Hello, this is my first post in this forum so far. Nethertheless I've read the books from Bob Cordell and Douglas Self and learned a lot about amplifiers along the way.
I've designed a discrete headphone amplifier using the classical "blameless amplifier" topology.
I am currently writing on an internship script for the students at our university. I am myself a student, so I have of cause still a lot to learn.
Ähm.. The problem is now. I designed the amplifiers gain crossover frequency f_c at around 360kHz for maximum stability. Using the formula given from
Bob Cordell Book.
C_dom = 1/(2*pi*R_LTP * f_c*A_cl)
with f_c = 360kHz, R_LTP=220Ohm, A_cl = 2 I get:
C_dom = 1/(2*pi*220*360k*2) ≈ 1nF
The amplifiers schematic is included below. The problem is: The simulation does not match the expected unity gain crossover frequency.
In the simulation f_c is more like 1.2Mhz...!
I am also including of cause the simulation and the real Meassurement I did with a scope.
On the scope I just looked for the frequency where V_out = V_in. This is my unity gain crossover frequency, right?
I know the output is heavily slew limited at those high frequencies, but this is of cause my first amplifier and its purpose wasn't best fidelity.
So do you guys have any ideas why I am encouraging this problem? I have not that many ideas at the moment and really does need some help
from you guys.
Thank you and
Cheers
Gruftgrabbler
Schematic (Spice)
Schematic of what I build in real Hardware:
J2 is normally bridged at pins 3 and 2 and its only purpose is to visualise crossover distortion when I short pins 2 and 1 and
connect a signal directly to it.
Simulation Result:
Scope Meassurement:
I've designed a discrete headphone amplifier using the classical "blameless amplifier" topology.
I am currently writing on an internship script for the students at our university. I am myself a student, so I have of cause still a lot to learn.
Ähm.. The problem is now. I designed the amplifiers gain crossover frequency f_c at around 360kHz for maximum stability. Using the formula given from
Bob Cordell Book.
C_dom = 1/(2*pi*R_LTP * f_c*A_cl)
with f_c = 360kHz, R_LTP=220Ohm, A_cl = 2 I get:
C_dom = 1/(2*pi*220*360k*2) ≈ 1nF
The amplifiers schematic is included below. The problem is: The simulation does not match the expected unity gain crossover frequency.
In the simulation f_c is more like 1.2Mhz...!
I am also including of cause the simulation and the real Meassurement I did with a scope.
On the scope I just looked for the frequency where V_out = V_in. This is my unity gain crossover frequency, right?
I know the output is heavily slew limited at those high frequencies, but this is of cause my first amplifier and its purpose wasn't best fidelity.
So do you guys have any ideas why I am encouraging this problem? I have not that many ideas at the moment and really does need some help
from you guys.
Thank you and
Cheers
Gruftgrabbler
Schematic (Spice)
Schematic of what I build in real Hardware:
J2 is normally bridged at pins 3 and 2 and its only purpose is to visualise crossover distortion when I short pins 2 and 1 and
connect a signal directly to it.
Simulation Result:
Scope Meassurement:
I just wanted to add that I just downloaded Bob Bordells LT-Spice Examples from http://www.cordellaudio.com/book/tutorial_simulations.shtml
I opened Example 7 and simulated the AC behaviour.
The calculated gain crossover frequency is at 540kHz. The measured frequency is pretty much the same.
However when I modify the example and add a current mirror to it like in Chapter 6: Fig6.6 - A Miller-Compisated Amplifier
f_c is now at around 1.4Mhz!
This is odd since on the same page f_c will be calculated at 500kHz (in the figure there is a 250pF instead of a 300pF cap used for C_dom)
So what am I missing here?
For reference the modified schematic:
I opened Example 7 and simulated the AC behaviour.
The calculated gain crossover frequency is at 540kHz. The measured frequency is pretty much the same.
However when I modify the example and add a current mirror to it like in Chapter 6: Fig6.6 - A Miller-Compisated Amplifier
f_c is now at around 1.4Mhz!
This is odd since on the same page f_c will be calculated at 500kHz (in the figure there is a 250pF instead of a 300pF cap used for C_dom)
So what am I missing here?
For reference the modified schematic:
Hi Gruftgrabbler,
Your calculations are correct however, the addition of the current mirror allows the effective use of both collector currents unlike the resistive load. This essentially doubles the gm in comparison. You therefore need to double the Cmiller to keep the same stability margin as before. Try increasing it from 300pF to 600pF and see if fc is around 540kHz.
Cheers
Paul
Your calculations are correct however, the addition of the current mirror allows the effective use of both collector currents unlike the resistive load. This essentially doubles the gm in comparison. You therefore need to double the Cmiller to keep the same stability margin as before. Try increasing it from 300pF to 600pF and see if fc is around 540kHz.
Cheers
Paul
I'm not sure where that equation comes from or what A_cl is, I don't have that book alas, but the way to calculate these things is to figure out the open-loop gain of each stage and multiply together. Then solve for unity gain in terms of frequency. The VAS term has local negative feedback from Cdom so its gain is a function of frequency.
You need the input stage transconductance (which will double with a current mirror, note), the VAS transistor beta will affect LF gain but not fc, in fact only the transconductance and Cdom affect the cutoff point for open-loop. Then you divide by the closed-loop gain to get the closed-loop cutoff freq:
fc = 1/(2pi) * (gm / Cdom) - open loop
fc = 1/(2pi) * (gm / Cdom / gain) - closed loop
The input stage transconductance will double with a current mirror, note, since the mirror matches the LTP's current swing.
For instance if gm= 10mS, Cdom = 1nF, gain = 6dB (x2), you get fc = 800kHz
You need the input stage transconductance (which will double with a current mirror, note), the VAS transistor beta will affect LF gain but not fc, in fact only the transconductance and Cdom affect the cutoff point for open-loop. Then you divide by the closed-loop gain to get the closed-loop cutoff freq:
fc = 1/(2pi) * (gm / Cdom) - open loop
fc = 1/(2pi) * (gm / Cdom / gain) - closed loop
The input stage transconductance will double with a current mirror, note, since the mirror matches the LTP's current swing.
For instance if gm= 10mS, Cdom = 1nF, gain = 6dB (x2), you get fc = 800kHz
Both of you are incredible helpful! Thank you very much! 🙂
I should've really checked this earlier. But at least I am glad to learn my lesson ^^
Increasing the capacitance in the example to 600p ends up on a f_c of about 600Khz, which is I guess near enough to the calculation.
@mark Thank you for the calculation. Very helpful! 🙂
So again thank you
and Cheers
Gruftgrabbler
I should've really checked this earlier. But at least I am glad to learn my lesson ^^
Increasing the capacitance in the example to 600p ends up on a f_c of about 600Khz, which is I guess near enough to the calculation.
@mark Thank you for the calculation. Very helpful! 🙂
So again thank you
and Cheers
Gruftgrabbler