Q17 Sigma22 Regulated Ultrastable Ultralow Noise Linear Power Supply +-50/60Vcc

[stef1777]

The simulation is almost over. I finally managed to create models for the E502/E-562 which seem quite faithful. I still have a few things to fix and I will upload it to my Github repository.

Stef.

 

[Moor]

Ok, great work ! ... TL431AB is SMD ... would it be really necessary ? it should have the same capabilities.

 

[stef1777]

It's just for simulation. The AB model is available and better. This remains the TL431B on the diagram and PCB.

I uploaded a new diagram to the Github with the real symbol for the CRDs.

 

[stef1777]

Voltages at 100mA and 3A.

Take with a grain of salt because the CDR models are “in-house”.

 

[stef1777]

Hello,

I uploaded the LTSpice simulation on my Github repository.

Total RMS noise on the simulation: 293.8nV from 1 Hz to 100 KHz at an output of 51 Vdc. These are of course "theoretical" measurements due to the fact that in a simulation the components are "ideal".

I'm waiting for new PCB version 1.1.1.

Stay tuned.

Stef.

 

[stef1777]

Hi Moor,

I tried the circuit with a single TO-247 MOSFET per channel (irfp140n and irfp9140n) and without 0.47R. The noise is higher (400nV instead of 293.8nV with 2 MOSFETs). I would have to try other TO-247 models to see if it is because of these particular models or if it is general.

Stef.

 

[Moor]

That is interesting ! Two mosfets worse than four ... ? hard to find an explanation. Very curious.

 

[stef1777]

This does the same thing with a pair of good IXTH80N20L and IXTH48P20P.

So confirmed, it is less good with a single pair of MOSFETs. Well, we're quibbling because it's already very low.

Stef.

 

[Moor]

Good to know this !

 

[stef1777]

Will the circuit work? The suspense is unbearable.

Stay tuned.

Stef.

 

[stef1777]

WORKS !!!



A pleasure to solder. The chips underneath are super easy to solder.

Just powered on without load. I'll let it heat up for now.

Max output with 60V input without load: 54.9V

 

[Moor]

excellent ! job well done!

 

[stef1777]

With a load of 50mA, the voltage rises to 55.07V.

55.07 on the positive rail and 55.06 on the negative rail.

The regulator takes 4 seconds to reach 50V and 11 seconds in total to reach 55V.

At 50mV everything is cold of course.

I'm going to bring out the heavy artillery soon to torture her.

 

[Moor]

ok ... the delay is normal, because you doubled the capacitors from the bases and emitters of Q1 and Q2. But this in turn offered a lower ripple in the simulation, so everything seems very good until now.

 

[stef1777]

I'm going to have to change the ignition delay of my amp.

White sugars are not all at the same temperature. Ideally, you would sort the MOSFETs into pairs.

This consumes a little more current on the negative side.

We will see what the high power tests give.

 

[Moor]

before changing the capacitors, first do a power test ... and measure the ripple. It is good to know it before going back with smaller capacitors

 

[stef1777]

I don't have a 220uF 80V so no test at 3A at the moment. I was supposed to order some but someone bought all the stock at Mouser yesterday.

 

[Moor]

so you tested with 100uF ?

 

[stef1777]

Yes.

There remains this one in 220uF but 5 mm higher (10x25:5).

661-EKYB800E221MJ25S

I will quietly do my tests up to 1A afterward we will see.

 

[stef1777]

Hello,

Kicad files and Gerber files uploaded to the Github repository.

Have fun.

Stef.