Detecting those little creatures from the bottom

When recently measured the performance of my CCS designs, I found that I wasn’t able to measure below -114dBV (2uV) and therefore limited my ability to measure CCS impedance. I was also keen to measure inductors, transformers and other reactive components to derive 2 or 3 component models for more accurate simulations during the design process. A preamplifier was in order, so I looked at options with some of the ICs I had at hand. A very nice suggestion from Burr Brown/Texas Instruments is shown below. The INA106 is a precision 10x differential amplifier. Not a cheap device, but quality does comes with it unfortunately. When coupled with a pair of OPA37, a very accurate differential preamplifier can be built with extremely low noise and distortion:

The resistor pair R1 and R2 set the gain. The output is Vo=10*(1+2*R2/R1)*ΔVin. A gain of 1,000 (60dB) will hopefully take the noise floor down to -160/170dB. For an accurate gain we need to use 0.1% resistors in particular for R2 to match the pair of them. Battery operation is mandatory for this preamplifier. The rejection of common mode noise is very good and with maximum common-mode input voltage of 10V, we can easily use it in the return leg of the measuring circuit. The decoupling capacitors in the diagram above are ceramic 1uF/50V ones.

Building the “Noise Inspector”

I found the mints candy box ideal for these devices. A tin box provides the necessary shielding an can fit a small PCB to accomodate this circuit. When drilling you have to take special care to de-burring as the tin holes can leave sharp bits. Some planning is required to ensure you can fit all the parts and given that the box is only 2cm max approximately, special care is needed to avoid components and connectors to make contact when the lid is closed.

PCB inside the tin box

The Noise Inspector is ready for action

The Noise Inspector

The performance is fantastic. I plugged in the Noise Inspector to the Pete Millet’s SC interface and got the following performance:

With the input in short-circuit, the noise floor goes down to about -175dB, however the 50Hz noise is -161dB. This FFT has a setting of 64 averages. -161dBV is about 9nV. Extremely low indeed! The other artifacts at 450Hz/25kHz are already part of my test set.

Now looking at the performance of the Rigol DG1022 Waveform Generator which the lowest output is 2mVpp we can see: 

The generator performs really well with a noise level of only 2.6uV (-111.7dB) picked up by the long cables running out from the BNC output from the generator to the Noise inspector.

Testing reactive devices

An idea I haven’t tried yet is to measure the frequency response of reactive components (e.g. inductors, transformers) and derive the 2 or 3-component models like suggested by Morgan Jones. The following setup will help measuring the response notch you will expect from the LCR circuit and by ensuring the right inductance is measured by using an CCS to set the choke or output transformer with the quiescent current:

The impedance can be derived by measuring Vo versus Vi at various frequencies. The diode can be a 600V TVS to protect the CCS, the noise inspector and the function generator from the back EMF generated by the inductor when the CCS is turned off. For most of the measurements we won’t need the noise inspector until we get close to the notch or the LCR system formed by the series LR plus the parallel capacitance of the inductor.

Author: Ale Moglia

"A mistake is always forgivable, rarely excusable and always unacceptable. " (Robert Fripp) View all posts by Ale Moglia