# Low power, low voltage, slow (0.1Hz) oscillator?

I had the opportunity to analyze a very interesting circuit today using a programmable unijunction transistor as a timer.

The power supply varies, and the circuit must operate on under 10uA of current (not counting charging the cap). It triggers an SCR every 10-30 seconds as long as the power supply is above 1.8VDC, and must operate across a range of 1.8 and 7.0 VDC.

The timing isn't critical - around 10-30 second intervals to trigger the SCR is fine (a short positive pulse). The lower the voltage, the longer the time interval is fine.

The kicker is the low current requirement (10uA or less), low voltage requirement (1.8V) and, as always, low cost (ie, replacing a 10 cent PUT with a 30 cent microcontroller would not be ideal).

What other options should I be looking at for a cheap, low current, low voltage, low accuracy timer design?

• That low current requirement is a real bind, otherwise I'd mention the good ol' 555 timer at 170µA@5V Commented Sep 16, 2011 at 20:46
• 555 timers are crude and all but obsolete. Cheap they may be, but there are more efficient ways to accomplish all of their functions. Commented Sep 17, 2011 at 7:08

Dynamic current calculation from datasheet. The RC will probably pull about 0.3 μA in addition to the Schmitt's static current.

• Oooh, I like this - I'll have to prototype it and see how it works out. With a little bit of work I might be able to get rid of the SCR, and the price for this part in quantity is the bee's knees. For those looking at this in the future, it's the NC7SZ14. Commented Sep 17, 2011 at 3:23
• Note: The Input leakage of the NC7SZ14 is +- 1 uA over it's temperature range. 1 μ A at 3 M$\Omega$ is 3V. You're probably better off with a much larger cap, and a smaller resistor. Commented Sep 17, 2011 at 6:19
• Yes, you need to test over temperature. 300kΩ / 330nF would work with correspondingly higher current draw. Also note that the NC7SZ14 is limited to 5.5V. The 1.8V - 7V voltage range is a hard spec to meet with CMOS gates like this. Commented Sep 17, 2011 at 14:18

I once did a micropower wakeup timer that took less than 1µA. I couldn't find the original circuit anymore, but this is pretty close to the topology as I remember it. It makes sense anyway, whether it's exactly like what I used before or not.

This version will draw a few µA, but the resistances are high enough that you shouldn't have to worry about board leakage and extra cleanliness.

During the timer off state, which is most of the time, all transistors are off. C2 slowly charges from the current thru R2. Eventually this brings the base of Q2 high enough to turn it on, which turns on Q1, which makes the collector of Q1 go high. This high-going edge does two things. First, it turns on the SCR. Second it temporarily turns on Q3. Q3 then dicharges C2 which turns it and Q1 off and the cycle starts all over again.

Transistors are poorly specified at these low currents, so you have to experiment. The output frequency of this circuit has a strong dependence on the supply voltage, but you said that was OK. I haven't tested this exact circuit.

• Be careful with part selection for the 47 μF capacitor. Some cheap electrolytics can leak a few μA, which would prevent this circuit from oscillating. Commented Sep 17, 2011 at 14:26
• @Markrages: Right, which is why I didn't specify a electrolytic. Commented Sep 17, 2011 at 16:14
• Just pointing out an obvious trap for the questioner, who is apparently rather cost-constrained on this design. (That cap will be the most expensive part of the circuit.) Commented Sep 17, 2011 at 16:51
• @Markrages: Yes. It doesn't have to be exact and R2 could be a little higher if a longer time constant is needed. I was thinking of two 22uF ceramic caps in parallel. It won't even see as much as 1 Volt and can vary with temperature, so the cheapest ceramic will be OK. Still, yes, that's a issue. Commented Sep 17, 2011 at 17:19

I'd look at using a micropower voltage detector/MCU reset chip (1-2uA) powered from an RC network as a relaxation oscillator.

• +1 -- I once used a 12C508 for something like this; initial gut feel was to use an HC4020/4040/4060 binary counter to divide down another clock, but the quiescent current of that chip is too high. Commented Sep 17, 2011 at 1:46
• This is my initial approach, but wanted to make sure I wasn't missing something more obvious. Commented Sep 17, 2011 at 3:10

Here's a single Schmitt gate - this one from TI - 74LVC1G125 - with claimed 10 uA max Icc across 1.65V - 5.5V - which is below your Vcc max spec of 7V. BUT the other one mentiuoned has the same limitation - so both are either unnacceptable or acceptable on this basis.

Cost is 6 cents in 3000 quantity from Digikey. Slightly less at higher volumes.

The NC7SZ14 datasheet is here Quiesecent leakage current appears similar to the above part.

EWhile Schmitt inputs notionally allow any input voltage without prpoblems it would be wortyh chercking whether Icc rises as Vi approaches the 1/2 Vcc area.