You might be able to run the symetri VAS from the second stage of
the K-V2. I have not tried it.
All my designs are modular (stage 1 / stage 2 / VAS). I can usually
mix/match/tweak ... and get it all to work.
This is how I have created new unique designs.
Edit - a resistive (and even a capacitive) VAS shunt can change harmonic structure.
"Voice" the VAS .....
OS
Last edited:
well its very nice to see kypton to give better output than the symmetri see the FFT at 0.3V input
An externally hosted image should be here but it was not working when we last tested it.
What's in KT Cordell models that the circuit needs? i.e. what have you added to Cordell models?
OK
found it in post17.
Last edited:
for the circuit can it drive lateral mosfet output stages with high bias like at 25V 1Amp bias per fet?
Run at 1.5A ? Oh ... the "wide" class A region FALLACY !
In my tests the BJT's have the lowest THD @60ma , generally. THD
actually increases >100ma (Bias). Some input stages - OPS combo's (Sankens) ,
test best <50ma - OMG !
FET's , not having GM doubling ... might make for a better class A result.
But , most of these IPS's "suppress" X-over distortion quite well (even
when a little under-biased).
OS
Better stability than a triple EF but this latter also allow low loading of the VAS and hence low distorsion, switching distorsion is not a problem even at very low bias if there s enough NFB, there s no way that 10uV of such distorsion could be more audible than mVs of regular THD.
I validated that the small differences between distortion levels at various even-odd harmonics is due to the transistor models. There are even harmonic profile differences between the KT-Cordell models from the Cordell models. Using Vendor models or LTC_Group models also generates large(to us) harmonic simulation differences. Careful selection of specific transistor models can win either the "big evens, smaller odds" or "big odds, smaller evens" sale.
How many Angels will fit on the head of a pin?
We probably put too much emphasis on tiny differences in a few simulations that will get swamped out in the physical construction. Sears' "Good" "Better" "Best" product grading system is probably a realistic philosophy.... or as AKSA commented these amps are "Good" and some even "Damn Good".
Start with the power supply and output stage to meet your requirements.... work forward.
Noticable differences between vendors and grades:
2SC3503_C
2SC3503_F
MJE340_C
MJE340_G
Qmje340
J2sk389
J2sk389_BL
J2sj109
J2sj109_BL
I cannot discuss this further as I do not play with the models, which are pivotal. However, I build a lot of amps and try hard to correlation the FFT using standard models and what I measure and what I hear. LTSpice is useful, and indicative, but not necessarily very accurate. Gross differences such as the profile of the harmonics do correlate with the listening experience; and I have noticed that at 20KHz fundament the overall subjective assessment of the amp not great, particularly with a push pull, global fb SS amp, where very few sound wonderful at the top end. However, for midrange and bass correlation you need to lower fundamental to 1Khz, 500Hz, and 50Hz.
While I'm not yet into the modelling of modern transistors, I'm interested........
Hugh
While I'm not yet into the modelling of modern transistors, I'm interested........
Hugh
You think so ... that is an amp me and vzaichenko both use , it is similar
to his valve hybrid creation.
(below 1 and 2). How about the best ... at 32db (150w) underbiased
at 45ma per device ?
1 is fully complimentary , but has equal 2/3rd at this high level. It even
beats the Doug self blameless (with TMC) at this level (and handicap).
2 is the "better" symasym , not quite fully symmetrical with a wilson
mirror level shifter creating the "symmetry".
More THD , but it is H2 heavy.
What is real strange ... you can hear this. #1 is absolutely perfect with
brushes,fingersnaps imaged and reproduced like no other design.
#2 has that added warmth with either male /female vocals .
PS - does not quite match the 1'st ones finger-snaps.
You have to do a lot of critical listening to hear this , but it reflects
on both the PC and spice FFT's.
Another strange thing is the plots are the same from 30-70ma bias for the
CFA , and the VFA is optimal from 50-100ma. The two designs interact
with the EF3 differently ... I suppose.
OS
Attachments
I see your point , the KT-cordell BJT models are
-@300beta for the BCxxx.
-250 for the 3503/1381's , I have to use D and E grade in the real world
and offset the beta difference elsewhere in the circuit to balance a symmetrical design.
-@200 for the 992/1845's.
Knowing this , is how I pick the degeneration in the simulation. I sometimes
need to adjust the real world values differently. Real world mouser semi's
are on the "high side" of what the models simulate.
This would throw many of the metrics off in the simulation , especially
measuring ppm distortion products.
OS
I agree 100% as one can definitely hear the difference between two provided you have the right speakers and also the right DAC n Cables.
how about the Kypton version how is that compared to valves?
-RNM
Here the question comes did u hear the brushes finger snaps better with a jfet input based amp or a Bjt? ones? Technically I think Jfet should be able to render it better but let me know your opinions.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.