>This design is simple to build, forgiving, well behaved, easily scaled for
>power and voltage and reasonably efficient.
>It SHOULD have use in many amateur applications.

At the start of this I should indicate for which parts of the circuit I
claim divine input. God doesn't need anyone to stick up for him but I should
point out where my responsibilities begin :-). The key "breakthrough" was
the use of the transistor "comparator" and "zener based voltage reference" .
I had been presented with a prior design which had used a standard
microprocessor reset generator IC as a switching regulator controller and I
had to produce one which was cheaper and very preferably functionally
superior in this application. So, the core transistor based "controller"
concept using (in this implementation) two transistors and no conventional
hysteresis control were the special input. How I implement and or extend the
design is my responsibility (to the extent that anything is).

I,  like Olin, was somewhat dubious that the circuit would work (let alone
work reliably) and was overjoyed to find out how well it did. Working out
just why came next :-)

/Since you've built this circuit and are using it, I'd be interested in
/seeing some performance data.

/What efficiency do you actually get, at whatever operating point
/you're using (Vin = _____, Vout = _____, Iin = _____, Iout = _____)?

I can give you rough answers from memory. For details I'll have a dig in my
design log or possibly it may be better to whip up a low power one as per
this
particluar requirement and measure them again.

/What frequency does it oscillate at?  What are the rise and fall times
/of the voltage waveforms on BUKFET's gate and drain?

Now you're asking useful questions :-).
The system is self oscillating and actual frequency is quite variable and
dependent on a number of factors.
The on time is a function of the time that it takes for the on state voltage
to rise to the trip point and the off time is a function of how long the
stored energy in the inductor will maintain the output voltage ABOVE the
trip point. Note I say 'a function of" - these are not the only factors.
Someone (Olin?) said that CBUK1 was unnecessary (actually he said it was in
the wrong place). This ws my attempt to control the minimum off time by
holding QB1 on when the output voltage had fallen below the transistor's
normal turn on point.
So, changing the inductance affects frequency significantly. Also CBUK2.
Load is also a major factor - more load = higher frequency.
Typically oscillation frequency is very low at very low load - under 10 KHz.
This is entirely expected as it is switching "on demand" and if there is no
demand it doesn't switch !. Adding a small permanment load increases Fmin.
Loading it up takes the frequency up rapidly. Depending on various
parameters this will be up to around 100 KHz or so.

I have basic spectrum analysis facilities available but my experience is
that a VERY good initial idea of emc problems can be had with the use of a
standard AM/FM radio. This circuit is about as quiet as a switcher gets. Not
perfect - a radio placed right on top of the coil certainly picks up signal
but the radiated harmonics are minimal.

FET drive waveforms are not nice. They are, rather to my surprise, much
nocer than for the system using the reset IC.
In my fulle rcircuit (posted ?2 days ago) I show an added FET gate turnoff
transistor. This very substantially improves the turnoff wavshape and time.
A similarly connected transistor for turn on produced minimal if any
improvement in actual performance.

Peak efficiency occurs for smaller values of Vin and is arounf 85%+ AFAIR.
This is not as good as some very optimised designs but entirly adequate in
many cases where this circuit would be useful. Having a 1 ohm Rdson FEt in
my case does not help!. eg at 10 volts out and 500 mA Pout = 5 watts and FET
Rdson loss would be 250 mW or 5% before any other losses are considered,
.
Efficiency drops off with increasing Vin (a major factor being that more of
the output is provided by flyback and less by forward conduction) and it
falls to around 55% AFAIR at 200 volts in. In my applicatio the 200 volts is
an extreme condition - normal values are in the 20 to 50 volt range but it
MUST be able to survive the occasional extreme excursion and work there.

For a system with less extreme max to min values a better result could be
obtained.

The PFET that I use has an Rdson of about 1 ohm - this is truly horrible
and is a significant contributor to losses. It was chosen because it
provided a
good mix of cost and performance in my application. 200 volt plus PFETS are
not very common!

>This circuit is special - it was provided by God (no kidding).

/Watch out, there!  One always needs to be aware that God has a sense
/of humor, and one of His more subtle pranks was to create Zener diodes/
/with weak knees.

Yes. I am well aware of the soft turn on of zeners.
The somewhat ill defined voltage characteristic in this case was very
tolerable because of the cost of a zener compared to a referenc ediode.

/With the resistor values shown, your circuit operates Zener diode
/ZBUK1 with about 60 microamperes of reverse current; while this may be
/enough for some diodes when operated at room temperature, it is
/definitely NOT enough for ALL diodes over any appreciable temperature
/range.
/Zener diodes frequently have a rather gradual transition between the
/non-conducting state (at voltages well below Vz) and reverse
/conduction beginning at voltages approaching Vz.  They generally
/should not be operated in this "knee" region because they can be very
/noisy, have wretched temperature coefficients, and appear to have
/breakdown voltages that are WAY out of spec.

I largely agree with these comments. I suggest that rather than having a
gradual transition what we are seeing is an exponential rise and moving the
curve onlyt a little either way on one axis can have a major influence on
the other - in this case voltage drop versus current.

I agree that the reproducibility of voltage over various bartches of zeners
and gereral zener characteristic swould be improved by increasing zener
current. This is difficult to do in this particular layout because of the
wway that the zener is used here which is quite different to when it is used
as a "reference' as in eg the Nat Semi circuit. This circuit could arguably
be 'improved" by substantially decreasing RBUK1 and PBUK1 to increase zenera
current. The extra power loss would be insigniicant. Notice that RB!/PB1 are
NOT intended to form a significant voltage divider (PB1>> RB1) - PB1 is
there to ensure QB1 is off in the absence of zenere conduction. In practice
PB1 could PROBABLY be ommitted but this would be risky.

The zener behaviour is better than might be expected because it is driven
from a very limited swing input voltage - Vout varies little over all
operating conditions.

/A good rule of thumb is that a Zener diode usually needs at least one
/milliampere of reverse current in order to reliably "do its thing".
/With only 60 microamps of Ir, you could run into troubles with this
/circuit- poor regulation, large changes in output with temperature,
/noisy operation, etc.- that show up in production, particularly when a
/new batch of diodes arrives.

Yes. Where it was reasonably possible to do so I would use a higher current
than this. Can you suggest a reasonable way of doing so here (apart from
reducing RB1 & PB! markedly)? There will be ways to do this but they are
liable to increase complexity and cost .
And, the circuit works very well for its purpose.
Moderate changes in operating point over temperature, load, Vin and time are
tolerable.


/I tried simulating this circuit in SPICE; it works, after a fashion,
/but it appears to operate more like a linear regulator that's decided
/it would be fun to oscillate a little, than a true switching
/regulator.

In practice it is VERY CERTAIN about what it is.
No matter how hard I tried to "fool" it (and I tried very hard indeed) it
always burst into full life cleanly at a certain point. One moment it is a
linear regulator and the next it is definitely a switcher.

Others have pointed out the problem caused by the weak
turn on/turn off drive to BUKFET

The resistor values were not meant to be applicable for all applications - I
should/could have reworked them for the actual challenge (should!) but they
actually come from the working 200 volt circuit).
By all means change the drive and turn off resistors. I was sloppy when I
snipped the simple circuit out of a larger diagram. RBUK4 and RBUK3 can be
much lower for a low voltage circuit., Also RBUK4 was actually being used to
drive a bipolar gate driver and not the FET directly.
For a say 30 volt maximum in circuit RB3 for 1/4 watt max dissipation could
be as low as V^2/P = 900/.25 = 3500r - say 3K9.
This would be OK for eg a 12 to 24 volt system with peak occasional
excursions to 30 Volt in and an SFR16 resistor as RB3 (Philips - 0.5 watt
rated).
In practice RB3 actually dissipates FAR less than its static DC value (due
to duty cycle aspects) and it also divides full Vin with RB4 so you could
use an evenm lower value in this case. probably eg RB3 = 1K, RB4 = 4K7 or
similar.
That would help immensely.

/, and it's a real killer: in
/simulation, at least, I get an efficiency that's not appreciably
/greater than that of a linear regulator, mostly because of BUKFET's
/long, long, LONG turnoff time.  You need sharp on/off transitions to
/avoid wasting power, and that means solid gate drive.

My real results AFAIK are say 85% at Vin = 15 and 55% say at Vin = 200 v for
a 10 volts out say. at 600 mA.
(The system works to 1 amp but peak real load was 600 mA and is now under
half that after I redesigned some peripheral equipment for another supplier
:-)).

A linear regulator would be 66% and 5% efficient respectively.

The buck dissipates 1 watt and 5 watts respectively.
A linear regulator would dissipate 3 watts and 114 watts respectively.
(!!!!)

In my application the switcher wins quite well.

The 200 volt in condition is very very very very rare - but it can happen.in
theory for very very short periods.
(This is used in excercise equipment and if Arnold S ever decided to set the
machine on Level 0 and pedal at super super lightning speed then this might
happen for a brief period.)

/Part of the problem is BUKFET itself; the IRF6215 you've chosen for
/BUKFET is a real monster of a FET, with a huge gate charge
/requirement.

Absolutely !!
Also rather high voltage for a PFET. I am actually using an ON Semi part
rated at 200 volts which is even worse !
I don't recommend this FET unless you need the very high voltage.
I am keen to extend this circuit to NFET use at some stage (SHOULD be easy)
as there are many more much better much cheaper NFETs available.
BUT, this one works well for the purpose.

/If this design is primarily for low-current use (a few
/tens of milliamps output), you could use something like a ZVP3310 from
/Zetex which would switch a lot faster.  Total gate charge for this
/device is about 0.8 nC, compared with 60+ nC for the IRF6215.

Agree entirley.
Or use a PNP bipolar.

/Got any measurement results to share?  I'd be interested in seeing
/them, particularly if they summarize results for a number of units.

Anon.
But even better would be if someone el;se gives results for versions of this
which are closer to more normal PICLIST use.
eg Byron Jeff has built a 10A 5 volt unit - it will be interesting to hear
actual results.
He says -

> I did some preliminary high current testing. Built a dummy load out of 4
> 0.7 ohm 25W ceramics, wired in series/parallel to give a 0.7 ohm 100W
load.
> Final result: Vout=4.75V @ 7.8A !!! A total success.

I notic ehe said it had Vout = 5 volts at no load so this was 1 0.3 volt
regulation drop at 8 amps. We'd have to know where he measured this to know
how well it was actually regulating.




regards


                Russell McMahon

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