I am building and learning about class AB amplifiers and I noticed that some designers include a low value resister in series with each complementary transistor and others leave the low value out. Some designer say that the emitter resisters are for limiting current.
It would be a great if someone could explain the benefits or disadvantages of using emitter resistors.
It would be a great if someone could explain the benefits or disadvantages of using emitter resistors.
There are 3 primary reasons to put emitter resistors in the output devices of a class AB amplifier:
1. Linearity: The resistors will ease cross-over distortion from the output stage. It will not eliminate it, but will reduce it substantially. Optimum condition for an EF output stage, RE = IQ/VT where IQ is the quiescent current of the device and VT is the thermal voltage, ~26mV at room temperature.
2. Thermal stability: Having RE when you have multiple output pairs will ensure that the output currents divides equally among the output devices. When you don't have an RE, if one output device gets hotter than the rest, it will try to take all the output current. This is bad since it may go beyond its power rating and go to heaven. RE solves this problem.
3. Current limiting: While RE won't do this by itself, it is usually part of the current limit circuitry which uses the voltage drop across RE as a sensor of an over current condition. I.e. if output current is too high, the RE voltage will too high, and if the current limit circuitry detects that this voltage is too high, it will do something to bring the output current down.
I hope it helps!
1. Linearity: The resistors will ease cross-over distortion from the output stage. It will not eliminate it, but will reduce it substantially. Optimum condition for an EF output stage, RE = IQ/VT where IQ is the quiescent current of the device and VT is the thermal voltage, ~26mV at room temperature.
2. Thermal stability: Having RE when you have multiple output pairs will ensure that the output currents divides equally among the output devices. When you don't have an RE, if one output device gets hotter than the rest, it will try to take all the output current. This is bad since it may go beyond its power rating and go to heaven. RE solves this problem.
3. Current limiting: While RE won't do this by itself, it is usually part of the current limit circuitry which uses the voltage drop across RE as a sensor of an over current condition. I.e. if output current is too high, the RE voltage will too high, and if the current limit circuitry detects that this voltage is too high, it will do something to bring the output current down.
I hope it helps!
Solid State Talk all about solid state amplification.
I thank you for you your great detailed answers to my question. I am not positive but I am thinking that the other complementary AB amplifier builders probably think that by connecting two 1N4148 series diodes to the bases this will solve most of their cross over distortion issues so there is no longer any need to use emitter resistors.
I thank you for you your great detailed answers to my question. I am not positive but I am thinking that the other complementary AB amplifier builders probably think that by connecting two 1N4148 series diodes to the bases this will solve most of their cross over distortion issues so there is no longer any need to use emitter resistors.
You are talking about two different things here. One is output stage biasing, the other is output stage emitter degeneration.
In an AB amp with BJT outputs, adding emitter resistors actually worsens crossover distortion. However, without any, the output stage tends to be very thermally unstable. (Keep in mind that power amps tend to have at least two cascaded emitter follower stages, at times three.) While thermal coupling to the bias diodes or bias spreader transistor will generally ensure a slight negative overall temperature coefficient if biasing is designed right, thermal conduction is slow. So any peak could momentarily increase bias current to potentially critical levels.
Output stage biasing is required for keeping output stage transconductance as flat as possible. If your output impedance varies a lot over output voltage swing, that is directly going to affect the voltage over the load (voltage divider), causing distortion that then has to be reined in by negative feedback. The worst case scenario is a vastly underbiased output with an outright dead zone in the middle, i.e. a region with output impedance so high that output voltage will remain at zero within a certain voltage range, not tracking that at all. In a scenario with feedback, the rest of the amplifier will then hurry to get past that zone on each and every zero crossing, putting high demands on amplifier slew rate and generally causing high levels of distortion in the process. A little example is found on this page.
I'd suggest going through Douglas Self's pages, and possibly his or Bob Cordell's amplifier design books as well.
In an AB amp with BJT outputs, adding emitter resistors actually worsens crossover distortion. However, without any, the output stage tends to be very thermally unstable. (Keep in mind that power amps tend to have at least two cascaded emitter follower stages, at times three.) While thermal coupling to the bias diodes or bias spreader transistor will generally ensure a slight negative overall temperature coefficient if biasing is designed right, thermal conduction is slow. So any peak could momentarily increase bias current to potentially critical levels.
Output stage biasing is required for keeping output stage transconductance as flat as possible. If your output impedance varies a lot over output voltage swing, that is directly going to affect the voltage over the load (voltage divider), causing distortion that then has to be reined in by negative feedback. The worst case scenario is a vastly underbiased output with an outright dead zone in the middle, i.e. a region with output impedance so high that output voltage will remain at zero within a certain voltage range, not tracking that at all. In a scenario with feedback, the rest of the amplifier will then hurry to get past that zone on each and every zero crossing, putting high demands on amplifier slew rate and generally causing high levels of distortion in the process. A little example is found on this page.
I'd suggest going through Douglas Self's pages, and possibly his or Bob Cordell's amplifier design books as well.
I thank you for you your great detailed answers to my question. I am not positive but I am thinking that the other complementary AB amplifier builders probably think that by connecting two 1N4148 series diodes to the bases this will solve most of their cross over distortion issues so there is no longer any need to use emitter resistors.
If you consider the equation given under the heading Linearity the value of RE is going to change directly in proportion to temperature rise (on the Absolute scale). The base emitter junction of a transistor has degrees of forward bias and the area of the die is relatively small so the transistor will heat up.
The emitter arrow points in the direction of the current while the negative charge carriers electrons in a NPN type travel in the opposite direction.
In this case the voltage drop across an emitter resistor in terms of forward bias is a negative feedback signal.
Basically you need emitter resistors for a BJT output stage, and often for MOSFETs too, in order to make the thing thermally stable. And you need them with multiple output pairs anyway to balance the currents (another thermal stability issue of a slightly different sort).
With certain devices like lateral MOSFETs which are thermally stable (increased temperature leads to decreased current), emitter resistors are not required.
With multiple BJT output pairs in a CFP configuration you need emitter resistors on each device as well as resistors to the output node (which are where emitter resistors for emitter follower output stages go, but aren't actually on the output device emitter due to the CFP topology)
The smaller the resistor value the more tricky thermal stability is to achieve, but the less cross over distortion you get, upto a limit (there is always cross-over distortion in a class B setup with any known device).
If you arrange the emitter resistor to have >= 0.7V across it at the max current you want your output devices to handle, it can be used as part of a simple current limiting circuit, since this can activate another transistor.
Emitter resistors also affect high frequency stability somewhat, although this is not a major issue to my knowledge.
Because emitter resistors are inside the negative feedback loop they have no real effect on the output impedance of the amp, 0.47 ohm emitter resistors doesn't mean the output impedance is >= 0.235 ohms, output impedances will be much less than this.
With certain devices like lateral MOSFETs which are thermally stable (increased temperature leads to decreased current), emitter resistors are not required.
With multiple BJT output pairs in a CFP configuration you need emitter resistors on each device as well as resistors to the output node (which are where emitter resistors for emitter follower output stages go, but aren't actually on the output device emitter due to the CFP topology)
The smaller the resistor value the more tricky thermal stability is to achieve, but the less cross over distortion you get, upto a limit (there is always cross-over distortion in a class B setup with any known device).
If you arrange the emitter resistor to have >= 0.7V across it at the max current you want your output devices to handle, it can be used as part of a simple current limiting circuit, since this can activate another transistor.
Emitter resistors also affect high frequency stability somewhat, although this is not a major issue to my knowledge.
Because emitter resistors are inside the negative feedback loop they have no real effect on the output impedance of the amp, 0.47 ohm emitter resistors doesn't mean the output impedance is >= 0.235 ohms, output impedances will be much less than this.
Last edited:
Thank you sgrossklass for suggesting to go to the to Douglas Self's website, great read. I learned a few things. I may just build one of his designs.
I did find a pdf that was written by Bob Cordell but a lot of it was way over my head.
I appreciate everyones generous help.
radiotek
PS: If anyone else has some good links please let me know.
I did find a pdf that was written by Bob Cordell but a lot of it was way over my head.
I appreciate everyones generous help.
radiotek
PS: If anyone else has some good links please let me know.
@Mark Tillotson: "Because emitter resistors are inside the negative feedback loop they have no real effect on the output impedance of the amp"
This is not 100% true. The amplifier output impedance is the the open loop output impedance / loop gain. So for a single output pair in an EF configuration, it will be:
ROUT = ROUT_OL / A_LG = (re + RE) / 2 / A_LG.
where ROUT_OL is the open loop output impedance, A_LG is the loop gain, re is the incremental emitter resistance of the output devices and RE is the emitter resistor. As the loop gain falls with frequency, ROUT approaches ROUT_OL. Thus, at 1KHz you are quite right, it does not matter much, but at 20KHz, if matters given most amps have loop gain cross-over frequencies of ~500KHz.
@sgrossklass: While I agree with the rest of your post, I disagree with: "In an AB amp with BJT outputs, adding emitter resistors actually worsens crossover distortion".
Using your argument about the amplifier output impedance, the output impedance with no RE and I_OUT = 0 is re/2. When I_OUT is large, output impedance approaches 0. Therefore, it is impossible to get a flat output impedance without RE.
When you have RE, when I_OUT = 0, R_OUT is (RE + re)/2. When I_OUT is large, output impedance approaches RE. Therefore, the optimum bias is when re = RE (Rout = RE at I_OUT=0 and ~RE when I_OUT is large). I agree that it is not possible to achieve this at all temperatures, actually, you can only do this at 1 temperature unless the output devices are biased PTAT (which as I write this, I am thinking why we actually don't do this.) But, this biasing condition achieves maximum flatness. I also agree that this is not a cure, but it is definitely better than RE=0. This was first published by a guy named Oliver. Attached is an excerpt from Cordell's book where he talks about this.
@radiotek: It seems to me you are fairly new to the art of amplifier design. I am actually putting a small course on Youtube (my fun side project) where I try to teach new guys about amp design and also designing what I hoping will be the lowest distortion amp around (I aiming high here 🙂 ). I am a bit behind because I am travelling, butt will get back to it soon. Here are some links:
SW-VFA-01: Audio power amplifier video series
YouTube
Let me know what you think.
Enjoy!
- Sandro
This is not 100% true. The amplifier output impedance is the the open loop output impedance / loop gain. So for a single output pair in an EF configuration, it will be:
ROUT = ROUT_OL / A_LG = (re + RE) / 2 / A_LG.
where ROUT_OL is the open loop output impedance, A_LG is the loop gain, re is the incremental emitter resistance of the output devices and RE is the emitter resistor. As the loop gain falls with frequency, ROUT approaches ROUT_OL. Thus, at 1KHz you are quite right, it does not matter much, but at 20KHz, if matters given most amps have loop gain cross-over frequencies of ~500KHz.
@sgrossklass: While I agree with the rest of your post, I disagree with: "In an AB amp with BJT outputs, adding emitter resistors actually worsens crossover distortion".
Using your argument about the amplifier output impedance, the output impedance with no RE and I_OUT = 0 is re/2. When I_OUT is large, output impedance approaches 0. Therefore, it is impossible to get a flat output impedance without RE.
When you have RE, when I_OUT = 0, R_OUT is (RE + re)/2. When I_OUT is large, output impedance approaches RE. Therefore, the optimum bias is when re = RE (Rout = RE at I_OUT=0 and ~RE when I_OUT is large). I agree that it is not possible to achieve this at all temperatures, actually, you can only do this at 1 temperature unless the output devices are biased PTAT (which as I write this, I am thinking why we actually don't do this.) But, this biasing condition achieves maximum flatness. I also agree that this is not a cure, but it is definitely better than RE=0. This was first published by a guy named Oliver. Attached is an excerpt from Cordell's book where he talks about this.
@radiotek: It seems to me you are fairly new to the art of amplifier design. I am actually putting a small course on Youtube (my fun side project) where I try to teach new guys about amp design and also designing what I hoping will be the lowest distortion amp around (I aiming high here 🙂 ). I am a bit behind because I am travelling, butt will get back to it soon. Here are some links:
SW-VFA-01: Audio power amplifier video series
YouTube
Let me know what you think.
Enjoy!
- Sandro
Attachments
- Home
- Amplifiers
- Solid State
- Low Pwr Class AB Complementary Audio Amp Question