I don't think they are necessary.
I do not know what good they do.
JFET transistors are different. They need such resistors.
I do not know what good they do.
JFET transistors are different. They need such resistors.
They are probably meant as base stoppers. If so, their purpose is to prevent VHF parasitic oscillations caused by the transistors and the wiring around them. Whether they are needed at all and if so, what value, is hard to predict, although a former colleague of mine had the rule of thumb to always use 47 ohm if there was no urgent reason not to do so.
A bit of pure L in in the base circuit, or pure C in the emitter. It could be the common mode C from the tail current source that ends up being the culprit. I had that happen on a tube LTP not too awfully long ago. The solution was some common R in series with the CCS. Soaks up some voltage compliance, but if you got it to spare it beats adding base resistance.
Input stage Degeneration.
For simplicity will improve slew rate and lower distortion.
Will make the stage far more linear across varying input voltages
Without degen expect current consumption of the input transistors
to only be slightly linear with voltages of around 40mV to 50mV
Past 40mV current will jump like wild and cause stability issues and distortion
With degen current consumption will be more linear over 400 to 500mV
and improve stability and reduce distortion
It will significantly lower the values of miller compensation capacitance needed
in the second gain stages. And very low currents of 1ma is all that is needed
in the differential to still have high slew rate.
Excessive current and excessive gain in the differential will cause stability issues.
Depending on emitter source impedance typical rule of thumb is
10:1 degen
Seeing the currents posted at 5ma, why ? dont know.
And minimal degen being used. looks already to be a stability nightmare.
If each transistor requires 2.5ma of current, whatever load it is driving is ridiculous.
And likely explains the high current if the goal was slew rate. Because the compensation
used is also high and ridiculous. No point in using dual diff, since the distortion will
be the same or if not worse than a more simplified topology with such excessive current.
For simplicity will improve slew rate and lower distortion.
Will make the stage far more linear across varying input voltages
Without degen expect current consumption of the input transistors
to only be slightly linear with voltages of around 40mV to 50mV
Past 40mV current will jump like wild and cause stability issues and distortion
With degen current consumption will be more linear over 400 to 500mV
and improve stability and reduce distortion
It will significantly lower the values of miller compensation capacitance needed
in the second gain stages. And very low currents of 1ma is all that is needed
in the differential to still have high slew rate.
Excessive current and excessive gain in the differential will cause stability issues.
Depending on emitter source impedance typical rule of thumb is
10:1 degen
Seeing the currents posted at 5ma, why ? dont know.
And minimal degen being used. looks already to be a stability nightmare.
If each transistor requires 2.5ma of current, whatever load it is driving is ridiculous.
And likely explains the high current if the goal was slew rate. Because the compensation
used is also high and ridiculous. No point in using dual diff, since the distortion will
be the same or if not worse than a more simplified topology with such excessive current.
Last edited:
This is the classical case:
A is a simple Colpitts oscillator: an inverting voltage-controlled current source with a CLC circuit as feedback network.
B is the same with a bipolar transistor instead of a controlled source. It's a signal schematic, so everything related to biasing is left out. The transistor could just as well be a FET or a valve.
C is the same, but now using the base-emitter capacitance (junction and diffusion capacitance) as the first capacitor.
D is C redrawn to make it look like a capacitively-loaded emitter follower with some inductance in the base lead.
E is two oscillators as in D that are coupled. They can still oscillate in common mode and they look suspiciously like a differential pair with a somewhat capacitive tail current source.
Be this as it may, I have had transistors in circuits that really didn't resemble Colpitts oscillators at all oscillate anyway. A current source transistor, for example, with its emitter going straight to a resistor. I had expected the resistor to give enough damping, but it didn't.
You don't need the base resistors, this configuration along with dual VAS transistors is a huge improvement on the badger front end.
Probably not in this circuit, stability wise anyway. Since the LTP itself is degenerated, keeping its differential gain in check. But they will offer some defense against excessive input currents if the pairs aren’t well matched or severely upset by abnormal operating conditions. The bias current out of the PNP’s base going into the NPN’s is limited by the resistors.
Q11 and Q12 in this circuit require base stoppers, and they are the only ones that need them. I never understood why. Their collector currents are fairly high, about 10 mA, so they do have a lot of transconductance.
I had 180 MHz oscillation on a cascode that did not have any degeneration in an amp I did about 12 yrs ago.
https://hifisonix.com/technical/cascode-amplifier-oscillation/
More recently I had a similar problem in a VAS stage using a beta helper where I used too high a value for the beta helper emitter load resistor and it oscillated between 30 and 40 MHz at 20-80 mV. The solution was to lower the emitter resistor from 10k to 1k.
https://hifisonix.com/technical/cascode-amplifier-oscillation/
More recently I had a similar problem in a VAS stage using a beta helper where I used too high a value for the beta helper emitter load resistor and it oscillated between 30 and 40 MHz at 20-80 mV. The solution was to lower the emitter resistor from 10k to 1k.
Q11 and Q12 from post #15 oscillated somewhere between 200 MHz and 300 MHz without their base stoppers.
It was the first discrete amplifier circuit I built after buying a 150 MHz analogue oscilloscope, before that I only had a 20 MHz scope. With the new scope, I could actually see the oscillation, even though its frequency was above the scope bandwidth. To locate the problem, I just connected the ground lead of the probe (set to 1:1) to its tip and moved the resulting wire loop above the circuit until I saw the largest VHF signal.
With a 20 MHz scope, I had never seen such oscillations at all. I could only infer that something must be oscillating from weird behaviour of the circuit under test. Hand effect, for example: bias points that change when you point at parts of the circuit.
It was the first discrete amplifier circuit I built after buying a 150 MHz analogue oscilloscope, before that I only had a 20 MHz scope. With the new scope, I could actually see the oscillation, even though its frequency was above the scope bandwidth. To locate the problem, I just connected the ground lead of the probe (set to 1:1) to its tip and moved the resulting wire loop above the circuit until I saw the largest VHF signal.
With a 20 MHz scope, I had never seen such oscillations at all. I could only infer that something must be oscillating from weird behaviour of the circuit under test. Hand effect, for example: bias points that change when you point at parts of the circuit.
That's very smart!
Smart circuit too - what was the purpose / initial thought for this design?
What I see is compensation of dc output (before the caps) balance through Q3/Q4, and LTP shift through Q7/Q6, and further output balancing through Q11/Q12, which are very close tied up between inputs & outputs (R15/R16--R14); hence your #11 post. Q1/Q2/Q3/Q4 have very large Re's (1k8)... Curious circuit indeed. Was it designed for controlling the Q5/Q6/Q11/Q12 (collectors only) dc output and lowering the output impedance?
It's for converting balanced signals to unbalanced and the other way around. The feedback via Q3, Q4, Q5 and Q6 ensures that the differential output voltage is (practically) equal to the differential input voltage. The slow common-mode loop via Q11 and Q12 keeps the output properly biased, but without making the common-mode output impedance at audio frequencies low.
As a result, the output more or less behaves as a floating voltage source. You can connect it to a balanced input, but you can also connect it to an unbalanced input, shorting the negative output to the local ground of the unbalanced input. It will work either way.
As a result, the output more or less behaves as a floating voltage source. You can connect it to a balanced input, but you can also connect it to an unbalanced input, shorting the negative output to the local ground of the unbalanced input. It will work either way.