Absolutely the BEST of the crop.
Nad-H or kypton ND may be the best IPS's , The slewbaby is even
perfect thermally between the 2 pairs.
Nice toner job !
OS
Only if you see one in a multi device setup that "hogs" Ic . Generally,
ON/Sanken bought in a group will be within .5mv across the emitter
resistors.
.22Re is pretty tolerant of slight mismatches. It is good to check every
Re after heavy use.
OS
Have you compared a IPS using a separately regulated supply? Possibly variable too, so you can adjust/determine what stage clips first, ips or ops.
Most do not like to do this added improvement because of? extra cost/complexity or of little/no benefit?
The IPS's always clip first. Most are cascoded. Thimios/still4given have
clipped them all. VERY tame ! Clip = rails - 5-6V (some soft clip with "rounded
edges"). Multiplier = -2V , Hawksford or Baxandall cascode = another - 3.2V.
Would we rather have the 5ma IPS enter into saturation (and be clamped) ,
or have that happen at the OPS ? We can better control how "dirty" the
clipping is at the VAS , as well.
This planned overload behavior will make an abused slewmaster literally
bulletproof - nothing is even remotely "strained". Most will actually clamp
and draw LESS current at a hard clip !
Edit - V1 and 2 can have a split IPS/OPS supply. Then I would recommend regulating
that supply to match the rails for that lower IPS voltage.
Edit 2 - With a separate IPS supply , even unregulated and fed to the multipliers -
almost dual mono PSRR would be had !! -90db+ is the worst with just one supply.
OS
Last edited:
2sc3503/2sa1381 replacement for Spooky IPS
Did anyone build the Spooky (or other) IPS using MJE340/MJE350 instead of 2sc3503/2sa1381 and compare the difference?
I'm unable to find proper 2sc3503/2sa1381, even Mouser stocks only KSC3503D and KSA1381E, which is not ideal either (I mean using different transistor grade in a pair).
Is the MJE340/MJE350 the only replacement for those?
And how much worse will the amp sound with the MJE pair?
Thanks
Did anyone build the Spooky (or other) IPS using MJE340/MJE350 instead of 2sc3503/2sa1381 and compare the difference?
I'm unable to find proper 2sc3503/2sa1381, even Mouser stocks only KSC3503D and KSA1381E, which is not ideal either (I mean using different transistor grade in a pair).
Is the MJE340/MJE350 the only replacement for those?
And how much worse will the amp sound with the MJE pair?
Thanks
Terry, you mean KSA1381E, right?
Did you do any selecting, matching, measuring to find similar units, or did you just use what arrived from Mouser?
Sorry, yes 1381E. No special matching. I believe it is more important in the CFA. The Spooky ran great right out of the gate. If it is important to you, there is a member selling genuine KSA1381-E KSC3503-E in the swap-meet forum for a very fair price.http://www.diyaudio.com/forums/swap...423y-2sa1209s-2sc2911s-ksa1381e-ksc3503e.html
Last edited:
mismatch madness.
In amps of old 😀 , the mismatch might of been a problem.
Cascoded VAS gain is determined by the to-92 (ksa/c 992/1845 or BC550/560)
active devices. A (slightly) mismatched cascode , no real issue.
With the cascode , even a MJE340 / KSA1381 pair would work !
Also , most IPS have both CCS adjustments and are servo'ed.
Non- symmetrical amps , like the wolverine .... you would get no performance
hit using a MJE(350) for the CCS half of the VAS.
The only amp that might be better with a closer matched pair is the Kypton V
(non-cascoded) .... but even here , the VAS gain is mostly determined by
the preceding stage.
A non-servo'ed classic "leach" amp with just a cap for DC feedback is
a "must match" topology to get acceptable offset and THD.
Slew IPS's are designed to compensate for the shortcomings of 21'st century sourcing.😀
OS
In amps of old 😀 , the mismatch might of been a problem.
Cascoded VAS gain is determined by the to-92 (ksa/c 992/1845 or BC550/560)
active devices. A (slightly) mismatched cascode , no real issue.
With the cascode , even a MJE340 / KSA1381 pair would work !
Also , most IPS have both CCS adjustments and are servo'ed.
Non- symmetrical amps , like the wolverine .... you would get no performance
hit using a MJE(350) for the CCS half of the VAS.
The only amp that might be better with a closer matched pair is the Kypton V
(non-cascoded) .... but even here , the VAS gain is mostly determined by
the preceding stage.
A non-servo'ed classic "leach" amp with just a cap for DC feedback is
a "must match" topology to get acceptable offset and THD.
Slew IPS's are designed to compensate for the shortcomings of 21'st century sourcing.😀
OS
Notice the "D" and "e" gains overlap on the fairchild VAS devices.
I've found many D's over 110 and many E's about 140+ , this would make
for quite an accurate cascoded VAS. Match your to-92's instead.
OS
Thanks Terry,
I sent a PM to him, I hope the shipping cost would be acceptable. If not, I will try MJE340/350.
Stefan
No ... don't. Why not be smart. You have a "weaker" 3503 ,pair it with a
higher beta super pair (bc550/560) or ksc 1845.
The cascoded VAS plus the EF3 eliminate the 20'th century matching BS
-period !!
Even if you did not do this , it would surely beat a high Cob 340/350 setup.
PS - the 340/350 would work , but slew and bandwidth would suffer.
The primary input pair (or pairs) are much more important to at least get
close.
As are the beta's of your groups of P and N output devices , to keep any
one of them from "hogging" current.
Edit - the kypton ND (best of the CFA's) shows 1ppm change using a ksa1381/ toshiba 2sc4793 pair
as VAS cascode ! Thd DOES rise with a mismatched input pair , as predicted.
OS
Last edited:
Well, unfortunately I do not have enough knowledge to alter the design.
I must admit that I do not fully understand your advice.
My situation is that I'm unable to find either of the pair (KSC3503 and KSA1381). I did not find a suitable replacement, so I'm looking for a solution. Either find the exact transistors, or to understand the design better.
So you say that if I had the correct devices but different grades, I can compensate by using higher beta cascade member (let's say if Q13 is "weaker" than Q12, I can use higher beta device for Q14) (Speaking of Spooky)?
It makes sense, but anyway, I do not have either of them so it would be easier to find the correct devices. I did not buy the non/matching pair from Mouser, because the shipping cost is way to high for me.
Thanks for advice anyway, I learned a lot from you guys already!
I must admit that I do not fully understand your advice.
My situation is that I'm unable to find either of the pair (KSC3503 and KSA1381). I did not find a suitable replacement, so I'm looking for a solution. Either find the exact transistors, or to understand the design better.
So you say that if I had the correct devices but different grades, I can compensate by using higher beta cascade member (let's say if Q13 is "weaker" than Q12, I can use higher beta device for Q14) (Speaking of Spooky)?
It makes sense, but anyway, I do not have either of them so it would be easier to find the correct devices. I did not buy the non/matching pair from Mouser, because the shipping cost is way to high for me.
Thanks for advice anyway, I learned a lot from you guys already!
Most of us don't completely understand what OS is saying.😀 You don't need to change any components to compensate for the mismatched gain ranks of the VAS transistors. OS took care of that when he designed it.
No , NO , you DON'T have to alter anything. When you are testing out
transistors .... just do a "controlled mismatch" on the VAS.
The gain of the cascoded VAS-hawksford ... is (active device + cascode) . So ,
offset the weaker 3503 or any other device (you could even use a
1837/4793 toshiba pair - reversed) in the VAS.
Just has to have a higher Ft and lower Cob - wide range will work.
Maybe a pix will show (below). SOOOOO simple a concept.
super-pair cascode is even more sensitive , just 10% more beta on
the weaker 3503 (for example) bc550/560's will counteract any imbalance.
You will end up with a perfect VAS , as good as a curve traced ,"anal"
matching. super -pair might not even matter .... just picoamps of
load on the input pair(s).
Edit - just save a few of your "oddball" higher gain 1845's to offset the 3503.
OS
Attachments
Last edited:
Thanks Rado,
good link, I may use it some time. I already ordered those VAS transistors from a member from India, the funny thing is that shipping from India is 5 USD, while that czech company charges 10 EUR for shipping. But that's common here and we can't do anything with it.
thanks again
