low quiescent current, 30v Vin, 5v out voltage reg?

[mab2]

Hi, having programmed a picaxe to control my solar tracker, and designed it to minimise power consumption, I now need to power it from the 24v battery (i.e. up to 30+ volts). The problem is that a standard reg like a 78L05 is still spec'd to draws quite a few mA, which kind of makes a mockery of my power saving efforts like putting the pic to sleep at night to save battery power.

I did look into a simple zenner, perhaps with an emitter follower transistor, but the 4.7v zenner I tried doesn't really hold it's voltage at <2mA.

having done a bit of a search for off the shelf low power regulators there are plenty, but they all seem to be low-dropout regs for battery power and the 'best' I've seen had a max Vin of 20v.

So the question is: Is there a simple way to get a stable supply in the range: 3.0-5.0v @ up to 2mA (though a few extra mA might be nice for future projects) from a 24v (up to 30v+) supply, with a low quiescent current (ideally <1mA)?

thx
m

 

[cct314]

Do you have a tap point on your 24 volt battery at a lower voltage - say 12 or even 6 volts - to which you may attach the low-dropout regulator?

If you are worried about unbalancing your battery bank, this may not be an issue: Some of the very low quiescent current regulators draw only a few 10's or 100's of uA - in the "noise" of the imbalance likely to occur between the cells of your battery system, anyway, and whatever controller you use, if sufficiently power-miserly need only draw a similar amount (or lower!) when asleep.

If you are still concerned by the operational power drain of the device unbalancing the system then one could actively switch using P-channel FETs, under processor control, to another, more conventional regulator operating from the 24 volt supply to get more current, perhaps a higher-voltage (15 volt) regulator diode-ORed with the low-dropout one: This would be done either during those short periods when one needed to wake up to perform some sort of management task, and/or the solar power budget was adequate for the task.

To go to an extreme - but it probably doesn't help with a PixAxe - if you have access to a point in the battery string where you can get just 3-4 volts after a diode drop you could run the PIC directly from that: A Nanowatt PIC would draw a negligible amount of power when asleep or ticking over at 31 kHz. Again, when full power was needed one could use some FET switches to connect to a higher-current regulator running from the 24 volt side.

 

[alec_t]

How about this?

With the component values shown you should get a fairly constant ~3.5V output for an input voltage from 5V up to 30V, at currents up to at least 1 Amp and for temperatures from -50C to +50C. Maximum no-load current is only ~250uA. The output voltage is the difference between the Vgs turn-on threshold voltages of the two FETs.

 

[mab2]

Hi, thanks for the ideas;

the battery's at the other end of the cable some distance away so tapping off at a lower voltage is probably not a solution i'd go for TBH.

I must admit I was hoping someone would know of a simple, low component count solution like a low QC equivalent of the 78L05, rather than having to actively control things and have multiple stages - otherwise it ends up being more complicated and power hungry than the old analogue tracker controller - surely I'm not the 1st person to want to run very low power, low volts off a 'high' voltage battery?

More to the point, I've already used up all my output pins (8pin picaxe 08M2 (12F1840 IIRC)) - so redesigning to control it's own power would be a faff.

I suppose I could run it of a 3AA NiCd and supply them from an old garden light solar panel or two but then that's another battery to die off and have to be replaced - seems a pity when there's 24v right there anyway.

cheers
m

 

[mab2]

oops, sorry Alec - I posted before seeing your post - just trying to get my head around your cct...

Hmmm- that looks just the sort of thing - in fact if it can provide an amp it's rather better than I was hoping for. I can see that they're both N fets; is there any reason why it wouldn't work OK with a couple of BS170's (up to a few mA) - I think I've got a packet of those somewhere...?

cheers
M

Found my bs170's - will give it a try...

 

[ronsimpson]

MIC2954-03

 

[mab2]

Aha! I knew it! that's exactly what I was looking for - but didn't find. Thanks.

(the BS170 version of your cct didn't work Alec-t - I think Vgs is too low on the 1st one to turn the 2nd on - but I think the concept could be tweaked to work with what I've got (saves waiting for an order of mic2954's to be delivered)...

cheers.

 

[blueroomelectronics]

Maybe Roman Black's (a member here) switching regulator might be an option.
https://www.romanblack.com/smps/smps.htm

 

[alec_t]

Here's an improvement on my first design, which doesn't depend on the difference between the two turn-on threshold voltages, so makes FET selection trivial.

The NFETs can be virtually any type with a Vgs(max) >30V. Mix and match. So BS170 should be fine (Idon't have a Spice model for that type).
The trimpot sets the output voltage. Output voltage is virtually independent of current over an input voltage range of 14-30V, and shows only ~5% change over a temperature range of -50C to +50C.
Max no-load current is only ~50uA.

 

[Tony Stewart]

This PMIC zener has a push pull driver up to 10mA and a quiescent draw of 1mA. Options are 5V or 10v.

http://cds.linear.com/docs/en/datasheet/lt1236.pdf

so choose LT1236ACN8-5
$4 (1)

Otherwise use a CC circuit with a HB LED using 3V drop for white at 1mA

Better yet forget about 7805 or even 78L05

use this LDO with <1 uA quiescent current
$2.26
10mA max.
**broken link removed**

Choose MIC5231-5.0YM5 ( SOT-23)

 

[mab2]

Heh! Spoilt for choice now.

I like those Roman Black SMPS (have bookmarked his page - thanks), but in this instance the power required at 5v (mA) makes me lean towards the linear options; I've added some micrel regs to my shopping list for when I next put an order in, but for now I think I'm going with Alec_t's improved nfet reg as I have all the bits to hand. After I'd turned off last night I figured out it was the identical Vth of the BS170's and had though of adding an led or two in series with vref fet as an experiment, but your solution is better Alec.

Many thanks.

M

 

[cct314]

The FETs that I had in mind were P-channel rather than BS170's, which are N - a bit of a sticking point, perhaps. Only the "high side" 24 volt transistor would need to be P-type, but if you weren't worried too much about wasting a tiny bit of base current a PNP bipolar could be used, instead! (You'd use an N-channel FET to drag the P device's base/gate down toward ground via the appropriate resistance value/dividers)

In the attached diagram the resistors in the FET version would be selected to prevent no more than the maximum gate-drain voltage from ever appearing across the FET (typically 20 volts) at the highest battery voltage- a more typical gate/drain voltage when the N-channel transistor at the bottom (which could be a BS170) was on would be something in the 10 volt range. The upper resistor needs to be there to make sure that the FET gets turned off when the N-channel is turned off as well. The typical values of the resistors would be in the several k-ohm to 10's of k range as little current need be consumed for this. Some values might be 22k for the upper resistor and 15k for the lower one, which would be "safe" to at least 30 battery volts for a FET with a 20 volt gate-drain rating. (If I've done my math right...)

In the PNP version the lower resistor would pull through the base just a few percent of the maximum expected collector current through the transistor for a small signal device with a beta comfortably above 100 (e.g. a 2N3906 - a value of 4.7k would be overkill and could easily handle a few 10's of mA) and in that case the upper resistor's resistance could be pretty high (say, 10k-100k) just to make sure that it was completely off when the FET was off.




The FET-based regulator by Alec_t looks clever, although I'd be interested in getting it hot and cold and putting capacitive loading on it to make sure that it behaved itself (e.g. voltage tempco, didn't oscillate...)



Having said all of that, the LT1236 looks like it could fit the bill!

 

[Tony Stewart]

Typical BJT switches need 10% of collector current on base to achieve spec'd Vce(Sat). More expensive super low Rce switches are rated for 2% of collector current, which tend to have hFE worst case>500.

just a FYI.

 

[alec_t]

I'd be interested in getting it hot and cold and putting capacitive loading on it to make sure that it behaved itself

Simulation shows it's stable with a capacitive load anywhere in the range 1pF to 1F, at temperatures between -50C and +50C. Tempco is ~2mV/degree C. Mind you, that's all just simulation .

 

[Tony Stewart]

There's no reason a logic level switch should be unstable unless there was coupling between the load supply (24V) and the logic supply (5V)

LDO's only become unstable at very low Iq and low Vi-o , when the sense voltage feedback ripple is suppressed by Cap ESR being excessive or too low, so specs show the acceptable range for Cap ESR. Higher Iq LDO's are inherently more stable but have inherently huge dropouts ( 2V) not Low at all.