I would like to amplify a difference signal from two electret microphones.

I would like to use an instrumentation amplifier with dual supply (AD620 for testing, probably INA103 for the final product).

Using dual power supply close to the maximum rating e.g -15V and +15 V will allow higher gain, and higher gain will result in much higher CMRR. I need high CMRR because the common mode signal will be far more louder than the differential signal.

I'll use a factory calibrated instrumentation amplifier instead of a simple differential amplifier for the same reason.

In the environment where this will be used, there is only a (noisy) +12 V power supply. My idea is to use to DC/DC converters, create +15 V and +30 V, and use the middle point as a new ground.

Suppose that the whole circuit will be isolated from the common ground, what are the disadvantages of using 0, +15, +30 V instead of -15, 0, 15 V?

(The DC/DC converters can do at least 1000 mA, much more than I need)

UPDATE: modified the requirements to +/-5V for making the design simpler. Would this circuit work, or should I use 7905 instead of 7805? I'm not sure.


simulate this circuit – Schematic created using CircuitLab


1 Answer 1


While this is certainly a viable design, it might turn out more frustrating than other approaches:

The "virtual" ground at +15 V is actually the output voltage of your "lower" step-up converter instead of the usually very well-grounded actual ground. That means that unlike "true" dual supplies, your approach means that the stability and accuracy of your power supply defines the 0V point – which isn't free of effect on your measurement. You can of course stabilize that by overdimensioning that supply, making sure that there's no oscillation of energy between the two supplies, large decoupling networks...

But in the end, you want to not go through extra lengths in an attempt to make a circuit less complicated or error prone.

Really, let's have a look at what we're talking about. This is the CMR curve from the AD620 datasheet:

CMR over parameters

So we're talking about the difference between 90 dB CMR and let's say 130 dB CMR, or rather 80 dB and 100 dB for a 1 kHz signal. 80dB+ are definitively impressive rejections already (assume you have a common-mode voltage of 1 V, how much of that do you get on the output? How far is that above Johnson-Nyquist thermal noise voltage over your next stage's input resistance?), and I'd be very surprised if adding a second supply to your problem wouldn't introduce more noise than you gain rejection …

Really, for least-noise, high rejection, not using your 12V SMPS to use two further SMPSes to generate a dual supply might be a pretty good idea, even cost-wise.

However, you say there's only one noisy 12V supply – so try to add as little additional noise as possible.

I'd personally go for something relatively simple, maybe a flyback supply where the secondary side has a center tap to provide your virtual ground, and use relatively massive capacitors to flatten the output voltage, followed by positive and negative linear voltage regulators (negative ones are a bit harder to get, but do exist).

Or, really, generate e.g. +10 V from your +12 V using a positive regulator, and a well-regulated +5 V – and really check if your system really benefits from the additional CMR in a real-world environment.

  • \$\begingroup\$ You meant +10V and +5V (not +5), right? Well, okay. Maybe using 10V range instead of 30V will not make a big difference in CMR. The flyback is a good idea too. \$\endgroup\$
    – nagylzs
    Dec 11, 2016 at 14:17
  • \$\begingroup\$ > not using your 12V SMPS to use two further SMPSes to generate a dual supply might be a pretty good idea BTW, this is an automotive application, the 12V is from the generator, not an SMPS. \$\endgroup\$
    – nagylzs
    Dec 13, 2016 at 11:45
  • \$\begingroup\$ ?! The generator in a car certainly won't be producing a noise-free voltage. Really, getting your system to a noise level where a CMRR >80dB matters is the real challenge here! \$\endgroup\$ Dec 13, 2016 at 14:37
  • 1
    \$\begingroup\$ yep, a lot of DSP, because noise sources typically aren't point-shaped, and neither are microphones :) but don't despair! your average smart phone has active noise cancellation in its microphone, too, and that is basically done with pretty simple DSP (so you don't need to buy expensive instrumentation or diff amplifiers) \$\endgroup\$ Dec 13, 2016 at 14:47
  • 1
    \$\begingroup\$ You can do the same with any sound card with a stereo line-in, and a bit of GNU Radio (see the tutorials for how to get started; maybe download the live DVD image to avoid all the installation \$\endgroup\$ Dec 13, 2016 at 14:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.