Hi all,Just been asked a question.An article on the TNT website about increasing an amplifier's bias current,gives you more Class A.I have posted this at the Audio Asylum,ages ago.What do you lot think about this?My amp is Class AB and if I increased my bias from 45ma to 120ma,say,do we get more Class A? Yes it runs hotter,but I've been told of increased distortion,somthing about harmonic distortion(I must buy a scope).The original article "Put a Tiger in your amp"gives a few numbers,but I havn't a clue about whats being calculated,and I've always assumed a Class A amp uses one power transistor.I would like to know how you would know if your listening in class A or AB.Am I to assume you would need sensitive speakers and the volume to be low,and wouldn't a transient signal send it into AB?
Can anyone explain this? I'd like an LED to show me when I'm listening in Class A.Do you think I'm barking up the wrong tree?At the end of the day,some sensitive speakers and an indicator and I've got me a Class A amp,right????
Please,can someone clear this up for me.Maths is not my strong point.
Thankyou.mikee55
Can anyone explain this? I'd like an LED to show me when I'm listening in Class A.Do you think I'm barking up the wrong tree?At the end of the day,some sensitive speakers and an indicator and I've got me a Class A amp,right????
Please,can someone clear this up for me.Maths is not my strong point.
Thankyou.mikee55
Mikee, for a commercial amp, in my opinion, it's probably not worth the bother. Class A running stresses the amp quite heavily, and unless it's designed for it, will greatly increase the risk of failure. If you want Class A, then build Class A, don't just muck around with the bias on an AB amp.
It is possible to put in an indicator circuit
that shows when an Class AB amp is running in A or B.
But it is not as clearly how the limit is to be defined.
At very low level a high bias Class AB can run most of sound in class A.
But some peaks in music would be in class B.
It is true, that with more sensitive loudspeakers
you will get more of the sound produced in the Class A domain.
Some Class AB can get better performance with increased bias.
Other may sound worse.
It is not that simple, that increasing bias always makes things better.
For a pure 100% class A, you wont need an indicator,
because you will hear when a class A amp goes into clipping.
Some class A amps have a RED LED indicator,
that give a warning when amp is -3dB from full power.
For a 30 Watt Class A amplifier, this is at 15 Watt output. (50% of power)
that shows when an Class AB amp is running in A or B.
But it is not as clearly how the limit is to be defined.
At very low level a high bias Class AB can run most of sound in class A.
But some peaks in music would be in class B.
It is true, that with more sensitive loudspeakers
you will get more of the sound produced in the Class A domain.
Some Class AB can get better performance with increased bias.
Other may sound worse.
It is not that simple, that increasing bias always makes things better.
For a pure 100% class A, you wont need an indicator,
because you will hear when a class A amp goes into clipping.
Some class A amps have a RED LED indicator,
that give a warning when amp is -3dB from full power.
For a 30 Watt Class A amplifier, this is at 15 Watt output. (50% of power)
In my opinion that particular article is one of the more notorious bits of audio info on the web. You could probably get away with it in the case of MOSFET output devices. In the case og BJT's, you are a lot more likely to harm the sound or even the amplifier itself. As I recall, the article doesn't even really consider that you need to know some specifics out the type of devices and the circuit configuration. Nor does it consider the heatsinking in the particular unit -- not a minor concern.
In general, for BJT's there is a more or less optimal bias that is hard enough to determine with a 'scope and/or distortion analyseer plus other equipment, let alone by a rule of thumb.
In general, for BJT's there is a more or less optimal bias that is hard enough to determine with a 'scope and/or distortion analyseer plus other equipment, let alone by a rule of thumb.
Actually, the same factors are in effect with Mosfets. It's just that some devices, like IR, sound awful at low bias currents. Depends heavily on the circuit too.
My advice, don't do that. Most amps do not sound better when you turn up the bias. The problem is, when you get a group of people in a room trying things, there is a danger of opinion going away from reality.
-Chris
My advice, don't do that. Most amps do not sound better when you turn up the bias. The problem is, when you get a group of people in a room trying things, there is a danger of opinion going away from reality.
-Chris
Thanks Y'all
Thanks for the replies,I have looked into building a Class A amp and have not been able to make my mind up as to building one.A few years ago I did make one,sort of,I used a TIP 141 and biased it into Class A and added another TIP 141 as an Emitter Follower I ran it of a 2.5amp transformer and although it sounded better than my AB amp,a JVC AX-2,it just burnt itself out.It hummed big time,I used a 4700uf cap on the speaker output to remove a big DC offset that pushed the drivers cone right out,OUCH,I guess my calculations were off.The transformer heated up,the transistors heated up,which in turn must have heated the transformer up some more.I had to heatsinks bolted together and blasted them with a 9" office fan.I figured that when I calculated the emitter current I forgot that I had to take into account the current drawn by the rest of the circuit,like base bias current used by both transistors.WHOOPS.I used BJT's because I couldn't figure out how to use mosfets.Maybe I'll have ago again one day.Yes TNT should go Boom ha!
Mikee55
Thanks for the replies,I have looked into building a Class A amp and have not been able to make my mind up as to building one.A few years ago I did make one,sort of,I used a TIP 141 and biased it into Class A and added another TIP 141 as an Emitter Follower I ran it of a 2.5amp transformer and although it sounded better than my AB amp,a JVC AX-2,it just burnt itself out.It hummed big time,I used a 4700uf cap on the speaker output to remove a big DC offset that pushed the drivers cone right out,OUCH,I guess my calculations were off.The transformer heated up,the transistors heated up,which in turn must have heated the transformer up some more.I had to heatsinks bolted together and blasted them with a 9" office fan.I figured that when I calculated the emitter current I forgot that I had to take into account the current drawn by the rest of the circuit,like base bias current used by both transistors.WHOOPS.I used BJT's because I couldn't figure out how to use mosfets.Maybe I'll have ago again one day.Yes TNT should go Boom ha!
Mikee55
MOSFETs are not linear devices. Their "gain" (transconductance actually) is rougly proportional to the drain current squared and is very low until the current has increased to 1A or so. That's why MOSFET amplifiers ask for a high bias, as it linearizes them.
Bipolar transistors do have a much more linear current gain and superposition (high bias) is not desirable, except to compensate for their somewhat lower gain at very low current levels, particularly at high frequencies.
Bipolar transistors do have a much more linear current gain and superposition (high bias) is not desirable, except to compensate for their somewhat lower gain at very low current levels, particularly at high frequencies.
It seems that everybody here , talk about the linearity of high or low bias , in the proverbial resistive load.
But the real advantaged of the high bias , both with mosfets or bipolars , came from the fact , that high bias , the open loop output impedance become lower and the performance of the amp become more unilateral with a real reactive speaker load and specially , the EMF almost "sees" a short at the output , with less chances of intermodulation with the input signal...
Sometimes , better than PC simulations , is nice , to make some real tests at the workshop ...
But the real advantaged of the high bias , both with mosfets or bipolars , came from the fact , that high bias , the open loop output impedance become lower and the performance of the amp become more unilateral with a real reactive speaker load and specially , the EMF almost "sees" a short at the output , with less chances of intermodulation with the input signal...
Sometimes , better than PC simulations , is nice , to make some real tests at the workshop ...
Is more important, how is bias stabil, when is floating rail voltage... Sometime I had one well known japanese amp on measuring....Bias, which was setuped in accordance with service manual ( it was class AB amp ) on correct value by AC 230 V, fall by 210 V ( which represent normal " floating " at rails by driwing with musical signal - - 10 % ) at zero = was at pure class B.... And it is feature of commonly used one transistor's biasing circuit, which you can see at mostly of amps....
Tube dude:
When the bias is set too high in a class-AB bipolar output stage, the output impedance jumps abruptly between the usual value and half that value, depending on whether both sets of output devices are or not conducting at the same time. That jump is a major source of intermodulation. The optimum bias for these designs should just provide a smooth transition between each bank of output devices, compensating for the gain loss when Ic is only a few mA.
Upupa Epops:
I have measured it and I'm fully aware of that issue.
First: relevant changes in bias with small changes in supply voltage should be taken into account (this should only happen in poor amateur designs, despite any marketing).
Second: when an amplifier is playing, the dies of the output devices are somewhat hotter than when it's idle. They may be, say, 20ºC or 30ºC hotter than the heatsink, but their temperature drops very quickly when you disconnect the load and the input signal in order to measure bias current. After a few seconds, the dies are already at the same temperature as the heatsink and your bias measurement has been fooled. To circumvent that, I leave the multimeter turned on and connected to the test points and I disconnect the input signal and the load very quickly while looking at the display. As expected, the multimeter shows a bias value that may be two or three times higher than expected (only 20% to 50% higher in clever designs with truly good compensation) and this value falls quickly to the expected value after a few seconds
Third: most amplifier circuits, both with MOS and bipolar output stages, suffer from current tails when they are asked to perform relatively quick zero-current crossing events. This happens because the gate or base drive circuit inserts charge into the bases/gates of one side faster than it extracts from the gates/bases of the other side. This in turn causes a dynamic increase in biasing only observed when playing significant amounts of high frequencies. Almost all bipolar CFP designs suffer from that phenomena, and also all MOS designs lacking gate buffers.
I have measured all that. In standard EF outputs, you may make the ground of either your oscilloscope or your amplifier floating. Then, you can just connect the oscilloscope ground to the output of the amplifier (before the inductor, if there is any) and each probe to the other end of one emitter resistor from each bank of output devices. As a result, you will see exactly how much current is being delivered to the load and cross-conducted. In order to get some RF isolation (desirable but not strictly required), I achieve common mode filtering by wrapping several turns of each probe wire around a ferrite toroid
Also, I'm currently working into a MOSFET amplifier that achieves stable bias not dependent on supply voltages or heatsink temperatures, and it's done without any thermal feedback. To achieve that, each output device is enclosed into a local current feedback loop with very high open loop gain. More about that may be read in the "Reinventing the N-channel wheel" thread : http://www.diyaudio.com/forums/showthread.php?postid=843895#post843895
When the bias is set too high in a class-AB bipolar output stage, the output impedance jumps abruptly between the usual value and half that value, depending on whether both sets of output devices are or not conducting at the same time. That jump is a major source of intermodulation. The optimum bias for these designs should just provide a smooth transition between each bank of output devices, compensating for the gain loss when Ic is only a few mA.
Upupa Epops:
I have measured it and I'm fully aware of that issue.
First: relevant changes in bias with small changes in supply voltage should be taken into account (this should only happen in poor amateur designs, despite any marketing).
Second: when an amplifier is playing, the dies of the output devices are somewhat hotter than when it's idle. They may be, say, 20ºC or 30ºC hotter than the heatsink, but their temperature drops very quickly when you disconnect the load and the input signal in order to measure bias current. After a few seconds, the dies are already at the same temperature as the heatsink and your bias measurement has been fooled. To circumvent that, I leave the multimeter turned on and connected to the test points and I disconnect the input signal and the load very quickly while looking at the display. As expected, the multimeter shows a bias value that may be two or three times higher than expected (only 20% to 50% higher in clever designs with truly good compensation) and this value falls quickly to the expected value after a few seconds
Third: most amplifier circuits, both with MOS and bipolar output stages, suffer from current tails when they are asked to perform relatively quick zero-current crossing events. This happens because the gate or base drive circuit inserts charge into the bases/gates of one side faster than it extracts from the gates/bases of the other side. This in turn causes a dynamic increase in biasing only observed when playing significant amounts of high frequencies. Almost all bipolar CFP designs suffer from that phenomena, and also all MOS designs lacking gate buffers.
I have measured all that. In standard EF outputs, you may make the ground of either your oscilloscope or your amplifier floating. Then, you can just connect the oscilloscope ground to the output of the amplifier (before the inductor, if there is any) and each probe to the other end of one emitter resistor from each bank of output devices. As a result, you will see exactly how much current is being delivered to the load and cross-conducted. In order to get some RF isolation (desirable but not strictly required), I achieve common mode filtering by wrapping several turns of each probe wire around a ferrite toroid
Also, I'm currently working into a MOSFET amplifier that achieves stable bias not dependent on supply voltages or heatsink temperatures, and it's done without any thermal feedback. To achieve that, each output device is enclosed into a local current feedback loop with very high open loop gain. More about that may be read in the "Reinventing the N-channel wheel" thread : http://www.diyaudio.com/forums/showthread.php?postid=843895#post843895
Nothing special happens. When the load is reactive, the zero-current crossing point just does not happen at 0 volts. Note that open loop output impedance of classic topologies with VAS is very high, since these circuits are current driven and their low output impedance is just an artifact of global feedback. Essentially they act as current servos, sinking or sourcing current from the output as required, until it reaches the preset voltage (following the input waveform). The zero-current crossing process is the same, despite at what voltage it happens, and current gain should be kept as constant as possible during that transition because low output impedance is just a side effect of current gain (and inversely proportional to it).
Eva said:
Very high open loop output impedance??
Not always...see some Accuphase amps and John Curl designed Parasound 1500...one mosfet driver with near , or less than 0,5 Ohms output impedance , followed bay a bipolar output , with a current gain of ~ 100...we get a open loop output impedance (independent of the overall feedback ) of ~0,5 /100 =0,005 open loop output impedance...
That I call extremely low open loop output impedance , independent of the VAS output impedance and clever design...
Could you explain how a non-saturated MOSFET may have 0.5 ohms of output impedance?
Maybe a class A source follower swinging betwem 4A and 5A at DC?
This is not practical neither real for AC operation, it's just conceptual over-simplification, because that MOSFET is going to have more than 1nF effective capacitance between gate and source, thus effectively shunting the input to the output. MOSFETs are evil devices, usually analysed in an oversimplified way. Their high input impedance gets actually very low at AC. Did you know that the gate may sink or source current without any Vgs change?
This is so true, that in the Mhz range you need more current to drive the gate than the current conducted by the device itself. It's just like a bipolar that runs out of gain when operated above its FT.
Maybe a class A source follower swinging betwem 4A and 5A at DC?
This is not practical neither real for AC operation, it's just conceptual over-simplification, because that MOSFET is going to have more than 1nF effective capacitance between gate and source, thus effectively shunting the input to the output. MOSFETs are evil devices, usually analysed in an oversimplified way. Their high input impedance gets actually very low at AC. Did you know that the gate may sink or source current without any Vgs change?
This is so true, that in the Mhz range you need more current to drive the gate than the current conducted by the device itself. It's just like a bipolar that runs out of gain when operated above its FT.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.