New Power Supply II

[redwire]

The other thread seems to be corrupted or locked so I will continue here

Although I had many wires and jumpers the PS seemed to work well and testing seemed good until i shorted the board and burned some traces.  Happy with the design, I had 3 new boards made and populated 2 of them.  Unfortunately something is not right.  When there is no load, the voltage would spike up and down about once a second.  If I disconnect the feedback from the power transistor to the board it will not spike.    I changed C5  from 100pf to 200nf and the output was steady without a load.  If set the voltage to 4V and connect a load, the output shoots up and oscillated between 35V and 17V at about 800Hz.    The voltage and current regulation did not seem to work.  Only if I set the current to minimum and then apply a load it will not spike and oscillate.  I changed the output tranistor from the darlington to a 2N5886 and had the same results.  Any suggestions. 

[Guest liquibyte]

I've been trying to fix it to no avail.  I can't even see the posts after #1018137, probably a database issue.  I'm going to try and put things back the way they were but I may lose it altogether.

Edit: Yep, lost the first post, it just disappeared and the thread is still cut off.  Sorry about that.

Edit II: I got them merged back but the end is still cut off.  I'll see if I can get mixos to fix it.

[audioguru]

The new circuit oscillates because the output amplifier transistors provide too much voltage gain.
The original circuit and all the improvements until recently had only one driver transistor as an emitter follower with a voltage gain of only 1 like a piece of wire. Your new circuit has Q1 and Q2 as common emitter amplifiers feeding series resistors that add with the following capacitance which causes phase shift. 

[redwire]

Would adding a resistor to the base of Q1 (BD140) and to the base of the power transistor be a solution?  What about changing to complimenary transistor set up (BD139) with a lower gain rather than the 2N2222 which has a much higher gain?

[Hero999]

As long as Q1 is a common emitter amplifier there's a risk it'll oscillate, especially under certain conditions (a large capacitor connected to the output with a low ESR). This is because you've introduced another pole with a phase shift influenced by the load impedance.

http://en.wikipedia.org/wiki/Phase_margin

[redwire]

The new circuit oscillates because the output amplifier transistors provide too much voltage gain.
The original circuit and all the improvements until recently had only one driver transistor as an emitter follower with a voltage gain of only 1 like a piece of wire. Your new circuit has Q1 and Q2 as common emitter amplifiers feeding series resistors that add with the following capacitance which causes phase shift.

But I didn't add another driver transistor unless you are referring to the use of the MJ11016 darlington output transistor.  I did try the circuit with a 2N5668 and it had the same output problems. 

As long as Q1 is a common emitter amplifier there's a risk it'll oscillate, especially under certain conditions (a large capacitor connected to the output with a low ESR). This is because you've introduced another pole with a phase shift influenced by the load impedance.

http://en.wikipedia.org/wiki/Phase_margin

The original design used a 100uf cap at the output.
Initially I thought this was too high and changed this to 10uf like the other design when I had oscillations I changed it back to 100uf, but now it seems I was  heading in the wrong direction and need to change it back to 10uf.

It seems having a slow output tranistor with a fast opamp can be an issue. So would using a faster power transistor help?  If so, how do I choose a faster transistor from the data sheet, by the bandwidth product? a 2N3055 is listed at 6MHz, but  a 2N3442G is listed at 80mHz.  Would that help?

I added a 100k resistor to the base of the MJ11016 power transistor and the oscillations did stop.  I will change the 100uf cap back to 10uf .

I didn't like the response on the current control and the setup with the led connected directly to the output of the opamp concerns me.  It seems this indicator should be driven by a transistor.    I noticed that when the inverting pin was higher than the non inverting pin the output voltage on Pin 6 would slowly decrease (over seconds) instead of nearly instant control.  I may need to change C4 to 100pf instead of 220nf but I don't see that as a total fix becasue of how slow it is responding. 

[audioguru]

But I didn't add another driver transistor

Your circuit has Q1 with high voltage gain and a phase shift from R15. You also have Q2 with high voltage gain and phase shift from R13. Then you feedback the high gain and extra phase shifts to the inverting input of the opamp so it oscillates.
The original circuit used only a single driver transistor as an emitter-follower (a voltage gain of only 1) with a very small phase shift.

[redwire]

Your circuit has Q1 with high voltage gain and a phase shift from R15. You also have Q2 with high voltage gain and phase shift from R13. Then you feedback the high gain and extra phase shifts to the inverting input of the opamp so it oscillates.
The original circuit used only a single driver transistor as an emitter-follower (a voltage gain of only 1) with a very small phase shift.

This is not a revision of the Power supply with 163 pages.  This is a different design  based on the original circuit here.  Which was posted on Youtube here    Although I changed some values I did not add any transistors or reconfigure the basic circuit design but I did add some caps for the op amps which gives is a design that appear to be similar to the long posted design.

Regarding high gain, if a BJT transistor is current driven wouldn't reducing the current by adding a resistor to the base also reduce the gain and negate the effects. 
  I found with a 10k resistor on the darlington MJ11016  and a 1.2K with the 2N5886 the oscillation ceased.  I found I still needed at least a 47uf cap so I decided to keep the 100uf.  So far I have gotten no other recommendation to attack the phase shift  issue except redesign.

[audioguru]

I turned off the video because I could not understand the very strong Engrish accent combined with the echoes in the room he is in. I think in part 1 he mentioned a lousy old LM358 opamp (it has poor high frequency response).

One of the comments at the end of the video mentions poor transient response and/or oscillation.

The 100k series base resistor for the 2N2222 adds tremendous phase shift (it is feeding stray and transistor capacitance) that you do not want. The series 1k base resistor feeding the BD140 transistor also adds phase shift. But the opamp is designed so that its gain is less than 1 at a frequency where its internal phase shift causes negative feedback to become positive feedback.
The phase shifts from the two added transistors add to the opamp's internal phase shift and causes the opamp to oscillate at a lower frequency where the entire amplifier still has some gain.

[redwire]

The 100k series base resistor for the 2N2222 adds tremendous phase shift (it is feeding stray and transistor capacitance) that you do not want. The series 1k base resistor feeding the BD140 transistor also adds phase shift. But the opamp is designed so that its gain is less than 1 at a frequency where its internal phase shift causes negative feedback to become positive feedback.
The phase shifts from the two added transistors add to the opamp's internal phase shift and causes the opamp to oscillate at a lower frequency where the entire amplifier still has some gain.

Would using a voltage divider from the output of the Opamp to the 2N2222 allow a significantly lower base resistor and reduce the phase shift?  Would a different transistor also help?

[redwire]

Another question:  Is the 100K resistor base resistor for Q2 (2n2222 voltage control transistor)  really necessary?

[Hero999]

Would using a voltage divider from the output of the Opamp to the 2N2222 allow a significantly lower base resistor and reduce the phase shift?  Would a different transistor also help?

The only real solution to the stability problem is to use an emitter follower on the output.

[redwire]

How about this redesign ?

Circuit_II.pdf

[Hero999]

How about this redesign ?

That won't regulate the voltage though, will it?

What's the output of the current amplifier when the the non-inverting input > inverting input? You need to design it so the current amplifier can only reduce the output voltage, not increase it.

[redwire]

Yes I did forget to show the diode from pin 3 on the voltage Opamp to prevent positive voltage  from the current op amp.    The op amp (TLE2141) can handle 0-44V and sink up to 80ma.    I estimate there will be  no more than 8ma to sink on the current opamp and about 3ma output on the voltage op amp using the darlington transistor. 

[Hero999]

The op-amp will need a negative supply to compensate for the diode voltage drop and that of the op-amp's output stage.

[redwire]

The op-amp will need a negative supply to compensate for the diode voltage drop and that of the op-amp's output stage.

Yep, your right.  could you add another diode on Pin 3 of the Voltage op amp after  the other diode leading to the current op amp.  This way the voltage set for 0 V output is actually  0.7 on the line leading to the non inverting pin.  Then the current op amp would only need to bring pin 3  the line down to 0.7? 

.

[redwire]

Here is the sketch with the diode on Pin 3 to compensate for the diode drop on the current op amp.  Would this work?

[audioguru]

Hi Redwire,
On the original and fixed version your diode D2 was a resistor with almost NO voltage drop. Now the D2 has a voltage drop that changes when the temperature changes then the output voltage changes when the temperature changes.

Without a negative supply for the current control opamp the output current will not be regulated when it is set to less than about 0.8V/0.27 ohms= 2.96A. The output voltage of the current control opamp must be able to go to -0.7V so that regulation at low currents works. Then its negative supply must be at least -0.9V so use two diodes like I did.

The value of your R9 is MUCH too low.

The current sertting pot should have a low value resistor in series with its lower end so that any input offset voltage of the current control opamp does not have a huge effect.

[redwire]

Hi Redwire,
On the original and fixed version your diode D2 was a resistor with almost NO voltage drop. Now the D2 has a voltage drop that changes when the temperature changes then the output voltage changes when the temperature changes.

Without a negative supply for the current control opamp the output current will not be regulated when it is set to less than about 0.8V/0.27 ohms= 2.96A. The output voltage of the current control opamp must be able to go to -0.7V so that regulation at low currents works. Then its negative supply must be at least -0.9V so use two diodes like I did.

Thanks for the reply.    I guess there is no getting around not having a negative charge pump with this type of design.  I'm curious to see how this guy gets around the problem.  Considering he is using a .1 ohm resistor, the theoretical load would be substantial before there would be stability.  I think from a practical standpoint if there is a heavy load and 0.47 ohm sense reisistor, the ground for the current op amp is pushed low enough down to provide some control.

I am also going to re-look at the design on the first post here.  I seemed to have the oscillation under control and I could change the voltage smoothly.  I was having some current control issues in that the opamps seemed to be fighting each other until the current setting was set to its lowest setting.  If I set the voltage at 14V and the current at 100mv, then connect a load of 1.5 A the current control would never kick in.  It was like in a runaway mode.