Krell KSA 100mkII Clone

Most of it's mentioned in the FAQ on the Wiki, it should be more than enough to get you going. The only adaption I would make would be a strong recommendation for 2SC228&2SA968 for the cascaded predrivers.

I haven't installed the boards into the commercial amp yet awaiting feedback from the owner, so the only operational model I know of is Mark's.

The resistor value in the place of the jumper is very flexible. he current drawn through it will be around 40mA so 10R will not even be a 1V drop. One thing to bear in mind though is that the higher the resistor is, the slower the front-end will turn on and thus be more likely to produce turn-on thump. A muting relay will therefore be a good idea. Instead of a timer controlling the relay, Krell used a DC protection circuit since turn-on thump is essentially DC offset misbalance, and thereby got two birds with one stone.

Regarding output transistor choice, the general rule is that if you go for higher-fT devices than the original MJ15003/4, minimize the wire lenghts accordingly. Krell got away with lengths >6" on the rear transistors and asymmetrical wiring schemes, but with faster devices that may be troublesome. Also remember to adapt the drivers' speed accordingly.

The KSA design isn't bothered much by PSU rails within reason. A few Volts here and there will not be a problem so 60V will be fine as long as you stick with 80V caps - most countries' allowed mains tolerance is in the region of 6% over and 10% under rated voltage so 63V will be cutting it close. Anything from 45-65V should work with the same component values.
 
No Brainer, resistors are fine

PWatts said:
Most of it's mentioned in the FAQ on the Wiki, it should be more than enough to get you going.

I think the wiki for the KSA100 could be “cleaned up” and “re-linked”. I think the last revisions to the nice documents you made were in two zip folders. I don’t know if those documents have been tweaked since? I don’t think they are pointed to in the wiki? Also it may be nice for future builders to have a visual of the mysterious resistor positions added in to the BOM pdf files so they can have all the info in one place. I don’t currently have a functioning pdf generator, otherwise I may give it a shot but I’m not even certain the stuff I have is the latest revision etc.

None the less, the "resistor thing” was well discussed at the beginning of the thread and I skipped over it, oops! I will post pictures of my resistors & boards. As it turned out, it was a “no-brainer” and the boards are indeed perfect! Thanks again guys; MarkG, PWatts, Flodstroem and everybody else for making it happen.:drink:

The only adaption I would make would be a strong recommendation for 2SC228&2SA968 for the cascaded predrivers.

I think it would be better for me to stick with slower devices to avoid nasty problems, so I’ll keep the 2SC2240 and 2SA970 and use the traditional MJ15000X type outputs.


I haven't installed the boards into the commercial amp yet awaiting feedback from the owner, so the only operational model I know of is Mark's.

WOW! We are well over due to see some more KSA100 building.

The resistor value in the place of the jumper is very flexible. he current drawn through it will be around 40mA so 10R will not even be a 1V drop. One thing to bear in mind though is that the higher the resistor is, the slower the front-end will turn on and thus be more likely to produce turn-on thump. A muting relay will therefore be a good idea. Instead of a timer controlling the relay, Krell used a DC protection circuit since turn-on thump is essentially DC offset misbalance, and thereby got two birds with one stone.

I’ll try the 10 ohm resistor and when I get the basic bench amp running I’ll work on the delay-on relay thing.

Regarding output transistor choice, the general rule is that if you go for higher-fT devices than the original MJ15003/4, minimize the wire lenghts accordingly. Krell got away with lengths >6" on the rear transistors and asymmetrical wiring schemes, but with faster devices that may be troublesome. Also remember to adapt the drivers' speed accordingly.

This is a big issue for me so I will definitely not use the fast outputs.

The KSA design isn't bothered much by PSU rails within reason. A few Volts here and there will not be a problem so 60V will be fine as long as you stick with 80V caps - most countries' allowed mains tolerance is in the region of 6% over and 10% under rated voltage so 63V will be cutting it close. Anything from 45-65V should work with the same component values.

I’m certain my rails with drop below -+60VDC when the amp is biased so definitely not an issue. Thanks for reinforcing that, though I see I already asked about this a long time ago and was told then it was OK. :ashamed:

I'll post a few more pics and then get back at it. Very exciting to be back to building!

Cheers!

Shawn.:)
 
Re: No Brainer, resistors are fine

PWatts said:
...........The KSA design isn't bothered much by PSU rails within reason. A few Volts here and there will not be a problem so 60V will be fine as long as you stick with 80V caps - most countries' allowed mains tolerance is in the region of 6% over and 10% under rated voltage so 63V will be cutting it close. Anything from 45-65V should work with the same component values.
TomWaits said:
..............I’m certain my rails with drop below -+60VDC when the amp is biased so definitely not an issue..........
Hi,
keep in mind that if the rail fuses blow, the bias drops to just 40 to 50mA and the voltage on the smoothing caps could rise by between 3V and 5V.
If you go for high supply rail voltage then the quiescent power goes up significantly, requiring a much bigger sink. If the ClassA bias is also increased to bring the ClassA 8ohm power up to match, then the quiescent power goes up even more.
+-50V @ 2.5A = 250W (100W of ClassA, 150W ClassAB into 8r)
+-60V @ 2.5A = 300W (100W of ClassA, 210W of ClassAB into 8r)
+-60V @ 3A = 360W (144W of ClassA, 210W of ClassAB into 8r)
 
Re: Re: No Brainer, resistors are fine

AndrewT said:

Hi,
... then the quiescent power goes up even more.
+-50V @ 2.5A = 250W (100W of ClassA, 150W ClassAB into 8r)
+-60V @ 2.5A = 300W (100W of ClassA, 210W of ClassAB into 8r)
+-60V @ 3A = 360W (144W of ClassA, 210W of ClassAB into 8r)

Looks good Mr. T I'll leave the windows open this winter.;)

PWatts said:
Shawn,
Don't confuse the LTP's 2SA970/2SC2240 (Q1-Q4) with 2SA968/2SC2238 - the latter is what Krell had originally used for Q9-Q12 so it'd be best to stick with them. The drivers (Q14-Q17)are still MJE15030/1.

I see...but I'm using the components suggested in your BOM. The 2SA970 and 2SC2240 look a lesser part than the suggested OnSemi devices. Why do you suggest the other devices, hfe or just a closer match to the original build? :confused:

Thanks for the input BTW! Here is how the simple resistor task turned out on my boards:

Cheers,

Shawn.
 

Attachments

  • missing_rs.jpg
    missing_rs.jpg
    95.1 KB · Views: 1,112
The original KSA100 didn't use MJE15030/1 for Q9-Q12; it came along with the KSA50 clone project since they're easier to obtain than the 2Sx series as used originally.

- For Q1-Q4, use 2SC2240 & 2SA970
- For Q9-Q12, use 2SC2238 & 2SA968
- For Q14-Q17, use MJE15030 & MJE15031

If you use these along with MJ15003 & MJ15004 on the outputs you have the exact same amp as the original Krell except for the mosfets (and according to Mark there's no sonic difference between the various types I catered for on the board).

Tip: If you look at C4 you'll notice an extra pad next to it, I put it there so you can fit in a larger single bipolar cap diagonally.
 
Originally posted by PWatts The original KSA100 didn't use MJE15030/1 for Q9-Q12; it came along with the KSA50 clone project since they're easier to obtain than the 2Sx series as used originally.

- For Q1-Q4, use 2SC2240 & 2SA970
- For Q9-Q12, use 2SC2238 & 2SA968
- For Q14-Q17, use MJE15030 & MJE15031

If you use these along with MJ15003 & MJ15004 on the outputs you have the exact same amp as the original Krell except for the mosfets (and according to Mark there's no sonic difference between the various types I catered for on the board).

How about:
- For Q1-Q4, use 2SC2240 & 2SA970
- For Q9-Q12, use MJE15030 & MJE15031
- For Q14-Q17, use MJE15034 & MJE15035

OK...I think this is what I've got to do unless some one sends me a pile of free 2SC2238's & 2SA968's. It would cost a bunch of money to get those...and perhaps fakes. Yes/No? So we're talking sonics and an attempt to obtain the orriginal sound? I'll do some more sniffing around on the net though. Do any of you folks have such Toshiba's to spare or trade? I doubt you'd let them go.:D

Tip: If you look at C4 you'll notice an extra pad next to it, I put it there so you can fit in a larger single bipolar cap diagonally.

Yes. :nod: To replace C2, C3 with a high end bipolar? I have the two 1000uF caps for these positions but am searching for a nice substitute so the spot is still open on the board. Or are you suggesting C4 be replaced with something a little more on the boutique side of things?

Cheers,

Shawn.
 
For Q9-Q12 the MJE's are an inferior, albeit simple, choice. I'm not sure about Canada or the USA but here in 3rd world South Africa genuine 2238&968's are available at less than $0.5 each.

For C2-C4, the best would be a single 470-1000uF (size not critical) Black Gate N/NX. The voltage doesn't need to be 16V, it can be 10V or even 6.3V. If you use the BG's you don't have to use C4. Anything else i.e. normal bipolars or back-to-back polars and you must use C4. Since I don't know of any quality bipolar caps besides the BG N-series, I'd suggest going for 2x Elna Cerafine if you can find them, or at least something with a pronouncable name and known pedigree. For C4, an excellent choice is the Vishay MKP1837 100nF and not even expensive too. Use them wherever you can.
 
I can't even remember what price I thought was a bit high last time around, but I'm willing to buy at $150 each, no problem. Even at $200. A GB would be excellent!

And then I just need to solve the heatsink problem. Get the rest of the components. Etc, etc, etc.. ;)

Flodstroem said:
Hi folks.
The original Krell KSA-100 transformer from Avel was quoted for approx. $295 (each unit if buying more than 5) This meant a total costs of $ 590 which is a lot of money. If we could reach a price level of say, $150 for e single unit including all extra Aux windings then it might be a GB. But I think if much higher a price it would not be a GB, Its my qualified guessing.

Folks, You should be aware of this: if buying an extra Aux-transformer you must calculate an extra cost for that. If looking at transformer supplier here in Norway, the cost for one unit (25-50VA, 2-3 secondary windings) would be as calculated:

transformer: 180NOK + VAT: 45NOK + shipping/handling: approx 65NOK would yield a totals of 290NOK which is approx. $53.20. :xeye:
But this is only applicable for us in Norway, and if you dont live right besides an electronic shop.

I cant see any benefits at all (for me) if going for a smaller (and probably a bad audio) aux transformer. Also, the whole amplifiers mains circuit will be more complicated (H.V. circuits, Fuses etc) and an extra transformer also do need a chassis area to be mounted at. :cannotbe:

I believe, if going for only one transformer unit (1kVA) included all extra aux-windings, such a unit wont be more expensive compared to a simpler mains + extra Aux transformer. :rolleyes: (of course, more expensive for those already sitting on a bunch of small transformers)

Regards :cool:
 
First Green, later production ones with the twisted T logo are black.
The suffix versions have a higher voltage rating.

Toshiba has an arm up KEC in Korea, these devices can also be Korean made with K A968 and K C2238 markings.
Manufactured under Toshiba license, equal devices, i can post a piccy of them too if desired.

(No touchy Chinese versions, make you sick.)
 

Attachments

  • tosh968.jpg
    tosh968.jpg
    59.1 KB · Views: 900
I also thought of that, but there must have been a reason why Krell used the 2238&968 for the predrivers and not the drivers? There isn't a big cost difference so it can't be that, or maybe there was a big difference back in the 1970's and 80's?

The 15030/1's have been used for drivers for most good amps on the market for several years now and they're quite good, but if Krell had used 2238&968 for the pre-drivers and not the driver stage too they must have had a reason.. Jacco?