locating the stoppers at the base/gate reduces the length of the connection.
That reduces the inductance.
With the reduced inductance, the stopper can now be more effective.
If the wire is just as long after you put the resistor in, it can't have reduced inductance. If the wire is shorter after putting the resistor in, then the same effect on inductance could be achieved just by shortening the wire. Putting the resistor closer will certainly reduce the stray capacitance depending on how long the wire is between the resistor and base.
- keantoken
- keantoken
Member
Joined 2009
Paid Member
Yes my thinking is the same, the purpose of base stoppers is nothing to do with the length of the connection, it's about adding real (as in non-reactive) impedance to the base as seen by the stage driving into the base, of the output devices. The base input impedance is complex (as in resistive and reactive components) and depends on the load that the device is faced with too.
I am way out of my depth here, but both are misunderstanding my message.
The internal capacitance from base/gate to other parts of the semiconductor junction, combined with an external inductance of the bond wire length + the lead leg length + the PCB trace length + the resistor lead length creates an LC that can become an oscillator.
By omitting the PCB trace and reducing the resistor lead length to 0.5mm and reducing the gate lead leg to just where it exits the package reduces the L part of the LC oscillator.
Then the resistance can be designed/found by experiment to quench the oscillation (low Q)
Attaching the stopper resistor directly to the gate/base lead right at the package reduces the inductance and that changes the LC to small L C.
Why that is easier to stabilise, I have to leave to the Members that understand AC electrics.
The internal capacitance from base/gate to other parts of the semiconductor junction, combined with an external inductance of the bond wire length + the lead leg length + the PCB trace length + the resistor lead length creates an LC that can become an oscillator.
By omitting the PCB trace and reducing the resistor lead length to 0.5mm and reducing the gate lead leg to just where it exits the package reduces the L part of the LC oscillator.
Then the resistance can be designed/found by experiment to quench the oscillation (low Q)
Attaching the stopper resistor directly to the gate/base lead right at the package reduces the inductance and that changes the LC to small L C.
Why that is easier to stabilise, I have to leave to the Members that understand AC electrics.
There certainly aren't many references to even base stopper resistors. Bob Cordell though, also supports Kean's view,
stating that the stopper places a minimum value on the resistance looking into the base. This needs to be be sufficient
to neutralise any negative impedance at the base due to a capacitive load on the emitter. It stops oscillation.
(Designing Audio Power Amplifiers p194f)
He's quite specific in requiring Mosfet gate resistors be placed close to the gate terminal, though I think his
expression "to damp out resonant circuits that may be formed by inductances operating with gate capacitances"
(p225) seems a bit vague as an explanation.
stating that the stopper places a minimum value on the resistance looking into the base. This needs to be be sufficient
to neutralise any negative impedance at the base due to a capacitive load on the emitter. It stops oscillation.
(Designing Audio Power Amplifiers p194f)
He's quite specific in requiring Mosfet gate resistors be placed close to the gate terminal, though I think his
expression "to damp out resonant circuits that may be formed by inductances operating with gate capacitances"
(p225) seems a bit vague as an explanation.
All of these statements are following the same line
but none of them come close to explaining why the location of the stopper is critical to it's effectiveness.
There is something else going on that none of you are mentioning. It has something to do with the LC local to the output device.
1. 220-330R (with a ferrite right on the gate lead)
2. 470R (ferrite optional)
3. 2.2R - 6.8R (depending on how many pairs)
OS
AndrewT,
What L? This link is just 1 of hundreds you get when you google for "emitter follower" and oscillation: http://www.analogzone.com/col_1017.pdf Clearly the circuit can oscillate without any inductance.
Rick
What L? This link is just 1 of hundreds you get when you google for "emitter follower" and oscillation: http://www.analogzone.com/col_1017.pdf Clearly the circuit can oscillate without any inductance.
Rick
Member
Joined 2009
Paid Member
A recommendation I read somewhere was to size the base stopper so that there is around 10mV drop across it. This means no one value will suit every circuit.
However, Hugh never asks a simple question. I suspect he has some ideas about how the base stoppers can influence the subjective sound.
However, Hugh never asks a simple question. I suspect he has some ideas about how the base stoppers can influence the subjective sound.
Thanks Pete. The values you give closely align with my experience too.
Gareth,
Actually, all my questions are simple (though the technical reasons might well be complex) and they all come back to sound quality. I found on one of my mosfet designs that I could actually tailor the sound, particularly the bass, by simply altering the gate stopper. This despite a measured, flat FR, very strange. I was using a FQA36P15, a 36A continuous 19.5S Fairchild 294W beast with 144A surge capacity. Cgs is around 3nF, a huge device. I found 120R to be the best.
Hugh
Gareth,
Actually, all my questions are simple (though the technical reasons might well be complex) and they all come back to sound quality. I found on one of my mosfet designs that I could actually tailor the sound, particularly the bass, by simply altering the gate stopper. This despite a measured, flat FR, very strange. I was using a FQA36P15, a 36A continuous 19.5S Fairchild 294W beast with 144A surge capacity. Cgs is around 3nF, a huge device. I found 120R to be the best.
Hugh
There is inductance internal to any device in the real world.
I think the reason stoppers should be closer to the device is that the total surface area of the lead connecting to the base determines it's susceptability to magnetic and static inteference. If we instead place the resistor close to the base/gate, we have relocated the susceptable area to the other side the of the resistor, and so to interact with the base/gate, the resistance must be included in the network.
120R and 3nF gives about 400KHz corner. My thoughts are that if the FET can oscillate, it really has it's own feedback loop. If changing the resistor changes IMD, then perhaps it changes how the IM products interact with the bass? If this is in a feedback amp, then the feedback loops may be interacting. 400KHz sounds like a typical unity-gain point for a feedback amp. I don't really know much about IMD, so maybe I am out of my depth here.
- keantoken
I think the reason stoppers should be closer to the device is that the total surface area of the lead connecting to the base determines it's susceptability to magnetic and static inteference. If we instead place the resistor close to the base/gate, we have relocated the susceptable area to the other side the of the resistor, and so to interact with the base/gate, the resistance must be included in the network.
120R and 3nF gives about 400KHz corner. My thoughts are that if the FET can oscillate, it really has it's own feedback loop. If changing the resistor changes IMD, then perhaps it changes how the IM products interact with the bass? If this is in a feedback amp, then the feedback loops may be interacting. 400KHz sounds like a typical unity-gain point for a feedback amp. I don't really know much about IMD, so maybe I am out of my depth here.
- keantoken
Or every device ! As an example , I use 1R's on a NJW21193/4 pair and 2.2 - 4.7R on my favorite NJW0281/0302 pair. AKSA's observations on sound vs. stopper also hold true for BJT's , but for different reasons ("droop" - Hfe non-linearity ). The newer 2sk/j to-3p - flat pairs (laterals) will also require a different stopper than the old hitachi TO-3's.
My next personal amp will be lateral , so I am doing the R and D beforehand.
OS
These values look correct.
But do you agree that if zero GNFB is used the base stoppers can be omitted?
They give me worse simulated distortion (bipolars, case 3), unless extremely low (1ohm).
Exactly, though I believe it is mainly relevant to Mosfets. The high gate capacitance of ~500 pf, in series with L undefined lead inductance you refer to,
forms a resonant circuit. If the lead length, hence L is minimised, in adding a significant series R element (empirically 100-500 ohms),
oscillation of the lumped series LCR is effectively damped. At 20-100 MHz, a little L goes a long way!
Equally important is the degree of isolation the gate resistors give to parallel circuits in larger amplifiers where these problems and interaction is more common and severe.
I have used 680E with 2SK1058 and 470E with 2SJ162 and found this combination to sound the best, irrespective of the circuits I have tried them in.
After wrapping the gate stopper resistor lead on to the Gate lead and soldering the joint, I snap off the remaining portion of the Gate lead. I am usually left with about 4-6mm of Gate lead, closed followed by the resistor body with the other end of the resistor lead on the PCB. Its a long time since I soldered both ends of gate stopper resistors on to PCBs. I began doing this after reading that resistor leads have some inductance. I have never used a ferrite bead and have rarely encountered oscillations, except when the circuit itself is not adequately compensated.
Output stage topology with voltage gain has a tendency to oscillate.
After wrapping the gate stopper resistor lead on to the Gate lead and soldering the joint, I snap off the remaining portion of the Gate lead. I am usually left with about 4-6mm of Gate lead, closed followed by the resistor body with the other end of the resistor lead on the PCB. Its a long time since I soldered both ends of gate stopper resistors on to PCBs. I began doing this after reading that resistor leads have some inductance. I have never used a ferrite bead and have rarely encountered oscillations, except when the circuit itself is not adequately compensated.
Output stage topology with voltage gain has a tendency to oscillate.
I talked about why in post 295, but I don't think I was clear enough. The surface area of any lead between resistor and device becomes a parasitic capacitor. The length becomes part of the local magnetic coupling. Decreasing this surface area/length will decrease local reactivity. Moving the resistor closer to the device will relocate parasitic capacitance to the other side of the resistor, including the resistor in the capacitive network, rather than leaving no resistance between this parasitic and the device.
Inserting a resistor in the wire between driver and device cannot decrease L, because the path the current takes is not any shorter. Any time electrons move, there is inductance. It depends simply on the permeability of the medium, and the length of the movement. In fact, carbon film resistors are tiny coils, so they add about a cm to an inch of trace inductance (my estimate, I have never seen a reasonable explanation of film resistor inductance). Inserting a resistor I don't think will decrease net capacitance either, because the resistor sheath probably has a higher dielectric constant than air.
I think ferrite beads could be used to eliminate oscillation without impacting the transfer curve (distortion in non-GNFB circuits).
ferrite bead
Normal stoppers for bipolars peak at about 3.3R, right? And would it be a stretch to say local BJT oscillation occurs above 10MHz? I believe this hints at the proper bead.
How about this:
2661000101 Fair-Rite EMI/RFI Suppressors & Ferrites
- keantoken
Inserting a resistor in the wire between driver and device cannot decrease L, because the path the current takes is not any shorter. Any time electrons move, there is inductance. It depends simply on the permeability of the medium, and the length of the movement. In fact, carbon film resistors are tiny coils, so they add about a cm to an inch of trace inductance (my estimate, I have never seen a reasonable explanation of film resistor inductance). Inserting a resistor I don't think will decrease net capacitance either, because the resistor sheath probably has a higher dielectric constant than air.
I think ferrite beads could be used to eliminate oscillation without impacting the transfer curve (distortion in non-GNFB circuits).
ferrite bead
Normal stoppers for bipolars peak at about 3.3R, right? And would it be a stretch to say local BJT oscillation occurs above 10MHz? I believe this hints at the proper bead.
How about this:
2661000101 Fair-Rite EMI/RFI Suppressors & Ferrites
- keantoken
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.