Conrinous DC versus pulsed DC sweep of RF transistor - question

Table .2 on page 6 lists the collector base capacitance as typically 1.1pF at Vce at 5V. It's a varactor and Figure 7 on page 11 shows how it varies with Vce.

Most of the charge is injected into the base when Vce starts rising, and progressively less - in Coulombs per volt - as Vce goes up.

A good Spice model for the transistor might capture that.

The differences between DC and pulsed curves are usually thermal. That transistor is so fast that your pulses (milliseconds or microseconds?) are effectively DC electrically.

If you make long pulses, into the 10s or 100s of milliseconds, you can see the thermal effects on the curves. Until it blows up.

There are tricks to actually measure junction temperature while pulsing. Mosfets use the substrate diode and bipolars can use the b-e junction as the temperature sensor.

As Phil H has noted, the Early voltage of SiGe parts is huge. In other words, the Ic vs Vc curve is very flat absent thermals.

If you see a drooping Ic/Vc (or Id/Vd) curve, a reverse Early voltage, that's usually thermal.

I have tried both pulsed DC Vce and pulsed DC Vbe. With pulsed DC Vce, DC sweep of Vbe is done over the permitted range(as per datasheet), and with pulsed Vbe, DC sweep of Vce is done over the permitted range(as per datasheet). The Vbe|Vce pulse amplitude is varied within the applicable range(as per datasheet). The results are always the same: The pulse widths are 3 - 5 ms and periods are 6 - 10 ms.

1. With pulsed Vbe|Vce, the Ice values are always the same, no matter what the amplitude of the Vbe|Vce pulses are. or what the pulse widrh or period are, for each case.
2. With continuous Vbe|Vce, the familiar Ice -Vce|Vbe family of curves is obtained. _

Thermal time constants are complex, but a first guess of 100 milliseconds first-order is a reasonable place to start.

Pulse for a few hundred milliseconds at some fixed Vc and observe the Ic waveform during the pulse. Wait a minute before pulsing again.

Everything electrical, namely non-thermal, will be settled in nanoseconds in an SiGe transistor.

It sounds to me like his pulsed sources aren't doing what he thinks they should be doing.

Alternatively, his pulse sources may have enough parasitic reactances to make the transistor scream at some high frequency.

Jeroen Belleman

Good point about oscillation. Fast transistors are tricky in wideband circuits.

He could touch various nodes with a pencil or a small screwdriver and see if anything changes.

Base resistors sometimes help.

I am using a Keithley 6842 picoammeter and a Keithley 2230G triple output DC and programmable power supply. The same transistor is used for continous DC Vbe|Vce and pulsed DC Vbe|Vce. So if the continous DC Vbe|Vce method generates the correct family of curves, why would it fail in the pulsed DC Vbe|Vce case ? There are collector and base resistors as well.