"The Vp distribution of the J113's is between 0.9 and 2.8 volts"
So yield for matching to 0.1V is ~ 5%.
Or you need a matching pool of 20,000 for 1000 matched devices.
That must be real fun. 😀
Patrick
So yield for matching to 0.1V is ~ 5%.
Or you need a matching pool of 20,000 for 1000 matched devices.
That must be real fun. 😀
Patrick
Naaah, the news from Monte Carlo simulations of matching is surprisingly positive.
Resistor matching error starting with six 1% resistors
I encourage you horse around with MC simulations and find out how many matches-within-0.1V you typically get, from a pool size of 4000, 6000, 8000, and 10000 starting devices.
Resistor matching error starting with six 1% resistors
I encourage you horse around with MC simulations and find out how many matches-within-0.1V you typically get, from a pool size of 4000, 6000, 8000, and 10000 starting devices.
You can fiddle the MC result one way or another by assuming a certain distribution.
Buy 10000 and measure. That is reality.
We did that for ~6000x Toshiba K170/J74's.
Cheers,
Patrick
Buy 10000 and measure. That is reality.
We did that for ~6000x Toshiba K170/J74's.
Cheers,
Patrick
Naaah, assume the worst case distribution, the one guaranteed to give the smallest number of matches. MC with that and be pleasantly surprised.
Then buy a smaller pool, save time, spend less money, and get your 1000 matched pairs anyway. If you're unhappy with the number of matched pairs this yielded, you can buy some more units and do some more measurements.
Then buy a smaller pool, save time, spend less money, and get your 1000 matched pairs anyway. If you're unhappy with the number of matched pairs this yielded, you can buy some more units and do some more measurements.
Sorry for the simple question, but are we trying to find, say, 1000 pairs where for each pair, the Vp of the transistors are within 0.1V apart, or are we trying to find 1000 pairs where all the Vps lie within 0.1V?
The reason behind matching in this case is equal current sharing.
So if the bias current differs by 20%, there will be 20% more heat, and hence 20% higher dT.
If the JFET has -ve tempco at bias, it is self regulating, then not critical.
If it has +ve tempco, it tends to run away thernally (gets hotter and hotter).
So you are then trying to get 2000x within 0.1V Vp.
For the circlotron, from purely distortion point of view, you want 1000 pair, each pair well matched (to probably much better than 0.1V).
But you still need to solve the thermal issue somehow.
Patrick
So if the bias current differs by 20%, there will be 20% more heat, and hence 20% higher dT.
If the JFET has -ve tempco at bias, it is self regulating, then not critical.
If it has +ve tempco, it tends to run away thernally (gets hotter and hotter).
So you are then trying to get 2000x within 0.1V Vp.
For the circlotron, from purely distortion point of view, you want 1000 pair, each pair well matched (to probably much better than 0.1V).
But you still need to solve the thermal issue somehow.
Patrick
Just to be clear, this is NOT on our list of projects in any foreseeable future.
My interest stops at Spice simulations.
So this will remain Nelson's exclusive.
We are working on something similar.
More down to earth (single devices), but no less exciting. 😉
Cheers,
Patrick
My interest stops at Spice simulations.
So this will remain Nelson's exclusive.
We are working on something similar.
More down to earth (single devices), but no less exciting. 😉
Cheers,
Patrick
Just to complete the collection, why not just use ONE pair of matched Semisouths ?
Super simple to build, can even do P2P.
And since Semisouth is known to be about zero tempco, you can take full risk and use no source resistor.
Great performance as well. 😉
Patrick
PS Unzip the LIB file and place in same directory as .asc.
Super simple to build, can even do P2P.
And since Semisouth is known to be about zero tempco, you can take full risk and use no source resistor.
Great performance as well. 😉
Patrick
PS Unzip the LIB file and place in same directory as .asc.
Last edited:
But then it would only have 2 Jfets. What kind of beast would that be?
Scary: The Beast With Two Jfets
Scary: The Beast With Two Jfets
The beast is inside the TO247 package when you inspect the bare die with a scanning electron microscope.
🙂
Cheers,
Patrick
🙂
Cheers,
Patrick
Attachments
Last edited:
Well, this is possibly a ”stupid” and often answered noob question.
But: Here it goes: What are the actual or expected SONIC benefits with 1000 jfets? 🙂
But: Here it goes: What are the actual or expected SONIC benefits with 1000 jfets? 🙂
Okey, thats interesting, if what you are telling me is true.
Because low distortion does not seem to be the issue, that makes true musicality, of what i know to this date.
But i am going to build the H2 generator, and the Korg preamp, in the fall. And doing some more cool hifi testing.
But i am truely skeptic. Better then Aleph J. Naaah. Seems extremely unlikely, to be really honest 😇
And if it possibly sounded better: The marigin from Aleph J to Full Nirvana is slim to none. 💕😎
But unbeliveable things sometimes happens.
"Extremely low distortion without negative feedback"
Perhaps you would like to show us some measurements ?
Just a FFT with 1W 8R 1kHz would be lovely.
Patrick
Perhaps you would like to show us some measurements ?
Just a FFT with 1W 8R 1kHz would be lovely.
Patrick
well, we al know who made uberspecial one, The Beast with Thousand Names
all other amps are as poor nameless orphans, comparing to https://www.diyaudio.com/community/...ds-or-not-your-fathers-j2.343073/post-5923308
(though, can't find defining sentence, via phone)
"all other amps are as poor nameless orphans"
Too complicated for stupid me to understand.
Cheers,
Patrick
Too complicated for stupid me to understand.
Cheers,
Patrick
