Here is a short write up on cascode oscillation I did back in 2012 when designing and developing the e-Amp.
Cascode Oscillation in Audio Amplifiers.pdf
I recently (2017) had a recurrence of the problem on another high power design – some pictures are shown below. When I went back and looked at the notes above, I realised I had not followed my original advice, and the problem had returned to plague me – clearly a case of ‘those that fail to learn from their mistakes are condemned to repeat them’.
This is what you get when you place the scope probe on the emitter of the cascode transistor. The probe capacitance is probably contributing to the problem and causing it to break into oscillation, or it may be increasing the level of existing oscillation. Either way, its not acceptable, and especially so if you are trying to design a circuit to deliver single digit ppm distortion performance.
Here it is with the time scale expanded:-
If you want to prevent or limit the probe from affecting the circuits behaviour, one trick is to look at it with a 10x probe – the input capacitance is much lower. Another option on 1x, is to place a 50-100 Ohm resistor in series with the probe – this helps to isolate the probe capacitance although you will still have some attenuation because of the scope probe and scope input capacitance. Use a 1206 surface mount device and solder it upright on the node you want to probe. Note that a 1x scope probe input capacitance is about 50 pF//1MEG Ohm and a 10x probe is 15 pF//10 MEG Ohm.
Its very important to note that you can form Colpitts oscillator structures in the base, emitter and collector circuits of transistors. Small signal audio transistors have fT’s of 100 ~ 300 MHz so all you need is some inductance (on a PCB this is often in the 40-60 nH range corresponding to 4-6 cm trace lengths) and then capacitive coupling (10-30 pF – layout dependant) from each end of the inductance to a non-inverting terminal on your amplifying device along with some gain. Below is a screen shot of three circuit examples. They all show HF instability and oscillation to some degree with 10’s to 100’s of mV at 20 to 100 MHz frequencies, but with some value combinations, it is quite easy to get volt level HF oscillation. As you can see, the LC networks that lead to problems can arise across any two terminals.
Although the PCB traces in a conventional analog amplifier are unlikely to be long enough to qualify as antenna’s at the frequencies mentioned, you will still couple a lot of this garbage capacitively into other small signal parts of your circuit. As noted above, if you are working on high performance audio circuits, problems like this will quickly put paid to any ppm or sub-ppm distortion aspirations you may harbour.
The high voltage PNP MMBT5401 and its NPN counterpart the MMBT5551 are often used for level shifters and feature an fT of 100 MHz to 300 MHz and a Cob of 6pF – they are fast and in the cascode configuration will easily oscillate if the conditions are right.
The following preventative measures (not an exhaustive list) provide a good starting point:-
- Place a 470 to 1k SMD (1206 or 0805) resistor as close as possible to the base of the cascode transistor. This lowers the Q of any inductance (i.e. ‘dampens’ it) in the base circuit and swamps any -ve resistance reflected into the emitter.
- In some cases, a SMD ceramic capacitor from the base of the cascode transistor to ground may help. I’ve found values between 10nF and 100 nF work well. Do not use film or anything else exotic – XR7 dielectric rated at 3-4 times the voltage on the base is about right. The capacitor ESR is also part of the fix.
- Make sure the overall loop area from the cascode base reference voltage to ground and the driver transistor in the emitter is small. If not, you will simply be adding inductance in the base circuit and will exacerbate the problem – loop areas have to be kept small to minimize inductance.
- Following on from (3) above, recall that the output at the cascode transistor collector is a current, so you can run fairly long traces from the cascode collector to the next part of the circuit – typically a common emitter stage referenced to the supply rails. However, you must minimize any capacitive coupling from the cascade collector circuit to its emitter – the best way to do this is through attention to layout.
- If the signal currents are low (1~10 mA), the propensity for the cascode circuit to break into oscillation can be further reduced by inserting a resistor of 100 Ohms to 1k in the collector of the cascode, located as close as possible to the device. This technique also helps by the way in emitter followers or beta helpers.
- If your circuit currents are low enough to allow, insert a low value resistor (100~200 Ohms) in the trace close to the cascode transistor emitter – this will help reduce the Q of the trace inductance.
- Pay attention to layout during the design stage – keep the cascode, driver transistor and associate circuit compact and with short traces. Keep loop areas small.
One final point about using Zener diodes as the reference voltage to the base of the cascode transistor. Zeners above about 7 V generate a lot of broadband noise right up into 100’s of MHz. Without filtering, damping and careful layout as described above, this noise can promote instability in cascode circuits.
Cascodes emitter impedance works like an inductance. This was used in early Tektronix oscilloscope for fast amplifiers to increase bandwidth (Peaking). Before using a cascode,
a real coil was used. The mathematics is (was) well understood, and your experimental cure
is ok. The details are explained in books by Dennis Feucht and “Wide Band Amplifiers” by Staric.
Regards,
UdoK