[essoess]

Well, I understand that it's an amplifier, and that it's probably more complex than the AKSA and Lifeforce, and that it seems to sound good. But... what design tweaks were used to make it better than the rest? Cascodes, complementary, bridge tied load/x-amping, complementary feedback pair? Something else? Secrets?

[AKSA]

SOS,

I've been keeping this quiet since making the discovery about 10 months ago because I wanted to thoroughly road test it.

Topology:  CFP front end, with grounded base current drive to the VAS, which uses buffering to deliver very low impedance drive (around 3R) to the output stage.  Lag compensation is used only across the current gain sector of the gain;  OLG voltage gain is not fettered at all.  The output stage is a DEF, with charge suckout and very closely matched output devices.  Two pole phase lead is used.

Because of the grounded base current drive, it's very fast and extremely quiet.  Fb factor is around 90dB up to 20KHz.  CLG is 30dB.  This is the best amp I've done, with extraordinary resolution.  Power supply is interesting;  two trafos and four full wave bridges are used, with Star Earth electrically decoupled from the trafos;  viz, no centre tap.

I hope this partly describes it.  The rest is jealously guarded.  I will be fitting this design to all Soraya retail power amps sold and in fact have rebuilt a number of Soraya Mark Is in the last three months.

I hope to publish a full web description, with cost, very soon.

Thank you for the inquiry,

Cheers,

Hugh

[essoess]

I almost guessed the CFP input stage. I've found it to improve the sound with no compromise, more than any other tweak I've tried (besides selecting "right sounding" transistors). I'm glad someone like you have come to the same conclusion I just wonder why it's not wider spread.

[AKSAphile]

Why did you use four full wave bridges? 


Is the electrical decoupling of the star earth the only reason?

[AKSA]

SOS,

The true advantage of the CFP input stage is the straightening of the transfer function from a skewed S shape to a straight line.  This is possible because the input and feedback devices are now run in constant current, and very nearly constant voltage, meaning that their Vbe never changes, and input impedance is FAR higher since collector is now operating at constant current.  Since the connection between the two devices is at emitter level, NOT at base level, then any alteration in Vbe will manifest as differential error at the output, increasing with greater diff input.  This is very non-linear, and while a high feedback factor can ameliorate this non-linearity, the best option is to reduce it to negligible levels BEFORE applying feedback.

A wonderful benefit of this configuration is also the very low impedance output, as it comes off the slave emitters.  In turn this means variations in the target impedance do not degrade LTP output, unlike the VAS, where output is taken from a very high impedance point - the collector, which is thus highly susceptible to variations in target impedance as the output stage cycles through the waveform, and drives non-linear loads like speakers. 



Hi Andrej,

Yes, each rectifier bank has its own rail, directly connected to the first filter cap.  Then there are small resistor decouplers on each electrode connecting the second cap, one side of which then goes to rail and the other to star earth.  This approach has the additional benefit of coralling most of the charge pulse current into the first cap, whilst leaving the second cap with speaker earth return current duties, minimising intermodulation.

It never made sense to me connecting star earth, a supposedly quiet point, to a thumping great AC point in a trafo.  Better to float it above the fray.....  this power supply is extremely quiet.

Cheers,

Hugh

[essoess]

Thank you for your replies AKSA. It's generous of you to share your findings since I know you make a business out of this. I have yet to try CFP with folded cascode, and that will be my next project with JFET/bipolar input stage. I think it'll be "it".

[AKSA]

Hi SOS,

It's a pleasure.  This forum is largely for my customers, all good people who know that the amps are good, and deliver what I say they deliver.  When I'm approached with courtesy I am most responsive;  but I am often pilloried on DIYaudio for my 'lack of formal engineering training and obvious technical shortfalls' to quote one colourful fellow, and there my responses have been far blacker.  I do not share any of my secrets on that forum because it is used as an R&D workshop by many.

My thoughts are this:  a good SS amplifier can be developed two ways;  one, like the AKSA, accepting that distortion is inevitable, and thus manipulating for best effect, chiefly by playing up H2 and H3 (veritably the AKSA is a 'tube' amp using solid state!);  and the second, by going 110% for a zero distortion paradigm.  I do not believe there is anything in between;  it's all or nothing.  I have found, since the Lifeforce, that once you pass a certain point of resolution, a threshold is crossed, and the 'surgical incisive sonics' of SS suddenly transform into a high detail intimacy with the recording, which is quite extraordinary, and astonishes with the depth of information actually recorded onto commercial CDs/DVDs.  Power supply design is crucial, and usually poorly done.  Taking it even further beyond this point, with something like the Soraya, is a further revelation - I have been amazed at the tiny details I've heard, sharp inward breaths, movement of the tongue on the vocalist's lips, sounds of the key buttons on saxes, quiet thuds as the harp is moved slightly on the concert stage, spaces around instruments, bowing technique on violins, chambering technique with the hands around the harmonica, rimshots on snare drums, decay of the skin on a large timpani, decays of cymbals, all these sounds become so much more detailed if the feedback signal is more accurate and the distortion is reduced below almost immeasurable levels.

Cheers,

Hugh