Well, I was not aware of this design until a week or two ago, for, as soon as I see words "gain clone" anywhere in title, I am skipping that for sure. This time I was wrong.
Interesting modification of very recognizable old Walker's Current Dumping concept. That has inspired me to investigate some further possibilities, but that doesn't necessary mean they are improvement, even thou some of them might turn out like that. Here are a few ideas as a humble contribution to respectable Widowsmaker's efforts. Please note that they are still JUST IDEAS and none of them has been built and tested by me yet! But some will be!
My thoughts were mostly oriented around three issues:
1.Good quality, very linear and low noise small signal BJT-s should perform at least equally well in differential amp stage, and they are more readily available (and affordable) than good quality JFET-s. That could make life easier for most DIY enthusiasts.
2.I am not very sure about sonic qualities of 78L15 in that position, which shouldn't be taken as my disrespect of otherwise quite clever idea to solve elegantly two functions with one component!
3.More chance should be given to the Class A stage to process and provide power for a first few watts, because up to a moment when Class C dumpers "jump into the game", amp is practically working as a pure Class A. Many people are praising importance of those "first watts" hence, it might prove beneficial to let them to be processed by Class A stage.
Interesting modification of very recognizable old Walker's Current Dumping concept. That has inspired me to investigate some further possibilities, but that doesn't necessary mean they are improvement, even thou some of them might turn out like that. Here are a few ideas as a humble contribution to respectable Widowsmaker's efforts. Please note that they are still JUST IDEAS and none of them has been built and tested by me yet! But some will be!
My thoughts were mostly oriented around three issues:
1.Good quality, very linear and low noise small signal BJT-s should perform at least equally well in differential amp stage, and they are more readily available (and affordable) than good quality JFET-s. That could make life easier for most DIY enthusiasts.
2.I am not very sure about sonic qualities of 78L15 in that position, which shouldn't be taken as my disrespect of otherwise quite clever idea to solve elegantly two functions with one component!
3.More chance should be given to the Class A stage to process and provide power for a first few watts, because up to a moment when Class C dumpers "jump into the game", amp is practically working as a pure Class A. Many people are praising importance of those "first watts" hence, it might prove beneficial to let them to be processed by Class A stage.
Attachments
I like the second proposition 
Somewhere in the beginning of the thread I suggested the use of small MOSFETs (like IRFD110, 120, DS107, BS170) instead of the input FET pair.
They may have higher noise (but should still be very small in this application!) but being enhancement devices, it is easily possible to increase their gm over that of the FETs by increasing the tail current of the LTP, subject only to thermal limitations.
This has a twofold advantage:
1) higher gm increases OLG in the first stage (a good discussion why this is interesting can be found in the 'simple killer amp' thread), in fact, it would even be possible to increase it sufficiently to degenerate the MOSFETs some and linearize the LTP.
2) Higher tail current also means higher current available to drive the MOSFET 'VAS' stage. This is quite important as the LTP + current mirror sees the whole Cgs of the MOSFET, PLUS the Cgd of the MOSFET subject to miller effect! Both are quite large even using a IRF9610.
I have used small MOSFETs in amp input stages a few times with great success, I think this could be a very interesting alternative to the FET input pairs.
I also happen to share Bora's misgivings re 78L15, although I have to admit without reserve, the setup is clever! LM317L may be one alternative, but even it is not close to a good discrete CCS.
Finally, having first constructed my own variant of the Walker circuit some 14 year ago, I have to agree with Bora that bringing the output stage closer to class B or even into class AB has benefits. Many who have built QUAD 405 clones know this as one of the proposed mods. The problem with this is, however, implementing it simply but splitting one leg of the bridge (C2, R14 in Bora's second schematic) into two parallel components and putting a Vbe multiplier between them, may be one way to do it. That being said, the amp loses it's appeal of having no adjustments, that way...
Somewhere in the beginning of the thread I suggested the use of small MOSFETs (like IRFD110, 120, DS107, BS170) instead of the input FET pair.
They may have higher noise (but should still be very small in this application!) but being enhancement devices, it is easily possible to increase their gm over that of the FETs by increasing the tail current of the LTP, subject only to thermal limitations.
This has a twofold advantage:
1) higher gm increases OLG in the first stage (a good discussion why this is interesting can be found in the 'simple killer amp' thread), in fact, it would even be possible to increase it sufficiently to degenerate the MOSFETs some and linearize the LTP.
2) Higher tail current also means higher current available to drive the MOSFET 'VAS' stage. This is quite important as the LTP + current mirror sees the whole Cgs of the MOSFET, PLUS the Cgd of the MOSFET subject to miller effect! Both are quite large even using a IRF9610.
I have used small MOSFETs in amp input stages a few times with great success, I think this could be a very interesting alternative to the FET input pairs.
I also happen to share Bora's misgivings re 78L15, although I have to admit without reserve, the setup is clever! LM317L may be one alternative, but even it is not close to a good discrete CCS.
Finally, having first constructed my own variant of the Walker circuit some 14 year ago, I have to agree with Bora that bringing the output stage closer to class B or even into class AB has benefits. Many who have built QUAD 405 clones know this as one of the proposed mods. The problem with this is, however, implementing it simply
but splitting one leg of the bridge (C2, R14 in Bora's second schematic) into two parallel components and putting a Vbe multiplier between them, may be one way to do it. That being said, the amp loses it's appeal of having no adjustments, that way...
ilimzn said:I like the second proposition
Somewhere in the beginning of the thread I suggested the use of small MOSFETs (like IRFD110, 120, DS107, BS170) instead of the input FET pair.
They may have higher noise (but should still be very small in this application!)
.............
1) Higher gm increases OLG in the first stage
(a good discussion why this is interesting can be found in the 'simple killer amp' thread)
.............
I have used small MOSFETs in amp input stages a few times with great success, I think this could be a very interesting alternative to the FET input pairs.
ilimzn said:
Higher gm increases OLG in the first stage
ilimzn is right
... about it is good having high OLG - Open Loop voltage Gain
in first, input stage
I have said at this board before:
In an amp with normal or high amount of global negative feedback
I consider INPUT STAGE being THE HEART of AMPLIFIER.
This means in practice,
I want my input stage as perfect as possible.
While if I change output power transistors, it is not the end of world.
I see Nelson Pass sometimes recommend several different MOSFETs, as working satisfactory when used in output of his constructions.
As what I know, and what I consider, when designing my preamplifiers
where noise REALLY is an issue,
I want to put as much gain in first stage, as my input devices can do with good result.
There is a limit in all transistors, where things gets un-linear and things add to distortion.
There is even a formula, to try to predict the resulting overall noise in an amplifier.
I do not remember exact formula ( help me mikeks )
but I can describe it basically with my own words.
Say you have a normal 3 stages amp.
--------------------------------------
Input differential stage:
Vgain= 20,
Noise= 2 (lownoise transistors)
VAS, Voltage Amplification Stage:
Vgain= 100
Noise= 5 (fast rise transistors)
Output push-pull follower stage:
Vgain= 0.9
Noise= 20 (high power NPN/PNP pair)
--------------------------------------
To find out final total noise of amplifier (remember, as told, Noise is NOT EVERYTHING that counts)
in round figures, we add noise in these 3 stages.
a) 20x2 = 40
b) 100x5 = 500
c) 18
The formula is NOT: a + b + c (=40+500+18) ~558
Because:
First stage will dominate result very much, second stage will add a bit.
Here is where some nice fellow member
can tell you this noise prediction formula.
If you need it.
I make good amplifiers, as I know basically how it works.
I can do without spice and formulas, for sure.
Without a good working heart (center of amplifier)
anyone would soon have near-to-death-experience!
Or is it life/music-difficult-to-enjoy experience .....
tschrama said:
Actually, this is putting it a bit broadly although for the most part accurately - class AB output stages will contribute most distorttion, but it is a rather different kind compared to that of a, say, unbalanced input stage. Also, the OLG magnitude and phase vs frequency is a VERY big player here, especially for HF distortion, and this is mostly hanled in the stages preceeding the output. This particular amp may be somewhat of a special case, as the bridge is designed to take the output stage nonlinearity completely out of the picture (of course, that would be in a perfect world, but it comes close). In this case, distortion happens in the preceeding two stages, LTP and 'VAS'. I am deliberately putting 'VAS' in quotes as MOSFET 'VAS' is actually rather different in operation compared to BJT - a LTP driving a BJT VAS sees very little voltage swing on it's output, and current is 'where it's at'. Driving a MOSFET is a different story, it's voltage in (with current due to Cin of the MOSFET). It should be noted that Ic/Ib is much more consant (even with early effect etc) compared to Id/Ugs, so we are talking rather different 'control laws'. Of course, at the same time this 'VAS' drives the output through the bridge, so it's really a class A amp on it's own - one more reason for the quotes.
This is something off-topic and probably quite unusual, but I believe it should be said about ILIMZN. He posses a remarkable gift of such a vivid introspection of inner "circuit life" but something is even more fascinating... what ever sub-circuit he "dissect" he is never loosing a "big picture" of the circuit as whole! All this might sound insignificant, but it is quite opposite. Not many people have that capability! Most younger members of DIY "brotherhood" should pay attention and read ILIMZN's posts. They can learn a lot!!! There are also many of more "knowledgeable" members so keen to rush into various discussions and "teach others a lesson" that will benefit from reading his posts!
Remarkable... I am glad to know you!
Remarkable... I am glad to know you!
I've been reading the thread with great interest - the thing looks promising.
Because unfortunately I have neither time nor skills to make a PCB: is there any way to get it readily made in one of the versions ? Even better with components (I can solder)?
widowsmaker, I've taken a look on your web site, but didn't find anything to buy. Do you still make kits or have discontinued ?
Thanks for any suggestion
Because unfortunately I have neither time nor skills to make a PCB: is there any way to get it readily made in one of the versions ? Even better with components (I can solder
)? widowsmaker, I've taken a look on your web site, but didn't find anything to buy. Do you still make kits or have discontinued ?
Thanks for any suggestion
No! As I stated before, I haven't tested that yet but now that I developed PCB it will happen in a few weeks. I have no doubt that it will work properly, for, I can see no obvious reason why shouldn't it. Frankly, I believe it should work even slightly better by all means, but, am I too optimistic or not, we will see!?
Knowing that sooner or later some younger, always power hungry enthusiasts, will come to idea to ask for mod's to reach higher power levels, I've done another PCB to accommodate 2 pairs of output BJT's. Layout drawing will come soon... maybe even today!
I hope that widowmaker would have no objections for my intrusion into his design, for there were no other intentions but to make one really good design maybe slightly better if possible!
Knowing that sooner or later some younger, always power hungry enthusiasts, will come to idea to ask for mod's to reach higher power levels, I've done another PCB to accommodate 2 pairs of output BJT's. Layout drawing will come soon... maybe even today!
I hope that widowmaker would have no objections for my intrusion into his design, for there were no other intentions but to make one really good design maybe slightly better if possible!
Thanks ilimzn. I was thinking about that issue, but on the other hand, I presumed that someone serious and determined to make really high quality amp will (to my opinion) at least try to obtain well matched output devices, for there are sources-suppliers this days that will match on demand, components in your order for some reasonable additional charge. Your point is still more than valid for all others attempting to do the amp without matched output devices! They will have to have emitter resistors to avoid disappointments and costs! Therefore, they will either have to do modifications of that PCB by themselves or, if there is any higher demand, I will do that and publish it here, again free for everyone.