Krell KSA 100mkII Clone

[jacco vermeulen]

Originally posted by PWatts
one of the most effective solutions would be an integrated low-power switching supply

That's a very innovative idea, quite a novelty on a class A amplifier. Sounds like a good way to solve the bandwidth versus dissipation issue. I've had to raise the bias of the drivers for the KSA50 output stage considerably.

[AndrewT]

Hi,

I've had to raise the bias of the drivers for the KSA50 output stage considerably

why? to solve a problem, to meet a unique need, or something else?

[jacco vermeulen]

See the MJE15034/5 datasheet.
(question : why the 25 Ohm emitter resistor on the KSA50 driver ? Think Stasis amp)

A regular KSA50 with 4 TO3s is likely to have a driver bias of ~70mA.
Doubled that with a 15 Ohm emitter resistor(2 parallel 4.5W Resista 30 Ohm metalfilms)

[Mark A. Gulbrandsen]

What problem did you see or hear without raising the bias. From an audible standpoint you'd have an extremely hard tme convincing me that it needs to be raised...... Aren't you using those Saken outputs on yours? The ones the size of a Volkswagon.....

Mark

[jacco vermeulen]

Yeah, that's the reason why.
On the verge of crossing the class A region the OnSemi driver still has enough bias to do 40MHz.
At high output currents the OnSemi devices are pretty neat.
It just sounds better.(shoot me)

[PWatts]

Okay I made the changes. The driver power can now be jumpered or separately powered by the jumpers that separate the two bottom layer copper pours, although a solid wire will probably be better. Given all the capacitance, a resistor in the region of 10-50ohms can also be experimented with if the powers are to be shared.

Conserning the decoupling arrangement, I gave the drivers one of the big electrolyitcs, an MKP and the axial cap, and added an axial cap and MKP to the rest. That should be enough capacitance and variety to please pretty much anybody.

I also changed the position of ZD1 in order to accommodate 7.5 and 15mm lead spacings for C1, and added/changed a few silkscreen things.

Lastly I fixed a serious mistake - when I changed one of the copper pours I forgot to reconnect Q17's collector

Flodstroem, either I cannot find it or P-CAD doesn't support it, but I cannot do a colour printout. The easiest would be to take the Gerber files and use a CAM viewer such as CAMTastic! or similar, where you can superimpose the layers on top of each other in the colours of your choice. We don't need a lot of people, only one with some time will do as I'm fairly certain that all is correct. Any person who's done PCB layout should have a CAM viewer. I attach the Gerber files and revised BOM.

Concerning the KMA-style regulated power supplies, I took a look at it and made some adaptions and simulations. With 5 resistors, three low-power transistors, two zeners and a number of electrolytics per rail, a very low-ripple and low-noise regulator can be built at very low cost and board space. I would however suggest that the driver and LTP sections both use their own supplies right down to the same rectifier (shared transformer secondaries should be OK, but it depends on how way-out you want to go. Even just using the main supply to subregulate will make a big difference. The mentioned circuit simulated less than 4uV ripple when I fed its incoming DC with 50 and 150Hz 4Vpp AC superimposed on it, with the amp running at full power. This is substantially less than if it's merely fed straight from the main bus with the 39V zener regulators.

We can make these regulator boards an optional part of the group buy depending on interest. It's simple enough to be built on veroboard, but will look so much more professional. Once again I suggest that it first be tested before being mass-produced, but given that it's just an enhancement of the original it shouldn't be a problem.

If I have time I can also try to evaluate the LM5007 500mA switching regulator as a pre-regulator for the whole story; it's a small 8-pin chip and only needs a few small external components. The 500mA capability should be more than enough, but I'll check for beefier ones to accommodate people who'd want to experiment with higher driver bias.

The only part that will take a little more space is the voltage bus for the incoming DC; but even that will not be too large is a smallish bank of 470uF caps as in the original is used. A small 20VA or so transformer and two of these boards per channel will offer signifantly cleaner supplies, at very low cost since all the components can be cheap. The big capacitors offered on the main board means that no big post-filtering caps on the regulator boards are necessary.

[PWatts]

2nd part of the Gerbers, and BOM.

[Zen Mod]

my vote for PS for input and drivers will be shunt reg;
anyway-I'm weird and I always choose-IF ANY- shunt reg.

[PWatts]

No problem with shunt regulators besides that they're very inefficient - OK to use for the LTP but maybe a little impractical for the drivers. But, depending on the interest it could be included.

The LM2951 seems to be a good choice as preregulator, 150kHz switching frequency and 1A current ability. Otherwise a LM317 would also work, but the switching regulator will be much more efficient in dropping the voltage with less dissipation and thus a smaller required transformer. Its 150kHz switching frequency will also lead to a very clean DC output especially after the rigorous subsequent regulation and filtering.

[Mark A. Gulbrandsen]

The fact that Dan The Man prefered the KMA-100 to the KSA-100 tells me that the old power supply is still a good place to start. Sure it's old but lets face it... it does make an improvement that people on this thread have reported hearing and... its VERY INEXPENSIVE to build! I think a seperate power supply thread is a good idea and those knowledgable in power supply design might want to start one. But for now please lets ust keep this thread going forword so we can get the boards made and at least one amp up and running to prove the board.

Mark