I'm designing a preaplifier for a geophone. The amplifier uses OPAMPs which require +/-15V power supply for optimal operation. How can I generate such voltage rails without inserting excesive noise into circuit? Power supply should be as quiet as possible, with 9V unregulated input. The signals from the microphone are very weak.
\$\begingroup\$
\$\endgroup\$
9
-
1\$\begingroup\$ Filter, filter, filter. \$\endgroup\$– Ignacio Vazquez-AbramsCommented Jun 9, 2018 at 13:41
-
1\$\begingroup\$ What is common-mode range of your op amps ? I'm curious about that. \$\endgroup\$– Long PhamCommented Jun 9, 2018 at 13:41
-
4\$\begingroup\$ This is sadly not a proper requirement doc: "Quiet as possible": Well, then get a giant solar cell, cool it as far as you possibly can, and expose it to a perfectly regulated light source. Or, you could actually try to figure out how much noise your system inherently has, thus setting a lower limit for what your power supply needs to fulfill :) Then, you'd actually have anything to work with instead of us just giving you answers that are either waaaay over the top or insufficient. \$\endgroup\$– Marcus MüllerCommented Jun 9, 2018 at 13:50
-
3\$\begingroup\$ Use different op-amps that run on 9 volts. \$\endgroup\$– Andy akaCommented Jun 9, 2018 at 14:53
-
2\$\begingroup\$ Batteries are pretty good low noise sources. Use four 9V cells to make two 18V rails, then drop to 15V with linear (not switching) regulators. \$\endgroup\$– Jack BCommented Jun 9, 2018 at 16:36
|
Show 4 more comments
2 Answers
\$\begingroup\$
\$\endgroup\$
Use a linear power supply with a screened power transformer, good filtering, followed by linear low-noise regulators (eg. TI) followed by a bit of RC+ ferrite bead filtering. Or batteries + linear regulators etc.
If you really need to work from a 9VDC battery you would be better off redesigning the circuit to work from something like +/-2.5V than trying to insert a switching regulator and then cover up the noise. It's not impossible to do it as you suggest, it just requires a lot of care, to the point where most experienced consultants (ie. folks who have an idea of what's involved) are going to no-bid or greatly increase the cost to cover all the work. Geophones are some of the most demanding signal sources around, and if you want to get low nV/sqrt(Hz) noise spectral density down to fractions of a Hz a lot of care is required.
Your signal bandwidth will probably be < 10Hz but you do need to worry about higher frequencies, including RF, causing subtle low frequency noise through nonlinear behavior.
answered Jun 9, 2018 at 16:14
\$\begingroup\$
\$\endgroup\$
The LT1763 and LT1962 are very low noise parts ~20uV-rms and have a higher PSRR than that of other companies at 65dB, which will give you better noise rejection if operating from a noisy source. The source 500mA and 300mA respectively.
If you need more current it is possible to parallel them (which will also give you lower noise due to the averaging of noise sources).
They are in a SOIC8 package which is easy to prototype with.
If you want to move to an SMT device the LT3045 and LT3042 are lower noise at 0.8uV-rms and have a kelvin connection to reduce source resistance and other noise sources.
answered Jun 9, 2018 at 20:12