Earlier in this thread I gave my impression of the v1.5 comparing to other regulators I tried.
The very well implemented Lm317/337 sounded overall OK but was in lack of resolution and air. It was no HIGH-END. I added a JLH C-multiplier / Ripple eater after it. There were some improvements to the sound, but only marginally better. Next I tried a series regulator with emitter follower. It was no better than a very well implemented LM317/337.
Next I tried the ALW Jung Supereg. You can searched for my recent thread dedicated to it. I did a number of "enhancements" to it. It was definitely the next level up. Most impressive was the very fine level of details, resolution and air it delivered, which rivaled all the best systems I auditioned. However, there was a level of high frequency hash I could not get rid of no matter what I tried. I suspected that it might mildly oscilate.
I also had a css-shunt regulator kit I tried a few years ago which I did not like the sound. This one is simular to the Borley's shunt reg but used a LM431 Vref and bipolar transistors. Implementation played a very important role here. After I started Salas v1.5, with the newly acquired knowledge and experience, I revisited the kit. I replaced the rectifier / RAW DC stage with the one I used for the Jung Supereg, and replaced the wrong and low quality output capacitor with an appropriate one. Now it sounds excellent (this kit is now for sale at cost price. If you are interested, contact me).
Then I tried the Salas v1.5. It is marginally better than the other "renovated" shunt kit, possibly due to lower noise. It is more simular than different.
Comparing the Jung Supereg to the shunt regs, I still found the Jung Supereg provides more details, resolution, air and impacts. However, there is a certain level of uneasiness when listening to the Jung Supereg. When coming to a very designed and implemented css-shunt, the shunt reg sounds more "RIGHT". This is a totally subjective description. I guess measurements should reveal what is going on. I also understand that some odd order harmoniecs may sharpen the sound. I guess that it is very difficult to get a regulator with an opamp driver to work without some level of ringing or bad transient response at higher frequencies.
After playing for a couple of weeks, I found the v1.5 delivers the music with relative ease. Overall it still compares well to all other regulators I tried. There is a tiny bit of harshness in the higher frequencies, but this may well due to the other parts in my audio chain, or perhaps the new buffer stage oscillate mildly?
But we need to understand that implementation plays a very important role here. The wiring example I gave in above says all. Even the best designed regulator can sound bad if not implemented correctly. So comparing regulators without noting the implementation is completely pointless.
So the Jung Supereg did not sound optimal to me may very well due to my poor implementation. It may sound superb if implemented properly by some one else.
In the same way, for the Salas v1.5 I built and liked, if you build it in your own way, it may sound excellent or bad depending on how you build it.
The very well implemented Lm317/337 sounded overall OK but was in lack of resolution and air. It was no HIGH-END. I added a JLH C-multiplier / Ripple eater after it. There were some improvements to the sound, but only marginally better. Next I tried a series regulator with emitter follower. It was no better than a very well implemented LM317/337.
Next I tried the ALW Jung Supereg. You can searched for my recent thread dedicated to it. I did a number of "enhancements" to it. It was definitely the next level up. Most impressive was the very fine level of details, resolution and air it delivered, which rivaled all the best systems I auditioned. However, there was a level of high frequency hash I could not get rid of no matter what I tried. I suspected that it might mildly oscilate.
I also had a css-shunt regulator kit I tried a few years ago which I did not like the sound. This one is simular to the Borley's shunt reg but used a LM431 Vref and bipolar transistors. Implementation played a very important role here. After I started Salas v1.5, with the newly acquired knowledge and experience, I revisited the kit. I replaced the rectifier / RAW DC stage with the one I used for the Jung Supereg, and replaced the wrong and low quality output capacitor with an appropriate one. Now it sounds excellent (this kit is now for sale at cost price. If you are interested, contact me).
Then I tried the Salas v1.5. It is marginally better than the other "renovated" shunt kit, possibly due to lower noise. It is more simular than different.
Comparing the Jung Supereg to the shunt regs, I still found the Jung Supereg provides more details, resolution, air and impacts. However, there is a certain level of uneasiness when listening to the Jung Supereg. When coming to a very designed and implemented css-shunt, the shunt reg sounds more "RIGHT". This is a totally subjective description. I guess measurements should reveal what is going on. I also understand that some odd order harmoniecs may sharpen the sound. I guess that it is very difficult to get a regulator with an opamp driver to work without some level of ringing or bad transient response at higher frequencies.
After playing for a couple of weeks, I found the v1.5 delivers the music with relative ease. Overall it still compares well to all other regulators I tried. There is a tiny bit of harshness in the higher frequencies, but this may well due to the other parts in my audio chain, or perhaps the new buffer stage oscillate mildly?
But we need to understand that implementation plays a very important role here. The wiring example I gave in above says all. Even the best designed regulator can sound bad if not implemented correctly. So comparing regulators without noting the implementation is completely pointless.
So the Jung Supereg did not sound optimal to me may very well due to my poor implementation. It may sound superb if implemented properly by some one else.
In the same way, for the Salas v1.5 I built and liked, if you build it in your own way, it may sound excellent or bad depending on how you build it.
After playing with v1.5 for a couple of weeks I now believe I can improve on my original PCB layout.
In series regulators, we only need to worry about the AC loop. With a shunt regulator, The DC loop may run at much higher current than the AC loop. I now even think that I may need to adjust for the exact CCS currents and regulated voltages on both the positive and negative rails so that I can eliminate any DC offset on the ground line. I am not sure if this is a worry or not but I guess I will do it this way, anyway.
The large shunt current would create some voltage difference on the ground. So the AC sensing point should be moved away from the shunt ground point. Previously, I used a star ground.
I don't want any newbies to copy my previous design as it is. For this reason, I have come up with a PCB layout to illustrate.
The previous version allows using large heatsinks or aluminium flats/angles.
The new one assumes that the each MOSFET dissipate less than a few watts and uses standard, widely available 11 degree / C TO-3 type small heatsinks. The PCB is therefore made smaller.
Best regards,
Bill
In series regulators, we only need to worry about the AC loop. With a shunt regulator, The DC loop may run at much higher current than the AC loop. I now even think that I may need to adjust for the exact CCS currents and regulated voltages on both the positive and negative rails so that I can eliminate any DC offset on the ground line. I am not sure if this is a worry or not but I guess I will do it this way, anyway.
The large shunt current would create some voltage difference on the ground. So the AC sensing point should be moved away from the shunt ground point. Previously, I used a star ground.
I don't want any newbies to copy my previous design as it is. For this reason, I have come up with a PCB layout to illustrate.
The previous version allows using large heatsinks or aluminium flats/angles.
The new one assumes that the each MOSFET dissipate less than a few watts and uses standard, widely available 11 degree / C TO-3 type small heatsinks. The PCB is therefore made smaller.
An externally hosted image should be here but it was not working when we last tested it.
Best regards,
Bill
Bill, thank you for such a detailed report! Very interesting.
You've been referring to v1.5 of salas. Perhaps you meant v1.1.
Your comments about the Jung regulator somehow echo my impression. Indeed, perhaps we both implemented it wrong. I too noticed a certain edge in the sound, and it was subtle.
Something is not clear to me. You say you've replace some wires with thicker 15A rated cable. Were those between the regulator output and the powered circuit? If so, we've talked about this at some length in the thread. Long thread, so it's certainly understandable you might have missed it. To achieve a low output impedance at higher frequencies, both the regulator layout AND the connection to the next stage are very important. More than two, three centimeters of wire is too long. We've been talking about remote sensing, and came up with plans, but I don't know if anybody implemented it that way so far. Safest bet: place the regulator right next to the next stage and couple them with solid thick wire.
Other people who implemented the shunt regulator might have something to say about the harshness in the high frequency. This is a bit surprising. Do you remember if there was similar harshness when you tried v1?
You've been referring to v1.5 of salas. Perhaps you meant v1.1.
Your comments about the Jung regulator somehow echo my impression. Indeed, perhaps we both implemented it wrong. I too noticed a certain edge in the sound, and it was subtle.
Something is not clear to me. You say you've replace some wires with thicker 15A rated cable. Were those between the regulator output and the powered circuit? If so, we've talked about this at some length in the thread. Long thread, so it's certainly understandable you might have missed it. To achieve a low output impedance at higher frequencies, both the regulator layout AND the connection to the next stage are very important. More than two, three centimeters of wire is too long. We've been talking about remote sensing, and came up with plans, but I don't know if anybody implemented it that way so far. Safest bet: place the regulator right next to the next stage and couple them with solid thick wire.
Other people who implemented the shunt regulator might have something to say about the harshness in the high frequency. This is a bit surprising. Do you remember if there was similar harshness when you tried v1?
1.1 is even smoother in the highs than 1 in my experience. But I use the 550buf directly P2P dead bug. What about a lower hfe npn/pnp try Bill? BC184/BC214 for bufs. That would ease it. Maybe you got some tiny riding oscillation.
I did it deliberately.
How can the regulator be used for, say, Linkwitz Orion or John K's NaO? It is not possible to have <3cm wires for every opamp among up to 2 dozens.
What I am planning to do in my final XO/EQ circuit, is to have a local Star Ground. I will use 10-15A wire soldered at the opamp power supply pins run directly straight a few mm above the PCB to the Start Ground. I will then connect the PSU output to the Star Ground with high power high current really thick wire.
Any better idea than the above? I would love to hear.
Your figure is spot on.
The 27k||1u1 is approximate. I guess I have probably used 20k. The current was 0.61mA/0.63mA. So for 27k 0.48 would be very close.
If using electrolytic capacitor, I would definitely have the -3dB point less than 1Hz. For film capacitor, I would be happy to go a bit higher. Bear in mind that the reg drives opamps, not phono.
2u2 or 3u3 MKP may be bulky. I would say 1uF with a 47k resistor may fit nicely on a PCB.
Sounds a great idea. I have been thinking about it, just don't know exactly which complementary transistors suit the best.
However, I have just got a bag of 2sk170 from Spencer so I am ready to build v2. I may come back to it later.
184/214 bufs will keep same currents and noise, only GBW product will go less than half. Interesting experiment at a point.
44 wires? I'm not familiar with the setup you need for the Orion, sorry.
Remote sensing just adds two wires to each regulator. Of course, it's just an idea. In theory it's how it should be implemented when short proximity is a problem
It is only a tiny dose of harshness and on this aspect it still betters others. I am not even sure if it is due to other components in the audio chain. I have an extensive range of CD/Super Audio CD sources and used the extreme samples for testing.
The Orion uses twenty something opamps. Each opamp has two PSU pins. If each opamp has its own PSU and GND lines then you need sixty something wires, which is impossible to implement.
If using only two wires, each opamp would need to share some common PSU and GND wires, which will completely defect the purpose of remote sensing.
I have not built the Orion, which many people raved to be one of the most outstanding loudspeakers ever designed and produced.
I did build John K's NaO, and in my opinion, at least equally compentent.
I studied and learnt from the above two speakers, and did my own. In my case, I have only 3 single opamps for the tweeter, the lower number should provide better sonic. Mine is a 4 way active including subs so it does use quite a few opamps. Further more, I persue very flat frequency response and will have response measured at listener's position from 200Hz to 20kHz within +/-1.5dB, apart from a narrow band BBC dip at 2.7kHz.
I did build John K's NaO, and in my opinion, at least equally compentent.
I studied and learnt from the above two speakers, and did my own. In my case, I have only 3 single opamps for the tweeter, the lower number should provide better sonic. Mine is a 4 way active including subs so it does use quite a few opamps. Further more, I persue very flat frequency response and will have response measured at listener's position from 200Hz to 20kHz within +/-1.5dB, apart from a narrow band BBC dip at 2.7kHz.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.