You could do all of us a great favour, and possibly to yourself a favour too, to scale up and replace your electrolytic capacitors with film capacitors, then let us know if you like the shunt reg better or the Jung SR.
Before I did my upgrades, although the JSR sounded more details than the LM317/337, I preferred the sound of the LM317/337 as it was more "musical" and easier on the ears.
It all changed now after I used the film capacitors, removed C2, and used proper output capacitors (currently 10uF/50V Nichicon Gold, and a 1.1uF MKP in series with a 0.5R film resistor). It is a night and day difference if comparing to the LM317/337.
I reckon I still need to spend at least this weekend to optimize it. The sound was a bit bright and harsh at higher frequencies before I added the output 1.1uF+0.5R. Now it is much cleaner but the harshness was pushed up higher in the frequency. I will try different padding resistance for the film cap and get the optimal value. The same tunning need to be applied to the pre-regulator, so that it can provide lower impedance at higher frequencies to the SR.
The shunt reg you attached looks very interesting. I wonder why the CSS is made from a LM317 not a more serious CSS. Anyway, I think it should sound very good. But I am not convinced that it would sound better than my current implementation of the JSR. Of course, I am open minded. I am thinking about building the "best" shunt reg and compare it to the JSR.
I am happy to build all of the "best" ones and compare them. It is a hobby and I enjoy doing it.
I spent half an hour last night browsing through some randomly picked pages of that long thread. I reckcon it would take a few days to go through them. Has it evolved to V2? On which page can I find the latest schematic?
There are both v1.5 and v2 finished, built, and tested, by me and some others. V1 is the very well established. v1.5 extends the output impedance bandwith of v1. v2 does that and further lowers it. v1.5 and v2 are both targeted at people who would want the output impedance (and phase) to be both low and flat in the audio freq. We're having some guests arriving momentarily, but I'll chase the right schematics for you after that. Can you tell me approximately what load you have in mind?
Build the Salas Reg. The first version is a guaranteed winner. The other versions are higher performance versions.
Build all of them.
Build all of them.
I've been trying to read up on shunt regs... still a newb tho...
can these be used to source currents for Class A amps?
sorry if this sounds stupid.
can these be used to source currents for Class A amps?
sorry if this sounds stupid.
So, the semiconductors will have voltage drops across them? Given that adequate heat sinking is provided, will it be possible to power say 3A per channel using this?
there are so many of these out there... can you please post a link with one that can achieve the above.
thanks in advance.
there are so many of these out there... can you please post a link with one that can achieve the above.
thanks in advance.
At least one of the members built one for his power amp (v1, and v1.5, and v2). Not a stupid question. I'm working exactly on this, to develop a shunt reg for a power amp, anything up to 50W for now. See the link I posted before.
I have a prototype that works very well at 3-4A at 20V output. Low output impedance, very low noise (order of tens of microvolts). I'll post it when it's ready.
I have a prototype that works very well at 3-4A at 20V output. Low output impedance, very low noise (order of tens of microvolts). I'll post it when it's ready.
Thanks!
I've been thinking of going through that long long thread. But like you suggested, maybe I should just put what output voltage and current I want.
Will try to keep up from now on.
I've been thinking of going through that long long thread. But like you suggested, maybe I should just put what output voltage and current I want.
Will try to keep up from now on.
Ikoflexer,
I am not sure how much current the final circuit will draw. As of the design today, which may be subject to change after final measurements, there are about 4 single and 18 dual OPA627 / LM4562 opamps. The maximum current draw of an OPA627 is 45mA, but I am not sure if it uses that much current under normal load. It may use perhaps 8mA each? I don't know. It can be difficult to calculate because it is frequency dependent and the opamps are running in class A/B.
Let me be conservative and assume that the circuit will draw 350mA. The rail voltages are +/-15VDC.
It would be good if you could point me to the V2 schematic.
Regards,
Bill
I am not sure how much current the final circuit will draw. As of the design today, which may be subject to change after final measurements, there are about 4 single and 18 dual OPA627 / LM4562 opamps. The maximum current draw of an OPA627 is 45mA, but I am not sure if it uses that much current under normal load. It may use perhaps 8mA each? I don't know. It can be difficult to calculate because it is frequency dependent and the opamps are running in class A/B.
Let me be conservative and assume that the circuit will draw 350mA. The rail voltages are +/-15VDC.
It would be good if you could point me to the V2 schematic.
Regards,
Bill
I have no doubt that the Jung Supereg can sound optimal if implemented correctly. It is very well designed and measured. What I am trying to do is to implement it optimally under real load condition.
The first thing I tried was to get rid of the electrolytic capacitors in the signal path and this has resulted in dramatical improvements over the original I built.
The next thing is to optimize the remaining supply capacitors.
1. The opamp within the SR needs a very low impedance supply across a wide bandwidth in order to work properly to produce a wide bandwidth low impedance regulator. It is a catch-22. I can not see how at very high frequencies the supply, as in my current implementation, is good enough for the opamp. Adding a 0.1uF high Q film / ceramic capacitor to the supply pins, which most manufacturer datasheets recommend, will be problematic because it forms a LCR resonance circuit with little damping. But without this high frequency bypass, I can hear the harshness from the regulator. What I am trying to do now is to add in a film capacitor for bypass and pad it with just sufficient R to damp any possible resonance. It is difficult without measurement equipment. If the R is too high (i.e. higher than the regulator output impedance at high frequencies), it won't be effective. If the R is too low, it causes ringing.
2. Since the Jung SR sources currents from the pre-tracking regulator, and the impedance of the LM317/337 is rising above 1kHz and becomes quite high above 100kHz, I guess this will severely limit the performance of the JSR. A regulator like LM317/337 is intended to be used at lower frequencies and at higher frequencies it relies on effective bypassing. So what I am planning to experiement is to add high frequency capacitors in series with damping resistors to the output of the PR, and hopefully this may help reduce some harshness in the treble of my current JSR.
I will report back when I get the results.
The first thing I tried was to get rid of the electrolytic capacitors in the signal path and this has resulted in dramatical improvements over the original I built.
The next thing is to optimize the remaining supply capacitors.
1. The opamp within the SR needs a very low impedance supply across a wide bandwidth in order to work properly to produce a wide bandwidth low impedance regulator. It is a catch-22. I can not see how at very high frequencies the supply, as in my current implementation, is good enough for the opamp. Adding a 0.1uF high Q film / ceramic capacitor to the supply pins, which most manufacturer datasheets recommend, will be problematic because it forms a LCR resonance circuit with little damping. But without this high frequency bypass, I can hear the harshness from the regulator. What I am trying to do now is to add in a film capacitor for bypass and pad it with just sufficient R to damp any possible resonance. It is difficult without measurement equipment. If the R is too high (i.e. higher than the regulator output impedance at high frequencies), it won't be effective. If the R is too low, it causes ringing.
2. Since the Jung SR sources currents from the pre-tracking regulator, and the impedance of the LM317/337 is rising above 1kHz and becomes quite high above 100kHz, I guess this will severely limit the performance of the JSR. A regulator like LM317/337 is intended to be used at lower frequencies and at higher frequencies it relies on effective bypassing. So what I am planning to experiement is to add high frequency capacitors in series with damping resistors to the output of the PR, and hopefully this may help reduce some harshness in the treble of my current JSR.
I will report back when I get the results.
Its inherent design, as you have noticed, has certain limitations. But it seems you're squeezing every ounce of performance from it, which is good. Thanks for sharing this with us.
My suggestion would be to implement the excellent v1 because it's the easiest to get right and it has the shunt reg sound. It can be easily setup for your load current by adjusting one resistor. You can always post specific questions in that thread, pointing to a specific schematic.
However, if you'd like a bit of challenge, have a look at the schematic I will post in a few minutes.
My suggestion would be to implement the excellent v1 because it's the easiest to get right and it has the shunt reg sound. It can be easily setup for your load current by adjusting one resistor. You can always post specific questions in that thread, pointing to a specific schematic.
However, if you'd like a bit of challenge, have a look at the schematic I will post in a few minutes.
Ikoflexer,
Thanks. I think I will give it a try. I guess I had better go and read through the entire thread first. Meanwhile, I will focus on implementing the Jung SR, after which I will get my 4-way active OB speakers up and running, then they will provide an excellent base to compare regulators and alike. Will get to it eventually.
Regards,
Bill
I spent the weekend doing what I planned in post #32.
Things were never easy. My PCB made from veroboard was quite compact. Replacing various capacitors and padding them with various resistors took a bit of efforts so some longer wires were used for quicker replacements. But the lengths of wires were critical so I could never get truly reliable results.
Wirewound resistors are useless in this application because of the high inductance. Other low ohm resistors are difficult to find and the only one I could think of is the SMD think film. They are not cheap at above $1 each with minimum order of 10. I had some 2W 1ohm metal film so I just wired them in parallel to get 0.25R, 0.33R and 0.5R.
I can appreciate how much efforts the original authors put in to design and testing of the SR.
In one instance I mistook two 4.7k resistors as 1R resistors so the supereg could not regulate, sending +/-20V to the power pins of my two OPA627, which has an “Absolute Maximum Rating” of 18V! I thought $50 was gone, but in a miracle, the OPA627s survived!
Things were never easy. My PCB made from veroboard was quite compact. Replacing various capacitors and padding them with various resistors took a bit of efforts so some longer wires were used for quicker replacements. But the lengths of wires were critical so I could never get truly reliable results.
Wirewound resistors are useless in this application because of the high inductance. Other low ohm resistors are difficult to find and the only one I could think of is the SMD think film. They are not cheap at above $1 each with minimum order of 10. I had some 2W 1ohm metal film so I just wired them in parallel to get 0.25R, 0.33R and 0.5R.
I can appreciate how much efforts the original authors put in to design and testing of the SR.
In one instance I mistook two 4.7k resistors as 1R resistors so the supereg could not regulate, sending +/-20V to the power pins of my two OPA627, which has an “Absolute Maximum Rating” of 18V! I thought $50 was gone, but in a miracle, the OPA627s survived!
OK, Here are some of the things I did.
Since the SR, from the rail to ground, only has a high ESR capacitor, I did not understand how the opamp PSU could have a low impedance path. So I gave a try on the “pre-bootstrap” version, i.e. to bypass the supply pins with a 22uF/0.1uF.
Some resistance must be added or the SR is guaranteed to oscillate. There are a number of ways to do it. The resistor can be added on the rail side which also form a low pass filter. The original version used a 22R. For my test I used 1R, which should be just sufficient to prevent resonance. Another way to wire it is to place the resistor on the ground side, so the power supply would be tied to the rail. They might have the same effect as “bootstrap” except that the noise would be common mode instead of differential mode. I was not too sure if it could achieve the same result.
Anyway, this gave a very good bypass to the opamp.
The small amount of high frequency harshness, which is the only remaining imperfection of my SR, was reduced.
However, the sound became a bit dull and the sound stage became obviously flatter. I have been playing with opamp bypassing a lot and tried many times but never liked adding any resistance to PSU.
So this was a NO and I reversed the change.
I have tried a few combinations of output capacitors + resistors, but did not get better sound than what it was when I reported it earlier.
So unless I start playing with the tracking pre-regulator, I think it is as good as it is.
Since the SR, from the rail to ground, only has a high ESR capacitor, I did not understand how the opamp PSU could have a low impedance path. So I gave a try on the “pre-bootstrap” version, i.e. to bypass the supply pins with a 22uF/0.1uF.
Some resistance must be added or the SR is guaranteed to oscillate. There are a number of ways to do it. The resistor can be added on the rail side which also form a low pass filter. The original version used a 22R. For my test I used 1R, which should be just sufficient to prevent resonance. Another way to wire it is to place the resistor on the ground side, so the power supply would be tied to the rail. They might have the same effect as “bootstrap” except that the noise would be common mode instead of differential mode. I was not too sure if it could achieve the same result.
Anyway, this gave a very good bypass to the opamp.
The small amount of high frequency harshness, which is the only remaining imperfection of my SR, was reduced.
However, the sound became a bit dull and the sound stage became obviously flatter. I have been playing with opamp bypassing a lot and tried many times but never liked adding any resistance to PSU.
So this was a NO and I reversed the change.
I have tried a few combinations of output capacitors + resistors, but did not get better sound than what it was when I reported it earlier.
So unless I start playing with the tracking pre-regulator, I think it is as good as it is.
OK, I think it is time for me to close this thread.
Why I posted it here?
A) If by any chance the experts came to point out my mistakes it would help me to learn and to get the most out of it.
B) I am a newbie but am happy to share what I found with the newer “newbies”.
Summary:
1) I tried OPA627 in place of AD825 and found the sound to be much more to my liking.
2) The original SR was designed using the OPA797 (bipolar). Low impedance to the inputs was essential for low noise so 500R was chosen to be the input impedance and two 100uF electrolytic capacitors were used (in the signal path). With AD825 or OPA627 (jfet) low impedance to the inputs is no longer necessary for reduced noise, so there is a benefit to scale the resistors up and use film capacitors. Film capacitors are a lot accurate than electrolytic capacitors. They are also substantially better at higher frequencies so filtering can be much improved. Subjectively, this gave a large improvement to the sound.
3) I found removing the capacitor around the zener also removed the colouration from the electrolytic capacitor and I did not hear obvious noise increase.
4) The output capacitor and its ESR, also the output capacitor at the tracking pre-regulator and its ESR, have some impacts on the sonic. This may be load dependant so what suits me may not suit you. I have a 1.1uF MKP in series with 0.5R, wired in parallel to the output capacitor, and I found this helped.
After I implemented my JSR in the way described above, the JSR is a big step forward from my previous very well implemented LM317/337, and sonically it betters a number of other regulators I tried.
Have fun.
Regards,
Bill
Why I posted it here?
A) If by any chance the experts came to point out my mistakes it would help me to learn and to get the most out of it.
B) I am a newbie but am happy to share what I found with the newer “newbies”.
Summary:
1) I tried OPA627 in place of AD825 and found the sound to be much more to my liking.
2) The original SR was designed using the OPA797 (bipolar). Low impedance to the inputs was essential for low noise so 500R was chosen to be the input impedance and two 100uF electrolytic capacitors were used (in the signal path). With AD825 or OPA627 (jfet) low impedance to the inputs is no longer necessary for reduced noise, so there is a benefit to scale the resistors up and use film capacitors. Film capacitors are a lot accurate than electrolytic capacitors. They are also substantially better at higher frequencies so filtering can be much improved. Subjectively, this gave a large improvement to the sound.
3) I found removing the capacitor around the zener also removed the colouration from the electrolytic capacitor and I did not hear obvious noise increase.
4) The output capacitor and its ESR, also the output capacitor at the tracking pre-regulator and its ESR, have some impacts on the sonic. This may be load dependant so what suits me may not suit you. I have a 1.1uF MKP in series with 0.5R, wired in parallel to the output capacitor, and I found this helped.
After I implemented my JSR in the way described above, the JSR is a big step forward from my previous very well implemented LM317/337, and sonically it betters a number of other regulators I tried.
Have fun.
Regards,
Bill
Perander,
It is the OPA627 in the load, not in the SR. +/-20VDC, and the OPA627 survived.
Regards,
Bill
It is the OPA627 in the load, not in the SR. +/-20VDC, and the OPA627 survived.
Regards,
Bill
- Status
- Not open for further replies.