My observations agree with what you have stated here. When designing power supplies with multiple stages of capacitors, I use higher voltage rated caps upstream (where incoming voltage may have a higher degree of variation), and lower voltage rated parts closer to the output stage.
Does the simulated reference voltage get noticeably worse when you include realistic, nonzero ESR and ESL of the VREF capacitor "C7" ? Do ESR/ESL have any effect upon the simulated ripple rejection at 120 Hz?
Are you happy with the amount of non-DC that is present on the VREF node?
_
Are you happy with the amount of non-DC that is present on the VREF node?
_
Attachments
Last edited:
I had a similar thought last night which works well:
I even have some ~ 2mA Idss parts, but not enough to build all the different variations. J113 might be the ticket....
PS: I just looked at the spec sheet of the CR160. It's just a diode-connected JFET too, so it's actually exactly the same thought.
I even have some ~ 2mA Idss parts, but not enough to build all the different variations. J113 might be the ticket....
PS: I just looked at the spec sheet of the CR160. It's just a diode-connected JFET too, so it's actually exactly the same thought.
Attachments
Only if necessary, ideally just choose lower bandwidth parts if bandwidth is too high.
One nice thing about parts such as the CR160 series is that they are already bin sorted for current. The J113 will probably need source degeneration resistance to achieve the desired current.
Jam also mentioned a good technique to improve performance of the voltage source – which is to use an RC dropping network to supply power to the current regulator diode. Once you have that in place, you can also try connecting the emitters of Q5 and Q6 to that same filtered power node. The emitter resistors can probably be eliminated if you do this.
Jam also mentioned a good technique to improve performance of the voltage source – which is to use an RC dropping network to supply power to the current regulator diode. Once you have that in place, you can also try connecting the emitters of Q5 and Q6 to that same filtered power node. The emitter resistors can probably be eliminated if you do this.
Jam would have warned you about it eventually. So would the other commenters in this thread who offer ideas on how to reduce it. Personally I am fond of the two-zeners approach because a zener diode costs ten cents while a current regulator diode costs two dollars from Mouser. If you buy them on eBay, you pay less, but what exactly do you receive?
I first saw the two zeners approach in the John Curl / Mark Levinson ML-2 power amp. Nelson Pass used it fairly often too, back when he was designing with BJTs. I put a minor variant of it in "Mountain View," and it works quite acceptably well there.
_
Last edited:
