Another Jung (ALW) Supereg built – It sounds excellent

[HiFiNutNut]

After two years since I was first interested in building a Jung Supereg, I have now finally built one.

Comparing to other regulators I tried, such as the LM317/337, a shunt regulator, a series emitter follower, etc, the Supereg is most obviously in the next class.

I spent quite some time looking for the “best” regulator for my line level XO/EQ (yet to be completed) circuit that uses up to 22 opamps in total for a pair of 4 way active loudspeakers (under the final stage of development). In the past a few months I started realising that the limiting factor of my system was the line level regulator.

Now on getting the satisfactory result from the Supereg, I could not help letting you guys know of this.

If you are looking for a good sounding regulator and don’t know which one to build, build this one – The Jung Supereg. I regret I did not build it earlier and wasted too much time on trying others.

Regards,
Bill

[HiFiNutNut]

I would be happy to purchase the ALW kits but I don’t have Paypal and I am in OZ so I got the schematic from this site: http://www.at-view.co.uk/alwsr.htm and referenced ALW’s PCB design, then used a veroboard to build one. I read the entire 3 part series (only 3 parts are available) articles from Jung’s site. The ALW’s schematic is basically identical to the 2000 version of the Jung’s Supereg, except perhaps only one component difference. So I would still call this the Jung’s Supereg.

I initially built it exactly as specified. I used AD825. I wired up a simple passive XO for the MTM part of my MTMWW-Sub speakers. The frequency response was not ideal but the speakers sounded quite reasonable with those reference grade drivers.

I built a single OPA627 buffer / preamp and used the Jung Supereg to supply the dual OPA627. No local bypass capacitors were used, as they were not necessary with the Jung Supereg. The source was an upgraded Marantz SA11 Super Audio CD player. The power amps were Randy Slone’s Optimos mono-blocks, which are truly remarkable in performance and accuracy.

Since there are electrolytic capacitors used (in the signal path) in the Supereg, some run-in is necessary. However, I believe they usually require a couple of hours to get at least 90% sound out of them.

Initial impression was that I could hear a lot more details comparing to other regs. However, it did not sound 100% right to me in my subjective experience. I can not describe the sound easily so I would give it a miss here.

[HiFiNutNut]

I would not dare to “modify” the Supereg, as it is really a perfected design. But I thought, without altering the design aspect of it, I could “implement” it better and hopefully got the sound that I liked.

First I was trying to see if a different Opamp would make any difference. I did not expect much difference if there were any, but the result was otherwise. I took the AD825 out and inserted the OPA627AP. To my ears the OPA627 sounded a lot better. I did this swap a couple of times and confirmed that I would choose the OPA627. This was confirmed again after other improvements were made, described below.

After listening to it for a few more days I still found something not quite right with the regulator. I started my suspicion on the electrolytic capacitors. From my previous experience of designing and playing with line level circuits and power amps, I found electrolytic capacitors, no matter how good they are, veiled the sound, comparing to good film capacitors.

I studied the schematic carefully. I believed that I could scale R7, C4, R8, R9, C7 so that I could use film capacitors.

In the schematic, R7 is 499R. Obviously low input impedance was considered in the design to minimize noise. Given that if there is no radio frequency interference, I felt comfortable to scale it up because the extremely low bias currents of the OPA627 would permit using larger values with little noise problem. I had some high quality 1.1uF MKP, so C4=C7=1.1uF, replacing the Rubycon ZA 100uF/25V, for +/-15 VDC, I made R7=22k, R8=47.47k and R9=41k.

This made a large difference. Not only details and resolutions were enhanced, the sound started sounding right to me.

[HiFiNutNut]

I gave it a listen for a few more days. I sounded quite good but I thought it was still not 100%. I was then thinking about how to replace the last remaining electrolytic capacitor in the signal path – C2 Rubycon ZA 100uF/25V. I re-read Jung’s article and noticed that this capacitor was a late add-on in the 2000 version and Jung said that it was not really necessary. Above 1kHz the capacitor had a lower impedance than the zener. But I thought without this capacitor the impedance of the Supereg was already very low! I guessed the use of the capacitor there is to reduce the noise of the zener, but I should see if I preferred the lowered noise with the colouration of an electrolytic capacitor or higher noise without the colouration.

So I removed C2. There were no longer any electrolytic capacitors used after the Supereg except the output capacitor. I could not hear more noise. I was very sure that I am not going back with C2.

The regulator now sounds excellent!!!

[HiFiNutNut]

I have also played with different output capacitors, from 10uF to 100uF, from ESR as low as 0.13 to as high as 1R, and sometimes padded with 0.5-1R resistance to increase the "ESR", with different brands - Elna garden variety, Nichicon Gold, Panasonic FC and Rubycon ZL. I could hear the different signatures of the caps, but no so dramatic as the changes described in above. They make only a small difference. I will further play with this when my line level XO/EQ is done.

[HiFiNutNut]

Now the question:

With the removal of C2, how much higher is the impedance of the regulator output?

The reg is driving a dual opamp at the moment so the impedance is slightly less important. When it starts to drive my real circuit with 22 opamps, the low impedance will become very important.

[HiFiNutNut]

I think I will ditch the cramping diodes and I guess it will sound better still! The OPA627 has a differential common mode input of +/-11VDC, same as the original AD797, so I guess the diodes can go.

I guess the diodes must degrade the sound. The leakage current they created is a theorectical 1000 times larger than the input bias current of the opa627!

[ikoflexer]

Congrats! The important thing is that you're happy with the results!

Which shunt reg were you comparing it with, if you don't mind me asking?

[HiFiNutNut]

ikoflexer,

A no-name shunt reg that possibly copied from some older Mark Levinson circuit. I once showed it to J. R who is the agent for Allen Wright's Supereg (shunt) and built hundreds of them. He told me the circuit is simular to his.

I am happy to build another shunt reg. I have been spending my time doing my active 4 way open baffle loudspeakers and have not checked the PSU forum. I have just learnt that you and Salas have a good shunt regulator.

Please point to me which shunt reg I should consider building. It will be used to drive about 22 opamps (opa627 + lm4562) with +/-15VDC rails.

Regards,
Bill

[peranders]

Originally Posted by HiFiNutNut

I think I will ditch the cramping diodes and I guess it will sound better still!

The purpose of those diodes is to protect the opamp in case of short circuit. This is also dependent of opamp also. I omitted also those but in my latest design I have those since I have room for them.