Best low noise regulator?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Regarding again post #214, I find the performance to be measurably superior with the 2k resistor instead of J2. With a triangular 600Hz ripple on the input, the ripple rejection is 65dB with the JFET and 71dB with the resistor. This is measured, not simulated.

Note that this compares unfavorably with LM317. The performance deficit could be chalked up to breadboarding technique, perhaps. Frankly I'm surprised the patient even survived :hot:

Wideband output noise is impossible for me to measure. In simulation it's about 1nV/rthz.

Here's what happens when a fast glitch come in on the input rail. It passes right through with some ringing.
 

Attachments

  • shuntglitch.jpg
    shuntglitch.jpg
    92.4 KB · Views: 711
www.hifisonix.com
Joined 2003
Paid Member
Have you tried removing C1 and then repeating the experiment? I get the feeling that there is some inter dependance between the value of R4, C1 and J2 - i.e. the values have to be selected carefully. I read Jung's notes - I would not have guessed adding th e cap would have improved performance - looks counter intuitive, but, you can not argue with his measured results.

J1 and J5 are also interesting features on this circuit. What is the function of J1, rather than a higher value for R5?
 

iko

Ex-Moderator
Joined 2008
jwb said:
Regarding again post #214, I find the performance to be measurably superior with the 2k resistor instead of J2. With a triangular 600Hz ripple on the input, the ripple rejection is 65dB with the JFET and 71dB with the resistor. This is measured, not simulated.

We're very fortunate to get real results like these. I'm curious though, how do they compare in the critical region of 50 to, say, 180Hz? For those in North America 60 and 120 Hz will be very important, as is 50 and 100 for those how get that on the utility line.



Note that this compares unfavorably with LM317.

This would be surprising and disturbing, because the lm317 does not come anywhere near this regulator, as I have it working here. Perhaps I made a mistake somewhere when using the lm317.
 
We're very fortunate to get real results like these. I'm curious though, how do they compare in the critical region of 50 to, say, 180Hz? For those in North America 60 and 120 Hz will be very important, as is 50 and 100 for those how get that on the utility line.

Unfortunately because of the way I am generating the ripple with a 555 timer, I'm not at liberty to dictate the ripple frequency. The closest I can come is 160Hz, where the rejection appears to be 75dB (400mV RMS ripple on the input, 70uV RMS ripple on the output.)

I'm a bit confused. Is the upper trace the one with the resistor, or the one with the jfet?

In both scope pictures, the upper trace is the output at 50mV/div and the lower trace is the input at 500mV/div. Both circuits were using the resistor rather than the JFET.
 

iko

Ex-Moderator
Joined 2008
jwb said:


Unfortunately because of the way I am generating the ripple with a 555 timer, I'm not at liberty to dictate the ripple frequency. The closest I can come is 160Hz, where the rejection appears to be 75dB

160Hz ok. If I understand correctly, the 75dB is for the resistor version. If it's not too much trouble, would you please try this with a jfet (best of 7-8 mA Idss)? I need to know if I should throw the simulator in the garbage. I'd try this test, but I can't yet measure uV
:bawling:

Are you using the Dennis Colins noise amp for measuring uV?
 

iko

Ex-Moderator
Joined 2008
jameshillj said:
It's great to see all these nice simple ccts that do the job well and also sound good.

Perhaps the R4 could benefit from some more attention.

Add a small resistor (0.1 - 1 ohm) from o/p rail to M2 so you can actually measure the shunt current - won't effect the result.

IRFP9610, etc?

Have youb tried feeding this with a simple series reg (low ripple), or alternatively, with a Cmultiplier (high Z)?
Generally, good sound results for small number of extra components.

Did play quite a bit with the value of R4, if that what you mean. Good suggestion about the small resistor, will do on the prototype.

IRF9610 in simulation has similar results. I don't have a real IRF9610 to try yet.

I have tried to feed this from a series reg (lm317) and it seemed to me that the sound suffered. Since it's an easy thing to try, this is an option everyone should test and decide on their own whether it's worth or not doing.

Didn't try a Cmultiplier yet, but will.

Thank you jh!
 
ikoflexer said:


160Hz ok. If I understand correctly, the 75dB is for the resistor version. If it's not too much trouble, would you please try this with a jfet (best of 7-8 mA Idss)?

Using a BL grade 2SK170, the output ripple in the same test is 200uV, which is not as good as the plain resistor. Note that there's not enough headroom in this implementation for the FET to reach it's Idss. With the 2K resistor and a 12V input, it can only theoretically reach a current of maybe 6mA. Maybe I should lower it to 1k?

Are you using the Dennis Colins noise amp for measuring uV?

I'm using a Tektronix 2465A with the multimeter option. It's good down to 10uV.
 

iko

Ex-Moderator
Joined 2008
jwb, in these tests, were you using the exact circuit from post 214? It is important if we compare, to compare apples with apples. That includes the +30V input, zener, etc. If you're telling me that the the circuit in post 214 is flatly superior with a resistor replacing j2, it's significant. Most of all, salas wants to know :D (just kidding), I too want to know :)

Though I'm not ready to believe everything the simulator says, relative type of tests likely do make sense, even in simulation.
 
www.hifisonix.com
Joined 2003
Paid Member
I get the feeling that J1 and R5 actually do not do much in this circuit. The base of Q1 is sitting at about 0.6V above gnd. Flowing through R7 is the base current - which will be very small - but lets say the resultant drop across R7 is 50mV. So, the bottom of D8 and C2 are sitting at 0.65V above ground. Is the JFET doing anything here with just 0.65V across it?

However, you will have very high loop gain in this first stage because D8 and C2 are effectively coupled straight into the base of Q1.

Have you tried th e circuit without J1 and R5? Maybe the result is the same.

Normally, with this type of circuit, there is no J1 - just R5 which is sized for 0.6/Iz where Iz is the desired zener current -10mA is a good number so R5 would be 56 or 60 Ohms.

Watch out at turn on that the base current of Q1 is not exceeded - R7 is a good way to do this.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Bonsai said:
Is the JFET doing anything here with just 0.65V across it?


Have you tried th e circuit without J1 and R5? Maybe the result is the same.

.

Yes it does. The 2SK170 especially if a rather low IDSS one is selected has very low pinch off and it works as a CCS freely in a 0.65V margin.

Yes I have tried it. Without J1 the Zener cycle pumps at VLF. J1 locks the ZVref quite nicely in practice. R5 lifts it just a bit from GND and results in faster transient response. (jwb has shown a step test).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.