Rheostat Audio Interference

I struggled with a problem like this on a Glassair III some friends
built.

Had a transistor in an Darlington emitter follower configuration. The
dimmer pot
drove the base. The lights were from the emitter to ground.

Trouble was, the thing was oscillating when at mid brightness
positions. Too much
capacitance on the output. A known problem with emitter followers.
One forgets
that they still have gain at a few tens of mhz. When it took off you
could hear it
in several of the radios. Darlington configurations have worse
stability problems.

I solved it by puttting about 100 ohms in the base right at the
transistor.

These circuits are designed by people not very skilled in the art.
They also suffer
the problem that if a bulb burns out shorted or there's an inadvertent
short on the
string of lights, the transistor fails. There is nothing to limit the
current.
That will usually take the pot too, especially if
it's near the high end of its range. The 100 ohm resistor will solve
that, too.

If it's not a darlington, the resistor will have to be smaller.

The cool way around all this is to design it with a P-FET power device
configured like an op-amp.

Bill Hale

wrote in message
ups.com...

Trouble was, the thing was oscillating when at mid brightness
positions. Too much
capacitance on the output. A known problem with emitter followers.

Horsefeathers. Emitter followers have less than unity voltage gain and are
stable as rocks. And how does a resistive LED load become capacitive?
Difficult to imagine.



One forgets
that they still have gain at a few tens of mhz. When it took off you
could hear it
in several of the radios. Darlington configurations have worse
stability problems.

I solved it by puttting about 100 ohms in the base right at the
transistor.

Tens of millihertz? Try again. And the base of the transistor at mid-gain
has more than 100 ohms of resistance in the control pot.




These circuits are designed by people not very skilled in the art.
They also suffer
the problem that if a bulb burns out shorted or there's an inadvertent
short on the
string of lights, the transistor fails. There is nothing to limit the
current.
That will usually take the pot too, especially if
it's near the high end of its range. The 100 ohm resistor will solve
that, too.

Unmitigated horsepoop. Bulbs don't burn out shorted. Bulbs burn out open.
If the transistor fails ( a million to one odds), the pot is open-circuited
and will not be damaged. You have absolutely no experience in the matter,
so why waste our time and bandwidth with your ignorance?





If it's not a darlington, the resistor will have to be smaller.

Don't apply for an engineering job at my company.



The cool way around all this is to design it with a P-FET power device
configured like an op-amp.


Why not an N-FET, or an NPN, or a PNP, all of which will solve the problem
elegantly.


Jim

RST Engineering wrote:

wrote in message
ups.com...

Trouble was, the thing was oscillating when at mid brightness
positions. Too much
capacitance on the output. A known problem with emitter followers.

Horsefeathers. Emitter followers have less than unity voltage gain and are
stable as rocks. And how does a resistive LED load become capacitive?
Difficult to imagine.

True for a real emitter follower, however if you have an N-stage
darlington with N=3, it can oscillate.

This was the subject of an IEEE Transactions article in the early 80's
and a source of great embarrassment for an engineer I worked with that
didn't read the article until after putting the design in production.

snip

--
Jim Pennino

Remove .spam.sux to reply.

Might want to read up.

If emitter followers are driving a load with a capacitive component, it
is reflected
to the input as a negative resistance. You can prove this with only
the simplest
hybrid pi model. It's a parasitic oscillation, not a loop oscillation
which couldn't
happen with gain 1 as you point out. You add enough series base
resistance
to swamp out the negative resistance, making it stable. There is
enough capacitance
in the wiring to excite the phenomena. This is why you see resistors
in the base
circuits of emitter followers all the time.

Darlingtons are much worse. You can show that with the hybird pi model
as well.
The oscillation will be near fT. That's a few mhz for 3055 type
devices.

Bulbs usually burn out open, true. What about some other short? The
circuit has NO short current protection other than the beta of the
transisitor. The base current becomes 1/beta of whatever short circuit
current flows. If the pot is set very near the max end of it's range,
dissapation will destroy the upper part of the resulting divider stick.

I have replaced enough of the panel mount edge adjust pots in Bonanzas
which have
this exact setup to know. Know what those cost?

The P-fet works great because you can get the output clear to the rail,
not the rail - 1 diode drop
as you are limited to with the 3055 approach or 2 drops in the
darlington approach.
It's inherently current limited by IDss as well. Draw it out: The
source goes to +14,
the drain to the lamps to ground. The control pot goes with the hot
end to +14, the
wiper to the gate, and the other end to the bulbs. So it also has a
small amount
of loop gain-- makes the adjustment very smooth. The huge gate-drain
capacitance
of the v-fet structure miller multiplied by the gain of the FET ensures
stability under
all conditions.

Maybe you should write it up for kit planes.

I won't be applying to RST anytime soon, but I did think you were
better than this.

Bill Hale

wrote in message
oups.com...


Might want to read up.

Have. Lots.


If emitter followers are driving a load with a capacitive component, it
is reflected
to the input as a negative resistance. This is why you see resistors
in the base
circuits of emitter followers all the time.

I really don't want to get into the whys and wherefores of parasitic
oscillation in this ng. Hell, the guy only wants his dimmer fixed. And
yes, a 100 ohm base decoupling resistor has been standard for me for the
last 40 years, mostly because I really can't control how well the collector
is bypassed for DC, audio, and RF simultaneously. It is difficult (not
impossible) to get parasitic oscillation where the collector is bypassed
from DC to daylight (or at least out to Ft).

The only place I *won't* use a base decoupling resistor is in a VHF power
amplifier where the input impedance is an order of magnitude lower than the
decoupling resistor.



Bulbs usually burn out open, true. What about some other short? The
circuit has NO short current protection other than the beta of the
transisitor. The base current becomes 1/beta of whatever short circuit
current flows. If the pot is set very near the max end of it's range,
dissapation will destroy the upper part of the resulting divider stick.

How do we know that there is no short circuit current protection? Did you
take it apart or do you have a schematic of this particular dimmer? We sure
could have saved a lot of wild ass guessing about the problem. He MIGHT
have it overloaded, but without knowing the particulars of this
installation, we are doing rectorandom guesses at the problem.



I have replaced enough of the panel mount edge adjust pots in Bonanzas
which have
this exact setup to know. Know what those cost?

No, and I have a hard time believing that Beech, the overdesigner of the
industry, put something out without short circuit protection of some sort.
However, if you have replaced them, then you are one up on me. No, I don't
know what they cost, but a simple current shutdown (with or without
foldback) is less than half a buck's worth of parts at the front end.


The P-fet works great because you can get the output clear to the rail,
not the rail - 1 diode drop
as you are limited to with the 3055 approach
Maybe you should write it up for kit planes.

You probably want to look at January, April, May, June, July 2001 Kitplanes
where I used N-channel, P-channel, NPN, and PNP transistors as the output
devices, explaining exactly what the tradeoffs were between each of the
devices. All in all, about ten designs.



I won't be applying to RST anytime soon,

And I thank you kindly.


but I did think you were
better than this.

When I'm in sci.electronics.design, I'm really quite careful about the
nuances of design. When I'm trying to get some poor guy's lamp dimmer to
work on RAH, RAO, or RAP, I'm a little less careful about being precisely
technically correct. For a dimmer that apparently worked correctly once
upon a time, poor design is about the last place I try and look. I'm not
above reengineering a crappy design, but if it really does have a parasitic
oscillation at the top end of the range, you'd have thought that in my last
45 years in this biz I'd have come across one, no?

Now, let's get back to fixing this sucker with what we DO know. Did the
fact that he can't key his radio transmitter or hear his radio receiver when
the unit was acting up mean anything to you? I doubt parasitic oscillation
keeps the transmitter key line from kicking the transmitter on. Use ALL the
clues, not just the one you are most comfortable with.

Jim

"RST Engineering" writes:



Unmitigated horsepoop. Bulbs don't burn out shorted. Bulbs burn out open.

I wish.... I had some hard experience with shorted ?317's I think
they were -- the 14v version of 327's. And lots of owners of
early-generation X10 modules discovered they would fail into full-on
when their 120vac lamp failed into full-off.

The clear solution for the OP was what the NASA LeRC 10x10 Supersonic
Wind Tunnel used for ""dimmers"" for the seven 40KHP drive motors
-- large glass tanks of salt water. The electrodes were cranked in
deeper to speed things up....

Oh, if you go that route, be sure and stick to positive G maneuvers
or you'll have a mess....
--
A host is a host from coast to
& no one will talk to a host that's close........[v].(301) 56-LINUX
Unless the host (that isn't close).........................pob 1433
is busy, hung or dead....................................20915-1433

Describe the failure mode that lets a bulb short out.

Jim



"David Lesher" wrote in message
...
"RST Engineering" writes:



Unmitigated horsepoop. Bulbs don't burn out shorted. Bulbs burn out
open.

I wish.... I had some hard experience with shorted ?317's I think
they were -- the 14v version of 327's. And lots of owners of
early-generation X10 modules discovered they would fail into full-on
when their 120vac lamp failed into full-off.

"RST Engineering" writes:

Describe the failure mode that lets a bulb short out.


I'm not an expert on same; just a victim. But from conversations
with someone at GE Nela Park decades ago; the filament breaks, and
can fall down from both gravity and err "sprong"ing when it lets go...

If the shortened filament end touches the OTHER post, it will draw
lots more current since it is shorter. It very soon burns out, but
in the meantime....

Or a length comes loose at both ends; it falls across the posts at the
bottom and.....


--
A host is a host from coast to
& no one will talk to a host that's close........[v].(301) 56-LINUX
Unless the host (that isn't close).........................pob 1433
is busy, hung or dead....................................20915-1433

We must not have been talking to the same people at GE, because one of the
things I did in my former life was do reliability studies on some of the
stuff we threw up into orbit -- like annunciator light bulbs (WAY, WAY
before LEDs could remotely be considered reliable enough for space flight).

If you look closely at an incandescent light bulb (especially the aviation
versions like the 327-28 volt and 330-14 volt) you will see that the
internals of the bulb start with a little dot of ceramic called the "bead".
The wires that come out of this bead are called "spreaders" or "stringers";
there is an optional electrically inert third wire called a "support" that
we can talk about later.

The spreaders are angled out at a fairly precise angle to keep a broken
filament from coming into contact with the other spreader and causing
exactly the failure mode you describe. It is geometrically impossible for a
dangling filament to come in contact with the other spreader. THe support
does exactly the same thing for a long-filament bulb -- holds up the
filament while it is still a lamp and keeps the broken filament from
touching the other spreader when the lamp burns out.

The only possible failure mode would be for the filament to break at both
ends simultaneously and drop down onto the bead in such a manner that the
slightest vibration would not cause the filament to drop harmlessly into the
bottom of the lamp base. While there is a mathematical probability that
this could happen, there is also a mathematical probability that the lamp
could disassemble itself and reassemble itself in a far corner of the
universe. I'm not sure which one is more probable.

Anyway, it seems the OP has found the "ground wire that ain't a ground wire"
and solved the problem, which is a good thing.

Jim



"David Lesher" wrote in message
...
"RST Engineering" writes:

Describe the failure mode that lets a bulb short out.


I'm not an expert on same; just a victim. But from conversations
with someone at GE Nela Park decades ago; the filament breaks, and
can fall down from both gravity and err "sprong"ing when it lets go...

If the shortened filament end touches the OTHER post, it will draw
lots more current since it is shorter. It very soon burns out, but
in the meantime....

Or a length comes loose at both ends; it falls across the posts at the
bottom and.....


--
A host is a host from coast to
& no one will talk to a host that's close........[v].(301) 56-LINUX
Unless the host (that isn't close).........................pob 1433
is busy, hung or dead....................................20915-1433

"RST Engineering" writes:

If you look closely at an incandescent light bulb (especially the aviation
versions like the 327-28 volt and 330-14 volt) you will see that the
internals of the bulb start with a little dot of ceramic called the "bead".

Well, I was not aware we were talking exclusively about aviation/space
rated lamps, but as it turns out while working for LeRC; I almost
emptied Building 142 with those unshortable lamps. I was driving
them with NE555's which turned out to be effective, if somewhat
smoky, fuses. The technician and I just looked at each other while
the cloud rose toward the smoke detector....the one going to the
sprinkler system and Evac alarms.

The NE555 fuse has an audible annunciator as well; at least when
the charred pieces fall down into the vent fans that spit them out
with a clatter...

In one of those "I'm not making this up.." aspects, my then-boss is
now a participant in this newsgroup, and I'm sure he remembers the
design/assembly in question. {But I ..cough... don't think I ever
bothered him with this particular FUBAR at the time...}

As for the X10/120v lamp aspects, also circa 1980, I'd looked into
it when my BiL wanted to know why his kept failing. I suspect
triac/et.al designs have progressed since then and maybe it's no
long an issue.



--
A host is a host from coast to
& no one will talk to a host that's close........[v].(301) 56-LINUX
Unless the host (that isn't close).........................pob 1433
is busy, hung or dead....................................20915-1433