Got a design here that contains a few 4060 long timers with RC oscillators. Battery operated so consumption counts. Some datasheets state that the maximum value for any resistor in that area is 1M. Since Rs (feedback resistor) must be at least twice Rt (timing resistor) this leads to rather lowish timing resistor values. Since this is a logic gate oscillator that requirement causes quite some current at voltages around 10V.
Anyhow, the TI datasheet does not seem to state that maximum:
... while ON-Semi does state a 1M maximum:
What gives? Why do some specify that max value as low as 1M for really low leakage CMOS logic? A remnant from the olden days when this stuff was leaky at times?
I usually use my own oscillators around CD40106 but this time those are all used up and there ain't no space no more :-(
The leakage current isn't through the CMOS gate, but through the gate- protection diodes. And the manufacturers can't afford to test their chips for long enough to be able to guarantee sub-microamp leakage currents.
The TI datasheet probably doesn't specify a maximum timing resistor because the TI marketing department was a crew of lying cheats back when the data sheet was written, in the early 1970's. For all I know, they may still be a bunch of lying cheats - the last time I fell foul of their sins of omission was in the mid 1990's, but they just might have cleaned up their act since then.
I'm pretty happy with TI, always was. Except once when they did a Texas style obsolescence on a chip. Aim -> pull trigger -> bang -> gone. The data sheet for the CD4060 seems to be a really old scanned copy.
I'm sure they know it's less but the automated testers can't go that low.
As for myself, I've used up to 10M with the '4060, and didn't feel too bad about it. But what I do worry about is the RC time constant. As the charging node approaches the threshold, small supply spikes, etc., could cause mishaps, as they shift the logic thresholds. Most '4060 chips consist of an inverter for the oscillator, followed by a second inverter, (for gain and to provide the rest of the oscillator), followed by a third-stage flip-flop or some other type of hysteresis stage. So there's plenty of room for count-advance trouble if the oscillator input's dV/dt is too slow.
I'd say TI has substantially cleaned up their act. I for one have certainly changed my opinion about TI, for the most part. Now I'm a dedicated customer.
I built a number ( 35-ish) of long-duration timers (20-24 hours) using the Philips 4541, which has a very similar oscillator. After mulling over the exact problem in the datasheet you note, I did some experiments with both Rs and Rt at 1M and Ct at about 1.4uF. (Actually
2 x Siemens poly 0.68 in parallel) It worked fine, and never caused any problems to my knowledge in the 10+ years they have been in service. I did note the time duration shifted by a little more than the specs suggest over the range 0-50 Deg C, but precision was not called for in the application.
I don't think you need to worry about Rs > 2Rt, I think that is just if you want their frequency formula to work. As long as Rs is "high" I expect it will still oscillate even if Rs < 2Rt, just at a different frequency.
Anyway I expect that most of your supply current might not be going into charging/discharging the cap at all, quite likely a lot of current is going through both the P and N channel FETs in the inverter where its input voltage slowly approaches the inverter's threshold, and both N and P channel devices are partly on.
To cure this, my only suggestions are to try not to make the supply voltage exceed the sum of the N and P channel threshold voltages by too much, or choose a different kind of oscillator.
You can determine (roughly) the sum of the N- and P channel threshold voltages by getting an inverter and tying its output to its input, and then feed the supply pin from a small current source or high value resistor. Basically an unbuffered inverter with output tied to input is just two diode-connected devices in series across the supply, and so with a really low constant current into the supply pin, the input or output pin of the inverter should be just over the N-ch threshold voltage, and the voltage between the positive supply pin of the chip and the inverter input/output will be just above the P-ch threshold voltage. Of course a "buffered" inverter with its input tied to its output will really contain three inverters and may not be stable. You might be able to make it stable for this experiment by putting a big cap to ground on the input and output pins that are tied together, thus hopefully making a dominant pole.
TI gives the maximum leakage as +/-1uA at 85°C. If we assume a threshold of 0.5*Vdd, the unit would stop working entirely with a
2.5M resistor and 5V, so the 1M is more-or-less within reason assuming you could allow something like 30% change in timing and given that the threshold can be different from 50%, and not necessarily in the direction that's favorable.
Using >1M on regular CMOS is living in unspecified land, but you may well be able to get away with it for a non-critical application. Note that there's a 4 order of magnitude difference between typical and maximum at 25°C, which should track over temperature.
Best regards, Spehro Pefhany
"it\'s the network..." "The Journey is the reward"
email@example.com Info for manufacturers: http://www.trexon.com
Then again TI states a max of 20M for the timing resistor, backing off to 10M at full VCC. Page 3-161:
And Rs (the front feedback) should be twice Rx or higher .... yikes.
What really surprised me though was the cross current of the first stage, peaking to 400uA at 10V VCC. Since this is an unbuffered inverter the oscillator spends quite some time in that area and the cross current ramp duration is almost 40% of the phase.
Yes, that's correct, except when the timing resistor gets to be much below 100K. In that case its contribution to the total is larger. But for several hundred k the cross current dominates and that's not going away no matter what, unless you lower VCC.
Yes, but that doesn't seem right. Then the front feedback resistor would have to be 40M or higher. This datasheet is a scanned version back from the times when data sheet values often had to be taken with a grain of salt.
If ultra-low power consumption is what you're looking for then the 4060 is the wrong part. The inverter is deliberately kept unbuffered so the user has the option of using xystal feedback for oscillation. The datasheet for the Philips HEF4060 has more detailed formulas for power consumption versus frequency as a function of Vdd.
Correct. And I normally never use such internal "one size fits all" oscillators. In this case we didn't have a choice. The circuit had to fit onto a postage stamp and already contained over 50 parts. The timers cannot share a common oscillator, not even a common chip, because one is watching the other to prevent grief if the main timer stalls for some reason.
Maxim has a battery monitor comparator optimized for ultra-low feedthrough current at threshold. It's just an itty-bitty part. You can use that as a relaxation oscillator, with much better defined threshold levels, to drive the 4060. Or knock that Vdd down to 2V and level shift the 4060 output back up.
Yours isn't exactly an easy name though (Hungarian origin?). Neither is mine, at least when it comes to pronouncing it.
Most newsreaders don't do well for specialty characters. Mine also showed a zero for the degree sign in your post. The really weird thing: It did show the correct degree sign in Jim's post but not in yours.