Super Regulator

If I do the layout for the option with the 2N2222 and 2N3906 instead of GLED431 I believe either option could be assembled with different components and links (on the same layout). Should I just do the layout for the 2N2222 and 2N3906? Is that superior to the GLED431? So far I understand that noise and dynamic impedance are key. With the 2N2222 and 2N3906 I think the dynamic impedance is lower but the noise is a little higher. In this application I am not sure if absolute voltage really matters?

Similarly I could place a spot for a JFET and resistor next to the Cascode Current Source to leave that as an available option. I don't have many JFETs but I see that I can add BF245B to an order that is planned soon. BF245B is listed as 6 to 15mA IDSS on the datasheet. The other option is BF245A which is listed as 2 to 6.5mA IDSS

The 1.6Ohm rbb of ZTX851 (951) is certainly very nice. That is 1/10 of the 17Ohm rbb of my 2N4403. On a couple of my protoboards for references I paralleled three 2N4403 (in one case) and four 2N4403 (in another case). Also in my 60 dB LNA there are three 2N4403 in parallel. Perhaps 17/3 = 5.7Ohm will be sufficient for my purposes.
 

Attachments

  • Reference Options.png
    Reference Options.png
    287.9 KB · Views: 293
Ok.

I have attached an annotated layout from the DIYAUDIO store for the super regulator. It seems like the reference is running off of the positive sense line?

So I should not do this and instead run the reference and opamp directly off of the regulator output from the pass transistor?
 

Attachments

  • Reference Running From Sense Line.png
    Reference Running From Sense Line.png
    227.4 KB · Views: 290
I have also attached the schematic from the DIYAUDIO store. It seems like the reference is on both sense lines (Vout and Return sense lines).

So I should not do that? (Correct?)


While I was doing my layout I was using the DIYAUDIO store schematic and layout as my "good example" reference.
 

Attachments

  • Schematic Reference Connected to BOTH Sense Lines.png
    Schematic Reference Connected to BOTH Sense Lines.png
    105.8 KB · Views: 286
I'd really recommend the cap multiplier for the opamp anyway and you might as well at least allow for a link from its output to the Vref feed as well. There's not much in it between the 2N4403 and the 2SA1015 but the 2SA1015 appears to have slightly higher gain yet higher capacitances. On balance I'd go for the higher gain of the 2SA1015 as the capacitances aren't significant within a cascode.

John

Could you please explain where your capacitance multiplier is connected? Is it fed directly from the output of the pass transistor? And where are you grounding the capacitor of the capacitance multiplier?

I am having trouble understanding how the capacitance multiplier can be connected and used for both the opamp and the reference if the opamp & capacitance multiplier should not be run off of the sense lines yet the reference runs off of the sense lines in the layouts and schematics that I have found.
 
I would not run anything from the sense (and sense return) lines. Any current going through those lines degrades the regulator performance.

Jan

If possible please take a look at my attached schematic (first attachment). The red box surrounds the capacitance multiplier that powers the cascode current source for the GLED431 reference.

All of the above is run from the sense lines which is what I see in the article for the LM329 reference.

Am I doing this wrong? If I replace the LM329 with the GLED431 and the cascode current source should I still run them from the sense lines?

Should I not use the capacitance multiplier? Or should I connect it differently?


[The second attachment is the reference schematic that from the DIYAUDIO store.]
 

Attachments

  • Jan Question.png
    Jan Question.png
    84 KB · Views: 307
  • Schematic.png
    Schematic.png
    284.5 KB · Views: 289
AX tech editor
Joined 2002
Paid Member
I am not absolutely sure as I haven't tested this. But you are running some load current for the cap mux and the ref network from the sense line. That causes some voltage drops across the sense line so these drops will be directly reproduced by the regulator on the output.

In the original circuit, I made the tradeoff: running the ref from the output gives a chance that any ac on the output ends up in the ref voltage, while running it from the sense may cause a drop that ends up on the output as described above. In the end, because the ref load is pretty much DC only, it would only change the output with a minuscule DC value.

In your case with the cap mux and the involved ref, I don't know if the load is also pure DC or that there will be an ac component / noise load.

My gut says don't but it may not be a problem. If you use LTspice, you may want to look at the noise in the cap mux & ref load current.

But to be realistic: this are all very much extremely low effects which most probably make no measurable or audible difference. IF there are effects they may be at high frequencies where wiring induction starts to come into play; when that happens you'd see it on the stability and/or excess noise.
Still an interesting design question ;-)

Jan
 
Thank you for taking a look at it.

I suppose if it is a problem (and if I can trace the problem to the capacitance multiplier) I could just jumper it out. I am hoping that the cascode current source plus the GLED431 alone will be sufficient as a backup plan.

The capacitance multiplier was suggested by John so I have added it as an option. But I am not clear on where he connects his capacitance multiplier.
 
Member
Joined 2011
Paid Member
FYI here's a decent constant current source with high output impedance (== excellent line rejection and load rejection) and good temperature stability. Two resistors, a diode, and two TO-92 devices. Current is constant for all CCS voltage drops greater than 2.0 or 2.2 volts, see my lab measured results, below.

LM334 makes quite a nice 5.5 mA current source, works down to 2.0-2.2 volts


_
549171d1463243250-lm334-makes-nice-5-5-ma-current-source-2-0-2-2-volts-cascoded-png
 
I don't think the impedance of the cap multiplier is an issue as it is so much lower than the CCS that follows it. I initially used a cap multiplier in place of an RC filter to the opamp only as I had found this extra isolation of the opamp helpful. It was only then that I thought that it can do no harm and possibly be of benefit to use the same cap multiplier to feed both the opamp and the reference. I wonder what possible harm could it do apart from reduce voltage headroom?

As for loading the sense terminals, I have recently started to run additional wiring to the load alongside the sense leads to prevent the additional current flow in the sense leads from degrading the overall regulation. Much depends upon the location of the load. If it is all on a single PCB then the additional returns from the load are not an issue. Where distance and a larger wiring "loom" is an issue then it is better to feed any opamp/reference circuit cap multiplier directly from the force outputs rather than the sense, especially when they are so well regulated already.

The cap multipliers main role is the opamp feed which does have very slight current variations. The feed to the reference does not as it is via a CCS so it won't do any harm. The cap multiplier I have routinely used is a BC550c with 36k feeding a 10uF 50 Wima MKS (small) or a 10uF chip ceramic. I always use a base stopper resistor of circa 100 ohms. It directly feeds the opamp decoupling cap(s) and the the feed to the ref's CCS is taken from there.

John
 
I really like your two pin ref mock ups, very similar paths to mine. Would recommend you try the OPA1611 in place of the AD825 and try the cap multiplier on at least the opamp. I would be surprised if you didn't hear the difference. Also have you tried splitting your R3 and linking the join to the supply with say a 100uF cap??

john

I normally use the AD817 which is better, but in this case I had 4 pieces of the AD825 in stock. I'm not a fan of BB op amps(from some past times when I was still using an op amps on the audio signal path), I prefer AD and I think I'll try LT6200 to reg out up to 10V, for more voltages there will definitely be something from the ADA series. I currently have AD8091 in one WJ shunt regulator on testing and the regulator has beaten in performance ADM7150.
The AD8091 is a cheap high speed rail to rail op amp features similar to the AD817 so I will definitely go in the direction of fast op amps but with lower noise than the AD817 or AD825.
I tried to quickly insert into LTspice the idea with the capacity multiplier and I don't see any changes in the output. I think if the regulator works well that voltage must be 'flat' so that no interference should occur to the reference and the op amp. Any good op amp with a good PSR should be able to handle that, and for CCS I definitely recommend JFET and that even without a cascade it works great. The Mouser has SMD 2SK879 which corresponds to 2SK246 in TO92.
 
Member
Joined 2005
Paid Member
Is the 334 too noisey as a current source for this application? I use the Semitec E102 cascoded with the 2N5486 in 31 volt regulators (as per Walt’s current sources article), and a 2SK170 for low voltage applications where theres not enough voltage to cascode the current source. The 2N5486 doesn’t leave enough voltage for the higher current Semitecs - a E202 will measure 2ma idss on its own, but around 1.75ma when cascoded with the 2N5486. You can see from the Semitec chart what current to expect from the higher current jfet current sources with the lower voltage the 2N5486 leaves them with. Also the 2N5486 has a 25 volt limit, but worked well for my apps, or use a series resistor with it. Walt suggested the jfet /2N5486 current source worked best around 18 volts so its mainly for higher voltage applications.
 
Last edited:
Using Regulator at 5.2V for Rasberry Pi DAC.

I am reasonably confident in getting the reference to ~2.5V
I am trying to work out what OP Amp would be best in this low voltage application?
If anyone else has gone down this path. Please let me know.
Any suggestions are greatly appreciated.
 
Capacitor choice

I have been following the discussions on capacitor selection in this thread. However, not everything is clear to me. It is said that the Esr of the capacitor must not be too low for stability reasons, however, does this apply to all capacitors in the circuit or just the output capacitor?

Am I right in thinking that on the other positions (except the output) the lower the esr the better?

Are elna silmic II 100uf/25v capacitors a good choice (for output capacitor and on the other positions) ?
 
AX tech editor
Joined 2002
Paid Member
It is basically the output cap. That cap modifies the loop gain and if there is zero ESR you get too much phase shift and possible instability.

If you stay away from boutique film caps and use a regular elcap, not any that are advertised with very low ESR, you're most probably OK. I always use the Panasonic or Nichicon 'industry' series.

Exact numbers are hard to give because it also depends on the type of opamp.
Say between 50milliohm and 200milliohms is OK.

In the other positions it really doesn't matter too much, in theory the lower is better but the effects will be minimal.

Jan