Krell KSA 100mkII Clone

[TomWaits]

Originally posted by Mark A. Gulbrandsen
I took a look at it... Thats defiately NOT the Mk-2 version. You can build that amp on the KSA-50 baords that Pinkmouse did.
Mark

Absolutely, it's gotta be a copy of an earlier KSA-100. Signs of some very interesting activity out there.

Shawn.

[PWatts]

Pity it's a MK1 version.. seems very well designed.

Besides for the ideas already mentioned, here are a few suggestions I thought of:

1) Provision for both TO220, TO92 and the package used for the IRFD MOSFET's.
2) On-board resistors for RC filtering for the input and/or driver stage, since there are on-board supply decoupling capacitors.
3) Capacitors in parallel with the 1N914 diodes as in the KSA50.
4) Legend on the silkscreen indicating which way to turn the bias trimpot for bias increase/decrease (little nice-to-have but can save lots of anguish for those with a heavy screwdriver hand ;-) )
5) Is a separatable driver stage really necessary? I've achieved perfectly stable results without mounting the driver stage on the main heatsink, and it will reduce board size and complexity.
6) Power LED's - besides indicating the obvious, it will also quickly show if the bias is far too high or a heavy current fault or short.
7) 2.4mm board thickness, 70um copper thickness, board colour? (matt yellow/matt green/gloss red/gloss blue/gloss black). I am personally quite fond of the last two..
8) Gold plated over nickel traces (1um)?
9) I suppose I shouldn't even mention the possibility of SMD components.. although a proper COG or NPO ceramic for the 62pF's may work better than silver mica/polystyrene with a much easier layout and shorter signal path. SMD 150R resistors for the LTP pairs will also make them much more compact and efficient, and will make it easier to place the LTP transistors back-to-back for thermal coupling. 1206 form factor is really not that hard to solder. One admitted drawback is that you often end up having the whole board's low-power components in SMD One of the drawbacks if you work with 0603 SMD stuff all the time.
But that's all now unless people believe in the sound of resistor quality in a design with global feedback such as this one..

Keep 'em coming, these are just a few that popped into my head.

Pierre

[Flodstroem]

Pierre,

one question: whats your reason for choosing the 2.4 mm board thickness? isnt the 1.6 mm
an all-rounder (cheaper), especially if choosing 70µm Cu?
Just wandering.

The rest of your suggestions looks fine (but I will use the silvered mica 62pF because I already have them) .

But I was also thinking of this:
My chassis will be 150 mm heigh (internal), Is it possible to have the board not wider than, say 140-148mm (but it could be longer). If not, it will not fit standing in my chassis and that will make a fit-problem for my project.

Regards

[PWatts]

Flodstroem,

The width you mention is unlikely to be exceeded, in fact I think it will be more or less the same size as the PinkMouse KSA50 boards. Especially if the SMD resistors are used, the board can become very compact and even possibly be smaller than the KSA50 ones. Of course it depends on the application, those using fan-blocks as in the original would like a short, yet not so wide limitation

I suggested 2.4mm since in my experience people, especially amateurs, tend to treat their boards rather roughly and a thicker board will be harder to bend or break. The added thickness will not make an electrical difference at the low frequencies used. I don't know about overseas production houses, but the local ones I deal with charge little or no extra for thicker boards.

Don't look for exotic caps for the 62pF, but rather linearity and to a certain degree heat tolerance. Some super-exotic caps may be appealling, but their physical size counts against them. I personally prefer polystrene since they're generally smaller than mica and fits in underneath the TO220 casing instead of next to it. An SMD ceramic will of course be right next to the device leads but I'm sure not many people will use it.

[Mark A. Gulbrandsen]

1) Provision for both TO220, TO92 and the package used for the IRFD MOSFET's.

> TO-220? Thats going to take up some real estate on the board! TO-220 is pretty much an obsolete package in the US. I agree with the other two though


On-board resistors for RC filtering for the input and/or driver stage, since there are on-board supply decoupling capacitors.

>Better yet lets take this one step further and include the extra power supply board to make this a KMA-100 project.


3) Capacitors in parallel with the 1N914 diodes as in the KSA50.

>K

4) Legend on the silkscreen indicating which way to turn the bias trimpot for bias increase/decrease (little nice-to-have but can save lots of anguish for those with a heavy screwdriver hand ;-) )

>K

5) Is a separatable driver stage really necessary? I've achieved perfectly stable results without mounting the driver stage on the main heatsink, and it will reduce board size and complexity.

> Stable results can be had but longer driver life will be had by properly heatsinking them. Even Krell's drivers ran way too hot because of too little sinking! On Pinky's version you didn't have to snap off the driver board if you didn't want to. I say make this part the same and you can leave it on or take it off.

6) Power LED's - besides indicating the obvious, it will also quickly show if the bias is far too high or a heavy current fault or short.

Come on now....
7) 2.4mm board thickness, 70um copper thickness, board colour? (matt yellow/matt green/gloss red/gloss blue/gloss black). I am personally quite fond of the last two..

>If everyone wants to pay for that. While I'm willing to front the $$ and have the boards done again we would have to be in aggreance on the pricing. There is a big price difference. I had the fist batch of Pinky's boards done in 2.4mm and the second batch done thinner with a considerable cost savings. Since I still sit on lots of those boards I vote to keep board cost down. Of course if someone else wants to front the $$ thats fine too.


8) Gold plated over nickel traces (1um)?

>Yes, the cost is minimal.

9) I suppose I shouldn't even mention the possibility of SMD components.. although a proper COG or NPO ceramic for the 62pF's may work better than silver mica/polystyrene with a much easier layout and shorter signal path. SMD 150R resistors for the LTP pairs will also make them much more compact and efficient, and will make it easier to place the LTP transistors back-to-back for thermal coupling. 1206 form factor is really not that hard to solder. One admitted drawback is that you often end up having the whole board's low-power components in SMD One of the drawbacks if you work with 0603 SMD stuff all the time.
But that's all now unless people believe in the sound of resistor quality in a design with global feedback such as this one..

Definatley no SMD. Very few can work with it properly and the parts cost is very high. As youmentioned above the whole amp will run quite warm and reach its own thermal equilebrim. Even the solder for SMDis rediclously expensive. Yes the brand of resistor does have a bearing on sound character even in this amp. I can speak with experience here since I've built 5 KSA-50's three of them utilizing different resistors. Vishays still provide the best character IMHO.

Are we designing a new amp or kloning the KSA-100 MK-2/ KMA-100? If the end result here is not at least close like the KSA-50 was then I'll just go ahead and go back to the Tektronix terminal strip method I was going to do. P to P has its merits too... especially when using silver contact points on ceramic strips!

Mark

[AndrewT]

Hi,

To-220 is pretty much an obsolete package in the US

what is the US using?

[Flodstroem]

What components do you mean will be located at this board area, the Q14-16/R39-41 only, or?

Regarding the schematics:
Could the double pair of MJ15030/31 drive 6 pair of MJ15003/4 or should I implement a third driver pair to drive the extra two MJ15003/4 pair? My reason for that is to rise the SOA and also for to drive even lower impedance speakers. If puting in a third driver pair, could I use the same value (75R0) as the originals or should I change the emitter resistors value (to a higher value ?)

Mark:

> Stable results can be had but longer driver life will be had by properly heatsinking them. Even Krell's drivers ran way too hot because of too little sinking! On Pinky's version you didn't have to snap off the driver board if you didn't want to. I say make this part the same and you can leave it on or take it off.

Is it a drawback or a benefit to locate the driver stage on same heat sink (and board) where the power stage/transistors are mounted (if not taking any practical issue in to account)?

Any wise answer will be appreciated.

Regards

[Stuart Easson]

An observation, both personal and reported elsewhere (Doug Self has good charts in his book)...

For any given level of output current, more output transistors sharing the load puts less of a load on the drivers. Basically the lower current in each transistor puts it at better point on the hfe curve, so the drivers have to supply less current. Thus if you add outputs, you pretty much don't need to add drivers. If you want three or four pairs of 15003/4's per driver pair, just make sure you use the right size emitter resistors and you are good to go...

IMHO more drivers are not necessary and will needlessly complicate the board, plus I hate having areas of boards not populated, makes things look unfinished.

Bearing in mind that 4 x MJ2119x pairs can already do >50A peaks, requiring a lot more output current (say 100A) always means more outputs, and can mean more drivers. If anyone has a special load they are planning to drive to extremes with these amps (or the KSA50) and wants to talk about how to ensure the output and driver stages can handle it they can post here or email me...

Stuart

[Flodstroem]

For any given level of output current, more output transistors sharing the load puts less of a load on the drivers. Basically the lower current in each transistor puts it at better point on the hfe curve, so the drivers have to supply less current. Thus if you add outputs, you pretty much don't need to add drivers. If you want three or four pairs of 15003/4's per driver pair, just make sure you use the right size emitter resistors and you are good to go...

Yes I agree to that.

Do you have any experience to this issue:
Is it a drawback or a benefit to locate the driver stage on same heat sink/power board where the power stage/transistors are mounted (if not taking any practical issue in to account)?

This location of driver stage will solve the issue regarding heat sinking of the driver transistors (no need for a special heat sink on the input/driver board).

Thanks for your advice.

Regards

[Stuart Easson]

I have built the amp both ways, and can't say that either solution struck me as technically better than the other, though less 'flying' leads seems neater to me. Having the driver heatsink away from other components might be good from the perspective of not cooking them...

For lower output levels, ~70W class A, I had the drivers on a different heatsink from the outputs. I arbitrarily decided I wanted them cooler than the relatively hot outputs, and sized the driver heatsink accordingly. For the bigger amp, I mounted the drivers on the same heatsink as the outputs, which had fan cooling. I used point to point wiring for the driver and output emitter resistors...

HTH

Stuart