Krell KSA 100mkII Clone

Jump to Latest
#1,165
AVWERK said:
What was the reasoning behind the original's 2sa968, 2sc2238) verses the BOM's MJE15034/35 change?
Click to expand...

Pierre picked them because the Toshibas are obsolete and the Onsemis are easily available. (read his posts somewhere between number 100 and 150)
Anything other than 15003/4s will lower the load on the drivers considerably.
I'd use the 968/2238s, if they are genuine Toshiba, otherwise dump 'm.
Other alternative is the 2SA1930/2SC5171 pair, also Toshiba.
These have a 20 watt power rating, opposed to 25 watts for the original drivers, means they'll certainly need additional heatsinking.
Advantages: 200MHz and a maximum collector current of 2 amps, the collector of the old ones handles 1.5
No problemo to use them in combination with high and constant hFe output devices.
No good for really going down under, Ic is only 100mA/1mS at 100V.
For linear high output current stuff i've yet to see something better than the 968/2238.
The 1930/5171 are working fine in the CKSA50, i sacked the initial 968/2238Bs.
(I went groupy buy with Markus Chocolate in China to buy the Toshiba numbers mundo cheapo)

I can read Mark's posts, probably a hickup from this week's blackout.
 
#1,166
Yep jacco's right - I picked the ONSemi's since they're used for the drivers and bias circuitry anyway, so it's less variation to source. However, use anything else of your fancy. I've looked at other alternatives from ONSemi and nothing's better than 34/35 e.g. BD139/140 and MJE172/182, with lower voltage margin as well. The toshibas are more appealling though.

Jacco, did you use toshibas for both the predrivers AND drivers on your KSA50?

I'd say the best would be to stick with the 968/2238 for the cascode, or 1930/5171, and keep the MJE's for the drivers but the 34/35 instead of 30/31. I think Krell wanted to make sure of reliability by using the 968/2238 only for the cascode and not the drivers too.

Concerning heatsinking, note that only Q9 and Q10 will dissipate enough power to require sinking, Q11 & Q12 are likely to be fine au naturel. However as I've stated before, the idea was to use a single sheet of aluminium for all four transistors.

Jacco, what would your choice be for the LTP? Safe and oldschool with the 2240/970 or something else?
 
#1,167
Pierre,

at first 2SA968/2SC2238 and MJE15032/33, the 2SA1540/2SC3955 and 2SA1930/2SC5171 after i managed to obtain them. Getting the 2SC3955 is easy, the 2SA1540 everything but.

No idea if there is a better BJT choice than 2SA970/2SC2240.
Very linear, low noise, hFe of the BL grade is high, and they're fairly easy to come by.
I don't really care much about fT numbers, linearity is the key factor.
Anything better means going a cascoded dual JFET input, but i rather wait for you to redesign before i start nagging. :clown:
ps: South African engineers in need of a few A968/C2238/A1930/C5171/J109/K389 can have them shipped free of charge, password is a name from an FD book.

(otr: Hitachi 2SB649/2SD669 drivers are by far the worst choice for driver devices of a KSA100MKII, imo, nice choice for the Vas and pre-drivers.)
 
#1,168
Right, more or less my thoughts. So you think the 1930/5171 would be a good bet for both the cascode predriver and drivers? What would Krell's motivation have been for the MJE's for the drivers, perhaps just to play it safe? Given Mark's glowing comments of his prototype it seems unlikely to screw up the sound going with the MJE's, but when there's the possibility for improvement.. Now that ONSemi has severely reduced the number of samples per order the MJE's appeal has started to wane a bit, thank goodness I've ordered lots of them when they were still giving away 25 at a time 😀

I didn't go through all the KSA50 posts once it became as numerous as a rockstar's illigetimate children, but what was the reason why people eventually went for the MPS series for the LTP? The only viable alternative to the 2240/970 is the BC546/556, and even there I'd prefer the 2S ones. Dirt cheap too and readily available even in far-flung worlds.

The dual JFET thought did cross my mind but left it just as soon afterwards due to mentioned availability issues 😉 Does FD en die Gori Lama; FD en die diepsee-duikbol qualify for those lovely 2SJ/2SK ones? :angel: (didn't like the last one too much though)
 
#1,169
I'm surprised Krell used Japanese semiconductors at all in the early 1980s.

The MJ15003/MJ15004 must have been the obvious choice for the output devices at that time, no Euro or Japanese TO3s that could compete and TO3s were the standard.
Apart from logistic issues, buying 100 or a 1000 makes a difference, especially with devices that cost up to 5 times as much.
Krell wanted to design power amplifiers that could handle tough loads, the MJE drivers must have been the rational choice.
Heat dissipation remains an issue for the drivers, even with stable hFe output devices as the modern Japanese, i've reduced the bias through the lower power drivers.

I used the MJEs too, in terms of money for value they were unbeatable in the 25 samples time window.
Same for the MPSAs, Toshiba vendors are not fully out in the open in the US.
Some outside of the US have more knowledge on what is obtainable in America and where than a lot of the US amp DIYs at the forums.


You were not supposed to say it in public, press @ :clown:
 
#1,170
Hi,
I would not recommend the 649/669 for the VAS transistor.

I was intending to use the 1360 in the 50 Klone and once I've checked through the numbers for the higher voltages in the 100Klone I will probably use them here as well.

Jacco,
2SB649/2SD669 drivers are by far the worst choice for driver devices of a KSA100MKII
Click to expand...
why?
What parameters let them down for this duty?

I will never use them into less than 4ohm loads.
 
#1,171
hFe of the Hitachi devices sky dives beyond 100mA.

At a collector current of 500mA theý may be good for just 30.
Reach the limits of current amplification due to dropping hFe values and the sound becomes stressed, is my experience.
Given the choice between the original Toshiba numbers and the 2SB/2SD, i'd prefer the linearity of the 968/2238 any day.
 
#1,172
Hi Jacco,
four output pairs shared between two pair of drivers gives an output current of about 100mA*2 *80=16A. That's maximum power into about 3r and just as effective into 3ohm. That's equivalent to 384Wac into 3r. and the 649/669 is not even round the gain droop corner yet.

Push the driver pair to 150mA and output current is now upto 20A (5Apk per output device) and into 2r that would require the rails to maintain 40Vpk into the load. Well the Krell just cannot go much higher in power output of 400W into 2r. This assuming the output devices maintain hFE of about 67 @ 5Apk

But as I said, if the load is restricted to 4ohm and above and upto 60degree phase angle then the 649/669 is man enough for the job and will not suffer a gain droop problem.
It's just as likely the output devices will suffer gain droop on these lower loads, then it's a toss up to which device is limiting sound quality.

It's worth repeating, for the users who have speakers that present less than 3ohm and severe phase angle, that the very robust 15034 is the way to ensure longevity into low impedance loads.
 
#1,176
I guess there is stilll alot of stuff to fix on the server..... Still no edit buttons... and I make tons of splling mistales...er typin mistales. Oh well. I;m sure the main man will get things hammered(thats missin too!) back in place pretty soon. I guess when the server crashed they lost my e-mail address... no e-mail address and I won't appear.

Mark
 
#1,178
Hi Gabor,

No servo needed. Only chance of needing one is if you remove the two large electrolytics in the feedback. Then you might need a servo. Just built it up the way it is.... you will be very pleased! I am glad that you got your boards before you have to move.

Mark

P.S. Hey guys... I wonder where the WIKI went to? All the links and such are there for assembling this monster. Perhaps Al or Sy can see if it can reeled back in place and I can put the link back in place below my signature......
 
You must log in or register to reply here.
Top Bottom