gerhard,
Here's a nice paper on thermal drift issues and 1/f noise with some measured data. MAT03's at .5 Hz or so corner. This is the first time I saw a theory vs. data on the 1/f^4 spectra due to thermals even though I have observed similar things over the years.
http://arxiv.org/pdf/physics/0503012.pdf
Here's a nice paper on thermal drift issues and 1/f noise with some measured data. MAT03's at .5 Hz or so corner. This is the first time I saw a theory vs. data on the 1/f^4 spectra due to thermals even though I have observed similar things over the years.
http://arxiv.org/pdf/physics/0503012.pdf
Last edited:
Yes, bootstrapped cascode, BF862 and ~PN4391. Mostly eliminates the variable voltage across the BF862, and with it various distortions, while also lowering Vds to about 4V and reducing dissipation and signal-induced dissipation shifts. The current source loading of the source and light loading by the op amp input means 862 Cgs is also rendered of little effect.cascoded source follower?
regards, Gerhard
It eats up common-mode range, but for phono is not a concern.
Borbely has done this sort of cascoding with 2SK170s and 2SK246 parts.
With a 1k loading, how much of what you're seeing is OPS and how much of that is common mode error? It's also one of the lower bandwidth parts on that list (not by such a huge margin as to explain the 20x difference, however).
Edit to add: John Caldwell (BB/TI App engineer) has said that the 1642 moved to a different process to help with common mode issues. Forgot the thread. Not sure if your part is before/after that change (not sure if that'd affect this load condition, however).
(And thanks for that data!)
Last edited:
That's interesting. They didn't look very hard for lower-noise JFETs, and their Wilsonian mirror could be quieter (although its noise is largely nulled). But I like the discussion about the LF noise.
"Common sense is the collection of prejudices acquired by age eighteen." - Albert Einstein
Keep in mind the intrinsic noise of the cartridge and input network. How much below that do you need to be for an uncorrelated noise source?
Yes, I have a similar design. The principle is around since Precision Monolitics times.
< https://picasaweb.google.com/lh/photo/AWKl7tzKFwX5wMVbdLFvrNMTjNZETYmyPJy0liipFm0?feat=directlink >
And somewhat cheaper:
< https://picasaweb.google.com/lh/photo/Bz6NkIibR7_yQyuKdwpPzNMTjNZETYmyPJy0liipFm0?feat=directlink >
Rubiola is a big shot in the time/frequency scene. His website is Enrico Rubiola home page.
Lots of interesting stuff. His book is excellent, also.
regards, Gerhard
ps: does anybody still make these nice plated-through boards with
ground, formerly Vero Powerplane boards?
Those from Roth are not plated-tru and are un-tinned.
Last edited:
Yes the circuits not that special, but I just assumed noise from air currents/thermal fluctuations would be 1/f sort of but 1/f^4 is a big difference.
Twin Industries, tinned, ground plane and plated-through pad-per-hole. 4" x 10" part number is 8100-410. Two more of them will arrive tomorrow, according to UPS
If the JFETs had a lower output conductance it would help, so perhaps that is part of the process change.
PS: Mouser shows the LME49990 as at an imminent end of life. But TI doesn't mention this on the datasheet. Anyone know what is up?
From the data sheet, I suppose the 1/f frequency is 99.5 Hz or so
Not really interesting when you are into 1/f.
JFET manufacturers seem to abhor data sheets. Yes, that's all.
regards, Gerhard
Attachments
I seem to recall that Danyuk measured a rather lower corner for the BF862, but I would have to dig out the article he did to verify this. But you are right, specs on FETs are meager.From the data sheet, I suppose the 1/f frequency is 99.5 Hz or so
Not really interesting when you are into 1/f.
JFET manufacturers seem to abhor data sheets. Yes, that's all.
regards, Gerhard
- Status
- Not open for further replies.