Interesting John. Sorry to put you to a task (again).
Your output resistance results are better (lower) than I might have guessed from squinting at the IRF820 data sheet. And, as you say, the fact that output resistance is nearly constant across the audio band bodes well for its good sound. As you well know, this is in stark contrast to most high-loop feedback regulators that wimp out in the upper octaves of the audio band, even though they may achieve lower output resistance figures at low frequencies (similar to the behavior of most solid-state amps, not coincidently). Signal-current transients will cause overshoot or damped ringing, and the output Z takes on reactive qualities at upper frequencies in these high-feedback designs. That shouldn’t be an issue with your simpler series-pass design.
I’ve used variants of the LM317 Maida design a lot, and these tend to sound better to my ears when lightly decoupled and then shunted by serious composite capacitors at the load, where the effective power supply Z at high frequencies becomes capacitor dependent, for better or for worse.
I had shunned the simple pass MOSFET at these lower currents for fear that transconductance would be so low that the output resistance would be quite high. It isn’t just that it could be high, but that it is bound to be very non-linear (as your data bears out). I seek to keep any non-linear resistance as low a ratio of the plate resistance values as possible, as a rule of thumb, so that non-linear current-induced voltages will factor very little in the output voltage. In your case, a nominal 14mA draw might give an output resistance of roughly 18 ohms. I don’t know your Artemis circuits, but that must be at least 1000 times lower than any plate resistance, which is probably a safe-enough ratio. If the Artemis both “pushes and pulls”, keeping power supply current draw fairly constant, so much the better. Have you attempted to measure distortion on the B+ line with your Audio Precision? No, please don’t do it on my account
I just bought some IRF820s to repair a Tek scope and have leftovers, so maybe I will try this series-pass design next time. Also, while wasteful of power and more demanding on the raw supply upstream, a current loading power resistor across the B+ would both lower the output resistance, and lower the ratio of signal current to quiescent current. Probably not a great idea for commercial designs, but perhaps this would be an easy and temporary test to see how much difference it makes, and whether there is further “gold to be mined” in power supply improvements.
Your output resistance results are better (lower) than I might have guessed from squinting at the IRF820 data sheet. And, as you say, the fact that output resistance is nearly constant across the audio band bodes well for its good sound. As you well know, this is in stark contrast to most high-loop feedback regulators that wimp out in the upper octaves of the audio band, even though they may achieve lower output resistance figures at low frequencies (similar to the behavior of most solid-state amps, not coincidently). Signal-current transients will cause overshoot or damped ringing, and the output Z takes on reactive qualities at upper frequencies in these high-feedback designs. That shouldn’t be an issue with your simpler series-pass design.
I’ve used variants of the LM317 Maida design a lot, and these tend to sound better to my ears when lightly decoupled and then shunted by serious composite capacitors at the load, where the effective power supply Z at high frequencies becomes capacitor dependent, for better or for worse.
I had shunned the simple pass MOSFET at these lower currents for fear that transconductance would be so low that the output resistance would be quite high. It isn’t just that it could be high, but that it is bound to be very non-linear (as your data bears out). I seek to keep any non-linear resistance as low a ratio of the plate resistance values as possible, as a rule of thumb, so that non-linear current-induced voltages will factor very little in the output voltage. In your case, a nominal 14mA draw might give an output resistance of roughly 18 ohms. I don’t know your Artemis circuits, but that must be at least 1000 times lower than any plate resistance, which is probably a safe-enough ratio. If the Artemis both “pushes and pulls”, keeping power supply current draw fairly constant, so much the better. Have you attempted to measure distortion on the B+ line with your Audio Precision? No, please don’t do it on my account
I just bought some IRF820s to repair a Tek scope and have leftovers, so maybe I will try this series-pass design next time. Also, while wasteful of power and more demanding on the raw supply upstream, a current loading power resistor across the B+ would both lower the output resistance, and lower the ratio of signal current to quiescent current. Probably not a great idea for commercial designs, but perhaps this would be an easy and temporary test to see how much difference it makes, and whether there is further “gold to be mined” in power supply improvements.
Here's a sim -- with and without a 1R resistor in series with C100 and an additonal 5uF on the output -- i think I ran it with a 40ma current draw:
An externally hosted image should be here but it was not working when we last tested it.
Brian Beck said:
Ooops -- I was thinking of low voltage electrolytics -- the 250VDC and higher rated ones have ESR tens to hundreds higher. i.e. the low value, high voltage Mallory's have ESR around 2.4 ohms.
The resistor on the capacitor is going to lower the Q.
Here's the phase-gain plot of the Maida regulator -- simmed exactly per the Linear Brief 47 parts:
An externally hosted image should be here but it was not working when we last tested it.
Regulator distortion
The DP-2 is back in its box, getting ready to be shipped back to Sean Ta in Simi Valley. However, I remember this distortion on the regulator output being around 4-6% with a 10mA load - nearly entirely 2nd order. It would be nice if this was lower. In the LA-1, the load is a choke-loaded triode (and is mostly choke-loaded triode in the phono preamps and power amps), so the current fluctuations seen by the regulator are minimized.
- John
The DP-2 is back in its box, getting ready to be shipped back to Sean Ta in Simi Valley. However, I remember this distortion on the regulator output being around 4-6% with a 10mA load - nearly entirely 2nd order. It would be nice if this was lower. In the LA-1, the load is a choke-loaded triode (and is mostly choke-loaded triode in the phono preamps and power amps), so the current fluctuations seen by the regulator are minimized.
- John
Jackinnj,
When we think of regulator output impedance, we expect to see a rise in the magnitude at higher frequencies, and the phase therefore would go inductive. I'm not sure that I know what you're plotting in the magnitude plots from your simulations, particularly in the Maida case. What is the Y axis variable? I know it's log-encoded in dB, but, dB of what divided by what? The X axis is clearly log frequency from 1 Hz to 20MHz.
When we think of regulator output impedance, we expect to see a rise in the magnitude at higher frequencies, and the phase therefore would go inductive. I'm not sure that I know what you're plotting in the magnitude plots from your simulations, particularly in the Maida case. What is the Y axis variable? I know it's log-encoded in dB, but, dB of what divided by what? The X axis is clearly log frequency from 1 Hz to 20MHz.
question to Wavebourn
Hi Wavebourn
Concerning your regulator, do you allow me the question 'is it possible to substitute some parts on it'?
First, I already have some IRF820's, can they be used as pass element? And about the TIP129, I could not find them anywhere, not even a datasheet. Can you recommend me which other part I may put in place there?
I do thank you very much for your attention.
Erik
Here is a shunt regulator with mosfet. So simple even I understood how it works: includes delay!
http://www.diyaudio.com/forums/showthread.php?postid=1007648#post1007648
Hi Wavebourn
Concerning your regulator, do you allow me the question 'is it possible to substitute some parts on it'?
First, I already have some IRF820's, can they be used as pass element? And about the TIP129, I could not find them anywhere, not even a datasheet. Can you recommend me which other part I may put in place there?
I do thank you very much for your attention.
Erik
Here is a shunt regulator with mosfet. So simple even I understood how it works: includes delay!
http://www.diyaudio.com/forums/showthread.php?postid=1007648#post1007648
Re: question to Wavebourn
You are welcome!
Yes, you can use IRF820s if they fit maximal parameters you require from it. Break-down voltage (zener inside), maximal current (actuallt, limited by a resistor in source), and dissipation power.
About TIP129, I used such a monster to survive in case of surge when output capacitor is discharged by a screwdriver. Actually, it needs to sustain maximal current that may be rdaws from the regulator, it amplifies current of the Zener. You may use a rtansistor of a different polarity, in such case you have to mount on the opposite side of the Zedner, and collector must be isolated from ground.
Here is the pic from your link:
It's stability depends on stability of a current source (voltage drop on resistors in series with it defines regulated output voltage). Delay without resistors from current source to ground may be calculated as T=V*C/I where I is current of the current source (Amperes), C is capacitance of the capacitor from gate to ground, V is regulated voltage (voltage), T in seconds.
With resistors it will be shorter, and depends on 2pi*R
ErikdeBest said:
You are welcome!
Yes, you can use IRF820s if they fit maximal parameters you require from it. Break-down voltage (zener inside), maximal current (actuallt, limited by a resistor in source), and dissipation power.
About TIP129, I used such a monster to survive in case of surge when output capacitor is discharged by a screwdriver. Actually, it needs to sustain maximal current that may be rdaws from the regulator, it amplifies current of the Zener. You may use a rtansistor of a different polarity, in such case you have to mount on the opposite side of the Zedner, and collector must be isolated from ground.
Here is the pic from your link:
It's stability depends on stability of a current source (voltage drop on resistors in series with it defines regulated output voltage). Delay without resistors from current source to ground may be calculated as T=V*C/I where I is current of the current source (Amperes), C is capacitance of the capacitor from gate to ground, V is regulated voltage (voltage), T in seconds.
With resistors it will be shorter, and depends on 2pi*R
Hi Wavebourn
Thanks for your reply. I have a bunch of MJE340, bought for HV CCS duties, and from your reply I understand I can use those (they will survive way more than lets say 200mA of current).
The output voltage is limited to the max allowable voltage across the 2N3439, which is about 350V. Is this true?
About the shunt regulator. I thought that the delay would be simply calculated by the RC time constant of the 10k resistor and 47muF capacitor?! I once built a regulator like this one
http://www.diyaudio.com/forums/showthread.php?postid=1112167#post1112167
but put a way too large capacitor on the base of the mosfet. The time constant was surely something in the minutes range.
Erik
Thanks for your reply. I have a bunch of MJE340, bought for HV CCS duties, and from your reply I understand I can use those (they will survive way more than lets say 200mA of current).
The output voltage is limited to the max allowable voltage across the 2N3439, which is about 350V. Is this true?
About the shunt regulator. I thought that the delay would be simply calculated by the RC time constant of the 10k resistor and 47muF capacitor?! I once built a regulator like this one
http://www.diyaudio.com/forums/showthread.php?postid=1112167#post1112167
but put a way too large capacitor on the base of the mosfet. The time constant was surely something in the minutes range.
Erik
ErikdeBest said:
Now, multiply max possible current by a voltage on this current and see if it will be able to dissipate a heat of such power.
Yes!
You've asked about MOSFET, so I answered. Any transistor (strictly speaking, P-N junction) has own voltage limit when it starts acting like a Zener. That's why I protected base - emitter junction of that 2N3439 using a diode: it breaks down at about 6V of a reverse voltage
Not exactly.
Output resistance of a voltage stabilizer (current source loaded by 2x25K resistor) is 50K, so add it to your 10K.
Conclusion: sometimes white is wtite, black is black.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.