A couple of clarifications
1. I didn't contact Walt before I did my PCB's, but did afterwards, primarily to say 'thank you', partly to ease my conscience (I actually had no intention of selling them, but was encouraged by popular demand).
Were I to do the same again I would ask for permission first, but since I do not make any significant income from the sale of these units (I value the feedback more than any financial gain) I felt less guilty - it's not a good excuse though!
2. As is mentioned on my website, Walt may not necessarily agree with all the component choices I made for that circuit either, certainly they are different from his original choices.
3. I think it's only polite to thank those that publish such circuits for their efforts, there's often considerable work put into such ventures that the reader may not be aware of.
My interest in the design was primarily that whilst widely recognised as offering great performance, the circuits seemed to be buried in the mists of time a bit. The constant clamour for information here, that was not served by any of the online resources I could find, swayed me to take a greater interest.
There was also a reputation for stability issues with the AD797 versions - these were addressed by Walt in his follow up 'Improved Regulator' article and in my experience the instability tag is no longer deserved.
There's an increasing number of commercial products starting to use what I strongly suspect is Walt's topology, with modification in many cases, but it seems to have taken some time for it to happen - companies such as Audiocom International who are marketing 'Super Regulators' or Trichord Research who are using Jung-type topologies in their phono stages.
I agree with Jan's comments - it's been very obvious to me, from the initial attempts by others to produce PCB's for the Jung reg's, that many seem to think that a 'PCB is simple'.
Design at this level requires serious attention to detail and Jan, like me, will have spent a long time developing that circuit board, I'm sure. Equally Walt will have expended great effort and time on the original circuit development, I am certain.
One can never assume that something doesn't matter, in audio, it often does.
Andy.
1. I didn't contact Walt before I did my PCB's, but did afterwards, primarily to say 'thank you', partly to ease my conscience (I actually had no intention of selling them, but was encouraged by popular demand).
Were I to do the same again I would ask for permission first, but since I do not make any significant income from the sale of these units (I value the feedback more than any financial gain) I felt less guilty - it's not a good excuse though!
2. As is mentioned on my website, Walt may not necessarily agree with all the component choices I made for that circuit either, certainly they are different from his original choices.
3. I think it's only polite to thank those that publish such circuits for their efforts, there's often considerable work put into such ventures that the reader may not be aware of.
My interest in the design was primarily that whilst widely recognised as offering great performance, the circuits seemed to be buried in the mists of time a bit. The constant clamour for information here, that was not served by any of the online resources I could find, swayed me to take a greater interest.
There was also a reputation for stability issues with the AD797 versions - these were addressed by Walt in his follow up 'Improved Regulator' article and in my experience the instability tag is no longer deserved.
There's an increasing number of commercial products starting to use what I strongly suspect is Walt's topology, with modification in many cases, but it seems to have taken some time for it to happen - companies such as Audiocom International who are marketing 'Super Regulators' or Trichord Research who are using Jung-type topologies in their phono stages.
I agree with Jan's comments - it's been very obvious to me, from the initial attempts by others to produce PCB's for the Jung reg's, that many seem to think that a 'PCB is simple'.
Design at this level requires serious attention to detail and Jan, like me, will have spent a long time developing that circuit board, I'm sure. Equally Walt will have expended great effort and time on the original circuit development, I am certain.
One can never assume that something doesn't matter, in audio, it often does.
Andy.
Beyond individuals here at diyaudio, how about a big company like Linear Technology that might incorporate the idea into a new component say the LT1963, sell it for half the cost of ALW's PCBA, and make serious money with the idea. Who knows what is going on inside their parts. It may be an exact copy of Jung's circuit. Who in the public knows? Necessary capacitance values noted in the datasheet. I'm not saying this is the case here, but I'm sure this has happened more than once.
We on topic peranders?
JF
We on topic peranders?
JF
Don't think so. First of all, you design different when you use silicone, many transistors, current mirrors, multi emitter transistors and other "weird" parts , very few caps, etc.
Much of the new thinking has taken place in the semiconductor factories just because you can design circuits which aren't possible in an another way. This is only what I think.
Much of the new thinking has taken place in the semiconductor factories just because you can design circuits which aren't possible in an another way. This is only what I think.
AC?DC?NaC?
"I also wonder how long wires or pcb traces do you think is necessary in order to force use of sense wires? It's cool, but is it necessary if you talk AC performance and disregard a couple of uV is DC loss?"
The series resistance of the wires is added to the output impedance of the regulator. This is MORE OF AN ISSUE for AC than the DC drop because it affects the regulation of the voltage at the device drawing the AC current (i.e., music signal) . Remote sensing puts the wiring impedance inside the feedback loop so the regulator's low impedance appears at the load instead at the regulator output with the wiring impedance between the regulator and the load. Doing remote sensing with a regulator with a several tens of MHz GBW is very challenging engineering and beyond the scope of many analog designers. PCB layout is super critical. The error signals that the op amp in a well designed regulator is sensing in it's feedback loop are below the microvolt level. Without an absolute understanding of exactly where the signal currents run and where the feedback sense points are taken, a few millimeters of printed circuit board trace in the wrong place can greatly effect the regulation. Milliamps through Milliohms matter.
I think Andy is safe from any real competition in his design for a long time. Designing a very good regulator can be even harder than good amplifier or low noise preamp circuit. Most don't even know what the design problems are much less the solutions. This is real engineering with a vengeance at this level and why many find designing regulated supplies that sound good is a very frustrating task. Ever wonder why so many people use emitter followers and don't go very far beyond that for power supply regulators?
"I also wonder how long wires or pcb traces do you think is necessary in order to force use of sense wires? It's cool, but is it necessary if you talk AC performance and disregard a couple of uV is DC loss?"
The series resistance of the wires is added to the output impedance of the regulator. This is MORE OF AN ISSUE for AC than the DC drop because it affects the regulation of the voltage at the device drawing the AC current (i.e., music signal) . Remote sensing puts the wiring impedance inside the feedback loop so the regulator's low impedance appears at the load instead at the regulator output with the wiring impedance between the regulator and the load. Doing remote sensing with a regulator with a several tens of MHz GBW is very challenging engineering and beyond the scope of many analog designers. PCB layout is super critical. The error signals that the op amp in a well designed regulator is sensing in it's feedback loop are below the microvolt level. Without an absolute understanding of exactly where the signal currents run and where the feedback sense points are taken, a few millimeters of printed circuit board trace in the wrong place can greatly effect the regulation. Milliamps through Milliohms matter.
I think Andy is safe from any real competition in his design for a long time. Designing a very good regulator can be even harder than good amplifier or low noise preamp circuit. Most don't even know what the design problems are much less the solutions. This is real engineering with a vengeance at this level and why many find designing regulated supplies that sound good is a very frustrating task. Ever wonder why so many people use emitter followers and don't go very far beyond that for power supply regulators?
When "Sense"-inputs come up everyone talk how important it is what AC decoupling near the regulator in order to avoid unstability. Doesn't this apply here also?
Walter Jung from the EDN article.
Walter Jung from the EDN article.
It has VERY much to do with load. If you have long wires you have also "L". What happens in an amp if you put an inductor at the output and then connect feedback after this L = oscillations if you are unlucky.
So, Jim you say also that AC wise it's no (or little) use with remote sensing because you short circuit the signal near the regulator.
So, Jim you say also that AC wise it's no (or little) use with remote sensing because you short circuit the signal near the regulator.
Re: AC?DC?NaC?
Fred, I agree completely. I remember how challenging it is just to measure the Zout of this things. You need very carefull 4-wire impedance measuring techniques or you end up measuring the resistance of a piece of wire or PCB trace. I got some very nice flat traces of just 10 or 12 milliohms flat to 100kHz. Great supply? No. The supply had less than 1milliohms up til 20kHz, and rising. If you don't see the rising, you'r measuring the wrong thing, or you are a genius. I know I am not.
One of the initial problems we had to stabilise the supplies with 797's was due to remote sensing. The added inductance (and capacitance) inside the loop were just too much.
Jan Didden
Fred Dieckmann said:
Fred, I agree completely. I remember how challenging it is just to measure the Zout of this things. You need very carefull 4-wire impedance measuring techniques or you end up measuring the resistance of a piece of wire or PCB trace. I got some very nice flat traces of just 10 or 12 milliohms flat to 100kHz. Great supply? No. The supply had less than 1milliohms up til 20kHz, and rising. If you don't see the rising, you'r measuring the wrong thing, or you are a genius. I know I am not.
One of the initial problems we had to stabilise the supplies with 797's was due to remote sensing. The added inductance (and capacitance) inside the loop were just too much.
Jan Didden
Jan:
>There used to be a big price advantage in production if you could stay with SS boards, even if you had to use tens of jumpers.<
Depends strongly on the local labor costs. In the 1970s and 1980s, the yen was cheap relative to the dollar, and wages were increasing but were not that high. In a situation like this, where labor costs are low, components that are cheap to make (like SS pcbs) will likely reduce the total cost of manufacture, even if the human labor content is comparatively high.
From around the middle 1980s, the value of the yen rose steeply, and so did Japanese wages. But the majority of Japanese audio products still relied on jumpered single-sided PCBs and complex wiring harnesses that looked like a ball of yarn after a cat had played with it. Even though the pcbs were themselves cheap, once the labor costs were added to the equation, the total cost of manufacture (and service and repair) was undoubtedly higher than if the designers would have opted for multi-layer pcbs.
It took some years for Japanese designers and management to get accustomed to new ways of doing things. Today, multilayer pcbs are quite common in Japanese audio equipment.
>There are things you can do in an IC that would be impossible in discrete form. And probably vice versa.<
Agreed, but I also think that there is a middle ground. By combining advanced transistors (like multi-gate FETs), multilayer pcbs, smd, three-dimensional construction techniques, it is possible to build circuits that, although may not quite match an IC for thermal tracking and complexity, do allow a lot of IC-style topologies and techniques (i.e., base-current compensation) to be used. Something along the lines of hybrid circuits.
regards, jonathan carr
>There used to be a big price advantage in production if you could stay with SS boards, even if you had to use tens of jumpers.<
Depends strongly on the local labor costs. In the 1970s and 1980s, the yen was cheap relative to the dollar, and wages were increasing but were not that high. In a situation like this, where labor costs are low, components that are cheap to make (like SS pcbs) will likely reduce the total cost of manufacture, even if the human labor content is comparatively high.
From around the middle 1980s, the value of the yen rose steeply, and so did Japanese wages. But the majority of Japanese audio products still relied on jumpered single-sided PCBs and complex wiring harnesses that looked like a ball of yarn after a cat had played with it. Even though the pcbs were themselves cheap, once the labor costs were added to the equation, the total cost of manufacture (and service and repair) was undoubtedly higher than if the designers would have opted for multi-layer pcbs.
It took some years for Japanese designers and management to get accustomed to new ways of doing things. Today, multilayer pcbs are quite common in Japanese audio equipment.
>There are things you can do in an IC that would be impossible in discrete form. And probably vice versa.<
Agreed, but I also think that there is a middle ground. By combining advanced transistors (like multi-gate FETs), multilayer pcbs, smd, three-dimensional construction techniques, it is possible to build circuits that, although may not quite match an IC for thermal tracking and complexity, do allow a lot of IC-style topologies and techniques (i.e., base-current compensation) to be used. Something along the lines of hybrid circuits.
regards, jonathan carr
peranders said:
Sorry, couldn't find "coolness factor" in my Oxford Dictionary. Other than that, the question really is, do you want to go to great length to make a very quiet and stable supply and then throw it overboard with too long, uncompensated wiring. The improvement with remote sensing is unquestionably, repeatably demonstrable, and not subtle in technical terms. Can you hear it? Depends on a lot of other things.
Or is your question: how important is superquiet and stable supply? I can't answer that question for you. I can't hear the difference between a BG or a Siemens cap, except in my wallet.
I have pleasure in trying to wring the last drop of performance from a circuit, knowing that it will make the resulting piece of equipment come closer to a straight-wire-with-gain. I may not be able to hear that last drop, but everyone has a turn-on. This is (one of) mine.
Jan Didden
Hi Andy,
I was going to email some questions to you, but since this thread is going, decided to post them here, might help someone else.
I am planning to use your Jung regulator, with the tracking pre regulator (LM317) to generate +/- 15VDC for my FET preamp.
I was planning on feeding it 21 VDC unregulated (15 VAC rectified).
First question is how much does it care about the unregulated VDC going in? I was thinking about getting a split core, over rated transformer (maybe 100 VA), and using Jensen 4 pole electrolytics in the rectifier circuit. Since the Jung is so good, can I use cheaper caps in the rectifier, and not notice a difference? Does the transformer make a difference?
How much decoupling can the target circuit have? I saw references to not exceed 100 uf decoupling caps, and not to exceed 1000 uf total capacitance.
I also need to know how to build a -15 VDC regulator, all the info in the manual seems to be for a +15 regulator (I think this question started the thread).
I was thinking about building separate regulators for left and right channels, but is that overkill? I would need 4 regulators then, seems like a lot, but I might do it if it improves the sound. Any comments? I will probably build it with two at first, and think about adding two more.
Also wondered how the regulators perform powering digital and clock circuits, generating 3.3 or 5 VDC. Would that be a good application for a Jung? I understand I would need a 2.5 V reference for this.
Thanks for any information you can provide. Hope this was not too many questions.
Randy
I was going to email some questions to you, but since this thread is going, decided to post them here, might help someone else.
I am planning to use your Jung regulator, with the tracking pre regulator (LM317) to generate +/- 15VDC for my FET preamp.
I was planning on feeding it 21 VDC unregulated (15 VAC rectified).
First question is how much does it care about the unregulated VDC going in? I was thinking about getting a split core, over rated transformer (maybe 100 VA), and using Jensen 4 pole electrolytics in the rectifier circuit. Since the Jung is so good, can I use cheaper caps in the rectifier, and not notice a difference? Does the transformer make a difference?
How much decoupling can the target circuit have? I saw references to not exceed 100 uf decoupling caps, and not to exceed 1000 uf total capacitance.
I also need to know how to build a -15 VDC regulator, all the info in the manual seems to be for a +15 regulator (I think this question started the thread).
I was thinking about building separate regulators for left and right channels, but is that overkill? I would need 4 regulators then, seems like a lot, but I might do it if it improves the sound. Any comments? I will probably build it with two at first, and think about adding two more.
Also wondered how the regulators perform powering digital and clock circuits, generating 3.3 or 5 VDC. Would that be a good application for a Jung? I understand I would need a 2.5 V reference for this.
Thanks for any information you can provide. Hope this was not too many questions.
Randy
I leave the first question to the experts.. but a normal transformer will do and also normal caps but also allways you can do it more well done if you know what I mean? If it's a question of budget, you can start with simplier parts and then "upgrade".randytsuch said:
About the capacitance at the output, some regulators can't take huge capacitance because the go into oscillations. It depends have the regulator is designed but if you can't measure it you can allways insert 1-100 ohms in series. With this you isolate the regulator from difficult load. The regulation gets a little bit worse but it's strongly dependent of the situation.
Just change all polarized parts, transisitors (NPN->PNP), diods, caps and reference. Note also thatthe opamp allways gets negative voltage at pin 4 and positive at pin 7.randytsuch said:
IMPORTANT: The opamp must be able to work (or at least start) at common mode voltage near positive rail. Far from everyone can do this. Just test or choose an rail-to-rail opamp.
Overkill, maybe but this is my trademark.... It's not bad to have a "serious" power but you can have more serious to chosen circuits.randytsuch said:
For digital circuits you need low impedance at 1-500 MHz or so. The absolute value isn't very important. Serious decoupling is VERY impoprtant. Also proper grounding.randytsuch said:
Instead of a pre regulator you can also do as I have done. Put small resistors after the recifier and to get even more filtering use an another RC- filter. I have a R-C-R-C filter in order to get smooth voltage. This is a passive solution so you must have control over the max load. An active pre regulator is more flexible if you have varying (spelling?) loads.randytsuch said:
I leave the first question to the experts..
If only the rest of the questions as well....
'you can allways insert 1-100 ohms in series. With this you isolate the regulator from difficult load. The regulation gets a little bit worse but it's strongly dependent of the situation."
At low frequencies, several orders of magnitude. That is the first time that I have seen differences of thousands of times higher described as "a little bit worse"
"Just change all polarized parts, transisitors (NPN->PNP), diods, caps and reference. Note also thatthe opamp allways gets negative voltage at pin 4 and positive at pin 7."
NO! Besides too being vague for useful advice for anyone who doesn't understand to a degree that would make it unnecessary to ask the question, it is wrong. The circuits are different enough to require different PCBs for the negative and positive regulators. The circuits are similar and require PNP-NPN and capacitor polarity differences, but the circuit topologies are different as well and require a different schematic and PCB layout for positive and negative versions.
"IMPORTANT: The opamp must be able to work (or at least start) at common mode voltage near positive rail. Far from everyone can do this. Just test or choose an rail-to-rail opamp."
Please explain this one. After reading about the development of this circuit for several years I have never seen this claim made. The bootstrap circuit for startup forces the voltage within the common mode range while the circuit is powering up. the circuit has been used with a variety of op amps. One of the most popular has a common mode input range 1.5 V below the supply voltage at the op amps power pins which is hardly a rail to rail op amp.
I hope my fellow forum member Pan'andlers (I let you call me Phred...) will address these concerns regarding his advice.
If only the rest of the questions as well....
'you can allways insert 1-100 ohms in series. With this you isolate the regulator from difficult load. The regulation gets a little bit worse but it's strongly dependent of the situation."
At low frequencies, several orders of magnitude. That is the first time that I have seen differences of thousands of times higher described as "a little bit worse"
"Just change all polarized parts, transisitors (NPN->PNP), diods, caps and reference. Note also thatthe opamp allways gets negative voltage at pin 4 and positive at pin 7."
NO! Besides too being vague for useful advice for anyone who doesn't understand to a degree that would make it unnecessary to ask the question, it is wrong. The circuits are different enough to require different PCBs for the negative and positive regulators. The circuits are similar and require PNP-NPN and capacitor polarity differences, but the circuit topologies are different as well and require a different schematic and PCB layout for positive and negative versions.
"IMPORTANT: The opamp must be able to work (or at least start) at common mode voltage near positive rail. Far from everyone can do this. Just test or choose an rail-to-rail opamp."
Please explain this one. After reading about the development of this circuit for several years I have never seen this claim made. The bootstrap circuit for startup forces the voltage within the common mode range while the circuit is powering up. the circuit has been used with a variety of op amps. One of the most popular has a common mode input range 1.5 V below the supply voltage at the op amps power pins which is hardly a rail to rail op amp.
I hope my fellow forum member Pan'andlers (I let you call me Phred...) will address these concerns regarding his advice.
Re: I leave the first question to the experts..
I don't say this is a big problem but you should be aware of it.
Fred, you know perfectly well that it is the high frequencies that are important and also more difficult for the regulators. A couple of millivolts in ripple, hum isn't especially harmful in front of a Jung regulator or any for that matter. We talk about raw voltage fed to a regulator so thousand times better than what.... at the output? Fred, I don't say it's wrong to use a pre regulator, just that there might be other solutions not that bad.Fred Dieckmann said:
I don't know about Andy's pcb (haven't checked close enough) but As I see only the opamp must have it's pin 7 and 4 swapped (requires a litlle bit patching) otherwise I don't know what mean.Fred Dieckmann said:
Fred, some opamps have "properties" if they are used outside the common mode limits. If it's good or bad you can't tell until you have tested the type. One thing to remember also is that different brands of the same type can behave differently.Fred Dieckmann said:
I don't say this is a big problem but you should be aware of it.
"I have Jan article in front of me and the PCBs for the + and - look close enough.
I would be very interested in knowing how many iterations the ALW pcb required. IMO, the the bandwidth is low enough to be pretty forgiving."
Then why go though the time consuming task of laying out two different PCBs for the negative and positive regulator? Would someone actually look at a schematics and stop shooting from the hip. They are similar
but not identical. I think a board could be designed for both options but compromise for the op amp PS terminals would very likely compromise the design. I am sorry that this circuit looks so easy to a lot of you, but the PCB layout is critical to a state of the art power supply design like this. The bandwidth of the op amp is pretty high and the understanding of all the voltage drops in the layout for the correct location of the sense points is very demanding. There are any number of ways that the performance can be compromised by even small layout changes from the optimum. I just don't know why this is unclear. We're talking about signals and trace resistences SIGNIFICANTLY BELOW the microvolt, milliamp, and milliohm level.
Another real factor is that offering a kit with complicated stuffing options is a disaster waiting to happen with people unfamiliar the design, even with excellent documentation. A board with a silkscreen useful for both? I can't even imagine, and I have designed PCBs with stuffing options much simpler that have still confused people despite my best efforts.
I would be very interested in knowing how many iterations the ALW pcb required. IMO, the the bandwidth is low enough to be pretty forgiving."
Then why go though the time consuming task of laying out two different PCBs for the negative and positive regulator? Would someone actually look at a schematics and stop shooting from the hip. They are similar
but not identical. I think a board could be designed for both options but compromise for the op amp PS terminals would very likely compromise the design. I am sorry that this circuit looks so easy to a lot of you, but the PCB layout is critical to a state of the art power supply design like this. The bandwidth of the op amp is pretty high and the understanding of all the voltage drops in the layout for the correct location of the sense points is very demanding. There are any number of ways that the performance can be compromised by even small layout changes from the optimum. I just don't know why this is unclear. We're talking about signals and trace resistences SIGNIFICANTLY BELOW the microvolt, milliamp, and milliohm level.
Another real factor is that offering a kit with complicated stuffing options is a disaster waiting to happen with people unfamiliar the design, even with excellent documentation. A board with a silkscreen useful for both? I can't even imagine, and I have designed PCBs with stuffing options much simpler that have still confused people despite my best efforts.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.