In valves, the distortion, ignoring nonlinearities before clipping, is a result of the input voltage being too positive or negative. Too positive and the grid reaches 0V, where-in the cathode is now exposed to the full potential gradient from the anode; the grid looses control of the cathode current. Too negative and the effect of the anode is overriden, preventing any current from being pulled away. Only under extreme circumstances is the cathode overloaded to the point at which it can no long supply an increasing demand in current. To drag the cathode current to it's absolute maximum requires a high potential and no grid bias in most cases. So the distortion, clipping, is originating from the grid voltage.
I would like to ask for a similar explaination of where the distortion is originating from in the majority of solid state systems using something like an NPN.
Here you usually have quite high current handling ability and a lower voltage requirement. If the base becomes too negative, close to 0V, the depletion zone reforms and the device stops conducting. If the input swings too far positive it approaches the same voltage as the collector and so looses it's control over the current between the emitter and collector. However, there is also the potential for the device's current carrying capacity to be reached within that range.
When designing a SS system, which is most likely to cause the majority of the distortion? The device reaching it's maximum current handling ability or it's base / gate loosing control of the current? I've also heard of SS designers complaining that SS devices can be highly distorted at low volumes, entering a region of none distorted amplification over a certain power level. Is this genuinely a region of clipping or nonlinear operation in the silicon it's self? If so, why don't the devices behave linearly in the lower power regions? Why do they posses this form of 'minimum power' requirement?
I would like to ask for a similar explaination of where the distortion is originating from in the majority of solid state systems using something like an NPN.
Here you usually have quite high current handling ability and a lower voltage requirement. If the base becomes too negative, close to 0V, the depletion zone reforms and the device stops conducting. If the input swings too far positive it approaches the same voltage as the collector and so looses it's control over the current between the emitter and collector. However, there is also the potential for the device's current carrying capacity to be reached within that range.
When designing a SS system, which is most likely to cause the majority of the distortion? The device reaching it's maximum current handling ability or it's base / gate loosing control of the current? I've also heard of SS designers complaining that SS devices can be highly distorted at low volumes, entering a region of none distorted amplification over a certain power level. Is this genuinely a region of clipping or nonlinear operation in the silicon it's self? If so, why don't the devices behave linearly in the lower power regions? Why do they posses this form of 'minimum power' requirement?
> In valves, the distortion, ignoring nonlinearities before clipping,
Hard to ignore that.
> ...is a result of the input voltage being too positive or negative. Too positive and the grid reaches 0V, where-in the cathode is now exposed to the full potential gradient from the anode; the grid looses control of the cathode current.
Not so. Look at transmitter tubes rated for Class C service. The positive grid lines look just like the negative grid lines except they go to higher plate current. The problem is that the grid impedance drops from infinity to about 1K. It takes heroic measures to drive audio into such a varying impedance cleanly, and mostly we don't try.
> where the distortion is originating from in the majority of solid state systems using something like an NPN. .... When designing a SS system, which is most likely to cause the majority of the distortion? The device reaching it's maximum current handling ability or its base / gate loosing control of the current?
The transconductance of any device varies with current.
Gain is a function of transconductance (Gm).
Signal is a varying current.
So gain varies with signal. The tops of the waves come out bigger/smaller than the middle or center (depending how you do it).
In tubes, Gm varies roughly as square-root of current. If you go from 100mA to 200mA, Gm is roughly 1.4 times higher.
In a naked BJT, Gm varies directly with current. If you go from 100mA to 200mA, Gm is 2 times higher.
A single-ended tube swung nearly to clipping will have about 5% 2nd harmonic.
A single-ended BJT swung nearly to clipping will have about 26% 2nd harmonic.
BJTs have such high Gm that we almost always stuff a resistor under them. This reduces and stabilizes effective Gm. I have not seen a 26% THD tranny amp in a few decades....
Tubes may large-signal clip either by running out of current or by running out of voltage (plate bottoming, which may really be that the plate can't pass any more current without going positive-grid, which audio drivers usually won't do).
With Gm stabilization, BJTs and MOSFETs invariably clip by running out of voltage, collector bottoming. If you held them there and reduced the load impedance, power would rise. Then the limit is melt-down (or protection-trip). It is also possible to run short of Base current, but most amps have enough.
> ...SS devices can be highly distorted at low volumes, entering a region of none distorted amplification over a certain power level.
Sure. Gm is proportional to current. If current is almost-zero, Gm and gain is almost-zero.
This is not much of an issue for single-ended amps. But BJTs are mostly used push-pull and semi-Class B. At no-signal they run at very low idle current. This does wonderful things for heat and cost and size. It does terrible things to small audio. But "SS designers" have known this for 40 years. The fix is simple: run a little current through the devices even when there is no signal. The practical implementation can be tricky and has certainly inspired hundreds of clever designs. But a few silicon diodes and a couple resistors can take 99.9% of the low-level curse off of BJTs.
Hard to ignore that.
> ...is a result of the input voltage being too positive or negative. Too positive and the grid reaches 0V, where-in the cathode is now exposed to the full potential gradient from the anode; the grid looses control of the cathode current.
Not so. Look at transmitter tubes rated for Class C service. The positive grid lines look just like the negative grid lines except they go to higher plate current. The problem is that the grid impedance drops from infinity to about 1K. It takes heroic measures to drive audio into such a varying impedance cleanly, and mostly we don't try.
> where the distortion is originating from in the majority of solid state systems using something like an NPN. .... When designing a SS system, which is most likely to cause the majority of the distortion? The device reaching it's maximum current handling ability or its base / gate loosing control of the current?
The transconductance of any device varies with current.
Gain is a function of transconductance (Gm).
Signal is a varying current.
So gain varies with signal. The tops of the waves come out bigger/smaller than the middle or center (depending how you do it).
In tubes, Gm varies roughly as square-root of current. If you go from 100mA to 200mA, Gm is roughly 1.4 times higher.
In a naked BJT, Gm varies directly with current. If you go from 100mA to 200mA, Gm is 2 times higher.
A single-ended tube swung nearly to clipping will have about 5% 2nd harmonic.
A single-ended BJT swung nearly to clipping will have about 26% 2nd harmonic.
BJTs have such high Gm that we almost always stuff a resistor under them. This reduces and stabilizes effective Gm. I have not seen a 26% THD tranny amp in a few decades....
Tubes may large-signal clip either by running out of current or by running out of voltage (plate bottoming, which may really be that the plate can't pass any more current without going positive-grid, which audio drivers usually won't do).
With Gm stabilization, BJTs and MOSFETs invariably clip by running out of voltage, collector bottoming. If you held them there and reduced the load impedance, power would rise. Then the limit is melt-down (or protection-trip). It is also possible to run short of Base current, but most amps have enough.
> ...SS devices can be highly distorted at low volumes, entering a region of none distorted amplification over a certain power level.
Sure. Gm is proportional to current. If current is almost-zero, Gm and gain is almost-zero.
This is not much of an issue for single-ended amps. But BJTs are mostly used push-pull and semi-Class B. At no-signal they run at very low idle current. This does wonderful things for heat and cost and size. It does terrible things to small audio. But "SS designers" have known this for 40 years. The fix is simple: run a little current through the devices even when there is no signal. The practical implementation can be tricky and has certainly inspired hundreds of clever designs. But a few silicon diodes and a couple resistors can take 99.9% of the low-level curse off of BJTs.
- Status
- Not open for further replies.