Building A High Gain Instrument Amplifier - Part 1

Motivation

My original goal of this project started when I was dealing with a rather nasty bug in a colleague's power supply design. Startup and load response functionality seemed fine, but when the device was loaded at it's nominal rating, the supply would emit high-pitched inductor whine. After around 20 minutes under load, the supply would fail catastrophically and would destroy devices without any direct or regulated power path. Something strange was going on.

I wanted to see if there was a way to measure signals propagating along a ground plane. After some online research, I came across a post by Christer Weinigel over on his blog detailing a design for an Instrument Amplifier. I ended up basing by design very closely on his with a few key modifications I'll discuss a little later on.

Following very limited investigation into the nature of ground-plane noise, and after discussion with various colleagues, I decided to assume the input signal would be no greater than a millivolt in amplitude. This assumption was later modified, and the final device requirements were to be calibrated for a signal amplitude in the microvolt range; this ultimately required a 1000x differential gain and extremely high impedance as not to affect the behavior of the noise signal.

This is the circuit I designed in OrCAD Capture and later modified due to some miscalculations.

I then created the following layout in OrCAD Allegro and sent it off to Oshpark for manufacturing. Normally I would've used PCBway, but I was also using this product to satisfy a class requirement at my university and they were buying the boards.

Assembly

After receiving and inspecting the boards I was very pleased to find no errors in placement, routing, or footprints. I quickly organized my tools and set to work.

If you do end up using my design files, note that the design uses 0402 and SMD almost entirely. A good microscope, tweezer set, and reflow gun are critical.

Testing

My test equipment list is as follows:

[if !supportLists]· [endif]Tektronix MSO2024B Oscilloscope

[if !supportLists]· [endif]SMA to BNC Adapter

[if !supportLists]· [endif]BNC to BNC Cable

[if !supportLists]· [endif]1 PSU with >5V bipolar supply

[if !supportLists]· [endif]3 Banana-Grabber power cables

[if !supportLists]· [endif]Signal generator

[if !supportLists]· [endif]Signal generator probes

After hooking everything up and running a few basic power-consumption and continuity tests, I turned the signal generator on and adjusted the Oscilloscope. The measurements I was receiving were a little bit off, and following a re-review of my design and reading a few textbook entries, I discovered that the feedback circuitry had been set up slightly incorrectly.

The gain in the circuit had actually been set to produce a combination of three times the common-mode voltage as well as five times the differential! Although this matched the behavior I noted while performing my initial tests, I initially wanted ten times differential gain and no common gain.

Although impossible to see, the unamplified input signal (blue) has an amplitude of 100mV and its amplified output (yellow) has a peak to peak voltage of 1.28V with an RMS of 436mV. There is also a small amount of phase difference with the amplified signal being shifted by 17.5 degrees. The RMS voltage suggests an offset that was likely induced by the incorrect common gain used by this test circuit. The phase difference may have been caused by the reservoir capacitor array on each of the op-amp’s supply input. Because the capacitors are referenced to common, and because of the 5x common gain, it is reasonable to assume they may have affected the signal response of the circuit.

After modifying the circuitry and ramping up the gain to 1000x to account for the new microvolt signal target, I was ready for testing.

---

Schematic Files

University Report