Oscilator

Bob Pease in his notes:

formatting link

on page 4 looks at few low frequency oscilators. The goal is to have lowest current draw. His best result with CD4007 was 22uA, and best overall was low power comparator which gave him

1.4uA.

I tried classic circuit from AOE (p. 426 figure 7.2) using

74HC04. I used 20M resistor as R1, 10M as R2, 1nF as C1 and 20pF speedup capacitor C2. With 1.5V supply current draw is 70nA frequency 21Hz, circuit works at least down to 1V. At 2V current is below 1uA but grow quickly with raising voltage to 9.6uA at 3V. Without speedup capacitor current draw is much higher, for example at 1.5V it grow to 1.2uA.

Some remarks:

- Pease works at 6V. Working directly at 6V it seems hard do better than he did. OTOH low-speed level shifter is probably doable at cost of 200nA

- current seem to go down with increased resistors. 20M is largest resistor that I have (and largest that I seen at resonable price)

- my measurement includes extra buffer gate but with no load. Adding mulitmetr in frequency mode (via a capacitor) increased current draw by about 30nA.

BTW: I measured static current for those HC chips and got 0.1nA with difference form 0 within measurent error. So one can hope for some really low power circuits.

--
                              Waldek Hebisch
Reply to
antispam
Loading thread data ...

The power consumption of low power oscillator fundamentally depends on the frequency you are working at - going faster pushes more current through the capacitance being charged up.

At a sufficiently low frequency the leakage current through the active comp onent driving the oscillator is going to become higher than losses in the p assive components fixing the oscillation frequency.

We had a thread here a decade or so ago about making a low power 10kHz osci llator. If all it's doing is oscillating you can keep it going with a coupl e of uA. The two example circuits were

1)a two transistor astable built with 5GHZ broad-band transistors, which ha ppen to have very low collector to base and collector to emitter capacitanc es 2)a LC oscillator with 330mH inductor with relatively low parallel capacita nce. The initial proposal called for a 500mH inductor, but didn't make any allow ance for its parallel capacitance - looking at the the inductors that we co uld buy off the shelf the 500mH part had too much capacitance to work at 10 0kHz, but the 330mH part would have worked.

Getting enough inductance for something that might work at a few Hz seems u nlikely.

--
Bill Sloman, Sydney
Reply to
bill.sloman

I don't have AoE III, only the 2nd edition, so cannot comment on that schematic. In general though the low-current issue with digital logic oscillators is the transition time high-low and low-high on the input. The slower the voltage change on an input pin, the longer will be the cross-current flow in the first stage of an input. That can be several uA or even tens of uA. Maybe that 20pF cap is used as a "kicker" to speed that up?

Nowadays one way to achieve low-frequency low-power oscillation is to use a uC in "near sleep mode" where an internal timer wakes it up periodically and the code switches the state of a port pin every time it does. Of course, that method looks very decadent to the purist :-)

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

For low-ish current draws (tens of uA) I think uP is the way to go, for extremely low current draws (single digit to sub uA) I think analog is still the way to go, it's finicky to get uPs into well and truly low power states, you might not even know what uP is best until the design is finished and you just have to try a bunch, there are more variables, in worst case situation might even be compiler-dependent, yuk.

Reply to
bitrex

A,

eased

t

I don't see any mention of FPGAs. They exist in the power domain you menti on, maybe not single digit uAs, but certainly 10's. The Lattice iCE40 Ultr aLite has a static consumption of 35 uA and a very good power vs. frequency ratio. For a frequency counter I expect the total power in use would not be much more than the static figure because only a very minimal amount of c ircuitry would be running at fast rates.

Rick C.

Reply to
gnuarm.deletethisbit

They're cheaper than I thought they would be, doesn't seem like a bad option at all on small to medium quantity projects. For large volumes a bare bones 8 bit uP or CD4007 is hard to beat on price

Reply to
bitrex

This frequency seem to be quite low. With 500pF (and the same other values) I get 69.8nA and 42.65Hz, with 145pF as C1 I get 71.4nA and

125.7Hz which confirms that switching losses are quite small. Replacing 20M by two 20M resistors in series, that is using 40M for R1 (and keeping 145pF as C1 and 10M as R2) I get 41.4nA and 68.28Hz. If I drop buffering gate I get 39.1nA, which again confirms that switching losses are quite small. As I wrote static leakage seem to be negligible and total power is essentially current flowing trough resistor of RC circuit. I other words, power savings are limited by availability of large resistors.

I wonder if the LC oscillator was just a project or a working circuit? AFAICS impedance of the coil is about 20kOhm, so to get small current one would need reasonably high Q. Also, one needs significant amplification at low current.

My guess is that for 10kHz crystal oscilator (assuming one can get appropriate crystal) due to high Q would be much better. AOE disscusses one for 32.768 kHz and claims 2.4uA for best combination working on 1.8V (and less for lower voltage).

Yes, at few Hz it seems that only way is RC.

--
                              Waldek Hebisch
Reply to
antispam

Strictly an LTSpice simulation, but with tolerably realistic components.

The series resistance of the coil is part of the published data. The Q of t he tank circuit was of the order of 100, IIRR.

32.768 Hz is the standard watch crystal frequency. Broad-range distributors (like Farnell/Newark/Element 14) offer lots of different part, but the cry stal is used in XY length-width flexure - usually called tuning fork.

formatting link

Lower frequency oscillator crystals do exist but might be tricky to get hol d of.

ms unlikely.

Or some odd sort of "crystal". The sensing mass in an integrated circuit ac celerometer might serve the purpose, if you could get at it.

If somebody were kind enough to offer a null-sensing accelerometer, with el ectrostatic force feedback to keep the sensing mass in one place, you might be able to bodge it into some kind of low frequency oscillator, though pro bably not one that was all that low powered.

--
Bill Sloman, Sydney
Reply to
bill.sloman

IIUC the schematic is also in 2nd edition, but without speedup capacitor.

Yes, exactly. AoE III says that capacitor speed ups transition to prevent oscilation, but apparently faster transition also helps a lot for current draw.

I am not a purist. In fact I am more software guy than hardware guy. But this was to better understand what hardware can do.

--
                              Waldek Hebisch
Reply to
antispam

Analog is also my preference but with a client we have done some amazing low power stuff using Atmel uC where a system had to wake up every few seconds, check on something and go back to sleep. We got into the sub-uA range. Waldek would need even less funtionality, just a port pin would need to change state.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Sorry, forgot a link I wanted to add:

formatting link

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Most of them are statically clocked (or whatever the term is) so they can be run down to arbitrarily low frequency, back in the day you could run stuff off a watch crystal or somethin' for low power consumption.

In modern times with the variety of sleep modes and power saving options to turn off peripherals I think what you did is the way to go, use the one of the stock clocking options like the onboard 8MHz say and stay in bed most of the time, then hustle as fast as you can to get the work done every so often.

Reply to
bitrex

While it works well it feels like cheating when doing it just for LF clock generation. Especially for an analog guy like me. Too many folks who'd turn around in their graves if they saw it.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I use 8 bit 8 pin AVRs like the ATTiny as general-purpose building blocks, there's a software suite made by these Russian folks:

which lets you "write" code using a GUI and "wire" interconnects and then it will automatically generate C++, compile it and punch it to the chip via a USB stick programmer or header on the board.

Makes a nice FPGA-substitute for low-speed stuff, e.g. say you need a chip that reads several voltages on the ADC ports and spits them out over i2c, that's literally a ten minute project from concept to uploading a binary to the device.

Reply to
bitrex

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.