Oscillations Redux / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

Oscillations Redux

previous 1 2 3 4 next

Move Thread

LAN_403

Proud Mary

Thu Apr 19 2012, 03:04PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

Have you thought of applying an analogue voltage derived from a variable resistor to your FET gate? If this restores good order, you will know where the problem - or at least part of it - is lying low.

sngecko

Thu Apr 19 2012, 03:41PM

Registered Member #3447 Joined: Fri Nov 26 2010, 11:10PM
Location: North Jersey
Posts: 97

That certainly makes sense. Let's see, simulations suggest that at 4mA, the gate voltage on an IRF830, should be around 4.5V. I'll just build a voltage divider off a 9V battery and we'll see if the oscillation persists. Hopefully I can test this theory tomorrow... Thanks.

Steve Maurer

Fri Apr 20 2012, 02:03AM

Registered Member #133 Joined: Fri Feb 10 2006, 10:27PM
Location: Pensacola, Florida
Posts: 47

Common mode noise is likely getting into your 24 V SMPS and is affecting its feedback loop. I have encountered this when driving pulsed loads (even with expensive commercial power supplies). Place a common mode choke in line with your 24 V supply (i.e. â€“ wind the 24 V power lead and the return lead together on a high permeability toroid).

A few other notes:
1) 100 Ohms may be too large for the gate resistance.
2) Place decoupling capacitors across your op-amp power inputs. (Try 0.01 uF to start with.)
3) Keep your a) op-amp output-to-MOSFET gate path and b) MOSFET source-to-inverting op-amp path close to each other in order to keep this loop area small. If you are using wires to your MOSFET, then twist the gate and source wires across the entire length. (You do not want noise introduced here.)
4) Placement and proper layout for the op-amps is important for noise free operation.

sngecko

Fri Apr 20 2012, 11:28AM

Registered Member #3447 Joined: Fri Nov 26 2010, 11:10PM
Location: North Jersey
Posts: 97

Steve-

Thanks for the tip. I'll go ahead and get some ferrite and check it out, too.

Oh, troubleshooting!

EDIT: I just ordered four 1" ferrite toroids. I'm planning on rewiring the 24V SMPS 110VAC input wires to be twisted together and spiraled around one of these toroids.I eyeballed the 1.8A rating of my SMPS and guessed that this would be enough ferrite.

EDIT: The permeability of these toroids is 2400. I don't think that's as fancy as the nano-crystalline ones for sale now, but I figure that I'll just counter-wind two 18 gauge jacketed wires as far as I can and see what happens.

sngecko

Fri Apr 20 2012, 02:35PM

Registered Member #3447 Joined: Fri Nov 26 2010, 11:10PM
Location: North Jersey
Posts: 97

I just conducted the following experiment:

(1) Disconnected SMPS input lines (left the ground connected to chassis ground and output return to chassis ground).
(2) Connect 24V battery in place of SMPS.
(3) Everything else as was.

Results:
(1) Tube lit in similar steady manner and very low current (500uA).
(2) Fans again slowed a little, but very little audible oscillation (lower frequency).
(3) As I turned the current up to 2mA, the frequency of the audible oscillation increased, and the fans slowed to almost a complete stop.

It looks like the current control is working pretty well. The cascode circuit was designed to have a very high dynamic impedance because of the negative resistance characteristics of laser tubes, so it may be that the oscillations are kinda normal. Their effect on the fans is unexpected, though.

EDIT: And in response to above, I guess the power supply isn't tripping... Even batteries won't keep the fans running at full speed. They just slow down progressively as current is increased to 2mA from a few hundred microamps.

sngecko

Fri Apr 20 2012, 06:46PM

Registered Member #3447 Joined: Fri Nov 26 2010, 11:10PM
Location: North Jersey
Posts: 97

In the file shown below, you will see that the triode output (cathode) wire is bundled with a number of other wires on its way to and from the indicated fuse. The 24V fan's positive line passes through this bundle. Could they be coupled in some way? The current is so very low that it's hard to imagine that a significant induction could affect a battery... I'll keep looking for the source of the 24V screwup.

Steve Conner

Fri Apr 20 2012, 07:55PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

If your tubes are oscillating, the RF could be confusing the brushless motor control circuit in the fan. The average current may be only 4mA, but it could be in the form of nasty little RF bursts with high peak power.

It is possible to have RF and audible oscillations at the same time. The effect is called squegging.

Again I recommend adding a grid stopper resistor at least.

sngecko

Fri Apr 20 2012, 09:21PM

Registered Member #3447 Joined: Fri Nov 26 2010, 11:10PM
Location: North Jersey
Posts: 97

Do you think that a potentiometer would be handy in this application? I'm not sure what value would be best, guessing somewhere from 1.5k to 100k. As you can see, I'm got a 20kOhm resistor tied to ground in parallel with the decoupling cap and neon, but I don't think that forms the proper RC filter with the valves input capacitance... Am I right in thinking that this resistor should be in series with the other components?

I wish I had an oscilloscope to get an idea of what "kind" of RF noise I was dealing with.

sngecko

Fri Apr 20 2012, 09:40PM

Registered Member #3447 Joined: Fri Nov 26 2010, 11:10PM
Location: North Jersey
Posts: 97

Something like this?

Marko

Fri Apr 20 2012, 11:45PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi gecko

Look what your circuit does: it's a linear regulator with full gain of your amplifier (tens/hundreds of thousands) being wired in the negative feedback path. The opamp itself as well as the mosfet will always introduce some delay in the feedback loop which translates a phase shift which increases proportionally with the frequency - result of which is, that at a certain frequency this finally reaches 180 degrees, turning your negative feedback into positive. And if your system still has overall gain greater than 1 at this frequency, it will oscillate just like any other astable multivibrator, tesla coil or whatever.

Since you did nothing to lower the HF gain of your TL084 (which is a pretty wideband opamp with cutoff at 4Mhz IIRC), oscillation here is pretty much a certainty. High dynamic gain of the mosfet in linear range probably makes it even worse.

Hence the infamous RC network we always see on TL494's and whereever. There is a series RC going from error amp output back to it's - input, It's purpose should now be obvious - at high frequencies the capacitor acts as low impedance and the resistor alone is virtually connected between - and amp output. For this to help though we need one more resistor in feedback path which will form a divider with the first one.

Now we've gotten an amplifier which has it's maximum gain at DC for minimum error and very low gain at high frequencies to prevent oscillations.

There are various ways of optimizing this network (since it after all slows down the response), but if your regulator is only intended to work with static load and reference then you could probably get satisfied by just slapping a random big cap until oscillations stop.

I'd probably start with something like 1k resistor on feedback wire and a 1uF cap or something like that.

There will also be assorted other random things that help - such as, swapping your opamp for slower, crappier one like 741, which acts as a low pass filter in itself. Other things like increasing the mosfet gate resistor value, or filtering the voltage across the mosfet with a capacitor could happen to help in a similar way.

Mosfets are also somewhat troublesome parts for linear use. For a circuit like this I'd probably just use a 200V darlington instead (which I think is more than enough for the tube to be off).

Hope this helps somewhat :)

Marko

previous 1 2 3 4 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob