High efficiency DC boost converter for very low current output

[davegravy]

I'm looking for the best method to boost the output from a single cell Li battery to about 24V. I only need roughly 2mA output @ 24V, but I want the highest efficiency and lowest noise possible.

The application is a constant current supply for a microphone sensor (IEPE) similar to these:
https://sound-au.com/project134.htm

I've been looking at switching boost converters and while some boast 95% efficiency it is only at higher output currents like 100mA. I found some that have high efficiency at 1-2mA but their max output is well below 24V.

What type of converter topology should I be using and are there any specific model recommendations?

[Kleinstein]

2 mA is not realy low current. 2 nA would be very low current.
For a 3 to 24 V step a simple boost converter is already borderline, with a rather short current pulse at the output.
The input side current would already be in the 15-20 mA range and thus well in the range some of the smaller SMPS chips are made for.
With modern chips much of the noise / EMI depends on the layout. A relatively high switching frequency makes filtering a little easier, but needs more case with the layout.

A capacitive multiplier tends to produce a low of supply ripple from current spikes - here 2 mA are allready a rather high current. This would be an alternative to consider for some 10 µA or less.

The constant current regulation would already offer quite some noise filtering / regulation.

For good efficiency I would consider an adjustable voltage, so that the voltage is not much higher than actually needed.  Chances are the 24 V are by far not needed.

[tooki]

2 mA is not realy low current. 2 nA would be very low current.

Disagree. ”Low” is an inherently relative term, if not expressly defined in some standard. It’s not here.

But I think that in the context of power supplies in general, and in everyday electronics, 2mA would absolutely be considered “very low”, and 2nA might be described as “ultra-low”. I mean, 2nA so little that most multimeters can’t even begin to measure it (a typical handheld DMM can’t even resolve single-digit μA).

[NiHaoMike]

Perhaps have the switching synchronized to the ADC? Since the load is rather predictable, a fixed duty cycle along with an overvoltage cutout should do.

[davegravy]

Thanks all for the replies, I am still digesting them.

For good efficiency I would consider an adjustable voltage, so that the voltage is not much higher than actually needed.  Chances are the 24 V are by far not needed.

I think what I'm hearing is I should take a step back and look more at the current source circuit first, and be sure 24V supply is really a requirement.

I'm reading this paper:
https://iopscience.iop.org/article/10.1088/1742-6596/1804/1/012161/pdf

I don't understand a lot of it yet but it seems the gist is a high performance current source can be made with a reduced voltage supply.

[Marco]

I'm looking for the best method to boost the output from a single cell Li battery to about 24V. I only need roughly 2mA output @ 24V, but I want the highest efficiency and lowest noise possible.

Simply current source wise these are if perhaps not conflicting requirements, but they are complicating requirements.

The simplest low noise high side current source would be a capacitance multiplier followed by a two PNP transistor current source ... but that needs around ~3 diode drops (actually two diode drops + Vcesat minimum) so that's 7% down the drain from the start.

There are current sources with lower voltage drops, but keeping it just as low noise and high ripple rejection as that extremely simple circuit is not necessarily easy. How about designing for low noise and reasonable efficiency? Then you can use the above current source and shoot for say 80% efficiency overall (with a boost converter which keeps the supply voltage to ~2V above the current source output voltage).

Current loop sensors sacrifice efficiency from the start to begin with, so what's another 10%.

PS. unlike in your link you really need better ripple rejection, because of the switching power supply, that's why the capacitance multiplier.

[NiHaoMike]

Current loop sensors sacrifice efficiency from the start to begin with, so what's another 10%.

Isn't a 2 wire active microphone always a current loop? The port powers the microphone using a voltage source connected through an impedance and the microphone outputs its signal by varying the current it's drawing.

PS. unlike in your link you really need better ripple rejection, because of the switching power supply, that's why the capacitance multiplier.

If the switching frequency is synchronized to the ADC, ripple that feeds through just becomes an offset that can easily be canceled out. Thus the ripple only needs to be attenuated enough to not cause the amplifiers to clip.

It's also worth noting that a boost converter in DCM is a current source, hence why a fixed duty cycle would work for this application. The only additional parts needed would be some way to handle the load being disconnected (overvoltage cutout or zener clamp) and something to decouple the low impedance of the output capacitor so that it appears as a high impedance to the signal line.

[SiliconWizard]

You can have a look at LT's offering (now Analog), for instance: the LT8410.

[Marco]

Current loop sensors sacrifice efficiency from the start to begin with, so what's another 10%.

Isn't a 2 wire active microphone always a current loop?

Which is never efficient when a large part of the voltage is needed for a relatively high voltage bias for the condenser, the amplifier doesn't need all that voltage.

It's also worth noting that a boost converter in DCM is a current source

A really choppy one, goes to zero, with an unusable impedance. Regardless of ripple an extra active current source is necessary.

Not sure letting a large ripple through would be acceptable, even if IMD isn't large enough to be relevant it doesn't feel hifi

[davegravy]

Isn't a 2 wire active microphone always a current loop?

Which is never efficient when a large part of the voltage is needed for a relatively high voltage bias for the condenser, the amplifier doesn't need all that voltage.

Not sure if this has any actual bearing but the sensor device I'm targeting is this prepolarized one.  It calls for a bias voltage of 20V min however I wonder if that's only necessary to use the full dynamic range of the mic. I don't need to support SPLs above 110 or 120dB so I'm hoping it will still function normally at lower bias - I plan to test, I suppose it might provide no useable output until min reqs are satisfied.

I searched high and low for example circuits of IEPE implementations in microphones and came up dry - appears to be a closely guarded industry secret. I did find this circuit note for an IEPE accelerometer however which looks to use shunt regulators.

[Marco]

In one of their documents they still mention current limiting diodes, so the internal circuitry is probably going to be some discrete JFET amplifier.

It probably does have ~10V shunt regulator with a ~200 Ohm series resistor at least, given the 10V@2 mA and 14V@20mA biasing. But if the +/-7V peak to peak output is accurate the amplifier is actually "fed" from the full bias voltage, not the output of the shunt regulator.

[Terry Bites]

or LT1615