NOTES FROM THE TEST BENCH
By Bruce Hofer

In this edition I would like to share a pet peeve regarding published SPICE models for certain discrete small signal transistors. Some of these seriously understate the value for base spreading resistance, typically labeled Rb or Rbb’ in the model. I like to simulate my designs and see predicted results that match reasonably well with actual measured data from a hardware prototype. Thus I tend to get upset when a published model does not reflect the reality of the device.
Check me out here, but several manufacturers of the popular 2N3904 and 2N3906 publish SPICE models showing Rb as having an absurdly low value of 10 ohms! Many factors affect Rb such as device area, geometry, and fabrication process. The value of Rb in most discrete general purpose small signal transistors is 100-300 ohms. Now, before I get too many emails, let me quickly acknowledge that Rb can be outside this range if a transistor type has been optimized for low noise, switching, high frequency response, high power, or some other specialized application.
Why the big fuss over base spreading resistance?
First, Rb interacts directly with the two junction capacitances, Cje and Cjc, to cause additional phase shift within the transistor. Using too low of a value for Rb can lead to simulations showing a circuit to be stable and well behaved, when just the opposite may be true. This is particularly important in feedback amplifier and integrator designs.
Second, Rb can be a significant noise source. As suggested previously, transistors can be optimized for low noise operation by having a relatively large area to minimize Rb. However, this usually comes at the expense of higher junction capacitances and slower speed. The SSM2212 super-matched NPN pair is just such an example, with an Rb of 28 ohms (56 ohms total for the pair). This gives it a residual noise floor of about 0.96 nV/√Hz ignoring other sources. A pair of general purpose 2N3904s with a typical Rb = 150 ohms (300 ohms total) would give a much higher noise floor, about 2.23 nV/√Hz.
So, to those of you who design amplifiers with discrete transistors, do NOT ignore or use an incorrect value for base spreading resistance in your models. If you want to explore the stability margin in a new design, try increasing the value for Rb in your simulations until instability is encountered. Then, you may want to investigate whether any tweaks in your design could further improve stability. This extra effort during the design phase might just save you from a nasty production line shutdown in the future.
Happy Holidays, and have a “Stable” New Year!
-Bruce Hofer
Audio Precision co-founder and Chief Analog Engineer
|