Low noise current source (and possible bypass element)

[ricko_uk]

Hi,
I thought of the attached two-stage circuit to implement a low noise current source with minimal power dissipation over wide load and voltage input. (Schematic now updated with synchronous regulator)

Any critique of the circuit itself? Things to watch out for or improve especially when it comes to keeping noise low?

The idea is to have the linear regulator working as a 2A constant current source as well as rejecting the noise from the switching regulator. And for the switching regulator to take care of the high voltage drop (from 24Vin) and to keep the LDO input voltage always 2V above the output.

The load can be as low as 3R or as high as 20R. Obviously at higher resistance values the input voltage would max out and the current would not reach 2A but that is acceptable and a there would be an udible warning so the user can lower the load.

The reason for using that regulator is that it has a high PSRR cleaning any noise from the switcher. But for a higher current version (or to use a lower current regulator), I was wondering about any additional noise that could be introduced by adding a classic parallel power pass element across the LDO. Would a BJT introduce less noise than a MOSFET or viceversa?
And how do I get at least a rough idea of the additional noise introduced?

Thank you

[bdunham7]

I haven't evaluated your entire circuit--seems like a neat idea--but if you truly need low noise then keep in mind that the PSRR of the linear regulator is not as good at the switching frequency of the preregulator.  If it is 50kHz, you have about 40dB of PSRR.  And that's assuming the rest of your design doesn't exacerbate that somehow. 

[ricko_uk]

Thank you bdunham7,

Why 40dB? In figure 16 on page 14 of the datasheet and also in the tabular specs) it shows a minimum of 70dB from 0 to 3KHz: https://www.mouser.co.uk/datasheet/2/389/ld1085-1849674.pdf

Or am I reading the specs incorrectly?

Regardless, do you suggest to perhaps increase the switching regulator's frequency and also add a low pass filter before the linear regulator?

Thank you

[bdunham7]

At 3kHz, 70dB looks right.  What frequency does your switcher run at?  Isn't it 40-60kHz or so?  3kHz would require some big components.  How much filtering you'd need depends on your requirements. 

[ricko_uk]

I updated the schematic with the synchronous AP64350 which I can run at 500KHz.

I know that push-pull converters can generate sub-harmonics so I was wondering about this specific converter because, being synchronous, it has a push-pull like output stage (although I believe it works slightly differently). Can it generate subharmonics?

Also, are there any flaws with that circuit/solution especially when it comes to keeping the noise very low?

Thank you

[doktor pyta]

IMO
-As C23, C24 you should use good quality X7R or X5R MLCC capacitors. At 500kHz this is the best choice. The plus sign on the schematic suggest you were aiming for electrolytic/ tantalum caps.
-use ferrite bead + MLCC just before linear stage.
-take great care while routing the DCDC converter as this aspect can ruin the whole effort

[David Hess]

I was wondering about any additional noise that could be introduced by adding a classic parallel power pass element across the LDO. Would a BJT introduce less noise than a MOSFET or viceversa?
And how do I get at least a rough idea of the additional noise introduced?

The noise will be dominated by the noise of the bandgap reference inside the integrated regulator.

The reason for using that regulator is that it has a high PSRR cleaning any noise from the switcher.

Keep in mind that the PSRR falls with frequency.

[Terry Bites]

The output can be filtered prior to the linear regulator for low noise.
As ever put a CM choke on the input and output. This keeps CM guff from leaking out and keeps noise from circulating in the load.
If your ground scheme and power routing is good you can get sub mV noise.
A small cap across the load will shunt high frequency guff to ground.
A BJT power device won't add noise to any noticeable degree.
A MOSFET will limit your compliance voltage.
The balance between the load carried by the outboard bjt and the regulator can be used to make a stiffer source.
If the majority of the load is caried by the bjt it reduces the effects of changing voltage and temperature on the regulator.
If you need a heat sink you want to isolate the transistor (and regulator) ground the heatsink, use it as a ground plane.
Note that linear regulators in the CC mode can be paralleled for higher current.
Each regulator has its own independent control loop. They do not interact so their outputs simply sum.. Use a couple of LD1085s/ LT1085s  Preferable  solution I think.

You can get rid of the linear reg by using CV/CC switcher design. Just stick it CCmode.
see  TI's SNVA871 and SNVA829 notes.

[Benta]

Input: 24 V.
Buck converter.
Linear regulator.
2 A through 20 ohms is 40 V.
It doesn't seem to fit.

[ricko_uk]

Thank you all!
updating the schematic

@Terry, thank you also for the schematics. In you Clipboard_1.jpg schematic, C1, C2 and L1 form a filter, what is the function of R1 and how should I choose its value?

In the Clipboard_3.jpg schematic, to make best use of the CM choke on the output, shouldn't the ground connection be to the left of the CM choke (as per attached screenshot) or at least on both sides of the CM choke?

A MOSFET will limit your compliance voltage.

Do you mean my 2V across the voltage regulator that I control via the op-amp? If so that would be only for high current and quite high Rdson mosfets. So it sounds like you are referring to something else? Could you please explain a bit more.

You can get rid of the linear reg by using CV/CC switcher design. Just stick it CCmode.

Attached is a schematic for another project I'm working on. In that I have the the feedback pin controlled by the current through the load. I assume that classifies as CC, but I thought that solution will have higher noise than the schematic we are discussing for this project. But it sounds like that might not to be the case?

@David

The noise will be dominated by the noise of the bandgap reference inside the integrated regulator.

So am I correct in thinking that I could use a very low current regulator with even better PSRR than the current one and also a suitable large BJT bypass element and achieve the same or better performance?

@Benta,
that's correct, but that is acceptable because as mentioned in the OP an audible alarm informs the user to adjust the load accordingly.

Thank you all again!

[David Hess]

The noise will be dominated by the noise of the bandgap reference inside the integrated regulator.

So am I correct in thinking that I could use a very low current regulator with even better PSRR than the current one and also a suitable large BJT bypass element and achieve the same or better performance?

It would not make any difference if the low current regulator also used a bandgap reference, and if it operates the reference at a lower current, then the noise will be even higher.  The problem here is that bandgap references are just noisy.  Usually it does not matter, but you asked about low noise performance.

The reference impresses voltage noise across the shunt resistor which then becomes current noise.  To lower this noise, the reference can be filtered, which is generally difficult or impossible without access, or a lower noise reference can be used.  Some integrated regulators are lower noise than others, so that is one thing to look for.  The alternative is to use an external lower noise reference and operational amplifier to control the regulator or pass element.

[Haldor]

Most older linear regulator IC have terrible PSRR at higher frequencies.  They were designed back in the days of transformer based power supplies and ripple meant 60 Hz.

There are newer regulators that have good PSRR above 1 MHz. I you are planning on cleaning up after a buck or boost DC-DC converter, then pay attention to the switching frequency when selecting the linear to follow it.

Forget the PSRR blub on the first page of the data sheet, look at the PSRR vs frequency chart buried further back. If the data sheet doesn't have that chart, this is what is called a clue that you picked the wrong regulator.

TI has a decent presentation about selecting LDOs that goes into noise and PSRR.

https://training.ti.com/sites/default/files/docs/you_think_ldos_are_easy-detroit_tech_day_0_0.pdf