I don't know much about power supply rejection; the subject does not come up much in the discussion of circuit design. I also had trouble finding information (in this forum and elsewhere) about the calculation of PSRR from a theoretical standpoint given device parameters. Sure, I can simulate or measure it, but I don't quite understand the mechanisms controlling it. I see how it can be increased with feedback but beyond that I draw a blank slate when I look for the numbers.
I bring this up because I just finished a prototype active crossover which uses Borbely designed JFET buffers. I did not realize it before, but this circuit seems to have an extraordinarily low power supply rejection ratio. A quick LTspice sim indicated this ratio was around 15db in the audio range for each rail, for the entire crossover which consists of some filters and 2-3 of the buffers. I don't know how close that is to the actual figure, but it has been more important than I anticipated to have a good layout and decoupling.
I'm considering making some changes to the topology I use in order to decrease the dependence on the power supply. The problem is I have no idea where to start.
Can anyone point me towards some information that would help me understand this?
I bring this up because I just finished a prototype active crossover which uses Borbely designed JFET buffers. I did not realize it before, but this circuit seems to have an extraordinarily low power supply rejection ratio. A quick LTspice sim indicated this ratio was around 15db in the audio range for each rail, for the entire crossover which consists of some filters and 2-3 of the buffers. I don't know how close that is to the actual figure, but it has been more important than I anticipated to have a good layout and decoupling.
I'm considering making some changes to the topology I use in order to decrease the dependence on the power supply. The problem is I have no idea where to start.
Can anyone point me towards some information that would help me understand this?
Well, first start by reducing the need to worry about PSRR.
I.E., use regulators. A separate pair of regulators for each
amp is easy. the LM317/337 family (or the improved Linear
Technology versions) are quite good for the small cost One
can always get much fancier at much higher cost. But,
3-pin regulators make the issue of PSRR essentially moot.
I.E., use regulators. A separate pair of regulators for each
amp is easy. the LM317/337 family (or the improved Linear
Technology versions) are quite good for the small cost One
can always get much fancier at much higher cost. But,
3-pin regulators make the issue of PSRR essentially moot.
Kilentra said:
Douglas Self's book on power amp design has a chapter on PSRR. Some of the PSRR calculations are difficult to do by hand, since they depend on things like the Early voltage of BJTs for example. But he has practical info about how to design biasing for best PSRR, and tracks down the sources of PSRR in an example amplifier. Even though this is oriented toward power amps, the same ideas apply to low-power circuits. Except that for low-power, regulators are cheap and practical.
"Borbely designed JFET buffers" isn't very specific
Borbley's all fet "super buffer" circuit should have good psrr but it would require some effort to gather even aproximate fet models for all the types used with n/p idss matching
simpler resistively biased single ended fet buffers will have low psrr, try ccs bais
Borbley's all fet "super buffer" circuit should have good psrr but it would require some effort to gather even aproximate fet models for all the types used with n/p idss matching
simpler resistively biased single ended fet buffers will have low psrr, try ccs bais
I'm using the circuit of Figure 16A of Borbely's article on JFETs with 2SK170BL's and +/-24V supplies. The source of the follower JFET is connected to another JFET which functions as a current source modulated by the input signal. I am also using LM317/337 regulators with the adjustment pins bypassed by 120uf. The problem is most likely hum and noise pickup on the crossover board itself which is difficult to fix with external regulation. Local decoupling seems to help somewhat.
Does anyone know what the actual rejection ratio of this circuit shoult be? Perhaps my simulation isn't working right and I am really looking for a different problem.
Andy, thanks for pointing me to Self's book, I will try and find a copy (it sounds like something worth keeping).
Does anyone know what the actual rejection ratio of this circuit shoult be? Perhaps my simulation isn't working right and I am really looking for a different problem.
Andy, thanks for pointing me to Self's book, I will try and find a copy (it sounds like something worth keeping).
poor psrr is inherent in the "white follower" topology
Borbley’s own comments make me wonder why you choose the “white follower” version in fig16a – do you need the low output z and are you willing to pay ~4x increased distortion and very poor psrr for the low z vs the more common fet ccs biased follower of fig 15c
http://www.borbelyaudio.com/ae699bor.pdf
A quick sim shows the fig 15c fet ccs biased follower has ~ 40 dB psrr and cascoding 15c’s fets with a pair of properly selected fets raises psrr to 90 dB
in a crossover filter you should be able to measure the low (~60 Ohm) output R of the 15c follower and adjust filter series R to accommodate it
Borbley’s own comments make me wonder why you choose the “white follower” version in fig16a – do you need the low output z and are you willing to pay ~4x increased distortion and very poor psrr for the low z vs the more common fet ccs biased follower of fig 15c
http://www.borbelyaudio.com/ae699bor.pdf
A quick sim shows the fig 15c fet ccs biased follower has ~ 40 dB psrr and cascoding 15c’s fets with a pair of properly selected fets raises psrr to 90 dB
in a crossover filter you should be able to measure the low (~60 Ohm) output R of the 15c follower and adjust filter series R to accommodate it
Thanks for explaining. My use of the White (Borbely) follower is really just because of my inexperience and ignorance... to be honest, I just looked at Borbely's example preamp circuit and saw that this follower was used as a buffer. If I do put this circuit to a PCB for others to use I'll definitely switch topologies. In my own hand wired board I've gotten satisfactory results so I'm not worried. I am glad that I understand it a little better now.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.