Alt-BEAM Archive
Message #02037
To: "'Steven Bolt'" sbolt@xs4all.nl
From: Wilf Rigter Wilf.Rigter@powertech.bc.ca
Date: Mon, 5 Apr 1999 02:43:39 -0700
Subject: [alt-beam] Re: Better 74HC design
Wilf Rigter wrote:
> > the 74HC cmos oscillator could be made more efficient with
> > the addition of a 10M resistor at the input of the first stage
> > forcing a larger input voltage swing (2 times the power supply)
> > and longer time constant for a given RC.
---8<--- (snip that is)
> > I will test at different supply voltages and I believe there are
> > some other tricks to speed the transition through the linear
> > input region for ultra low power applications. I will report
back
> > with the results.
Steven Bolt wrote:
---8<---
> I look forward to reading about them. Note that 74HC micropower
> generally means a Vcc of about 2V.
> I had a look and was able to gain about 1uA. At 0.005, the
> pulse/pause ratio is already very small. The dominant factor is
> actually the chip manufacturer. With two of three I tried, the
> SunEater_II/IV approach results in a wasted current (not going
into
> the storage cap) of about 10uA. The third (Motorola) needed
> something like 36uA.
Steven Bolt is obviously very experienced with the low voltage/low
current applications of HC devices but it is a exciting new experiment for
me which I would like to share with others.
Even with the best choice of chip manufacturer some improvement can
be gained by adding the 10M resistor.
The experimental data below show the improved efficiency. So the
advantage of using a series 10M input resistor is real but justifying the
additional resistor depends on the operating voltage and the circuit
efficiency requirements. The experiment also shed light on the causes of
"linear" operating current. The basis for improved performance of the
oscillator is same as the sampled threshold detector: It is based on
increasing ratio of "digital" to "analog" time at the HC input. That is,
increasing the time a "digital" voltage present at the HC gate input
compared to the time the "linear" voltage is present.
The "linear" supply current occurs in HC logic when the input
voltage is between logic levels and this causes input circuit n and p
mosfets to turn on at the same time, allowing current to flow from +V to 0V.
Normally, these are short current pulses which occur during fast transitions
between logic levels and is referred to as "dynamic" current. Since this
dynamic current is the integral of the linear current pulses, it is
proportional to
the supply voltage (Vcc) as well as the frequency and rise/fall
times of the input waveform.
In BEAM Nv / Nu circuits, frequencies are low and transitions are
slow and "linear/dynamic" current pulses become "visible" instead of being
integrated as average current. This is particularly true of the voltage
comparator circuit where the input voltage duration in the linear region
approaches 100% of the time near the trigger point. Steven Bolt's innovation
was the use of an HC voltage comparator with a short analog sampling pulse
which decreased the comparator power dissipation by 3 orders of magnitude.
The sampling technique requires an oscillator which itself becomes
the dominant load with dynamic power dissipation during transitions when the
input is in the linear region. Steven Bolt's technique of using a Schmitt
trigger in a capacitive feedback CMOS oscillator improves the power
efficiency of this circuit compared to using linear inverters ie 74HC04.
Here are the results of testing a National 74HC14 CMOS oscillator of
SunEater_IV type described earlier by Steven Bolt with and without the
series input resistor (10M) at various supply voltages. The range of
voltages tested was from 5.2V to 1.2V. Below 2.0V, the operating current
drops almost to zero and there was no improvement using the 10M resistor.
The 74HC14 oscillator actually worked down to 800mV at which point the
current was less than 100nA.
The same frequency and pulse/pause ratio was used for all tests
although low duty cycle actually increased the current consumption of the
oscillator itself, with or without the series resistor. The sampling
threshold detector was not part of the test circuit since it is not directly
affected by the series 10M resistor.
Vcc w/o 10M (uA) with 10M (uA)
1.2 1.0 1.0
2.0 7.1 6.4
2.5 22.5 18.7
2.7 31.8 26.0
3.0 50.9 39.1
4.0 131 100
4.5 183 138
5.2 326 231
Interestingly, a 74C14 required only 10uA @ 3.0V and 1uA @ 2V
After the experiment, I am convinced of the benefits of operating
low-power HC BEAM type circuits at 2V or even 1.2V. I believe that for low
power 5-6V voltage HC applications, a 2V voltage "analog" core with 5-6V
digital outputs (like a Pentium 8^) would be a good solution.
regards
wilf
------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/list/alt-beam
Free Web-based e-mail groups by eGroups.com
Home