I'm thinking of building a MOSFET voltage regulator to supply a steady 400v, with load current varying between ~50mA and ~130mA and input voltage varying between ~430v and ~470v. Series regulators are sometimes criticized for their noise, so I was wondering, should a series MOSFET voltage regulator be less noisy using just a voltage ref. on the gate, rather than using an error amplifier? Because of the gain of a MOSFET, an error amp seems superfluous.
In theory if you had a infinite gm MOSFET, then the impedance looking into to Source (1/gm) would be 0 ...... Now if you used this perfect MOSFET for a series regulator with only a reference voltage at the GATE, then you would have no worries about regulation or noise and best of all you would not need to close the loop with feedback using an error amplifier....
In reality you will have an impedance looking into this SOURCE...
The feedback, error amp output , will wigle in opposition to the output at the SOURCE, providing a error correction at the gate...
This will lower the output impedance of the SOURCE , by a factor of the gain of the (ERROR Amp x Gain of the MOSFET stage)+1 ....
As for noise, this should lower as well by the factor of the famous feedback equation...
The error amp must by properly compensated to avoid instability around the closed loop..... A quick look at the BODE plot of the OPEN loop response combined with the ERROR AMP repsonse will show GAIN and PHASE MARGIN of the closed loop response......
Chris
In reality you will have an impedance looking into this SOURCE...
The feedback, error amp output , will wigle in opposition to the output at the SOURCE, providing a error correction at the gate...
This will lower the output impedance of the SOURCE , by a factor of the gain of the (ERROR Amp x Gain of the MOSFET stage)+1 ....
As for noise, this should lower as well by the factor of the famous feedback equation...
The error amp must by properly compensated to avoid instability around the closed loop..... A quick look at the BODE plot of the OPEN loop response combined with the ERROR AMP repsonse will show GAIN and PHASE MARGIN of the closed loop response......
Chris
Ray, from the requirements, it sounds like this is the B+ supply for a push-pull AB output stage? If so, why worry much about the noise?
If noise really is an issue, then by all means use an active loop, but you can also get great noise performance from a Maida-style regulator with the reference resistor and output bypassed. Then it really doesn't matter much what the pass device is, the noise is controlled by the three pin reg. You've seen Erroll Dietz's data on 317 noise and bypassing?
If noise really is an issue, then by all means use an active loop, but you can also get great noise performance from a Maida-style regulator with the reference resistor and output bypassed. Then it really doesn't matter much what the pass device is, the noise is controlled by the three pin reg. You've seen Erroll Dietz's data on 317 noise and bypassing?
Cerrem,
Thanks but I'm not sure if that answers my question or not (I couldnt' follow all of it!) I understand your point about impedance, but the noise thing is a mystery.
SY,
Screen + voltage amp stages for EL34 pentode PP amp in AB1. Screen current can vary with signal amplitude from 5mA to 25mA per tube. Noise could be a problem, I suppose, depending on whose opinion you listen to. Just thought I'd ask which approach is likely to be less noisy.
Thanks but I'm not sure if that answers my question or not (I couldnt' follow all of it!) I understand your point about impedance, but the noise thing is a mystery.
SY,
Screen + voltage amp stages for EL34 pentode PP amp in AB1. Screen current can vary with signal amplitude from 5mA to 25mA per tube. Noise could be a problem, I suppose, depending on whose opinion you listen to. Just thought I'd ask which approach is likely to be less noisy.
Well, in that situation, any noise is common mode, so you get some advantage there. The mu-g2 is also a very low number. And the output transformer steps it down. So you can have a surprising amount of ripple or noise and still not see it in the output; the stiffness at DC is more important at this spot.
As a reality check, the screen reg on my RLD amp is a very simple Maida (no bypass on the set resistor, 47u bypass on the output), and the entire amp shows noise better than 80dB down from 1V. I doubt that the limitation there is from the screen.
A properly bypassed Maida will give you about 2-3uV of noise independent of the pass device. That's 160dB below what the tubes are swinging...
As a reality check, the screen reg on my RLD amp is a very simple Maida (no bypass on the set resistor, 47u bypass on the output), and the entire amp shows noise better than 80dB down from 1V. I doubt that the limitation there is from the screen.
A properly bypassed Maida will give you about 2-3uV of noise independent of the pass device. That's 160dB below what the tubes are swinging...
Thanks, SY, that's encouraging. If I build a simple series MOSFET reg, the voltage reference on the gate will need to be 400v. I could get this using a stack of either zener diodes or VR tubes. Which do you think would be better? Another possibility might be a CCS from the unregulated B+ through a resistor to ground. The voltage ref would be across the resistor.
Yes, it would be filtered. The voltage from the zener/VR tube stack would go through a 1Meg res. to the gate (via a 22k stopper). At the junction of the 1Meg res. and the stopper would be a 4.7uF cap to ground, This would both filter the voltage reference and act as a delay while the tubes warm up.
SY said:
Horowitz and Hill have a very nice MOSFET regulator in "The Art of Electronics" You can analyze it in Tina or LTSpice by breaking the feedback loop with a 50 ohm resistor, inserting a signal source and drawing the phase-gain plot.
For reference, the Maida regulator is described in National Semi "Linear Brief 47" -- from March 1980 -- it uses a zener to limit Vin-Vout. while the regulator floats. I honestly don't know whether it is capable of providing a noise level as low as Sid suggests (microvolts). The reference in the LM317 just isn't noiseless enough for that level of performance. Jan Didden did a lot of painstaking work characterizing the LM317 regulators back in the last century
Here's a snapshot of the Last-PAS regulator performance -- (Audio Amateur 4/82) -- this too uses a zener to limit the difference across the regulator, but ties the ground to a zener string (bypassed). the annoying thing about the LastPAS regulator is that it takes up a bit of space (althoughh it conveniently fits exactly into the allotted real estate on the Dyna PAS preamps).
An externally hosted image should be here but it was not working when we last tested it.
i think you have to take a look at the charts in 2/95 Audio Amateur -- at any rate, here are some bandwidth limited results integrated over 10 seconds -- using a 5223A storage scope and a 5a22n differential amplifier -- it's a much lower noise setup than my TDS3012B -- (edit) I also use the Jim Williams "cookie tin" isolation method and zero out the charge on the preamp before each measurement:
BASELINE:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
BASELINE:
An externally hosted image should be here but it was not working when we last tested it.
From my experience: simple source follower is enough for grids of output toobs and pre powering.
I used an one transistor error amp in the power supply with positive feedback by current that applies voltage to tubes when they are hot and start consuming current.
I can repost the schematic if you want.
I used an one transistor error amp in the power supply with positive feedback by current that applies voltage to tubes when they are hot and start consuming current.
I can repost the schematic if you want.
Jack, I am totally wrong here. On reviewing the Dietz paper, he shows noise voltage versus frequency, which is in the microvolt range. But without knowing his measurement bandwidth, it's likely that the rms value is pretty high as you're seeing.
The interesting thing about his results is that the value of the output bypass controls that noise spectrum because of the interaction with the source resistance and inductance of the 317.
The interesting thing about his results is that the value of the output bypass controls that noise spectrum because of the interaction with the source resistance and inductance of the 317.
Hi, Wavebourn.
I am encouraged by your comments, as well those from SY.
Hi, SY,
Could an LM317 be used effectively as a current source, to develop a fixed voltage across a resistor to apply to the gate of a MOSFET series regulator, do you think? It would avoid having a string of high-voltage zeners, which I don't like the idea of, or a string of VR tubes, which would be unwieldy and take up space.
I am encouraged by your comments, as well those from SY.
Yes, that would be interesting, thanks.
Hi, SY,
Could an LM317 be used effectively as a current source, to develop a fixed voltage across a resistor to apply to the gate of a MOSFET series regulator, do you think? It would avoid having a string of high-voltage zeners, which I don't like the idea of, or a string of VR tubes, which would be unwieldy and take up space.
A first, it starts softly, raising a voltage up slowly, waiting for tubes to warm up and to start consuming a current. As the result of a current, due to a positive feedback it more quickly increases output voltage up to a regulated value, after that a positive feedback stops working, but a negative feedback starts to dominate, stabilizing output voltage on a given level, and decreasing a ripple level, without a costly and heavy chokes, without vacuum tubes to regulate output voltage.
Please be advised that no one polarity of a rectifier and filter cap must be grounded!
Also, it produces a negative bias voltage that may be used for good, otherwise leave it as is (it is a positive feedback loop for soft starting).
Vout=12V*(R5/(R6+R7)) so you may vary output voltage.
Minimum input voltage should be no less than 5V + ripple + Vout + 12V.
However, you may use it for another voltages. For example, 36V zener may be used for a bias stabilizer (it is a shunt regulator that takes the current from the entire amp), and resistors in feedback recalculated for 350V output (however, make sure that selected transistors and electrolytics can work on such voltages), to power 6L6 amp. 400V rectified in such case would be enough.
Please be advised that no one polarity of a rectifier and filter cap must be grounded!
Also, it produces a negative bias voltage that may be used for good, otherwise leave it as is (it is a positive feedback loop for soft starting).
Vout=12V*(R5/(R6+R7)) so you may vary output voltage.
Minimum input voltage should be no less than 5V + ripple + Vout + 12V.
However, you may use it for another voltages. For example, 36V zener may be used for a bias stabilizer (it is a shunt regulator that takes the current from the entire amp), and resistors in feedback recalculated for 350V output (however, make sure that selected transistors and electrolytics can work on such voltages), to power 6L6 amp. 400V rectified in such case would be enough.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.