I want a measurement amplifier where I can say "forget the preamp" when it
comes to voltage noise. Well, within limits, but the measured objects get better
with time, too.
My pain threshold is currently a dozen ADA4898 duals stamps in parallel.
That style of etched subcircuits is quite handy btw, I have designed a nice library
of modules that makes short development of prototypes.
I use those dielectrically insulated transistor arrays made by Intersil, but
only RF npn and pnps, no FETs. I could also use a DI fet opamp, or a normal
one that survives a 5 krad dose. Apart from LF411 there seems to be nothing qualified. :-(
Oh, pliers was not the right word, more like tweezers & a precision wire
cutter, but it looks that way under the microscope.
regards, Gerhard
(warmest xmas ever in history: 20.7°C in Munich)
Attachments
Last edited:
Gerhard, did you estimate input current noise? You can use this preamp only for Rs<10Ohm
no problem for my oscillator control voltages, supplies and references.
50 Ohms would be 1 nV/sqrtHz already...
(I'm currently not really into audio)
50 Ohms would be 1 nV/sqrtHz already...
(I'm currently not really into audio)
Could the indigent among us unwashed masses, too poor to afford books, beg a simplified version of these procedures? 🙂
OK, I don't think Jan would mind. The 604K resistor makes a 100nV/rt-Hz reference at the input this allows calibration without regard to bin width or windowing so even the most primitive FFT software should be usable. Just average lots of noise with the resistor switched in and make that reference line 40dB then short it and get the noise of the FET alone, 0dB equals 1nV.
This is all I used for those plots in the Pass forum.
Attachments
Last edited:
Much more sophisticated than my own crude lash up Guru Wurcer. 😱 We grovel at your feet.
What's the Pass forum?
Last edited:
Beware some low gm FET's might not generate enough gain at low currents to overcome your soundcard's noise. I did not consider these FET's as interesting to folks here. Flip everything for P channel.
Hello Scott,
Do you use Adobe Audtion, if it is may I ask why you use this package.
For SIM guys -->
All you SIM users looking for better models and more computing power -->
Check out my old alma mater.... LLNL in Livermore, Calif.... they've always had the most powerful scientific computer in the world and it is constantly being replaced/upgraded; What could you do with 512 bit word size and 20-Petaflop speed? Being installed is the "Sequia".... a Terascale Simulation Facility. And, try to imagine the memory part of it.
-RNMarsh
All you SIM users looking for better models and more computing power -->
Check out my old alma mater.... LLNL in Livermore, Calif.... they've always had the most powerful scientific computer in the world and it is constantly being replaced/upgraded; What could you do with 512 bit word size and 20-Petaflop speed? Being installed is the "Sequia".... a Terascale Simulation Facility. And, try to imagine the memory part of it.
-RNMarsh
Last edited:
Because my wife got the entire Adobe Suite as a student. Adobe's implementation of virtual memory and handling of HUGE audio files I find very useful. I can load and entire 10hr. MP3 as a wave and search visually for interesting sections. Converted to wav this is a 7G or so file, once loaded the zooming in and out and scrolling has no latency. Goldwave for instance choaked on this and started taking huge amounts of time for screen redraw every time I moved to another section. In general I do my own signal processing and use MATLAB or MATHCAD to control all aspects of the environment so I have no interest in any of the specialized programs available.
Actually people are exploring massively parallel video game graphics chips. We tested a prototype that easily was 10X as fast as a the best workstations available.
All the computers in the world won't help bad models.
We dont use no stinkin' bad models. [do we?]
Them bad ones are reserved for SPICE users. 🙂
-RNM
Them bad ones are reserved for SPICE users. 🙂
-RNM
Last edited:
Wiki --- IBM Sequoia. 1.572 million cores. 98,000 nodes are interconnected in a 5-dimensional Torus topology. The entire supercomputer (all 3000 sq ft of it) runs on Linux !!
All conceived, developed, built and used by real human being types.
All conceived, developed, built and used by real human being types.
Last edited:
Some of those GPU's cost more than my laptop all by themselves and then to run them massively in parallel, that aint't cheap! I would love one for when I am doing my CAD work and particularly when I am doing photo realistic rendering....... Of course HP decided that I didn't need an expansion slot or any other slot in my laptop, bad design in that respect. Next time, next laptop or my next desktop/workstation.
It's funny how terminology changes. Cores are CPUs (or "execution units" or something - a modern "microprocessor" is a "quad core" which I would think of as having/being four microprocessors on a chip), but when I was first learning about computers core was a type of memory.
Problems is nto the nr of cores, problem is how to get SPICE do parallel computing when doing transient analysis. AFAIK, it cannot be done today.
Ia m using MicroCap10 pro, and even though it does parallel computing by sharing the load amongst CPU cores when doing parametric sweeps or THD calculations, during TRAN it usus only one CPU. THD is done by running TRAN for several values of freq or amplitude.
What SPICE needs is a THz CPU instead of 1MEG of cores.
Sidebar: the world´s fastest computer, the Titan, draws a whopping 8 MW! That would take 2-3 win generator power plants 😛
Ia m using MicroCap10 pro, and even though it does parallel computing by sharing the load amongst CPU cores when doing parametric sweeps or THD calculations, during TRAN it usus only one CPU. THD is done by running TRAN for several values of freq or amplitude.
What SPICE needs is a THz CPU instead of 1MEG of cores.
Sidebar: the world´s fastest computer, the Titan, draws a whopping 8 MW! That would take 2-3 win generator power plants 😛