Tee Hee! Made a voltage regulator today.

e ? . g w e

right

ahh

Heh, quite

NT

--
Since the lamp exhibits hysteretic behavior, I suspect that the value
of the resistance can also be used to change the output frequency,
regardless of the effect of the negative resistance region.

--
Schematic?

Yes, changing the resistor will change the sawtooth frequency, but the value of resistance must be restricted to the range of values where the load line on the V-A graph intersects the glow lamp's transfer characteristic when the slope of the transfer characteristic is negative.

The equation for the oscillation frequency is given by:

1/[RC*log(Vs - Ve)/(Vs - Vf)]

where Ve is the lamp extinguishing voltage and Vf is the lamp firing voltage, and Vf>Ve. Also assuming that the frequency is low enough that the discharge time through the lamp is negligible.

What bothers me most about the above equation is that dimensional analysis says it doesn't work. You can't take the log unless it is a pure number. So do you mean:

1/[RC*log((Vs - Ve)/(Vs - Vf))]

That would result in Hz, just fine, dimensionally.

Jon

Right! Sorry about that!

Also, as I mentioned the equation 1/[RC*log((Vs - Ve)/(Vs - Vf))] neglects the discharge time of the capacitor through the lamp, so the equation is only approximately valid for frequencies less than 200 Hz. To calculate frequencies above that the text uses one of the famous copouts by modifying the equation with a correction factor alpha, so it becomes:

1/[alpha*RC*log((Vs - Ve)/(Vs - Vf))]

For an "average" lamp type 5AB alpha is 1 from DC to about 200 Hz, it then breaks upwards so it intersects 2 at about 10 kHz.

The book also has some interesting verbage on various external effects on glow lamps. Summarizing: When in total darkness, electrostatic fields of the proper polarities can manipulate the firing voltage of the lamp, but not the maintaining voltage. The firing and maintaining voltages have a negative tempco of about 40-50 mV /degree C. Sawtooth oscillators using glow lamps were exposed to gamma radiation at levels up to 120,000 Roentgen per hour without any change in the circuit's functioning, however prolonged high levels of radiation caused the glass to become brittle.

If you add an emitter-base resistor to the PNP pass element, your output voltage divider parts can be adjusted to make more sense. As is, the circuit seems to be running on leakage.

Some signal diodes to limit diff input voltages might also be prudent.

RL

Might also need a dominant lag capacitor (33-100pf) from C to B of the pass element, at some time.

I'd expected NE2 to look more like 75V, than 50V, hence the v-divider comment.

RL

These are the classic reg circuits:

John

Nope, it was 50V -- that surprised me, too. Of course, le olde junque box may have coughed up something that wasn't quite an NE-2.

The dominant pole comes from the 1uF filter cap on the +135V output (which I really should have included, since it makes such a difference -- funny that out of all the replies you're the _first one_ to notice that). It's like any other LDO -- leave off the output cap, and it oscillates like mad.

--
www.wescottdesign.com

Not really leakage, per se -- the diff amp has a 1.6mA standing current, so the transistors in the diff amp are dissipating about 1/10th watt each. (The pass transistor is dissipating something like 1/2 to 2W).

Granted, the current is pretty low -- but that's by design, since the overall current from the supply is on the order of 20-50mA. I certainly didn't want to load it too much.

--
www.wescottdesign.com

I would have used the transistor version with the NE-2 for a reference, except that the only HV power transistor I had on hand was PNP.

I did consider the all-tube version, but then I _would_ be using up my

30V of overhead, and more room than I had, and probably the filament current capability of my power supply transformer.

And besides, John, this is a Low Drop Out regulator -- that extra 1V of overhead might make a real difference y'know.

--
www.wescottdesign.com

e ? . g

w the

,

. ---/\/\---+---/\/\---+---- . | | . =3D=3D=3D =3D=3D=3D . | | . ----------+----------+----

Simple as that. R needs to not drop many volts, and R&C need to not exceed the permitted max load on a valve rectifier.

NT

So, can you make a light level detector with a neon relaxation oscillator and a frequency counter?

Or, can you stabilize your neon relaxation oscillator to 500kHz with a

500 meter long* reel of fiber optic cable and some lenses?

• Or whatever 600m * (velocity factor) comes out to.

--
www.wescottdesign.com

Those are classic! But I think the tube based one at least will have pretty poor performance. The low mu of the triode will not give much loop gain, and the Miller capacitance will take a dump all over the high frequency response. Probably best to use a pentode instead of a triode, or a cascode, or better yet a differential pair DC coupled to a pentode or cascode. But of course then it's not classic anymore...:(

Leakage ? It looks like a perfectly working circuit to me.

I like this myself

165VDC+20V Ripple NMFETS OUTPUT ----> +---+---------------++|++-------------+ | |V| | 93.3958V +/- 200uv | === | | 1M + | | ___ | | +-----+-|___|+-----+ + | z | A 100V ZENER + + | | +-|| | NMFETS ->|| | +-||+----------+ | | | | === .-. GND | | | | 1M '-' | === GND (created by AACircuit v1.28.6 beta 04/19/05

That has been tested up to 1 amp.. holds constant.

Jamie

Are you suggesting that the two BJTs in the diff amp are anywhere near equally sharing that current given normal load conditions? I hadn't imagined it possible.

Jon

A depletion mosfet would be the ideal pass element, like the tube but no filament supply needed. Dropout would be low and line regulation would be excellent, because the g-s resistor would present a constant-current load to the error amp.

John