It looks like they sold several hundred since I ordered them too... The funny thing is the amount on order... 2400. I think our purchase infulenced that order to some extent. Ya have to realize that they base that on sales over the last quarter... all of a sudden alot of diodes are moving and so on... I bet they sit on most of those diodes for many, many years. How much ya wanna bet the others went to China???
Mark
Mark
Boards for sale?
Hi
I am looking for another set of 100.2. boards. If anybody has any to sell, please contact me.
Mark-
Do you have the direct link Mouser link to the diodes? I need to buy some more.
Clone versus the Krell 50.2 "Upgrade Odessy"
At this stage my stock standard Krell 50.2 is still giving me much more listening pleasure (exceptional soundstage!) compared to my Krell 100.2 which has been fitted with new Sprague PSU caps, new output devices, Black Gate PCB and rewired Van den Hul wires from the PCB to output wires and Van den Hul D102.3 interconnect.
Jozua
Hi
I am looking for another set of 100.2. boards. If anybody has any to sell, please contact me.
Mark-
Do you have the direct link Mouser link to the diodes? I need to buy some more.
Clone versus the Krell 50.2 "Upgrade Odessy"
At this stage my stock standard Krell 50.2 is still giving me much more listening pleasure (exceptional soundstage!) compared to my Krell 100.2 which has been fitted with new Sprague PSU caps, new output devices, Black Gate PCB and rewired Van den Hul wires from the PCB to output wires and Van den Hul D102.3 interconnect.
Jozua
brilliant power delivery
Hi,
I have some numbers to cheer you up.
All done at 1kHz.
Mains 243Vac.
DC blocker on mains.
15ohm Thermistor on mains supply.
Bias set to Vre=24mV (optimum ClassAB).
Maximum outputs:
open circuit 34.0Vac PSU +-52.0Vdc
into 8r0 32.3Vac =130W =+21.15dbV PSU +-50.4Vdc
into 4r0 30.8Vac =237W =+20.74dbV PSU +-49.45Vdc
into 2r0 28.6Vac =409W =+20.10dbV PSU +-47.0Vdc
Mains 241Vac
Bias set to about 2.6A, Psupply =+-50.1Vdc
maximum ouputs
into 8r0 33.1Vac OCv 33.3Vac =137W =+21.37dbV
into 4r0 32.1Vac OCv 32.6Vac =257W =+21.10dbV
into 2r0 30.1Vac OCv 31.1Vac =453W =+20.54dbV
I note that dbV loss into 4r0 & 2r0 when biassed into ClassA is quite a bit less than when biassed to ClassAB. Is this and the lower power in ClassAB due to the effect of different Thermistor temperatures?
I'm particularly impressed with the 0.83dbV loss into 2r0 cf 8r0.
BTW,
the Open Circuit voltages (OCv) allow calculation of 1kHz output impedance. I used a full Pi filter based on Thiele. 8r0+47pF to ground, series combination 1uF//4r0, then 4r0 +47pF to ground feeding the speaker terminals.
I think they show Zout between 50mohm and 66mohm (33.3-33.1/33.1*8=0.048?). Can someone confirm if these are as expected?
Next to check stability. Then listening to the ceramic compensation before I build up the second channel.
Just a quick reminder to save you going back.
transformer 37Vac >900VA
smoothing +-75mF
6pair MJL4281/4302
drivers 2 pair 2sb649/c669
Now, if someone were to layout a 3pair of drivers PCB feeding into a 9pair output stage!!!!
Hi,
I have some numbers to cheer you up.
All done at 1kHz.
Mains 243Vac.
DC blocker on mains.
15ohm Thermistor on mains supply.
Bias set to Vre=24mV (optimum ClassAB).
Maximum outputs:
open circuit 34.0Vac PSU +-52.0Vdc
into 8r0 32.3Vac =130W =+21.15dbV PSU +-50.4Vdc
into 4r0 30.8Vac =237W =+20.74dbV PSU +-49.45Vdc
into 2r0 28.6Vac =409W =+20.10dbV PSU +-47.0Vdc
Mains 241Vac
Bias set to about 2.6A, Psupply =+-50.1Vdc
maximum ouputs
into 8r0 33.1Vac OCv 33.3Vac =137W =+21.37dbV
into 4r0 32.1Vac OCv 32.6Vac =257W =+21.10dbV
into 2r0 30.1Vac OCv 31.1Vac =453W =+20.54dbV
I note that dbV loss into 4r0 & 2r0 when biassed into ClassA is quite a bit less than when biassed to ClassAB. Is this and the lower power in ClassAB due to the effect of different Thermistor temperatures?
I'm particularly impressed with the 0.83dbV loss into 2r0 cf 8r0.
BTW,
the Open Circuit voltages (OCv) allow calculation of 1kHz output impedance. I used a full Pi filter based on Thiele. 8r0+47pF to ground, series combination 1uF//4r0, then 4r0 +47pF to ground feeding the speaker terminals.
I think they show Zout between 50mohm and 66mohm (33.3-33.1/33.1*8=0.048?). Can someone confirm if these are as expected?
Next to check stability. Then listening to the ceramic compensation before I build up the second channel.
Just a quick reminder to save you going back.
transformer 37Vac >900VA
smoothing +-75mF
6pair MJL4281/4302
drivers 2 pair 2sb649/c669
Now, if someone were to layout a 3pair of drivers PCB feeding into a 9pair output stage!!!!
www.stereophile.com/amplificationreviews/774/index6.html , bottom part.
PS: 90V rails for the output section, multiple higher rail voltages for the front end stages !
www.diyaudio.com/forums/showthread.php?s=&threadid=104399
PS: 90V rails for the output section, multiple higher rail voltages for the front end stages !
www.diyaudio.com/forums/showthread.php?s=&threadid=104399
a little OOPs
Hi,
I just noticed that all those power output tests were with the NFB only measuring the driver output. i.e. the main output line is still not connected back to the PCB! But it works.
I tried measuring the frequency response @ 2W into 4r0.
I used the scope to measure sinewave height (DC input setting).
I ramped up the frequency from 1kHz to 100kHz.
The voltage on the scope starts to drop from about 50kHz and is 8% down (5Vpp to 4.6Vpp) at 100kHz. That's just -0.8db @ 100kHz.
Is this method valid?
Still with the missing connection to NFB.
I connected 5n6F//4r0 to output both before and after the Thiele network, no visible response change with 10kHz square wave. Leading edge corners just slightly rounded.
Tried 100nF, again in both locations;
before Thiele, 1.5 cycles of damped ripple (~200kHz) about 0.5Vpk
after Thiele, 1 cycle of damped ripple (~200kHz) about 0.3Vpk
Tried 1uF;
before Thiele 2.5 cycles of damped ripple (~150kHz extending half way along the square top) about 0.7Vpk.
after Thiele 1.5 cycles of damped ripple (~150kHz extending one third way along the square top) about 0.6Vpk.
Do I need to alter compensation?
Let's not consider where just yet, more or less?
Turned the signal input down to zero, Rs~=300r, DMM measured noise 0.1mVac (-110db ref full output) and switched off power. The output ramps up to 5mVdc and settles back down to near zero as the caps discharge.
Which power output should connect back to the PCB?
the output line where it exits from the paralleled devices,
or
the output line just before the terminal mounted Thiele network,
or
the output line just after the terminal mounted Thiele network i.e. on the terminal?
Hi,
I just noticed that all those power output tests were with the NFB only measuring the driver output. i.e. the main output line is still not connected back to the PCB! But it works.
I tried measuring the frequency response @ 2W into 4r0.
I used the scope to measure sinewave height (DC input setting).
I ramped up the frequency from 1kHz to 100kHz.
The voltage on the scope starts to drop from about 50kHz and is 8% down (5Vpp to 4.6Vpp) at 100kHz. That's just -0.8db @ 100kHz.
Is this method valid?
Still with the missing connection to NFB.
I connected 5n6F//4r0 to output both before and after the Thiele network, no visible response change with 10kHz square wave. Leading edge corners just slightly rounded.
Tried 100nF, again in both locations;
before Thiele, 1.5 cycles of damped ripple (~200kHz) about 0.5Vpk
after Thiele, 1 cycle of damped ripple (~200kHz) about 0.3Vpk
Tried 1uF;
before Thiele 2.5 cycles of damped ripple (~150kHz extending half way along the square top) about 0.7Vpk.
after Thiele 1.5 cycles of damped ripple (~150kHz extending one third way along the square top) about 0.6Vpk.
Do I need to alter compensation?
Let's not consider where just yet, more or less?
Turned the signal input down to zero, Rs~=300r, DMM measured noise 0.1mVac (-110db ref full output) and switched off power. The output ramps up to 5mVdc and settles back down to near zero as the caps discharge.
Which power output should connect back to the PCB?
the output line where it exits from the paralleled devices,
or
the output line just before the terminal mounted Thiele network,
or
the output line just after the terminal mounted Thiele network i.e. on the terminal?
No comments!
An update.
Connected the output back to the output terminal on the PCB (now as designed).
tried connecting a variety of output caps//4r and //8r.
all after the Thiele network. slightly damped ripples, much worse in duration than previous testing.
Tried with them connected before the Thiele to output terminal. The ripples are almost undamped.
Tried 100nF from before the Thiele to audio ground while the output still had 8r connected. F5A fuses blew instantly.
Added 22pF to the 47pF compensation, no measurable differences. Disappointment. Didn't try the 100nF output to audio ground test again!
Looking at the 1kHz squarewave there is some peaking (about 10%) at the start of the waveform. At 10kHz this has become a slope to half way along the top. Is this an indication of the need for further compensation SOMEWHERE?
I can just find a 0.9mVpp 5MHz oscillation that submerges into the output noise rythmically over a repeat period of about 5seconds.
Wideband (1Hz to 30MHz) noise is about 1mVpp, with shorted input rather than Rs=300r, the DMM noise reads 0.0mVac cf 0.1mVac from before.
The smoothing cap ripple while quiescent is about 50mVpp @ Ib=1.5A. This rises to about Ib=2A as the amp warms up over about an hour or more.
DC output offset.
Looking at stability of output offset.
Shorted input (Rs=0r0+1k0) cold the offset starts at -3mV and moves towards 0mV+-0.5mV when hot.
Shorted out the NFB caps (DC coupled throughout).
reset the output offset. Now it goes from 17mV at cold to -5mV when warm (Ib=2A). Not a bad range and could easily be corrected with a servo for those that want to go that route.
BUT here is a BIG downside. At switch off the Dcblocking version has a short term pulse to about 120mV ( a second or less) and settles back down to <10mV as all the caps discharge.
The DC connected version sends a 1.5V continuous signal to the output that settles down to a volt or so and then starts rising again to 1.5Vdc at the output.
Since this is with power down, any servo is going to be inactive and may not be able to correct that large an error.
It looks like a fully DC coupled amp requires an output relay preferably with instant off and/or DC detect to remove the risk of excessive DC to the speakers.
I'm thinking about listening to it before altering the compensation.
Any thoughts?
ps,
the Terralec toroid transformer with the DC blocker is completely silent.
An update.
Connected the output back to the output terminal on the PCB (now as designed).
tried connecting a variety of output caps//4r and //8r.
all after the Thiele network. slightly damped ripples, much worse in duration than previous testing.
Tried with them connected before the Thiele to output terminal. The ripples are almost undamped.
Tried 100nF from before the Thiele to audio ground while the output still had 8r connected. F5A fuses blew instantly.
Added 22pF to the 47pF compensation, no measurable differences. Disappointment. Didn't try the 100nF output to audio ground test again!
Looking at the 1kHz squarewave there is some peaking (about 10%) at the start of the waveform. At 10kHz this has become a slope to half way along the top. Is this an indication of the need for further compensation SOMEWHERE?
I can just find a 0.9mVpp 5MHz oscillation that submerges into the output noise rythmically over a repeat period of about 5seconds.
Wideband (1Hz to 30MHz) noise is about 1mVpp, with shorted input rather than Rs=300r, the DMM noise reads 0.0mVac cf 0.1mVac from before.
The smoothing cap ripple while quiescent is about 50mVpp @ Ib=1.5A. This rises to about Ib=2A as the amp warms up over about an hour or more.
DC output offset.
Looking at stability of output offset.
Shorted input (Rs=0r0+1k0) cold the offset starts at -3mV and moves towards 0mV+-0.5mV when hot.
Shorted out the NFB caps (DC coupled throughout).
reset the output offset. Now it goes from 17mV at cold to -5mV when warm (Ib=2A). Not a bad range and could easily be corrected with a servo for those that want to go that route.
BUT here is a BIG downside. At switch off the Dcblocking version has a short term pulse to about 120mV ( a second or less) and settles back down to <10mV as all the caps discharge.
The DC connected version sends a 1.5V continuous signal to the output that settles down to a volt or so and then starts rising again to 1.5Vdc at the output.
Since this is with power down, any servo is going to be inactive and may not be able to correct that large an error.
It looks like a fully DC coupled amp requires an output relay preferably with instant off and/or DC detect to remove the risk of excessive DC to the speakers.
I'm thinking about listening to it before altering the compensation.
Any thoughts?
ps,
the Terralec toroid transformer with the DC blocker is completely silent.
Hi,
added 15pF // R15 (22k) .
Peaking on 1kHz and 10kHz squarewave seems the same.
100nF//8r at speaker terminals gives a low level damped ripple.
680nF//8r at terminals gives high level slightly damped ripple. The worst seen yet.
1uF//8r goes haywire. Ripple almost half height of square wave.
I cannot see that 5MHz oscillation that I had last night.
Do I add R value in series with 15pF or simply remove it?
It would be nice to see a nearly flat topped square wave.
added 15pF // R15 (22k) .
Peaking on 1kHz and 10kHz squarewave seems the same.
100nF//8r at speaker terminals gives a low level damped ripple.
680nF//8r at terminals gives high level slightly damped ripple. The worst seen yet.
1uF//8r goes haywire. Ripple almost half height of square wave.
I cannot see that 5MHz oscillation that I had last night.
Do I add R value in series with 15pF or simply remove it?
It would be nice to see a nearly flat topped square wave.
Heatsinks
Hi
I had some heatsinks that were meant for a 2nd Aleph X project that I have diverted to the Krell clone project. The idea is to operate completely fanless. The size and sheer weight of the metal implies a monoblock layout. It does however look as if I will be able to fit in a 50 watt clone as well.
My only dissapointment is that the metalshop workmanship is terrible. What does one do when people seem keen to assist but have no idea what accuracy means?
Once the box-and-output devices layout is finalised all the metal will annodised to a dark grey.
Jozua
Hi
I had some heatsinks that were meant for a 2nd Aleph X project that I have diverted to the Krell clone project. The idea is to operate completely fanless. The size and sheer weight of the metal implies a monoblock layout. It does however look as if I will be able to fit in a 50 watt clone as well.
My only dissapointment is that the metalshop workmanship is terrible. What does one do when people seem keen to assist but have no idea what accuracy means?
Once the box-and-output devices layout is finalised all the metal will annodised to a dark grey.
Jozua
Attachments
AndrewT said:
Andrew,
1/60s at f2.8 and a dim trace (otherwise it looks burnt out) is ok without a tripod, but usually the graticule is then too dark. If you need to see it, auto setting with a steady hand I find acceptable. All mine are done this way, but a tripod would be better.
Regards,
Brian.