JFET input, MOSFET VAS, LATERAL output = Perfect!!

[MiiB]

I totally agree on this....
This way we can also remove the input capacitor...with the risk to introduce DC offset if the source is poor...

This is also the version i currently have running in my simulator...(sims really good...2.harmonc at -110 dB..and higher order barely existing

[lineup]

MiiB
Nice. You have our circuit in mind.
It is good to have you. I have learnt a few things from you.

Hope to see someone bias 2SK170 from negative supply rail.

In the meanwhile a diagram for mikelm.
It was long ago I posted it for swordfishy
But it is so nice. Isn't it ...............???

mosfet VAS is back

[sonnya]

Originally Posted by AKSA

SWF,

I think the stability is quite good, but, the ongoing, undamped overshoot concerns me.

First up, try about 47pF from drain of the VAS to the jfet source, and if that is not sufficient, add a 10pF silver mica miller cap across gate/drain of the VAS.

I hope this scotches the overshoot... and it should also improve the sound quality further.

Thanks, saw the 3.42V bias voltage; that presumably is around 1.7V Vgs for each output device.

Congratulations to Lineup for a great topology, thanks mate!

Cheers,

Hugh

A second solution to stability issue is a cap from vas drain to output.

with resistive load you will have unity gain. but with a capacitive load it will start working as the gain fall below unity.

try 470pf - 2.2nf.

It works well.

[MiiB]

This is what I'am currently thinking....Higher power version Hawksford cascode..and two sets laterals..LED biased current sources...Simple circuit... hopefully some builders...

Distortion is so low it's hardly worth mentioning at -115 dB 2. order...at+/-25 V..output-swing..

[mikelm]

Originally Posted by lineup

Now, if you get your DC-coupled Fetzilla to work
I will be even happier.

I am looking for the parts now but may take a while, but I will definitely build the DC linked version.

thx for the spice model

mike

[lineup]

I see. Good.
I will wait. No hurry.

[gazzagazza]

Wise words clearly based on a lot of experience.

Originally Posted by AKSA

Lineup,

I appreciate your goodwill, and careful sponsorship of this thread. SWF is a 31 year old jumping out of his skin with different circuits and tweaks every hour of the day, hell, I was like that once... I see this as a learning exercise. His input is pivotal and much welcomed.

You correctly identify my approach, and fearlessly describe your own. Both approaches are valid and can learn from each other.

However, the low THD approach is the conventional engineering focus and has been with us for decades. Sometimes it works, sometimes it doesn't. Often very low THD amps - the Self Blameless for example - don't really appeal in a long listening session for most audiophiles. Tastes vary. The purists, often designers, will say that they want a straight wire with gain, the amp should add nothing at all. But the majority of audiophiles, it seems to me, often with little or no knowledge of the circuitry, don't seem to like the straight wire with gain, many of them drifting across to the highly coloured tube school of audio. We can sweep aside their assessment by saying human perception is unreliable, tastes vary, people wouldn't know a good amp if it bit them on the nose, and so on. But you will find a lot of knowledgeable, thoughtful people saying that you should never buy anything without listening to it carefully, with music you know well. I second that. After all, our food, clothing, artistic and entertainment preferences are formed over a long time, and until we hear/see/smell/touch something a few hundred times, we really don't know the experience well enough to form a preference.

Therefore, we might be measuring the wrong thing, trying to assess how fast the automobile can go by reading off the licence plate....

My thoughts are that the profile of the distortion is important. For example, if the distortion is 0.05%, a highish figure, but 99% of it is H2 and H4, it will probably sound rather better than 0.005% where 80% of it is H5 and H7. Jean Hiraga first noted this in the sixties. Earl Geddes with his Gedlee Index has formalised this observation, and takes weighted, high account of odd order distortion artefacts. Absolutely nothing is new under the sun.

Furthermore, lots of loop gain means that while the distortion artefacts, taken individually, are reduced in amplitude, they increase in number, and trail on until at least the 25th harmonic, particularly beyond a loop gain of about 40dB. Could it be that if we keep loop gain below about 45dB we might have better sound?

Lastly, the back emf and load phase shift of a speaker is a truly malevolent thing, and it seriously disrupts the feedback mechanism, causing correction to be phase shifted radically from the input. This confuses the amp, which cannot distinguish between its distortion/phase shift and that of the speaker to which it is corrected. You could argue this is the idea, but there is no doubt that the reactive speaker load seriously destabilises amplifiers. Again, this problem is ameliorated by reducing loop gain.

Hugh

[mikelm]

mmmm - interesting reading this again in light of swordfishy's latest version because that does have very low distortion indeed, so I expect the sound will be more neutral than warm . . . but the larger VAS mosfet does roll off OLG more than anything I've tested in spice to date - so perhaps it has the best of both approaches.

I never expected to see a feedback amp with a Push Pull o/p that can operate with normal loads without any compensation - truly a step forward.

p.s. does anyone know where I can buy 2sk170's or Lsk170's in europe ? the ones I have will probably only arrive in about six weeks - can't wait that long

[MiiB]

Mike....

the BF862 part in my circuit might be even better....it's easy to get...and has less noise and twice the gain...

[AKSA]

Wow, this has moved along......

Brief comment on the fb resistor: it does not destabilise anything at all. By using the series fb resistor to set up the DC operating conditions of the amp we actually simplify the circuit. In practice, it will have an almost constant DC current through it, fixed by R4, the source/gate resistor of the VAS.

Idss for the 2SK1780 is listed as ranging from 2.6ma to 20mA, in three distinct ranges. I'm hoping that SWF is using the BL series, the middle range, 8-12mA. 2/3 of the Idss is correct, but as pointed out the fact the gate bias is positive indicates either the stage current is too high or the grade of the jfet is inappropriate.

This is a low transconductance device with very low noise, around 22mS. That means Id increases by 2.2mA for each 100mV of gate potential increase. (22mA for each volt). A BF862 will therefore generate another 6dB of loop gain, and compensation might need revisiting. An increase of VAS degeneration could do this, of course.

I don't actually recommend a cap across the fb resistor to stabilise things. I prefer a combination of miller cap on the VAS, and phase lead from VAS drain to fb node. In my experience it's more predictable, more stable. It does work, but to me it's less elegant.

Mike, interesting that the mosfet VAS actually rolls off at HF!! This can only be the capacitance at the gate pulling back slew rate, but it's not what SWF found at the outset. This certainly explains why lag compensation is not required even though we have about 75dB of loop gain.

SWF, the rating of the S170 is 400mW. For reliability I gun for no more than one third of this, 135mW, so you are on the limit, particularly as your climate is subtropical. OTOH, you should try for 2/3 of Idss for best results. Fortunately, you only have low voltage across the device.

Cheers,

Hugh