Interesting. The problem with both issues (not being able to reduce 2nd diff current and not being able to avoid the output node with the driver emitters) turned out to be the same thing: some type of parasitics or feedback across the flying Baker clamps.
I added some reversed 1N4148s (and reduced the Zeners by a diode drop), and everything is ducky.
I'd love to know the reason behind this if anyone's seen something like it before....
Cheers,
Jeff.
I added some reversed 1N4148s (and reduced the Zeners by a diode drop), and everything is ducky.
I'd love to know the reason behind this if anyone's seen something like it before....
Cheers,
Jeff.
Well now, isn't Bob's speed up capacitor just the neatest little thing?
It increased my slew speed so much that I was able to pile on more phase-lead compensation. In fact, I fear I've taken the "buzz" out of BuzzBomb. In theory it is now eminently stable, with margins of 52º and 10dB.
I might have to leave the improved current mirror to ZM in case he wants to Babelfish it. I stuck a fork in this version and it's done.
It increased my slew speed so much that I was able to pile on more phase-lead compensation. In fact, I fear I've taken the "buzz" out of BuzzBomb. In theory it is now eminently stable, with margins of 52º and 10dB.
I might have to leave the improved current mirror to ZM in case he wants to Babelfish it. I stuck a fork in this version and it's done.
Hi Michael,
Welcome to the best thread on the forum........
Probably the mirror improvements, but the design is converging into a fine amplifier but is excellent as it stands.
I need to talk to you after the first of the year about that pre-amp topology we discussed some time back. I may have some boards for you.
Jam
Welcome to the best thread on the forum........
Probably the mirror improvements, but the design is converging into a fine amplifier but is excellent as it stands.
I need to talk to you after the first of the year about that pre-amp topology we discussed some time back. I may have some boards for you.
Jam
Speaking of current mirrors, devices, etc.:
Since a mirror attempts to duplicate the current you'd ideally like to use the same transistors, match them, and put them on the same heatsink. (The resistors should be matched too.)
Conversely, with a feedback current source, it's not important that the transistors be matched, or even be the same transistor.
Do I have that right?
Thanks,
Jeff.
Since a mirror attempts to duplicate the current you'd ideally like to use the same transistors, match them, and put them on the same heatsink. (The resistors should be matched too.)
Conversely, with a feedback current source, it's not important that the transistors be matched, or even be the same transistor.
Do I have that right?
Thanks,
Jeff.
The two transistors in a standard current mirror operate at vastly different Vce's**, so they dissipate vastly different amounts of power. Putting them on the same heatsink ensures their Tc's are about the same, but ThetaJC guarantees that their Tj's will be different.
The Wilson mirror, and other cascoded mirrors, brings the Vce's down to the same level. Now the power dissipated in the two devices is roughly equal.
Heavy emitter degeneration makes the 2T standard current mirror less sensitive to deltaVbe (caused by deltaTj). Remember deltaVbe =approx= -2.2 mV per degree C. If deltaTj is X degrees C, why not drop (10 * X * 2.2mV) across the degeneration resistors. Aaaah.
BTW this is cumbersome to simulate in SPICE, whose default assumption is that all circuit elements are at the same junction temperature. You gotta work hard to tell it otherwise. Simulation Program with Integrated Circuit Emphasis, after all....
**especially so in a power amp whose rails are many dozens of volts.
_
The Wilson mirror, and other cascoded mirrors, brings the Vce's down to the same level. Now the power dissipated in the two devices is roughly equal.
Heavy emitter degeneration makes the 2T standard current mirror less sensitive to deltaVbe (caused by deltaTj). Remember deltaVbe =approx= -2.2 mV per degree C. If deltaTj is X degrees C, why not drop (10 * X * 2.2mV) across the degeneration resistors. Aaaah.
BTW this is cumbersome to simulate in SPICE, whose default assumption is that all circuit elements are at the same junction temperature. You gotta work hard to tell it otherwise. Simulation Program with Integrated Circuit Emphasis, after all....
**especially so in a power amp whose rails are many dozens of volts.
_
Last edited:
........the best ways to improve accuracy a mirror is to reduce the early effect and reduce dissipation(heat) in the mirror devices. The other issue is peaking of the mirror response but this occurs at such high frequencies it is not much of an issue in our application.
The helper transistor works on the early effect and the cascode on dissipation.
The Buzzbomb currently uses a Widlar mirror which is the most primitive form of mirror, The Wilson mirror as used in the HPA-1 is a big improvement and will work well in the BuzzBomb if you want to reduce complexity.
Jam
The helper transistor works on the early effect and the cascode on dissipation.
The Buzzbomb currently uses a Widlar mirror which is the most primitive form of mirror, The Wilson mirror as used in the HPA-1 is a big improvement and will work well in the BuzzBomb if you want to reduce complexity.
Jam
Attachments
Last edited:
The warmer device is running at about 500mW, so its junction is going to be 9ºC above the case temp, giving a deltaVbe of about 20mV.
The current mirror is running at 11mA with 100ohms emitter degeneration, so I already have 50X.
However, in the Wilson current mirror the cascode is implementing feedback, so it doesn't need to match. And it's the only one that's getting hot, so it could go on the heatsink by itself, and the two standard current mirror transistors can even be TO-92s. Right?
The current mirror is running at 11mA with 100ohms emitter degeneration, so I already have 50X.
However, in the Wilson current mirror the cascode is implementing feedback, so it doesn't need to match. And it's the only one that's getting hot, so it could go on the heatsink by itself, and the two standard current mirror transistors can even be TO-92s. Right?