Thanks in advance for any info.
There are numerous references to Oliver bias set point in various threads, but so far I have not been able to find the original analysis.
Where might this information be found?
Also my SPICE simulations predict a much lower value for lowest distortion bias current for a given exterenal RE (93 ma @ .1 ohm)
Also, SPICE predicts higher optimum bias as temperature increases, suggesting that keeping the bias constant with temperature would cause increase in distortion as devices heat up.
There are numerous references to Oliver bias set point in various threads, but so far I have not been able to find the original analysis.
Where might this information be found?
Also my SPICE simulations predict a much lower value for lowest distortion bias current for a given exterenal RE (93 ma @ .1 ohm)
Also, SPICE predicts higher optimum bias as temperature increases, suggesting that keeping the bias constant with temperature would cause increase in distortion as devices heat up.
Last edited:
Hi Guys
Oliver's assertion was simply that the external RE should equal the internal re' in an emitter follower output stage. This results in a target voltage across RE of 26mV, related to the 26mV/I that the internal re' is.
Spice will often yield lower THD at higher than the "Oliver" current for a given design. This does not make the design "wrong" as Oliver only explored limited aspects of an output stage and amplifier, and came up with the erroneous term "gm doubling". That term only means that both sides of the circuit contribute to the gm of the output stage at low currents, in a range where the total gm of the output stage stays constant. After one side turns off, the gm of the other side rises with increasing load current. So, it might have been more relevant to refer to the "constant gm zone" instead.
Most error correction schemes effectively try to maintain a constant gm for the output stage despite varying current by using local feedback.
Following Oliver rigidly will result in output stages that do not idle extremely hot, which is a good thing in the summer or if you live close to the equator. Oliver's work was an important step in improving and understanding solid-state amplifiers.
Have fun
Oliver's assertion was simply that the external RE should equal the internal re' in an emitter follower output stage. This results in a target voltage across RE of 26mV, related to the 26mV/I that the internal re' is.
Spice will often yield lower THD at higher than the "Oliver" current for a given design. This does not make the design "wrong" as Oliver only explored limited aspects of an output stage and amplifier, and came up with the erroneous term "gm doubling". That term only means that both sides of the circuit contribute to the gm of the output stage at low currents, in a range where the total gm of the output stage stays constant. After one side turns off, the gm of the other side rises with increasing load current. So, it might have been more relevant to refer to the "constant gm zone" instead.
Most error correction schemes effectively try to maintain a constant gm for the output stage despite varying current by using local feedback.
Following Oliver rigidly will result in output stages that do not idle extremely hot, which is a good thing in the summer or if you live close to the equator. Oliver's work was an important step in improving and understanding solid-state amplifiers.
Have fun
SPICE predicts optimum bias at a factor 2.8x LOWER than Oliver. Predicted distortion at Oliver bias was extremely high.
Circuit is EF3 ("T" circuit) with MJ15022/23 outputs.
Hi Guys
Twice as high, half as high - it demonstrates my point that the Oliver guide is just that - a guide - and one that does not point to lowest distortion for all amps. So it is actually a guide to what?
Maybe not a guide but an illumination of the constant-gm region and the changing-gm regions to either side. The transition points are where a couple of distortion mechanisms make themselves known.
There are other examples of circuits that idle way below the Oliver value to attain best performance. It is more typical to be above the suggestion.
What range of numbers are you getting with your circuit so far?
Have fun
Twice as high, half as high - it demonstrates my point that the Oliver guide is just that - a guide - and one that does not point to lowest distortion for all amps. So it is actually a guide to what?
Maybe not a guide but an illumination of the constant-gm region and the changing-gm regions to either side. The transition points are where a couple of distortion mechanisms make themselves known.
There are other examples of circuits that idle way below the Oliver value to attain best performance. It is more typical to be above the suggestion.
What range of numbers are you getting with your circuit so far?
Have fun
As stated in original post, SPICE predicts 93 ma @ RE = 0.1 ohm. This suggests 9.3 mV / RE. SPICE also predicts the trend that lower V(RE) holds for various values of RE from 0.05 up to .33.
My experimental results based on measurement of IM distortion and listening confirm optimum bias in range 80 - 100 MA @ RE = 0.1 ohm
Lowest distortion occurs with lowest possible value of RE, with RE = 0.05 and bias @ 154 MA. In this case, V(RE) = 7.7 mV.
At RE = .33, I = 33 ma > V(RE) ~= 11 mV.
This demonstrates that optimum V(RE) actually increases progressively as RE increases.
Distortion is unacceptable with RE > 0.1 ohms.
All of these values are signifacantly lower than the low limit of gm/2 (13 mV) suggested by Oliver analysis.
Hi Guys
I meant what THD20 at what load?
Self demonstrated that trend for THD vs RE, but as usual, left out details like idle current - did he change it or not...
Have fun
I meant what THD20 at what load?
Self demonstrated that trend for THD vs RE, but as usual, left out details like idle current - did he change it or not...
Have fun
SPICE values for distortion are meaningless, only good to demonstrate relative values. In the analysis, current was adjusted for various RE to obtain lowest distortion. Rload = 8 ohms. In all cases, Lower RL required slightly higher Ibias with slight increase in distortion.
The combinations for lowest distortion at RL = 8 ohms were:
RE = 0.05 -- Ibias = 154 ma -- > V(RE) = 7.7 mV
RE = 0.10 -- Ibias = 93 ma --- > V(RE) = 9.3 mV
RE = 0.33 -- Ibias = 33 ma --- > V(RE) ~= 11 mV
The combinations for lowest distortion at RL = 8 ohms were:
RE = 0.05 -- Ibias = 154 ma -- > V(RE) = 7.7 mV
RE = 0.10 -- Ibias = 93 ma --- > V(RE) = 9.3 mV
RE = 0.33 -- Ibias = 33 ma --- > V(RE) ~= 11 mV
Last edited:
Hi Tim
Those are not THD values, just the RE vs current vs Vq reported above.
The voltage across the emitter resistor is called Vq.
Self uses the voltage across BOTH REs as Vq.
Have fun
Those are not THD values, just the RE vs current vs Vq reported above.
The voltage across the emitter resistor is called Vq.
Self uses the voltage across BOTH REs as Vq.
Have fun
Hello rayma and Struth. Thank you both for the pdf article [1971] which you've attached by Dr. Bernard M. Oliver. He was an accomplished scientist per his remarkable CV on the web which is a great read. I'll peruse his 60+ patents to look for other intellectual nuggets; his recorded thoughts.
Hi,
Gm doubling is not an erroneous term, it exists in all Class A and AB
designs. Class aB which is optimum bias is defined by no gm doubling.
Biasing cannot maintain constant Gm. It varies and setting bias
is optimising its variation. D. Self covers it in great detail.
rgds, sreten.
D. Self does not classify the optimum bias in the AB class.
Class AB gives gm doubling around in the 0 V zone.
The optimum bias belongs to the class B with some quiescent current, just as are most op-amps.
Class AB gives gm doubling around in the 0 V zone.
The optimum bias belongs to the class B with some quiescent current, just as are most op-amps.
M. Leach analysis concludes that gm doubling does not occur.
http://users.ece.gatech.edu/~mleach/papers/classab.pdf
http://users.ece.gatech.edu/~mleach/papers/classab.pdf
He also designed an audio amplifier that was built by many HP employees.
http://www.hparchive.com/Manuals/Barney_Oliver_Amplifier_Manual.pdf
Hi Guys
If English is not your first language, then it is easy to miss the difference between the definitions of the words "doubling" , "constant" and "contributing".
The "gm-doubling region is just where both sides CONTRIBUTE gm. A quick calculation for any two current points within that region will show that there is only one point - at zero signal output - where each side contributes half the gm, and you could at that point only say the total gm is double that of either side.
So, TOTAL gm is constant in the gm-doubling region for a class-AB amp, just as it is in a push-pull class-A amp.
Self has confused some matters of class naming inasmuch as he calls an optimally-biased class-AB amp class-B. The fact that idle is not zero makes it class-AB for the rest of the world. Calling the region beyond class-A class-AB is another thing, and quite common.
THD as a dB seems like an obfuscation. Percent or ppm makes more sense to more people.
-80dB = 0.01% = 100ppm
-100dB = 0.001% = 10ppm
-120dB = 0.000 1% = 1ppm
-140dB = 0.000 01% = 100ppb
-160dB = 0.000 001% = 10ppb
All zeroes in LTspice is 0.000 000% indicating <10ppb.
One thing I've noticed is that if you have a low THD number at a low idle current at 8R, THD will rise very fast as RL decreases, and that this will be a much faster rise than with idle set to a higher current..
Have fun
If English is not your first language, then it is easy to miss the difference between the definitions of the words "doubling" , "constant" and "contributing".
The "gm-doubling region is just where both sides CONTRIBUTE gm. A quick calculation for any two current points within that region will show that there is only one point - at zero signal output - where each side contributes half the gm, and you could at that point only say the total gm is double that of either side.
So, TOTAL gm is constant in the gm-doubling region for a class-AB amp, just as it is in a push-pull class-A amp.
Self has confused some matters of class naming inasmuch as he calls an optimally-biased class-AB amp class-B. The fact that idle is not zero makes it class-AB for the rest of the world. Calling the region beyond class-A class-AB is another thing, and quite common.
THD as a dB seems like an obfuscation. Percent or ppm makes more sense to more people.
-80dB = 0.01% = 100ppm
-100dB = 0.001% = 10ppm
-120dB = 0.000 1% = 1ppm
-140dB = 0.000 01% = 100ppb
-160dB = 0.000 001% = 10ppb
All zeroes in LTspice is 0.000 000% indicating <10ppb.
One thing I've noticed is that if you have a low THD number at a low idle current at 8R, THD will rise very fast as RL decreases, and that this will be a much faster rise than with idle set to a higher current..
Have fun
Hi Tom
Your THD plots suggest that you are using a too-low value for the feedback cap to ground. Anything <10mF is too low and larger values show marked improvement in low-frequency THD.
For example, on an amp I was working on recently, 10mF yielded 4.4ppm at 25Hz into 1R, where 47mF yielded 440ppb - a ten times improvement for a five-times value change.
Almost all of the low-f THD in most amps is from the electrolytic caps. These should not be used to shape the bass response. Even using good 10kHr caps, electrolytic DF goes crazy after a few years and distortion will rise.
Have fun
Your THD plots suggest that you are using a too-low value for the feedback cap to ground. Anything <10mF is too low and larger values show marked improvement in low-frequency THD.
For example, on an amp I was working on recently, 10mF yielded 4.4ppm at 25Hz into 1R, where 47mF yielded 440ppb - a ten times improvement for a five-times value change.
Almost all of the low-f THD in most amps is from the electrolytic caps. These should not be used to shape the bass response. Even using good 10kHr caps, electrolytic DF goes crazy after a few years and distortion will rise.
Have fun
- Status
- Not open for further replies.