Krill - The little amp that might...

Status
Not open for further replies.
Jump to Latest
#742
megajocke said:
If this voltage increases the transistors conduct harder, if it decreases they conduct less - stabilizing voltage between their emitters.
Click to expand...

Of course the voltage between the emitters is stabilized! There's a capacitor across the emitters, so they are effectively shorted together at signal frequencies. Voltage from Q8(e) to Q11(e) is constant.
 
#743
megajocke said:
Hi Glen,

Removing the capacitor won't make it any more switching either, look at the base-emitter diodes of the bias transistors which are in series with the diode string. This fixes the voltage between bias transistor emitters.
Click to expand...


Hi.

I agree 100%, but the capacitor is the most easy and obvious thing to point out!

Cheers,
Glen
 
#745
jwb said:


Of course the voltage between the emitters is stabilized! There's a capacitor across the emitters, so they are effectively shorted together at signal frequencies. Voltage from Q8(e) to Q11(e) is constant.
Click to expand...

Did you read the

The base current in the bias pair is equal in both transistors at all times. I is not, however constant. Changes in the emitter current cause changes in the base current. When one transistor conducts less, the other conducts less. If for example Q8 conducts less because the voltage on the emitter of Q7 goes more positive (causing more current to flow in the base of Q13) then Q11 conducts less causing a greater voltage drop from emitter to collector. This makes more current available to the base of Q15 causing greater conduction to maintain an on state.
Click to expand...

in the post before mine?

It makes My Head A Splode every time I read it and I've tried to keep away from this thread for that very reason. 😀 I had to comment it when it came up again though.
 
#747
Look at the base-emitter diodes.

The base and bias string current is set by emitter-to-emitter voltage, not by the collector current and current gain like suggested. That is only true if both transistors have enough Vce to work like current sources. In this circuit they are almost saturated.
 
#748
I think operating in saturation is the entire point.

For some reason, this makes sense to me but not to you. I think it's because I'm thinking of it in terms of the current loop from Q9 to Q12, via Q7/Q10 and Q13+14/Q16+15. Q8 and Q11 have the effect of steering base current into the non-conducting output power transistor, which prevents it from becoming reverse-biased.

If I'm wrong on that point, I'd like to know.
 
#749
By Megajocke -Bias transistor conduction is controlled by their common base current which is determined by the voltage over the series connection of upper bias transistor emitter-base diode, the bias string and lower bias transistor base-emitter diode. If this voltage increases the transistors conduct harder, if it decreases they conduct less - stabilizing voltage between their emitters.
Click to expand...

Well said, I thought about that (leakage currents) between bases. I found that slower , lower Hfe devices actually work
better (several mA on the "idled" side).

Using fast transistors (KSA1381/3503) ,the OPS still switches
unless you raise the CCS's to 10mA +, with the Mje's
7-8mA gives better results. It seems the choice of devices
is critical with this topology.

This are just what I have seen by simulating it (over and over) .
Listening to it..., running off another amp, and a/b' ing it against
that original source.. it has less "listening fatigue".
I built it with mje340/50's and the Ksa's, the mje's sound better
for the diamond/bias/CCS's , but for the darlington's, a
2SC4793/2SA1837 with njl0302/0281 OP's gives ample gain.

It (krill) is a tricky circuit but in trying to understand its operation,
I have learned SO much more ..(pass , tanaka)Thanks... Steve for
a "brainwarping" :headbash: :headbash: design !!

Q8 and Q11 have the effect of steering base current into the non-conducting output power transistor
Click to expand...
that is most evident by looking at the current signal through
the diode string while in simulation..

P.S. Another side effect of all this is my new interest in class A itself !

OS
 
#750
Hey Os,
You're probably the most familiar with this amp (apart from Steve) as you studied/simulated/built it. Do you have the equipment to measure & produce the sort of information being asked for here?

Interested in what you say about the slower transistors - are these better just for the bias transistors or all OPS transistors?

What were your other impressions of the sound?
 
#751
I could be wrong here, but I think the non turn-off property is not coming from the biasing section, it come from how the predriver drive the final transistor. The final transistor's base doesn't have any bleeder resistor to discharge it's charge.

If someone is still simulating Krill ops, could he try simulating with 20khz or 50khz tone, and then try to put 220ohm between base of Q14 to base of Q16 and see if there is something different with and without this 220ohm bleeder resistor? The trapped base charge will evident in higher frequencies like 20khz or 50khz instead of 1khz.
 
#752
By Jkeny,Hey Os,You're probably the most familiar with this amp (apart from Steve) as you studied/simulated/built it. Do you have the equipment to measure & produce the sort of information being asked for here?
Click to expand...

All I have is a PC scope and LT so my opinions are subjective
at most..:bawling:

It was easy to at least compare the OPS to a self or triple
by just prototyping the current amp and using a modular
approach to A/B it.

Subjectively , it SOUNDS like a refined ASKA type bootstrapped
amp with nested feedback,(I have 2 of those).

By luminauw -If someone is still simulating Krill ops, could he try simulating with 20khz or 50khz tone, and then try to put 220ohm between base of Q14 to base of Q16 and see if there is something different with and without this 220ohm bleeder resistor
Click to expand...

I did that on both the LT sim and the real thing. On the sim
with 1k/.1u and my already mentioned OPS this value seemed
to take care of bleeding the charge (and changing the gain
of the darlington). Different devices and
CCS Iq's required different values.

on the REAL OPS I dropped R to 680 because things were too
"bright" at 1K. The 3 factors to "get this thing right" seems to
be, 1. correct Iq of the CCS's 2. Gain of the darlingtons , and
3. Hfe for the biasgen. /diamond (choice of devices).

I wish I had a AP analyzer and/or a 20mhz "real" CRO.. I would of had this " licked" by now.. :headbash:
OS
 
#753
lumanauw said:
I could be wrong here, but I think the non turn-off property is not coming from the biasing section, it come from how the predriver drive the final transistor. The final transistor's base doesn't have any bleeder resistor to discharge it's charge.

If someone is still simulating Krill ops, could he try simulating with 20khz or 50khz tone, and then try to put 220ohm between base of Q14 to base of Q16 and see if there is something different with and without this 220ohm bleeder resistor? The trapped base charge will evident in higher frequencies like 20khz or 50khz instead of 1khz.
Click to expand...


I've actually been simulating it with a 22R resistor in that position.
 
#755
yes , it does.. and the switching performance is also affected.

As I said before, the value of the resistor AND the cap is
dependant on the driver (Cob/Hfe) and OP (Hfe) ,just
the same as when you optimize a double or triple EF OPS.

A near perfect balance has to be reached (CCS Iq/OPS hfe/
diamond Hfe) Steve seems to have added Re's on the diamond
to "trim" at least one of the variables in his latest schema.

OS
 
#756
Steve Dunlap said:
I see we both have handicaps. I am afraid I can't give you an answer that requires no actual thought.
Click to expand...

Sensitive area? I can think of one reason why you would respond in such an personally attacking manner to a request for explanation about a circuit you have gone public with. 😉

If anyone else can help me understand why this bias scheme "ought" to linearise the OS I would appreciate your ideas. Assuming that's what it is meant to be doing, of course. 🙂
 
#757
The biasing of certain components could be better optimised.

The quiescent current for Q13,15 is very low.

About 70 ohm could be added across the b-e junctions of Q14 and X13 output transistors to pull a more optimal idle current of about 10mA through the drivers Q13 and Q15 (I prefer much more). And/ or another resistor between Q13 and Q15 emitters to also bias the drivers up more optimally.
 
#759
Thanks for that link, jkeny. Here's Charles' post for all to see:

Charles Hansen wrote:
Hi Steve,

Thanks for sharing your output stage with us. Here are my thoughts:

a) All-in-all, the circuit is basically an emitter-follower triple. This is a good thing. Most people only use a double, and this loads down the front end far too much, causing lots of distortion.

b) You've come up with a clever way to drive the triple from a single point. This can be very useful in many situations. For example if one wants to use a tube front end, it is much easier if you can drive just a single point. With an emitter follower double, the classic "diamond buffer" allows drive from a single point, but has the problem that it is only a double.

c) The way I see it, the second set of transistors (the ones after the input emitter followers) are very clever level shifters that kill two birds with one stone. You've shifted the level up by one Vbe (needed to offset the two Vbe's going back down in the rest of the output stage) and you've done your thermal compensation at the same time. I've never seen anything quite like it, and it is very original. Congratulations on the good work!

The odd thing is that (unlike just about every other output stage around), there is nothing tying the two driver transistors together. Normally there would be a resistor from emitter-to-emitter (in a good design that keeps the drivers in class A) or resistors that tie to the output stage (in a less-good design that lets the drivers go into class B along with the outputs).

I've listened to differering resistor values at that location, and all I can say is that using no resistor there made for quite a strong (but pleasant!) sonic signature.

Cheers.
Click to expand...

So I can be satisfied if this sums it up. I got the impression from this thread that having floating/saturated transistors in the bias chain had something to do with improving linearity. Charles doesn't mention this. If no one is making this claim then I have no more questions.
 
#760
G.Kleinschmidt said:



That's a pretty average response, more so in light of the rants over the last page or so against those who have dared to question the theory of operation and performance claims you have made for this circuit.

People here how aren’t idiots have explained more that one issue with the claimed “non-switching” nature of this design (such as that erroneous capacitor placed across the driver transistor bases), which you pretty much refuse to acknowledge.

Oh well, each to their own
Click to expand...


"People here how aren’t idiots "

People have speculated more than one issue with the nature of my design. Some have even made unsubstantiated claims. So far, none of you have offered up the kind of proof you demand of me.

Rant or not, it would be far easier for someone to build an amp and test my claims than for me to attempt to build another and test it in a manner you would accept. You have already stated my testing was either done poorly/incorrectly, or I am lying.

It may come as a surprise to you, but the amps I refer to were built for customers that paid for them. I no longer have them here to measure.

Early on in this thread, which you probably haven't read, I offered to supply all parts and use of my shop, at no charge for either, to anyone that wanted to come over and build an amp for testing. I still have the parts to build one.
 
Status
Not open for further replies.
Top Bottom