Krell KSA 100mkII Clone

KSA 100 MK 2

Driver Board
 

Attachments

  • new.jpg
    new.jpg
    76.4 KB · Views: 1,643
Krell KSA 100 Mk2

The German Krell Distribution says:
The Bias across the 1 ohm emitter Dale resistors is 380mv – 430mv. The heat sink temperature is cirka 80°C.
The sound is much besser as with 620 or 690 mv.

The KSA 100 Mk 2 is better with the MJ15024/25 output transistors and has more resolution.
Best regards Alex
 
I agree to Zen Mod comments:

there is no better way than have drivers and bias transistor placed on same heatsink along with outputs

I know that every design and/or designer often must make more than one compromise-but this thing is obviously verrrrrry important in long term,or is it important for designer's peace in mind.........judging by numerous pro amps I have opportunity to see or repair.....

This is also my experience after 25 year in audio service business (repair-man). Best solution would be if the driver-part of board could be cut of (eg. easily "snap-of") for those who wants to do so for to place the drivers on the main power heat sink.. I think Mark mention this in another post or thread.

Regards :cool:
 
Actually the 2SC/2SA devices are a bit better quality I think. I have always used them for the most part for input stages and they are pretty easy to get here in the States.

I can say for sure that mounting the bias device and the drivers on the main sink in the KSA-50 helped its long term stability. I have built it both ways... seperate sink for the three devices, and mounted to main sink. With the latter I have absolutely no dc balance drift or bias drift at all even after months of operation its exactly where I set it. With them mounted thiss way if you set it after everyting thermally stabilizes it plum just stays there. I also have no worries about thermally stressing the drivers over many cold-hot-cold cycles.

Mark
 
Alex, you mention the Mk2 sounded better with MJ15024/25 transistors: was this a DIY upgrade or did the Mk2 come out with them at some stage?

I was actually hoping that we could avoid all the anguish of flying leads by doing the same as Krell did with everything on the same board, and just a larger heatsink as Mark is currently investigating. If I have to make provision for both a snap-off version as well as a bolt-down one, the board will get very cumbersome. BTW, buying on eBay is not as cheap if you have to figure in international shipping fees :dead:

Mark, if you want to mount the drivers on the main sink, what's the purpose of the extrusions? Ot are they for the pre-drivers?

I've already progressed quite a bit further with the design, streamlined a few things, routed all the grounds, fixed a few mistakes etc. As soon as the heatsink is decided on I can finish up. Unless something unforseen comes up I should have it done in a few days.

The cascode pre-drivers do not need a lot of heatsinking. Remember as I mentioned earlier, only two of these four really need sinking, and all four will dissipate about 2W in total. You don't need much to sink that. I'm thinking of a single large sink for the lot, it will be sturdier and look better than two or four separate ones. Isolaton is mandatory, unless you can find those MJF plastic tab version transistors.

I too think that TO220 is too big and overkill for the FET's. TO92 isn't necessarily better though, if you look at the Zetex datasheets you'll see a horribly high Rds of 10 and 20ohms respectively, and fairly high capacitance too. But, they take up space and that's one thing I don't really have to spare. The IRFD's are IMO a much better solution - why the hell do nobody take the time to check them out?!? They're cheap and easily available and a damn nice package. Added bonus is that they can be stacked on top of each other for extra power and lower Rds.

I should perhaps just mention that I'm unlikely to ever hear this clone in action since I have no intention of building one myself - I built two KSA50's for a friend and got intrigued by the design, but that's it. Maybe he'll decide to upgrade to a KSA100 though :D , but for the time being I'm doing this project out of goodwill and boring weekends at home - My day job research is on class-D and one has to remain loyal ;)
 
Well with number 2 you only have .870 inches to work with to mount the device to... not much. With number 3 you have 1.1 inches which is almost 1/4 inch more space. Also with #2 how will you attach it to the board? Number 3 has a mounting foot to tap holes into and screw it to the board.

I agree on one heatsink strap piece like Krell did in the original for the cascodes.

I would prefer to go with the one piece board at this point. It has worked fine for many years in all the KSA-100's out there. I just want to know how you are going to make #2 sturdy so the devices are not supporting the weight of the sink. If there is no adaquate way to mount the sink to the board then a seperate driver board is fine and prefered from a mechanical standpoint. Even Krell went to a seperate driver board in the KSA-80.

I will try both types of devices ultimately and leave the ones in that sound the best... could be the Zetex's....could be the IRF's. Only listening will tell that story. Specs mean little to me... only the final sound characteristic. We already kow the Zetex's work well in a similar situation in the KSA-80..... but I will still try both types before deciding. Also the multiple emitter devices MJ21193/94 are what I settled on for my KSA-50 after trying many types both TO-3 and TO-247.... The Multiple emitter devices sound really really good but the specs are only mediocre!

Mark
 
Official Court Jester
Joined 2003
Paid Member
PWatts said:
.....................
I should perhaps just mention that I'm unlikely to ever hear this clone in action since I have no intention of building one myself - I built two KSA50's for a friend and got intrigued by the design, but that's it. .................


and what is possible intriguing amp for you this days ?

for me -Krell regarding ,maybe just Krell KSA25 .....:devilr: ,just for nostalgia reasons .........
 
Okay, I've been swayed by the lovely Zetex 2110G - the A can be used in the TO92 I provided for, and I'll try to make provision for the G on the bottom side - however there isn't room for lots of vias or copper pour to sink it so don't complain if it overheats!

I would say to go for #3 heatsink - it will work the easiest and resembles the original the closest. Any idea if it will be enough though?

So what is the opinion on the predriver sink? Each man for himself for whatever he can cobble together? In the straight line configuration like I have it now, I think a nice thick piece of 2mm aluminium plate, perhaps bent at the top for a bit more area, should be more than sufficient.

As to which MOSFET sounds best, I won't be surprised if there would be little or no difference. The only thing bugging me about the Zetex ones are the high Rds(on) compared to other FET's. Admittedly it's not as much a loading problem to the preceding stage as with capacitance, but it may cause the FET not to turn on as hard as with others. However, the margins are so large in this application that pretty much anything will work. Listen and decide, there are enough to choose from.
 
Zetex is fine but

The IRFD110/9110 looks to be an good substitute, but I wouldnt be concerned if using Zetex
ZVP RDson= 8 ohm (VP=4 ohm) max
ZVN RDson= 4 ohm (VN= 2.5 ohm) max

If it works, make holes for the IRFD and a pattern on back for the SOT-223.

Another solution would be to place those Zetex on component side and mount(eg. solder) them to a standing position. If this is a simpler or easier solution, make it. Then there is also an option to pre-mount the Zetex to a small Al-strip (epoxy) and then solder all four transistors in a standing position on the board including some support for the Al-strip.

I have done it in SMPS´s so I know it works.

Regards :cool:
 
The straight line thing is ok but why not keep it as the original. Those small fined sinks are cheap even new only a couple of dollars each.

Then there is also an option to pre-mount the Zetex to a small Al-strip (epoxy) and then solder all four transistors in a standing position on the board including some support for the Al-strip.

I did this years ago on a project and it works fine. I would have no problem doing this.

Mark
 
OK here's the progress made so far:

1) Routed ground and power. Important grounds routed separately one their own line back to the star point.
2) Routed predrivers and their resistors and Miller caps
3) Added support for the Zetex SOT-223 MOSFETs at the bottom side (this was damn difficult). The board now supports three different packages for the FET's. Added vias and a large copper pour for proper heatsinking of them. They can be seen as the grey blocks.
4) Extended bypass caps for 5, 7.5 and 10mm pitch.
5) Added large electrolytics at the bottom since there was space anyway.
6) Enhanced silkscreen, added Reference Designators.
7) Fixed a MAJOR routing mistake

I tried going as symmetrical as possible as Al did, but here and there I had to shift a few things for better routing. Besides, rather symmetrical routing than symmetrical component placement.

The white line is where the driver stage can begin in order to allow enough space for a heatsink for the predrivers. The driver board will have the five transistors, the four emitter resistors, one resistor and trimpot for the bias, the 100nF CE bypass cap and the four power supply bypass caps.

One issue may be with the reference designators. I added them last, and it's quite difficult to know what belongs to what. I suggest that I add a second set of them INSIDE the footprint to aid during soldering.

Unfortunately this current design will be very difficult to adapt for a separate driver board since there is no space for the terminals. Those who would like to do so would have to hardwire it. Because the heatsink will be so large, it may also not be feasible to make the boards breakable in order to strengthen them. I guess this particular aspect would need to be discussed first.

The routed part of the board as it is measures 105 x 92mm (4.1" x 3.6")

What still needs to be done is:

1) Further improve silkscreen; make the Reference Designators easier identifiable
2) Route driver stage according to heatsink used
3) Ensure holes are large enough
4) Support for trianglular trimpot pinouts
5) Add mounting holes

Any other suggestions/comments?
 

Attachments

  • ksa100_prelim2.jpg
    ksa100_prelim2.jpg
    89.4 KB · Views: 1,448
PWatts,

The progress is not far from a finished board, and it looks very good so far

Here are some tips due to only practical matters (based on a very low resolution picture of the board and you have my apology if I have missed this):

1) make all donut pads as large as they can be due to nearby routes. This will raise the quality of the board, especially when member with limited knowledge to soldering works are going to solder the components.

2) the pads/footprints for the TO-220 transistors could also larger (rectangular pads ?). Those transistors will produce a lot of heat and this will weaken the solder pads mounting/glue to board when aging.

3) Now, if you route the last part of the board: the driver (and thermal tracking) transistors
be sure to have at least 1-2 mm space to the rest of the board. This is for to make cutting away the driver-stage easy for those who wants to do so. You can add some donut pads for the wire connection on each side of this cutting-line if this is practical. You could even make or print some reference markings on the up or the down side of board for to make cutting easier.

4) Dont know why you mention the reference designator dont you use same as in the original schematics?

Excellent job so far Pierre

Regards :cool:
 
OK, I will post a PDF next time, the 100kB maximum limit makes it difficult to post a proper bitmap. For the time being, here's the last bitmap, the next post will be PDF.

I'll try make the pads as large as possible and nice and wide apart; I keep having to remind myself that I too messed up many expensive boards when I was starting off.

The reference designators are different since this is a different schematic than the KSA50. I used the designators as used in the Mk2 schematic we all have. What I referred to was that with the tightly packed components it's sometimes difficult to see which designator belongs to which footprint. A solution to help the soldering process is to place a second set of the same designators inside the footprint, i.e. they will be covered up when the part is soldered on. I would still advise to check the traces too see what is connected to what to make dead sure before soldering instead of just blindly following designators.

Right, here's the next iteration. The blue border is roughly the size of the original board, and the mounting holes on the corresponding positions. The pads at the sides are for the output, power and transistor base connections to the output stages on the block heatsinks. People will have to decide whether they want it or not, but personally I think it would be nice to have the board resemble the original closely (esp with the driver heatsink) but with all the refinements. It also makes it an attractive drop-in upgrade for current KSA100 owners, especially for Mk1 owners wanting a hassle-free upgrade to Mk2 with the possibility of easily reverting back to the old board.

The board cost will not be much higher, only 36cm2 and there are no extra holes etc. I will still add the screw terminal connectors for those who want to connect to more other output stages.

I made a few routing changes, refined the silkscreen, and placed two more resistors on the bottom side. This makes heatsink placement easier and allows a more compact layout. Although a single heatsink for the predrivers is possible, a smaller one for each pair may work easier. "foldback" types are usable since I placed the resistor between the two sets on the bottom side.

The driver stage needs to go as low down as possible to have as much are above available for the driver heatsink. How low it goes down ultimately depends on how much space would be needed for the predriver heatsink. 5mm? 10mm? Unfortunately the nature of this layout caused the driver stage to be higher up than on the original board, so we need as little clearance as possible between it and the predrivers to limit board space and increase heatsink area.

I regret having to say this, but those who want a removable driver stage will have to do so at own risk. I'll try making the amputation process as easy as possible, but we're already pushed for space so I cannot do a proper distanced layout just for the sake of it. I will add some proper sized pads, but a connector is out of the question - that all is of course unless people want to abandon the whole on-board sink idea and go for the KSA50 clone's approach, but I like the old-school idea.

I've checked, it will be possible to add the fourth hole for the LTP's, but it will not be in a straight row. The third pin would therefore need to be bent a bit, but that's all that won't disturb the whole layout.

Still needs to be done:

1) Get more exact measurements of the original board diameters and mounting hole placements (I just extrapolated from pictures)
2) Add the LTP holes and triangle trimpot pad
3) thicken the power ground traces (no use having thick power traces and the return path is thin!)
4) Mounting holes for driver heatsink, new component footprint for the horizontal TO220 drivers

Somebody would also need to double-check the board for faults. Preferably someone who has P-CAD 2004, but with concentration a graphic approach would also work. Even though I very rarely make a routing mistake, accidents do happen and I don't want a lot of angry board buyers if something goes up in smoke. I also routed directly instead of first drawing up the schematic and importing the netlist in order to save time. The issue with the Delta Audio trimpots are a good example of what I'd like to avoid - and that was an easy mistake to fix fortunately. I have to say I'm quite chuffed that at my progress tempo, I placed the first component 30 hours ago and it was by no means the only thing I did this weekend.

Mark: take a look at the board and see if you can work with the space for the heatsink and whether I should lengthen the board more. I'll then add the heatsink and mounting holes, and give you the exact locations of hole positions and width needed.

Zen Mod: I forgot to mention earlier, I got intrigued by the design not because it's advanced or novel or anything, but in fact because of the very reason that it's so old and unrefined, yet it remains one of (sonically) best amps to this day.
 

Attachments

  • ksa100_prelim3.jpg
    ksa100_prelim3.jpg
    92.3 KB · Views: 1,260
Consider designing a single PCB layout optimized for a narrow monoblock using On-Semi output transistors soldered directly to this PCB.

The KSA 100 weighs over 100 pounds. I purchased the KMA 100 monoblocks both for ease of moving and the better electrical isolation.

I usually build a single PCB monoblock similar to the attached photo, but with the transformer mounted either vertically to the front plate for cooling and more efficient packing, or on the bottom of the chassis to raise the heatsink above the floor for better thermal air pumping. A single PCB allows the use of high Ft output transistors with less chance of oscillation than a separate driver PCB spaghetti wiring. It is much easier to find two modest VA transformers than one monster VA transformer.

There are large heatsinks like the KL271 profile with 0.3 C/watt thermal resistance that I've been using in my designs.
 

Attachments

  • sm70.jpg
    sm70.jpg
    46.8 KB · Views: 1,481