I am a fan of the design philosophy and expertise of the late Allen Wright, and over the years I have completed both of his hybrid valve preamplifiers: the single-ended Six-Valve Preamp (SVP) and the differential Realtime Preamplifier (RTP3) from his published circuits, with the help of his inspiringly off-piste Tube Preamplifier Cookbook (the TPCB). Both are original and fully developed designs, embodying Allen’s demanding philosophy of audio circuit design as presented in full in the TPCB, and their potential sound quality is very high. Brad Morrical, for instance, rated them both close to the top of his group test on Positive Feedback Online, and both preamps are generally highly regarded, with a consensus on their detailed, natural and fatigue-free presentation of music. Here is Roy Gregory’s HiFi+ review of the SVP, and here is Pat Kennedy’s Six Moons review of the RTP3D.
Allen always stressed the importance of power supply design, and in particular promoted choke regulation and active shunt regulators, giving plausible reasons why both ought to sound better than the default arrangements. The phono and line stages of the SVP, with their relatively high-valued anode load resistors, are rather susceptible to noise and non-trivial output impedance in the HT supply, as well as to feedback through the HT line. The RTP3, with its differential architecture, should be relatively immune to the details of the HT supply, but Allen nevertheless recommended the same PSU and regulation as with the SVP.
Both preamps as sold by Allen’s company, Vacuum State Electronics, whether in finished or kit form, include Allen’s SuperReg regulator, one per channel, and the SuperReg is also still available separately as a kit (debatably expensive at 150 euro each) from VSE. The circuit of the SuperReg is fully described in the TPCB; in brief, it has an op-amp based regulator with a MOSFET shunt element, and the regulator is fed by a constant current source with a MOSFET wrapped around an LM317 (though Allen latterly noted a sporadic instability with the LM317 and recommended its replacement, once the shunt current has been satisfactorily set for the given load, by an appropriately chosen resistor).
When I built my RTP3 I decided that a pair of SuperReg kits was too expensive to justify, and chose to build Emile Sprenger’s HPHV regulator, for which a detailed set of instructions is available online (e.g.<
redacted basaudio links>) instead.This in itself was a strong attraction, as the SuperReg circuit by contrast is only published in the TPCB or as a kit package from VSE. The HPHV is conceptually similar to the SuperReg, but is optimised to give maximum bandwidth by using a racehorse of an opamp (OPA655) and a carefully laid out PCB design, along with tweaks to the circuit such as current sources to stabilise reference voltages, and series chokes to reduce RF ingress. I made up some PCBs in the electronics lab at work from the manual, and built them up as per the instructions. I expected high performance from the regulators: from the start they appeared to work fine, apart from some HF instability as evidenced by interference to FM radio. I installed them in the first instance in my RTP3 and got good sound, but after a year or so the regulated voltage started to drift down, and in the end dropped low enough for the phono stage to drop out altogether. I never found out exactly why the regulators failed, but I suspect that it was a combination of building such a highly-tuned high-bandwidth circuit on an amateur-standard PCB, along with possible repeated overvoltages at its input. The PCB as presented in Emile's manual is a sophisticated double-sided thing with guard tracks around the op-amp inputs to reduce stray capacitance, but mine were significantly limited by the resolution of my home inkjet printer, and I did indeed have to clear one or two short circuits between tracks so I don’t have confidence that the PCB was up to scratch.
Still reluctant to fork out for the VSE kits, I did some research online and discovered Nikos Salas's SSHV project online: here is the schematic with a simulated response (which I don't quite understand, and am reluctant to believe). This is a much simpler approach to shunt regulation, replacing the op-amp with a pair of high-voltage NPN transistors, so reducing the gain of the regulator circuit, but retaining the current regulation at the input using a pair of cascaded DN2540 depletion mode MOSFETS. Although I had reservations about the low gain of the voltage regulator stage compared with the op-amp circuits in the SuperReg and HPHV (which had the potential to reduce the ripple rejection and regulation bandwidth, and raise the output impedance), I thought that this should be more robust and less finickety than the HPHV, and built a pair on matrix board. They worked as designed, and I installed a pair in each of my RTP3 and SVP. In the SVP - which is what I listen to most of the time - the result was a warm and musically involving sound, but there was an audible hum from the phono stage and after a while I started to notice a lack of definition and dynamics in the bass. After I had swapped in a passive pre for the line stage of the SVP and my old Audio Synthesis ADEQ phono stage I convinced myself that the SVP was the source of the deficiencies, and since the signal circuits were faithful copies of the ones Allen suppled the finger of blame hovered firmly over the voltage regulators.
I was finally at the stage where I was willing to try Allen’s own SuperRegs, so I overcame my natural parsimonious instincts and contacted Johanna at VSE to order a pair. These arrived in good time: the kits consist of a good-quality PCB, an almost-complete set of parts, and a good set of instructions. I say “almost complete” only because the kit doesn’t include a regulator for the op-amp supply, and I had to drill holes to add a simple LM317 circuit to do the job. Allen always recommended replacing the default op-amp in the SuperReg (actually an OP-07 in the kits I have) with something like an AD797, providing a suitable decoupling cap is installed between the power pins of the IC, so I ordered a couple of AD797s and some ceramic disc caps. The kit was very easy to build, and both boards worked straight away with no teething problems. The regulated voltage and current proved easy to adjust and very stable, and as a bonus it didn't induce any hissing or crackling on BBC Radio Three FM. I installed the new regulators in my SVP and, a couple of hours warmup later, noticed an immediate improvement in the areas I had found disappointing with the Salas regs: acoustic and electric basses sounded firmer and clearer, and when I turned the volume up the sound swelled with it, becoming more dynamic and palpable, with none of the sogginess I had got used to before. I can now hear further into the soundstage, and the hum and buzz from the phono stage has almost vanished.
So what are my conclusions? I still believe that Emile’s HPHV is technically the best design of the three, and would expect that one built on a professionally produced PCB - a group buy was organised on DIYAudio recently - and when protected from voltage transients at its input would perform flawlessly and possibly audibly better than the SuperReg: Emile’s design philosophy (similar to Allen’s) was that the wider the available bandwidth of the regulator the better the performance at the upper end of the audio spectrum, and his HPHV is a fully-realised incarnation of that. The simplicity of the SSHV2 is appealing, and it is certainly much more straightforward to build, as well as not needing the extra 17V op-amp supply of either of the other two. All the same, in my experience at least its performance is not on the same level as the others, although I can’t confirm my listening impressions with measurements, and it may suit other circuits better than it did mine. The VSE SuperReg design is not in the public domain, and the kit is expensive, but appears to work as intended and - in my system, at least - sounds better than with the SSHV2. There is of course the added glow of confidence that this is the very same circuit as used in Allen’s top-end commercial preamps. I am very happy with the sound of mine in my SVP, and have so far had no issues with it - and in fact I recently bought and built up a second pair for my RTP3.
Alex
Allen always stressed the importance of power supply design, and in particular promoted choke regulation and active shunt regulators, giving plausible reasons why both ought to sound better than the default arrangements. The phono and line stages of the SVP, with their relatively high-valued anode load resistors, are rather susceptible to noise and non-trivial output impedance in the HT supply, as well as to feedback through the HT line. The RTP3, with its differential architecture, should be relatively immune to the details of the HT supply, but Allen nevertheless recommended the same PSU and regulation as with the SVP.
Both preamps as sold by Allen’s company, Vacuum State Electronics, whether in finished or kit form, include Allen’s SuperReg regulator, one per channel, and the SuperReg is also still available separately as a kit (debatably expensive at 150 euro each) from VSE. The circuit of the SuperReg is fully described in the TPCB; in brief, it has an op-amp based regulator with a MOSFET shunt element, and the regulator is fed by a constant current source with a MOSFET wrapped around an LM317 (though Allen latterly noted a sporadic instability with the LM317 and recommended its replacement, once the shunt current has been satisfactorily set for the given load, by an appropriately chosen resistor).
When I built my RTP3 I decided that a pair of SuperReg kits was too expensive to justify, and chose to build Emile Sprenger’s HPHV regulator, for which a detailed set of instructions is available online (e.g.<
redacted basaudio links>) instead.This in itself was a strong attraction, as the SuperReg circuit by contrast is only published in the TPCB or as a kit package from VSE. The HPHV is conceptually similar to the SuperReg, but is optimised to give maximum bandwidth by using a racehorse of an opamp (OPA655) and a carefully laid out PCB design, along with tweaks to the circuit such as current sources to stabilise reference voltages, and series chokes to reduce RF ingress. I made up some PCBs in the electronics lab at work from the manual, and built them up as per the instructions. I expected high performance from the regulators: from the start they appeared to work fine, apart from some HF instability as evidenced by interference to FM radio. I installed them in the first instance in my RTP3 and got good sound, but after a year or so the regulated voltage started to drift down, and in the end dropped low enough for the phono stage to drop out altogether. I never found out exactly why the regulators failed, but I suspect that it was a combination of building such a highly-tuned high-bandwidth circuit on an amateur-standard PCB, along with possible repeated overvoltages at its input. The PCB as presented in Emile's manual is a sophisticated double-sided thing with guard tracks around the op-amp inputs to reduce stray capacitance, but mine were significantly limited by the resolution of my home inkjet printer, and I did indeed have to clear one or two short circuits between tracks so I don’t have confidence that the PCB was up to scratch. Still reluctant to fork out for the VSE kits, I did some research online and discovered Nikos Salas's SSHV project online: here is the schematic with a simulated response (which I don't quite understand, and am reluctant to believe). This is a much simpler approach to shunt regulation, replacing the op-amp with a pair of high-voltage NPN transistors, so reducing the gain of the regulator circuit, but retaining the current regulation at the input using a pair of cascaded DN2540 depletion mode MOSFETS. Although I had reservations about the low gain of the voltage regulator stage compared with the op-amp circuits in the SuperReg and HPHV (which had the potential to reduce the ripple rejection and regulation bandwidth, and raise the output impedance), I thought that this should be more robust and less finickety than the HPHV, and built a pair on matrix board. They worked as designed, and I installed a pair in each of my RTP3 and SVP. In the SVP - which is what I listen to most of the time - the result was a warm and musically involving sound, but there was an audible hum from the phono stage and after a while I started to notice a lack of definition and dynamics in the bass. After I had swapped in a passive pre for the line stage of the SVP and my old Audio Synthesis ADEQ phono stage I convinced myself that the SVP was the source of the deficiencies, and since the signal circuits were faithful copies of the ones Allen suppled the finger of blame hovered firmly over the voltage regulators.
I was finally at the stage where I was willing to try Allen’s own SuperRegs, so I overcame my natural parsimonious instincts and contacted Johanna at VSE to order a pair. These arrived in good time: the kits consist of a good-quality PCB, an almost-complete set of parts, and a good set of instructions. I say “almost complete” only because the kit doesn’t include a regulator for the op-amp supply, and I had to drill holes to add a simple LM317 circuit to do the job. Allen always recommended replacing the default op-amp in the SuperReg (actually an OP-07 in the kits I have) with something like an AD797, providing a suitable decoupling cap is installed between the power pins of the IC, so I ordered a couple of AD797s and some ceramic disc caps. The kit was very easy to build, and both boards worked straight away with no teething problems. The regulated voltage and current proved easy to adjust and very stable, and as a bonus it didn't induce any hissing or crackling on BBC Radio Three FM. I installed the new regulators in my SVP and, a couple of hours warmup later, noticed an immediate improvement in the areas I had found disappointing with the Salas regs: acoustic and electric basses sounded firmer and clearer, and when I turned the volume up the sound swelled with it, becoming more dynamic and palpable, with none of the sogginess I had got used to before. I can now hear further into the soundstage, and the hum and buzz from the phono stage has almost vanished.
So what are my conclusions? I still believe that Emile’s HPHV is technically the best design of the three, and would expect that one built on a professionally produced PCB - a group buy was organised on DIYAudio recently - and when protected from voltage transients at its input would perform flawlessly and possibly audibly better than the SuperReg: Emile’s design philosophy (similar to Allen’s) was that the wider the available bandwidth of the regulator the better the performance at the upper end of the audio spectrum, and his HPHV is a fully-realised incarnation of that. The simplicity of the SSHV2 is appealing, and it is certainly much more straightforward to build, as well as not needing the extra 17V op-amp supply of either of the other two. All the same, in my experience at least its performance is not on the same level as the others, although I can’t confirm my listening impressions with measurements, and it may suit other circuits better than it did mine. The VSE SuperReg design is not in the public domain, and the kit is expensive, but appears to work as intended and - in my system, at least - sounds better than with the SSHV2. There is of course the added glow of confidence that this is the very same circuit as used in Allen’s top-end commercial preamps. I am very happy with the sound of mine in my SVP, and have so far had no issues with it - and in fact I recently bought and built up a second pair for my RTP3.
Alex
Last edited by a moderator:
Any kind of regulated power supply, over the possibility of regulate the output between defined limits, and maintain it in time, and temperature, load and line variations; must offer as low output impedance to the loads as possible.
So, from this fact, I can't see why a shunt regulator can behave better than a series one, as both of them have the same characteristics, as it would be seen as a "black box" ignoring what is there inside it. Then, I can't understand why a shunt regulator will "sound better" than a series one with same output parameters. Is there any measurements over the "ear impressions"?
Cheers.
So, from this fact, I can't see why a shunt regulator can behave better than a series one, as both of them have the same characteristics, as it would be seen as a "black box" ignoring what is there inside it. Then, I can't understand why a shunt regulator will "sound better" than a series one with same output parameters. Is there any measurements over the "ear impressions"?
Cheers.
If there is inductance and a swinging current in the circuit under load a series regulator has less effective regulation that a shunt.
A shunt regulator has a clearly defined power send and return path. So the current circulation paths may be better defined and shorter.
The one few people understand is that a shunt regulator having a close to zero impedance to ground is in parallel with the power rails and thus will reduce the thermal noise. Note a series regulator regulation is modeled by a small resistor in series with the voltage source, assuming a voltage source with zero thermal noise.
Thanks for the explanation, but I continue not seeing any difference. A shunt regulator offering the same low resistance (High conductance) across its output, must behave in the same manner than a series regulator of the same output resistance, provided both have the same parasitics and both are in their linear regions. Again, regarding the PS as a black box, and "looking" from its output terminals, it must be indistinguishable between the two topologies, else they aren't similar. In such a case, they are't neither comparable.
A shunt regulator offers high conductance over a high resistance reference, a series one has a low resistance over a low conductance reference.
A shunt regulator offers high conductance over a high resistance reference, a series one has a low resistance over a low conductance reference.
Just look at the rise time of a series regulator under a noise spike compared to a shunt regulator. The should be the same. Now look at the fall time of the same spike. The series regulator can only have decay from the load. The shunt will drop the spike as quickly as it can slew the other way in a properly designed shunt regulator.
If you test your black box regulators by by imposing a sine wave on the power rails there will be more distortion with a series regulator than with a shunt. Now look at how the rails are bypassed and see if there is rail modulation! (Best example "Boot Strap" capacitors.)
I think you also missed the ability to place a shunt at the point of circulating currents. Try modelling the "Ground" with a series of small resistors.
A theoretical advantage of a shunt regulator with a current source at its input is that this guarantees a very high dynamic impedance relative to any practical value of fixed resistor, and so should in principle offer better isolation from the PSU than a simple series regulator.
How this works in practice, especially at high frequencies, is a different question.
Alex
How this works in practice, especially at high frequencies, is a different question.
Alex
OK, in case of noise ripple/spikes, the shunt "shunts" it to ground, the series blocks it to reach the load. The result must be the same.
There was an interview with Walt Jung in AudioXpress which is archived on his site. Comments on shunt regulators at the end: http://waltjung.org/PDFs/AX_WJ_Interview.pdf
For low impedance you need a fast error amplifier and a high gm regulating device.
EDIT: Oh, and by the way, there is no need of an external power supply and voltage regulator for the HPHV if one were to redesign -- one could use a boosted LR8N3 regulator for the error amplifier and voltage reference. For best PSRR, a DMOS current source for both.
For low impedance you need a fast error amplifier and a high gm regulating device.
EDIT: Oh, and by the way, there is no need of an external power supply and voltage regulator for the HPHV if one were to redesign -- one could use a boosted LR8N3 regulator for the error amplifier and voltage reference. For best PSRR, a DMOS current source for both.
Last edited:
I agree that in theory a perfect regulator, whether series or shunt, should have the same ripple rejection. On the other hand, what I think Simon is trying to say is that the load regulation (the response of the regulator to dynamic changes in the current drawn by the load - which may be resistive or reactive) is not the same between the two types.
Alex
Hi Alex. From a strictly theoretical point of view, the shunt regulator will operate better when the load releases current in the bus, deriving it to the return lead. The series will do it better under load demands of current, carrying it from the bulk main filter(s). So, one of them will affect the rise time of the load voltages, and the other, the fall time. What I don't understand how do this fact affects in a notable manner the "sounding" of each one.
From Walt Jung interview
Didden-Hawksford interview
https://www.google.com.ar/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwiO9a7OrOPTAhWDgZAKHYFLAAAQFggmMAA&url=https%3A%2F%2Flinearaudio.nl%2Fsites%2Flinearaudio.net%2Ffiles%2FDidden-Hawksford_Part1.pdf&usg=AFQjCNGHMuxBtGnD0_ZZDr_1Ymp0tU2qVw&cad=rja
Last edited:
In an amplifier with two power rails and a capacitor across them when one rail is loaded and sags that will push the other rail up. A series regulator has no way to pull it back down other than load demand.
Me too.
In general terms the design is almost good, except perhaps for the use of the TL431.
Using a 2SK369 (noise about 1nV/√Hz @ 1KHz) to feed a noisy TL431 (noise about 130nV/√Hz @ 1KHz) seems to me a total waste, and without a pass filter (a resistor was forgotten) is not my idea of a technically good design.
On the other hand, Salas shunt regulator is superb, its only "flaw" could be a little thermal drift which can be easily solved with a small isothermal enclosure.
BTW, the SuperReg seems the worst of the three, at least in simulations.
I would be very interested in your simulations. What bandwidth and output impedance do you estimate for the SSHV?
Alex
Hi, Salas Shunt Regulator 250V for 50mA load
PSRR
Output impedance
Transient response
Hard to beat...
Attachments
Very interesting, thanks! The simulations of the SSHV do indeed look pretty good. I'm surprised that a zero-feedback, low-OLG circuit like his can have such a low impedance.
I'm intrigued by the impedance plot for the SuperReg, though. It has a VERY low impedance over the most of the audio band, which is what you would expect from an op-amp-based circuit. But where does the time constant for that nasty peak at 30kHz come from? Is that the instability in the constant current circuit that Allen talked about? Could you repeat the sim for the case where the LM317 is replaced by a resistor chosen to give, say, a 50mA current?
What I didn't mention in my article is that in the TPCB Allen was vehemently against including electrolytic caps in any critical circuit, but there in the SuperReg is a 47uF cap in full view across the regulated output. In the HPHV Emile manages to get away with only 1uF PP caps across the HV line.
Alex
I'm intrigued by the impedance plot for the SuperReg, though. It has a VERY low impedance over the most of the audio band, which is what you would expect from an op-amp-based circuit. But where does the time constant for that nasty peak at 30kHz come from? Is that the instability in the constant current circuit that Allen talked about? Could you repeat the sim for the case where the LM317 is replaced by a resistor chosen to give, say, a 50mA current?
What I didn't mention in my article is that in the TPCB Allen was vehemently against including electrolytic caps in any critical circuit, but there in the SuperReg is a 47uF cap in full view across the regulated output. In the HPHV Emile manages to get away with only 1uF PP caps across the HV line.
Alex
Any regulated power supply should have a low output resistance, but what does that mean in practice?
It means only that the output voltage is stable for different loads from a static viewpoint. Make a variable load and measure the output voltage differences.
This by no means explains how this regulator reacts dynamicly and how it does affect sound.
In fact, I have not heard any series tube regulator that doesnt make the music sound slow in terms of dynamic. And thats where the parallel regulator has many advantages and why Audio Note uses parallel regulator I think.
Actually that is why the impedance is shown as a function of frequency (for example, see the simulation results posted by popilin above). For small changes, the impulse/time response is quite trivially related to the frequency response.
The "static" impedance you refer to is the impedance at 0Hz.
Alex
- Status
- Not open for further replies.