You need to experience enough of such stuff in good systems to know on your own. There are both excellent series and parallel regs and there are so so ones. Both families share some general subjective characteristics. Use the one of your design that gives you confidence in better spec IMHO. Because having the amp made, debugged, and listened to for days, tuning it for bias and passive parts even, in a great quality large scale system to push its high nominal power and then test pick regs to liking, its not feasible right now I guess.
I dont know, but mr Thel(Germany) seems to like the term "electronic choke", and also favours slow and soft regulation
but then, also claims nothing beats a good old fashioned ordinary supply, when sufficient
but then, also claims nothing beats a good old fashioned ordinary supply, when sufficient
I would agree. The closer you approach perfect regulation, the higher is the loop gain, and therefore the problems of most feedback circuits such as stability, overshoot, etc. Subjectively perfect regulation with tubes gives very, very good SS sound, if that is indication of what regulation can do; it points the way to tube interaction with the power supply being the source of the so-called 'tube sound'.
Hugh
Hugh
ich glaube dies ist es, Sie muessen aber selber uebersetzen, bitte 😀
Info Spannungsregler
wird wohl aber eine experte wie dich nur zum lachen bringen 😛 but I still think there are some good hints, between the lines 🙂
but a tube regler in SS amp, that would be something new, and old
Lamm does that. Has Mosfet preamps with tube regs.
P.S. The auto translation butchers the THEL texts. They probably had traumatic experiences with typical older chip regs it seems though.
P.S. The auto translation butchers the THEL texts. They probably had traumatic experiences with typical older chip regs it seems though.
A few years back, THEL used to swear on purely passive PSUs (either batteries or CRC or CLC).
He also did discrete circuits almost exclusively.
Now it seems that he changed his mind. Commercial reality ?
Patrick
He also did discrete circuits almost exclusively.
Now it seems that he changed his mind. Commercial reality ?
Patrick
kean,
You might be getting into the "spec" trap. In my opinion a good shunt always trounces a series regulator and as Hugh rightly pointed out there is the feedback question. Low output impedance in itself is not the end all.
Salas more than proved this with his simple shunt........try and make it too complex and you lose, same with the amplifier. If specs were everthing a Yamaha reciever by all rights should be the best sounding piece of equipment today.
I don't mean this to be a damper but ' if you live by the meter you die by the meter' most successful designers will agree with me on this.
Regards,
Jam
You might be getting into the "spec" trap. In my opinion a good shunt always trounces a series regulator and as Hugh rightly pointed out there is the feedback question. Low output impedance in itself is not the end all.
Salas more than proved this with his simple shunt........try and make it too complex and you lose, same with the amplifier. If specs were everthing a Yamaha reciever by all rights should be the best sounding piece of equipment today.
I don't mean this to be a damper but ' if you live by the meter you die by the meter' most successful designers will agree with me on this.
Regards,
Jam
Last edited:
If a meter is your only test equipment, then yes. If the only tool you have is a hammer then everything becomes a nail.
If this is so, it seems my philosophy regarding stability is close, and it is possible to tailor most regs (at least in simulation) to model an inductor accurately. It also suggests that a good reg will be stable (despite resonance) into high-Q loads.
I am somewhat puzzled at the preeminance of shunts, but one thing that strikes me is that the CCS provides high input isolation without having hooks in the regulator's AC characteristics (as opposed to series regs where Early affect and pass drive are both modulated by a usually larger degree by input voltage, and these parameters interact with how the rest of the circuit performs). The shunt approach seems to be a more conceptually cleaner approach.
So is it possible that the added isolation with shunts is what makes them superior to series regs with comparable specs?
- keantoken
kean,
You are correct, the current source in the shunt regulator is key to its superior performance IMHO.
Jam
You are correct, the current source in the shunt regulator is key to its superior performance IMHO.
Jam
BTW, Salas' circuits have far and beyond incredibly low output impedance, and amazing specs. Although SSHV would seem to trade specs for simplicity, as I gathered, which is not at all heinous. (I'm reading my first book in quite a few years, a biography of Einstein, which must be the cause of my suddenly increased language proficiency - I can hardly recognize my own writing!)
I've been investigating the role of the CCS in shunt performance, and as it turns out it affects DC input rejection more than I would have hoped. Using a MOSFET CCS doesn't help much, and so far I haven't improved on what I already use without unseemly complexity. I could really use a Jfet with 100mA Idss 😀. It is worth investigating constant-current diodes here?
- keantoken
I've been investigating the role of the CCS in shunt performance, and as it turns out it affects DC input rejection more than I would have hoped. Using a MOSFET CCS doesn't help much, and so far I haven't improved on what I already use without unseemly complexity. I could really use a Jfet with 100mA Idss 😀. It is worth investigating constant-current diodes here?
- keantoken
Thanks everyone for your insight. I've stopped to think more deeply about the shunt regulator. At first I was settled on it mainly because everyone agreed and I was confident it would perform to my standards as well (for however much they're worth). But after time to "chill", my eyes are growing accustomed to the light. From this discussion I think more could be desired from the CCS (Previously I have questioned why Salas and Ikoflexer use other types).
At first I did not plan to be so involved, and so I would have recommended Iko or Salas' circuits for the frontend. But I was encouraged to go on with my own process and it still seems that this is true. However I think Hugh knows that my preference is to err on the side of novelty, which is why I still pull in my direction when Iko and Salas seem to have already perfected the art of shunt. Even so, it helps no one to insist on a dead end, so if you have doubts I want to hear.
Must sleep now,
- keantoken
At first I did not plan to be so involved, and so I would have recommended Iko or Salas' circuits for the frontend. But I was encouraged to go on with my own process and it still seems that this is true. However I think Hugh knows that my preference is to err on the side of novelty, which is why I still pull in my direction when Iko and Salas seem to have already perfected the art of shunt. Even so, it helps no one to insist on a dead end, so if you have doubts I want to hear.
Must sleep now,
- keantoken
A depletion mode Mosfet can be your high current ''Jfet''. I have shown SSHV with Supertex DN2540 here for max simplicity, remember?
Oh, of course. Someone posted a Jfet CCS earlier I think which would be nearly perfect. Cascode this MOSFET, and I think we have it made, without increasing complexity.
do you have a good model for the DN2540? The supertexmodel I got needs a 180R resistor, giving 16V drop...
- keantoken
do you have a good model for the DN2540? The supertexmodel I got needs a 180R resistor, giving 16V drop...
- keantoken
Kean,
you can explain a shunt regulator to be like a lawn mover, it chops of all the junk above some set hight. Unlike a series regulator that just reduces everything, including the junk above a certain level. A series regulator is a more economical option when power consumption is the objective.
you can explain a shunt regulator to be like a lawn mover, it chops of all the junk above some set hight. Unlike a series regulator that just reduces everything, including the junk above a certain level. A series regulator is a more economical option when power consumption is the objective.
*DN2540 MODEL
*
.MODEL DN2540 NMOS (LEVEL=3 RS=1.05 NSUB=5.0E14
+DELTA=0.1 KAPPA=0.20 TPG=1 CGDO=3.1716E-10
+RD=11 VTO=-1.50 VMAX=1.0E7 ETA=0.0223089
+NFS=6.6E10 TOX=725E-10 LD=1.698E-9 UO=862.425
+XJ=6.4666E-7 THETA=1.0E-5 CGSO=2.50E-9 L=4.0E-6
+W=59E-3)
*.ENDS
Each one will have a different IDSS. Maybe 150mA is around max figure. Most real samples will need about 70R for 30mA. Don't trust the model to fix a current setting resistor in a real build. Also use a gate resistor, they do oscillate easy.
Nico, a series regulator will have trouble with back-EMF the same way a shunt will have trouble with forward-EMF. A series reg won't tolerate negative current, a shunt won't tolerate much forward current.
The frontend will never need to draw close to 100mA, unless we use a BJT output stage.
The CCS shown gives us 20db more input isolation. The increased CCS capacitance causes more resonance, but if it's a problem in real life I know where to put the RC network.
- keantoken
The frontend will never need to draw close to 100mA, unless we use a BJT output stage.
The CCS shown gives us 20db more input isolation. The increased CCS capacitance causes more resonance, but if it's a problem in real life I know where to put the RC network.
- keantoken
Attachments
Kean,
The front end only requires about 10mA. If you feel that 100mA is necessary, (10 times higher), there is nothing wrong with your choice. However, over specifying is normally dictated by the worst case calculations.
I apologise if the example a few hundred posts ago showing the transient response of the shunt regulator was confusing you. As you may well know, transient response relates to both bandwidth and impedance.
Nico
One could think, for some an inclination* to go for a push-pull regulator.
* odd disambiguation word of the day