeit not titally unexpected similarities to circuits
various people arrived at independently.

Circuits covered are:

3 transistor buck down conveter (eg 5V) to LED constant current
(ref Electronic Design Aug 7 2000 p130)

3 transistor boost up converter (5v to 12v example)

2 transistor buck down converter (20v to 12v example)
Understandably it has many similarities with Richard's and Roman's
circuits - you can only do so many things with 2 transistors..
ref Electronic Design, Feb 6 1995 p118.


2 transistor boost (1.2v to +/-5v (capacitor pump for -5)

2 transistor flyback inverting plsu isolated output
(+5 to -12 plus isolated +12 example)

Basic introduction ok for beginners.
Treatment of cores is very minimal.
Overall and excellent article and worth looking at by anone interested in
this area.




           Russell McMahon

--
http://www.piclist.com hint: To leave the PICList
RemoveMEpiclist-unsubscribe-requestspam_OUTspamKILLspammitvma.mit.edu>

More... (looser matching)
- Last day of these posts
- In 2003 , 2004 only
- Today
- New search...