Oscillations Redux
sngecko, Wed Apr 18 2012, 09:28PMSome of you will remember my gas laser/discharge power supply, current regulator postings in the past, which led to a nice, stable neon lamp glowing merrily.
Today I received two new-old-stock Siemens Helium Neon lasers for further testing. The are roughly the same geometrically and electrically as the Mells Griot HeNe that met its premature end a few months ago. After the success of the small neon lamp, I expected all things to go well. Not so fast.
The pre-strike current flow certainly was stopped by the grounded grid, thanks to Proud Mary. Upon striking, my instrumentation showed 4 mA through the tube and somewhere under 2kV dropping across it. That was just as it should be. There was no flicker on and off as in the initial experiment. The main problem, however, came again from a high-pitched noise accompanied by the cooling fans' stoppage. The thought of the anode seal overheating and the oscillation destroying my FETs led me to shut down the experiment after 10 seconds or so.
So, from whence comes this oscillation? If I had an oscilloscope (seems obvious, right?), I think I would probe (1) the op-amp output that is driving the MOSFET or (2) the FET drain.
When I first ran SPICE simulations, I saw that the TL084 JFET-input op-amp model provided by TI would oscillate very quickly around the set current value. When I used the similar Linear models (LT1001), the output was solid. The regulator is using TL084s. I thought the simulation oscillation was a fluke, and I don't want to buy LTs unless I know that the problem is related to my application of the TL084s.
Also, what is the link between the oscillation and the fan stoppage. The fans all run off of a +24V SMPS that is not connected at all to the drive/instrumentation circuits (they run off of a linear +/- 15V PSU), except through the main input and chassis ground... I think. I should also mention (in relation to a post by Forty earlier) that the power supply is referenced to earth ground. So, all the power supplies, the regulation circuitry below, and the HV power supply share a connection to chassis ground and, by extension, earth ground.
Here's the circuit,
Re: Oscillations Redux
Proud Mary, Wed Apr 18 2012, 11:27PM
What is the source of the 5V/14.5A filament supply?
Proud Mary, Wed Apr 18 2012, 11:27PM
What is the source of the 5V/14.5A filament supply?
Re: Oscillations Redux
sngecko, Wed Apr 18 2012, 11:31PM
Each filament has its own 5V, 20A transformer. The center tapped secondary is not connected. A simple relay switches on the transformer.
I forgot to mention above that there are two such circuits, but I can only use one at a time, as I have only one triode.
sngecko, Wed Apr 18 2012, 11:31PM
Each filament has its own 5V, 20A transformer. The center tapped secondary is not connected. A simple relay switches on the transformer.
I forgot to mention above that there are two such circuits, but I can only use one at a time, as I have only one triode.
Re: Oscillations Redux
Proud Mary, Thu Apr 19 2012, 12:22AM
I won't comment on the SMPS, and its failing fan, as there are many others here who know them far better than I ever will.
But as for the 3-500Z - triodes are always looking for an opportunity to oscillate, and will find feedback paths beyond the limited imaginings of circuit simulators unless we take definite steps to stop them.
I would start off by sticking some chokes in the anode, grid and cathode lines. The self-resonant frequencies of these chokes should all be different and not harmonically related. If you are unable to actually measure the self-resonant frequency, then choose values such as 2.2, 4.7, and 6.8 mH. These different values will stop the triode from implementing any Miller effect or tuned-anode-tuned-grid oscillatory stunts that it may have been planning.
Then I think you should decouple the line between the laser cathode and the triode anode. For example, wire a 4K7 wirewound from the laser cathode to the triode anode choke, and bypass the junction between resistor and choke directly to Earth/chassis via a suitably robust capacitor of 100nF. This will also make it harder for the laser - a gas diode - to get up to any negative resistance oscillator hi-jinx which it may have been planning with the triode.
Make sure that you use a common earth point for the connections around the valve socket, or you will have trillions of electrons running round in little loops all over your chassis. Each loop will have its own magnetic field, which will try to induce currents in whatever conductors will listen to them.
Proud Mary, Thu Apr 19 2012, 12:22AM
I won't comment on the SMPS, and its failing fan, as there are many others here who know them far better than I ever will.
But as for the 3-500Z - triodes are always looking for an opportunity to oscillate, and will find feedback paths beyond the limited imaginings of circuit simulators unless we take definite steps to stop them.
I would start off by sticking some chokes in the anode, grid and cathode lines. The self-resonant frequencies of these chokes should all be different and not harmonically related. If you are unable to actually measure the self-resonant frequency, then choose values such as 2.2, 4.7, and 6.8 mH. These different values will stop the triode from implementing any Miller effect or tuned-anode-tuned-grid oscillatory stunts that it may have been planning.
Then I think you should decouple the line between the laser cathode and the triode anode. For example, wire a 4K7 wirewound from the laser cathode to the triode anode choke, and bypass the junction between resistor and choke directly to Earth/chassis via a suitably robust capacitor of 100nF. This will also make it harder for the laser - a gas diode - to get up to any negative resistance oscillator hi-jinx which it may have been planning with the triode.
Make sure that you use a common earth point for the connections around the valve socket, or you will have trillions of electrons running round in little loops all over your chassis. Each loop will have its own magnetic field, which will try to induce currents in whatever conductors will listen to them.
Re: Oscillations Redux
sngecko, Thu Apr 19 2012, 12:57AM
How would I measure the self-resonance of this arrangement, if I had an o-scope? it would depend on the laser, right? If I knew this self-resonant frequency, how could I best attenuate it? I may look into buying a little 2MHz USB oscilloscope.
I certainly will look into the chokes on the triode, but I'm concerned about putting a resistor in the main current line (i.e., between the laser cathode and the triode anode). The circuit could be passing 80mA and I was going for efficiency with the whole active regulator thing.
sngecko, Thu Apr 19 2012, 12:57AM
How would I measure the self-resonance of this arrangement, if I had an o-scope? it would depend on the laser, right? If I knew this self-resonant frequency, how could I best attenuate it? I may look into buying a little 2MHz USB oscilloscope.
I certainly will look into the chokes on the triode, but I'm concerned about putting a resistor in the main current line (i.e., between the laser cathode and the triode anode). The circuit could be passing 80mA and I was going for efficiency with the whole active regulator thing.
Re: Oscillations Redux
Proud Mary, Thu Apr 19 2012, 01:22AM
The problem with trying to establish the self-resonant frequency of your whole arrangement is that it is a dynamic circuit, and even very small changes in parameters might caught instability.
Outside of the new wave of thermionic valve audio designers, the use of chokes as I have described is a common prophylactic measure in power valve design - especially where directly heated valves are used. You assume there will be instability unless you take steps to prevent it.
If you are not ready to add the series resistor I have proposed, then as second best, I would suggest bypassing the laser cathode to Earth via a capacitor of 100nF - 500nF so that unwanted blips and spikes that might trigger oscillation are sent directly to Earth. Clearly, this capacitor will add some lag into the circuit's response, so don't make the capacitor larger than you have to.
Proud Mary, Thu Apr 19 2012, 01:22AM
The problem with trying to establish the self-resonant frequency of your whole arrangement is that it is a dynamic circuit, and even very small changes in parameters might caught instability.
Outside of the new wave of thermionic valve audio designers, the use of chokes as I have described is a common prophylactic measure in power valve design - especially where directly heated valves are used. You assume there will be instability unless you take steps to prevent it.
If you are not ready to add the series resistor I have proposed, then as second best, I would suggest bypassing the laser cathode to Earth via a capacitor of 100nF - 500nF so that unwanted blips and spikes that might trigger oscillation are sent directly to Earth. Clearly, this capacitor will add some lag into the circuit's response, so don't make the capacitor larger than you have to.
Re: Oscillations Redux
sngecko, Thu Apr 19 2012, 01:28AM
Sounds like a plan. Before I rip this thing apart, two things occur to me. (1) It might work just fine with a huge CO2 laser head on there, like the much smaller neon lamp, and (2) it would be pretty informative to just see the noise in the line...
Also, any idea why the +24 V power supply fails as soon as the discharge lamp starts, but then comes back on when it's extinguished? That's weird to me.
sngecko, Thu Apr 19 2012, 01:28AM
Sounds like a plan. Before I rip this thing apart, two things occur to me. (1) It might work just fine with a huge CO2 laser head on there, like the much smaller neon lamp, and (2) it would be pretty informative to just see the noise in the line...
Also, any idea why the +24 V power supply fails as soon as the discharge lamp starts, but then comes back on when it's extinguished? That's weird to me.
Re: Oscillations Redux
Proud Mary, Thu Apr 19 2012, 09:30AM
When you say the 24 V (SMPS) supply 'fails', I take it you mean it trips. Is this a response to an over-current condition, or an anomalous condition caused by high voltage or unwanted circuit artifcats going back down the line? Stick a high voltage diode in the 24V line to block anything that may be trying to go the wrong way, and monitor the current to see what, if any, transition occurs at the moment of lamp starting.
Proud Mary, Thu Apr 19 2012, 09:30AM
sngecko wrote ...
Sounds like a plan. Before I rip this thing apart, two things occur to me. (1) It might work just fine with a huge CO2 laser head on there, like the much smaller neon lamp, and (2) it would be pretty informative to just see the noise in the line...
Also, any idea why the +24 V power supply fails as soon as the discharge lamp starts, but then comes back on when it's extinguished? That's weird to me.
Sounds like a plan. Before I rip this thing apart, two things occur to me. (1) It might work just fine with a huge CO2 laser head on there, like the much smaller neon lamp, and (2) it would be pretty informative to just see the noise in the line...
Also, any idea why the +24 V power supply fails as soon as the discharge lamp starts, but then comes back on when it's extinguished? That's weird to me.
When you say the 24 V (SMPS) supply 'fails', I take it you mean it trips. Is this a response to an over-current condition, or an anomalous condition caused by high voltage or unwanted circuit artifcats going back down the line? Stick a high voltage diode in the 24V line to block anything that may be trying to go the wrong way, and monitor the current to see what, if any, transition occurs at the moment of lamp starting.
Re: Oscillations Redux
Steve Conner, Thu Apr 19 2012, 10:26AM
Switched mode power supplies can do just about anything when hit with RF.
I agree that stoppers on the tube electrodes would be a good thing to try, but I don't see the need for chokes. The currents in the circuit are pretty small compared to a RF amplifier, so those cement block wirewound resistors should do fine. Put a 100 ohm one in series with the grid of the 3-500Z if you haven't already: right at the tube socket. If that doesn't help, you can also try one on the plate terminal, or a higher value like 1k.
Of course, the oscillations could also be coming from the op-amp or MOSFET stages, or they could be "Nyquist" oscillations encompassing the whole feedback loop. But the symptoms suggest a pretty high frequency, so the tube is the most likely culprit.
I also suggest not wasting your money on a low-end USB oscilloscope. Get a cheap second-hand analog one instead.
Steve Conner, Thu Apr 19 2012, 10:26AM
Switched mode power supplies can do just about anything when hit with RF.
I agree that stoppers on the tube electrodes would be a good thing to try, but I don't see the need for chokes. The currents in the circuit are pretty small compared to a RF amplifier, so those cement block wirewound resistors should do fine. Put a 100 ohm one in series with the grid of the 3-500Z if you haven't already: right at the tube socket. If that doesn't help, you can also try one on the plate terminal, or a higher value like 1k.
Of course, the oscillations could also be coming from the op-amp or MOSFET stages, or they could be "Nyquist" oscillations encompassing the whole feedback loop. But the symptoms suggest a pretty high frequency, so the tube is the most likely culprit.
I also suggest not wasting your money on a low-end USB oscilloscope. Get a cheap second-hand analog one instead.
Re: Oscillations Redux
sngecko, Thu Apr 19 2012, 01:25PM
It sounds like I should buy/borrow/steal an oscilloscope and try to determine the identity of these illicit electron trafficking oscillations.
I'm pretty sure that the dropping out of the +24V SMPS is tied directly to the oscillations, as everything seemed to work fine when the small neon discharge lamp was working. But as soon as the oscillations started acting up, it dropped out.
Also, during the last HeNe laser test, the frequency of the audible oscillation changed as I adjusted the set current. This seems to suggest, as is almost obvious, that the oscillation is in the op-amp driver or the current regulation chain.
Would it make any sense to try and shield the SMPS? Or its input and output lines with ferrite beads?
sngecko, Thu Apr 19 2012, 01:25PM
It sounds like I should buy/borrow/steal an oscilloscope and try to determine the identity of these illicit electron trafficking oscillations.
I'm pretty sure that the dropping out of the +24V SMPS is tied directly to the oscillations, as everything seemed to work fine when the small neon discharge lamp was working. But as soon as the oscillations started acting up, it dropped out.
Also, during the last HeNe laser test, the frequency of the audible oscillation changed as I adjusted the set current. This seems to suggest, as is almost obvious, that the oscillation is in the op-amp driver or the current regulation chain.
Would it make any sense to try and shield the SMPS? Or its input and output lines with ferrite beads?
Re: Oscillations Redux
Proud Mary, Thu Apr 19 2012, 03:04PM
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.
Proud Mary, Thu Apr 19 2012, 03:04PM
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.
Re: Oscillations Redux
sngecko, Thu Apr 19 2012, 03:41PM
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.
sngecko, Thu Apr 19 2012, 03:41PM
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.
Re: Oscillations Redux
Steve Maurer, Fri Apr 20 2012, 02:03AM
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. – 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.
Steve Maurer, Fri Apr 20 2012, 02:03AM
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. – 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.
Re: Oscillations Redux
sngecko, Fri Apr 20 2012, 11:28AM
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, 11:28AM
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.
Re: Oscillations Redux
sngecko, Fri Apr 20 2012, 02:35PM
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, 02:35PM
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.
Re: Oscillations Redux
sngecko, Fri Apr 20 2012, 06:46PM
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.
sngecko, Fri Apr 20 2012, 06:46PM
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.
Re: Oscillations Redux
Steve Conner, Fri Apr 20 2012, 07:55PM
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.
Steve Conner, Fri Apr 20 2012, 07:55PM
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.
Re: Oscillations Redux
sngecko, Fri Apr 20 2012, 09:21PM
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, 09:21PM
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.
Re: Oscillations Redux
sngecko, Fri Apr 20 2012, 09:40PM
Something like this?
sngecko, Fri Apr 20 2012, 09:40PM
Something like this?
Re: Oscillations Redux
Marko, Fri Apr 20 2012, 11:45PM
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
Marko, Fri Apr 20 2012, 11:45PM
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
Re: Oscillations Redux
sngecko, Sat Apr 21 2012, 01:49AM
As Watson once said, "I have no doubt that I am very stupid." I appreciate the insights that Proud Mary, Marko, Steve Conner, and Steve Maurer have brought to my little project. It's been a long time coming to fruition, and thanks to you, it has a good chance of getting done. I've learned a heck of a lot along the way.
This is what I've got to be going on for now:
(1) If oscillation is in the triode, then first try grid stopper resistance. Then anode, grid, and cathode chokes (in that order?) as well as bypass caps in anode chain (20kV, 100nF!)
(2) If oscillation is in the feedback loop vis-a-vis the op-amp and MOSFET, then try an RC filter across the output to the negative input.
Re: (2), I've done some RC calculations and think that R=100k and C=.1uF or so should give -20dB attenuation at 100Hz. Sound good? Am I right in thinking that this arrangement (kinda RCR) acts like an RC filter in both directions?
I'd like to try (2) first since its fairly easy to solder a couple little parts together and jam their little legs in those DIP sockets. We'll see.
sngecko, Sat Apr 21 2012, 01:49AM
As Watson once said, "I have no doubt that I am very stupid." I appreciate the insights that Proud Mary, Marko, Steve Conner, and Steve Maurer have brought to my little project. It's been a long time coming to fruition, and thanks to you, it has a good chance of getting done. I've learned a heck of a lot along the way.
This is what I've got to be going on for now:
(1) If oscillation is in the triode, then first try grid stopper resistance. Then anode, grid, and cathode chokes (in that order?) as well as bypass caps in anode chain (20kV, 100nF!)
(2) If oscillation is in the feedback loop vis-a-vis the op-amp and MOSFET, then try an RC filter across the output to the negative input.
Re: (2), I've done some RC calculations and think that R=100k and C=.1uF or so should give -20dB attenuation at 100Hz. Sound good? Am I right in thinking that this arrangement (kinda RCR) acts like an RC filter in both directions?
I'd like to try (2) first since its fairly easy to solder a couple little parts together and jam their little legs in those DIP sockets. We'll see.
Re: Oscillations Redux
Steve Conner, Sat Apr 21 2012, 05:51AM
Yes!
Marko's comments may well be valid too, but oscillations in the FET/opamp part of the circuit are a separate issue.
FWIW I've built similar current regulators, albeit minus the high voltage cascode, and they worked without compensation. As far as the op-amp's feedback is concerned, the MOSFET is functioning as a source follower, so the loop gain can never be more than 1.
Steve Conner, Sat Apr 21 2012, 05:51AM
sngecko wrote ...
Something like this?
Something like this?
Yes!
Marko's comments may well be valid too, but oscillations in the FET/opamp part of the circuit are a separate issue.
FWIW I've built similar current regulators, albeit minus the high voltage cascode, and they worked without compensation. As far as the op-amp's feedback is concerned, the MOSFET is functioning as a source follower, so the loop gain can never be more than 1.
Re: Oscillations Redux
Steve Maurer, Sat Apr 21 2012, 03:02PM
Steve and Marko both have good ideas. Definitely try installing the grid resistor first. If oscillation persists, then pursue the op-amp section also. The last op-amp stage may be modified to form a low pass filter if needed.
Clarification concerning the 24 V SMPS common mode filter:
If a SMPS happens to run erratically under certain dynamic loading conditions (due to high frequency noise presented to the output SMPS), then a common mode filter placed in line with the output leads of the SMPS (not the AC line cord input to the SMPS) will sometimes help. Note that the load on the power supply plays a part in the loop response of the SMPS.
(On a side note: A good way to verify SMPS stability is to test the output with a fast rise time pulsed load at several frequencies. I usually check my switch mode supplies at pulsed frequencies of 100 Hz, 1 kHz, and 10 kHz.)
Steve Maurer, Sat Apr 21 2012, 03:02PM
Steve and Marko both have good ideas. Definitely try installing the grid resistor first. If oscillation persists, then pursue the op-amp section also. The last op-amp stage may be modified to form a low pass filter if needed.
Clarification concerning the 24 V SMPS common mode filter:
If a SMPS happens to run erratically under certain dynamic loading conditions (due to high frequency noise presented to the output SMPS), then a common mode filter placed in line with the output leads of the SMPS (not the AC line cord input to the SMPS) will sometimes help. Note that the load on the power supply plays a part in the loop response of the SMPS.
(On a side note: A good way to verify SMPS stability is to test the output with a fast rise time pulsed load at several frequencies. I usually check my switch mode supplies at pulsed frequencies of 100 Hz, 1 kHz, and 10 kHz.)
Re: Oscillations Redux
Marko, Sat Apr 21 2012, 03:31PM
Hello -
nah that doesn't look right - you need some resistance in your feedback path (wire going from your sense resistor to opamp negative input) to form a meaningful divider. And for start I'd just replace your CRC-whatever with a single 1uF cap to make a simple I regulator. This should stop the oscillations and you can always upgrade it alter if you need better impulse response (don't know if you want to audiomodulate the lasers or whatever?)
Also, only now I've realized that the filter cap on the triode grid doesn't go to it's cathode, but all way to the ground. If you think the tube might be unstable, maybe it would be wiser to connect the cap straight to cathode (again it doesn't matter for static operation). This should nullify any chances of tube squeg or whatever better than any grid resistors, as I think.
Marko
Marko, Sat Apr 21 2012, 03:31PM
sngecko wrote ...
As Watson once said, "I have no doubt that I am very stupid." I appreciate the insights that Proud Mary, Marko, Steve Conner, and Steve Maurer have brought to my little project. It's been a long time coming to fruition, and thanks to you, it has a good chance of getting done. I've learned a heck of a lot along the way.
This is what I've got to be going on for now:
(1) If oscillation is in the triode, then first try grid stopper resistance. Then anode, grid, and cathode chokes (in that order?) as well as bypass caps in anode chain (20kV, 100nF!)
(2) If oscillation is in the feedback loop vis-a-vis the op-amp and MOSFET, then try an RC filter across the output to the negative input.
Re: (2), I've done some RC calculations and think that R=100k and C=.1uF or so should give -20dB attenuation at 100Hz. Sound good? Am I right in thinking that this arrangement (kinda RCR) acts like an RC filter in both directions?
I'd like to try (2) first since its fairly easy to solder a couple little parts together and jam their little legs in those DIP sockets. We'll see.
As Watson once said, "I have no doubt that I am very stupid." I appreciate the insights that Proud Mary, Marko, Steve Conner, and Steve Maurer have brought to my little project. It's been a long time coming to fruition, and thanks to you, it has a good chance of getting done. I've learned a heck of a lot along the way.
This is what I've got to be going on for now:
(1) If oscillation is in the triode, then first try grid stopper resistance. Then anode, grid, and cathode chokes (in that order?) as well as bypass caps in anode chain (20kV, 100nF!)
(2) If oscillation is in the feedback loop vis-a-vis the op-amp and MOSFET, then try an RC filter across the output to the negative input.
Re: (2), I've done some RC calculations and think that R=100k and C=.1uF or so should give -20dB attenuation at 100Hz. Sound good? Am I right in thinking that this arrangement (kinda RCR) acts like an RC filter in both directions?
I'd like to try (2) first since its fairly easy to solder a couple little parts together and jam their little legs in those DIP sockets. We'll see.
Hello -
nah that doesn't look right - you need some resistance in your feedback path (wire going from your sense resistor to opamp negative input) to form a meaningful divider. And for start I'd just replace your CRC-whatever with a single 1uF cap to make a simple I regulator. This should stop the oscillations and you can always upgrade it alter if you need better impulse response (don't know if you want to audiomodulate the lasers or whatever?)
Also, only now I've realized that the filter cap on the triode grid doesn't go to it's cathode, but all way to the ground. If you think the tube might be unstable, maybe it would be wiser to connect the cap straight to cathode (again it doesn't matter for static operation). This should nullify any chances of tube squeg or whatever better than any grid resistors, as I think.
Marko
Re: Oscillations Redux
sngecko, Sun Apr 22 2012, 01:03AM
Thanks Marko and Steve---
That contraption I put together there didn't make any sense to me either. Sorry, Marko. At least a 1uF cap along that path makes sense to me. I understand that the feedback signal will be slightly out of phase, so the cap sort of smooths the gap a bit, right? Since my input will be manual, fed from a pot off a linear 15v supply, there shouldn't be any weird HF inputs to interact with that cap.
I'll get to work at once on the grid stopper and see what happens, then try out the opamp cap.
sngecko, Sun Apr 22 2012, 01:03AM
Thanks Marko and Steve---
That contraption I put together there didn't make any sense to me either. Sorry, Marko. At least a 1uF cap along that path makes sense to me. I understand that the feedback signal will be slightly out of phase, so the cap sort of smooths the gap a bit, right? Since my input will be manual, fed from a pot off a linear 15v supply, there shouldn't be any weird HF inputs to interact with that cap.
I'll get to work at once on the grid stopper and see what happens, then try out the opamp cap.
Re: Oscillations Redux
sngecko, Mon Apr 23 2012, 06:10PM
THIS JUST IN: I attached a 1.5k grid stopper and fired up the laser. Unfortunately the symptoms persisted, but I at least learned that only the main 24V case fan is slowing in the presence of this oscillation. The other two 24V boost fans (one for each tube channel) seems to operate at or near normal parameters during the noise.
AGAIN: I just confirmed that with a .22uF cap across the output and negative input pin... no change. Still, an audible squeal increasing in frequency with current through the circuit. So, the last culprit is, in fact, the triode, right?
I guess the 1.5k stopper resistor didn't do the job. Also, I need to look closer at the input to that silly fan. I just can't see where it's coupled to the RF noise -- drive circuitry as someone suggested, I guess.
What would be the effect of tying the grid to the cathode with a .5uf capacitor, as suggested above?
sngecko, Mon Apr 23 2012, 06:10PM
THIS JUST IN: I attached a 1.5k grid stopper and fired up the laser. Unfortunately the symptoms persisted, but I at least learned that only the main 24V case fan is slowing in the presence of this oscillation. The other two 24V boost fans (one for each tube channel) seems to operate at or near normal parameters during the noise.
AGAIN: I just confirmed that with a .22uF cap across the output and negative input pin... no change. Still, an audible squeal increasing in frequency with current through the circuit. So, the last culprit is, in fact, the triode, right?
I guess the 1.5k stopper resistor didn't do the job. Also, I need to look closer at the input to that silly fan. I just can't see where it's coupled to the RF noise -- drive circuitry as someone suggested, I guess.
What would be the effect of tying the grid to the cathode with a .5uf capacitor, as suggested above?
Re: Oscillations Redux
Marko, Wed Apr 25 2012, 12:09AM
Not sere why, but you just seem to insist on leaving out that crucial resistor that would give sense to the whole thing. You need something to form a voltage divider on your feedback signal!
It's the RC time constant that slows down the response, and if you make it into like seconds (1 meg , 1uF) I have strong feelings that oscillations should stop on that part.
That's of course the most basic solution of all and you have some big homework to do on regulators if you wish to optimize your impulse response speed, overshoot and so on.
I also have no idea if such a simple solution is even proper for a negative resistance load such as a laser tube - you could have a well working current regulator which still creates a relaxation oscillator with the unintuitive aid of some capacitor. (needles to say, don't use a "filter capacitor" across your tube!)
To test the regulator, always use dummy resistance first (or just keep it shorted) until you've gotten it to work properly.
Marko
Marko, Wed Apr 25 2012, 12:09AM
Not sere why, but you just seem to insist on leaving out that crucial resistor that would give sense to the whole thing. You need something to form a voltage divider on your feedback signal!
It's the RC time constant that slows down the response, and if you make it into like seconds (1 meg , 1uF) I have strong feelings that oscillations should stop on that part.
That's of course the most basic solution of all and you have some big homework to do on regulators if you wish to optimize your impulse response speed, overshoot and so on.
I also have no idea if such a simple solution is even proper for a negative resistance load such as a laser tube - you could have a well working current regulator which still creates a relaxation oscillator with the unintuitive aid of some capacitor. (needles to say, don't use a "filter capacitor" across your tube!)
To test the regulator, always use dummy resistance first (or just keep it shorted) until you've gotten it to work properly.
Marko
Re: Oscillations Redux
sngecko, Wed Apr 25 2012, 02:41AM
You know? That's a really good point that I seem to have overlooked... test the regulation portion with a dummy load, then work back.
I just started a new thread that addresses what you put up there, essentially an inverting integrator, right? Well, the way my circuit is set up, I have a non-inverting, unity, closed loop feedback chain. I am learning more and more how this is an inherently metastable arrangement that often leads to RF oscillations of all sorts, especially with the TL080 series op-amps. They apparently have a relatively high output impedance that just keeps the oscillations coming. I've considered the following solutions:
(1) Twist the gate drive and source wires together all the way down to the off-board regulator elements
(2) Replace the TL084s with AD713s, with lower output impedance
(3) I like this one for some reason: Put a 30-something Ohm resistor in the feedback loop with several turns of magnet wire around it (parallel hookup) to block HF component, but pass the DC component
(4) Build up a non-inverting integrator!
sngecko, Wed Apr 25 2012, 02:41AM
You know? That's a really good point that I seem to have overlooked... test the regulation portion with a dummy load, then work back.
I just started a new thread that addresses what you put up there, essentially an inverting integrator, right? Well, the way my circuit is set up, I have a non-inverting, unity, closed loop feedback chain. I am learning more and more how this is an inherently metastable arrangement that often leads to RF oscillations of all sorts, especially with the TL080 series op-amps. They apparently have a relatively high output impedance that just keeps the oscillations coming. I've considered the following solutions:
(1) Twist the gate drive and source wires together all the way down to the off-board regulator elements
(2) Replace the TL084s with AD713s, with lower output impedance
(3) I like this one for some reason: Put a 30-something Ohm resistor in the feedback loop with several turns of magnet wire around it (parallel hookup) to block HF component, but pass the DC component
(4) Build up a non-inverting integrator!
Re: Oscillations Redux
Steve Conner, Wed Apr 25 2012, 06:38AM
What Marko said, your integrator is not "non-inverting", or the feedback would be positive.
Steve Conner, Wed Apr 25 2012, 06:38AM
What Marko said, your integrator is not "non-inverting", or the feedback would be positive.
Re: Oscillations Redux
Marko, Wed Apr 25 2012, 07:54AM
No... just add a 1 meg resistor into your feedback path.
Marko
Marko, Wed Apr 25 2012, 07:54AM
sngecko wrote ...
You know? That's a really good point that I seem to have overlooked... test the regulation portion with a dummy load, then work back.
I just started a new thread that addresses what you put up there, essentially an inverting integrator, right? Well, the way my circuit is set up, I have a non-inverting, unity, closed loop feedback chain. I am learning more and more how this is an inherently metastable arrangement that often leads to RF oscillations of all sorts, especially with the TL080 series op-amps. They apparently have a relatively high output impedance that just keeps the oscillations coming. I've considered the following solutions:
(1) Twist the gate drive and source wires together all the way down to the off-board regulator elements
(2) Replace the TL084s with AD713s, with lower output impedance
(3) I like this one for some reason: Put a 30-something Ohm resistor in the feedback loop with several turns of magnet wire around it (parallel hookup) to block HF component, but pass the DC component
(4) Build up a non-inverting integrator!
You know? That's a really good point that I seem to have overlooked... test the regulation portion with a dummy load, then work back.
I just started a new thread that addresses what you put up there, essentially an inverting integrator, right? Well, the way my circuit is set up, I have a non-inverting, unity, closed loop feedback chain. I am learning more and more how this is an inherently metastable arrangement that often leads to RF oscillations of all sorts, especially with the TL080 series op-amps. They apparently have a relatively high output impedance that just keeps the oscillations coming. I've considered the following solutions:
(1) Twist the gate drive and source wires together all the way down to the off-board regulator elements
(2) Replace the TL084s with AD713s, with lower output impedance
(3) I like this one for some reason: Put a 30-something Ohm resistor in the feedback loop with several turns of magnet wire around it (parallel hookup) to block HF component, but pass the DC component
(4) Build up a non-inverting integrator!
No... just add a 1 meg resistor into your feedback path.
Marko
Re: Oscillations Redux
sngecko, Wed Apr 25 2012, 11:31AM
Okay. I get that now. What if I'd just like to slow its frequency response down to around 3-5kHz, rather than almost DC?
EDIT: After looking at some similar gas discharge regulation circuits, I've determined that a basic inverting integrator with about 4700pF filter capacitor and 46.4kOhm resistor (and, of course, a similar bias resistor), should solve this problem admirably. Thanks all.
sngecko, Wed Apr 25 2012, 11:31AM
Okay. I get that now. What if I'd just like to slow its frequency response down to around 3-5kHz, rather than almost DC?
EDIT: After looking at some similar gas discharge regulation circuits, I've determined that a basic inverting integrator with about 4700pF filter capacitor and 46.4kOhm resistor (and, of course, a similar bias resistor), should solve this problem admirably. Thanks all.
Re: Oscillations Redux
sngecko, Wed Apr 25 2012, 02:28PM
[DELETED by author for crimes against rationality]
sngecko, Wed Apr 25 2012, 02:28PM
[DELETED by author for crimes against rationality]
Print this page