3.3V "Battery step up converter" and "timer enable" not fully powering up and starting

Thread Starter

Mahonroy

Joined Oct 21, 2014
287
Hey guys, I created this circuit that takes a single coin cell battery (panasonic lithium cr2477) and run it into a AAT1217ICA-3.3 which boosts the battery voltage to 3.3V, and maintains the 3.3V as the battery drains. The AAT1217ICA-3.3 also has a SHDN pin (high keeps the regulator running, low shuts down the regulator). I have a timer IC (TPL5111) connected up to the AAT1217ICA-3.3 SHDN pin.

This works by on initial startup, and every 3 minutes, the timer activates the AAT1217ICA-3.3 SHDN pin. The microcontroller does some work, then signals its done to the timer. The timer will then deactivate the AAT1217ICA-3.3 and wait 3 more minutes before re-activating it, and this is an endless loop.

If I remove the battery, and connect up a 3.3V power source, everything works just fine; 3.3v rail is regulated, microcontroller does work and signals to timer its done, and timer shuts down regulator. However... If I try to take a current draw measurement with a volt meter (by intercepting the power wire), the regulator won't start up and it continuously resets. If I use the coin cell battery like normal, I get the same problem... the regulator won't start up and it continuously resets.

Here is my circuit schematic:
schematic.jpg

Here is monitoring the lithium coin cell voltage right at the battery (continuously resetting):
battery_voltage.jpg

Here is monitoring the timers DRVn pin that activates the AAT1217ICA-3.3 SHDN pin (continuously resetting):
timer_drv_pin.jpg

Here is monitoring the 3.3V rail (continuously resetting):
3v3_power_rail.jpg

If anyone could offer any help or advice on how to troubleshoot this thing, it would be greatly appreciated, thanks!
 

ebp

Joined Feb 8, 2018
2,332
Increase the value of C15. It might take quite a lot of capacitance. It should be "good" capacitance - low equivalent series resistance (ESR) and low leakage current (to avoid wasting power from the cell). Polymer tantalum may be worth considering, though I don't know how good they are in terms of leakage. Ceramic are "best" in terms of ESR and leakage, but also larger for the same CV product. Do not consider Zxx or Yxx ceramic types - they are HORRIBLE!

Switchers have a negative resistance (slope) input characteristic. If the input voltage falls, the input current will rise. If the input power source is sufficiently low impedance (that is, it can deliver more current without a significant reduction in voltage), which the source cell most definitely is not, this isn't a big problem, usually. With high source impedance, the negative input resistance will "flatten" the input - which is exactly what you are seeing. You can't do much about the cell's impedance, so your only hope is to use sufficient capacitance to keep the converter running until the output voltage is up to where voltage regulation begins.

There appears to be a 12-13% overshoot in the 3.3 V rail at startup. I would regard that as intolerable, but I don't see any way of doing anything about it.
 

crutschow

Joined Mar 14, 2008
23,522
The reason it doesn't work with the meter in series is the internal resistance of the meter when measuring current.
Try a higher current scale and see if that helps, since that will reduce the series resistance.
 

Thread Starter

Mahonroy

Joined Oct 21, 2014
287
Increase the value of C15. It might take quite a lot of capacitance. It should be "good" capacitance - low equivalent series resistance (ESR) and low leakage current (to avoid wasting power from the cell). Polymer tantalum may be worth considering, though I don't know how good they are in terms of leakage. Ceramic are "best" in terms of ESR and leakage, but also larger for the same CV product. Do not consider Zxx or Yxx ceramic types - they are HORRIBLE!

Switchers have a negative resistance (slope) input characteristic. If the input voltage falls, the input current will rise. If the input power source is sufficiently low impedance (that is, it can deliver more current without a significant reduction in voltage), which the source cell most definitely is not, this isn't a big problem, usually. With high source impedance, the negative input resistance will "flatten" the input - which is exactly what you are seeing. You can't do much about the cell's impedance, so your only hope is to use sufficient capacitance to keep the converter running until the output voltage is up to where voltage regulation begins.

There appears to be a 12-13% overshoot in the 3.3 V rail at startup. I would regard that as intolerable, but I don't see any way of doing anything about it.
Thanks for the info! I soldered a 22uF ceramic capacitor ontop of C15 which was already a 10uF ceramic capacitor and this seems to fix the problem. I then desoldered the 22uF and tried adding a 10uF instead and this was not enough capacitance and the regulator could not start.

Do I just replace this with 33uF capacitor and call it a day (or 47uF to be safer?), or is there a way to determine what the proper value should be? The datasheet said all that was needed was 4.7uF so I'm a bit confused. Thanks again for the help!
 

ebp

Joined Feb 8, 2018
2,332
Datasheet circuits are typically typical, so they don't always apply. The caps shown at the input of switchers are often based on the assumption the source impedance is quite low and the cap isn't going to have to do a whole lot.

Unfortunately, there really isn't enough detail in the datasheet to make a good estimate of the capacitance required. What I would suggest is that you do some testing - and that won't be especially easy. What you need to do is test startup with a cell that is getting toward the end of its useful life and monitor the voltage at the input to the converter with your scope, looking at the minimum voltage during startup. Compare that with the specified "maximum minimum" startup voltage spec on the datasheet. By "maximum minimum" I mean the highest specified voltage at which the circuit is supposed to be able to start - there may be minimum, typical and maximum values e.g. (values from thin air) 2 V, 2.3 V and 2.5 V. You would use the 2.5 V value because it is the "worst case."

Having done that, I would chose a capacitance of at least a few tens of percent higher than what seems to work reliably. Tantalum caps are pretty good in terms of aging. Ceramic caps can be funny things. They can slowly age to a value considerably below nominal (which is reversed if they get hot enough, but then they age again) and most of the high capacitance types have a moderately large negative voltage coefficient of capacitance. For example, a 10 µF X7R rated at 6.3 V might be down to 8 µF at 6.3 V, but perhaps only down to 9 µF at 3.3 V. The reason I said to stay away from Zxx and Yxx types is that their voltage coefficient of capacitance is huge, and unless you run them at well below rated voltage they can be down to 25% or less of nominal value. The good thing about modern surface mount ceramics is that high values are very compact and remarkably inexpensive, relative to what through-hole types used to be. Tantalum don't suffer from the voltage coefficient problem and have low temperature coefficient, but the leakage current is higher than that of ceramics.

Don't forget to check the cell impedance curves to evaluate effects of temperature over your required range.

Using two cells in parallel would probably yield considerably more than double the operating lifetime, but there is the space & weight issue.
 

Thread Starter

Mahonroy

Joined Oct 21, 2014
287
Thanks again for the info! I had one more question if you guys don't mind. You mentioned the voltage overshoot on the 3.3V rail, so I measured it and there is about a 1-2 millisecond voltage spike of 3.7/3.8 volts before it stabilizes at 3.3 volts. What do you think might be a good way of resolving this without having to mess with a TVS or zener diode? Think adding extra capacitance on the 3.3V rail might solve it? As it is currently, there is a total of about 21uF worth of capacitance on the 3.3V rail.

Oscilliscope showing the 3.3V rail come online, microcontroller does work and signals done (roughly 60ms worth of work), then you can see the 3.3v rail go to sleep:
3v3_power_rail_fixed.jpg
 

ebp

Joined Feb 8, 2018
2,332
I see no easy mechanism for dealing with overshoot with that part. Usually overshoot results because of lag in the control loop - it doesn't throttle back the duty cycle fast enough. Often a problem at startup like that is dealt with by using "soft start" - relatively slow increase in the duty cycle to allow frequency compensation capacitors in the error amplifier to get charged up to normal operating range. But there is no soft start control available and there are no external compensation caps.

One thing I rather wonder about is that the output switch FET is between the of the IC and the output filter capacitor. I wonder if there is something odd going on due to that, although I see the circuit is much the same as an ap circuit in the datasheet. I would be tempted to try adding at least some capacitance right at the output of the IC - perhaps 1 to 5 µF (totally guessing). It might make things better - or worse.

The problem with adding a lot of capacitance at the output is that it works against you with regard to the start-up problem. Until the output cap is up to operating voltage the input current of the switcher will be high and drag down your input voltage for longer.

The overshoot may not be a problem if it doesn't exceed the absolute maximum operating voltage for the parts in the system. Zeners and TVS diodes are difficult in that voltage range because they tend to start conducting below nominal voltage.
 

Thread Starter

Mahonroy

Joined Oct 21, 2014
287
So I left this thing running over the weekend to measure battery life... sure enough when I checked on it Monday morning it was constantly resetting again... the voltage regulator could not start back up. It looks like as the battery drains, there is a certain point where the problem comes back. So as a test I touched a 100uF electrolytic capacitor across the battery terminals, and it started right up and worked normally again. So I figured it just needs a bit more capacitance to make it work at lower voltage levels. I had to add about an additional 80uF to make it start working again, for a total of about 140uF. Does this seem right? I found a 220uF, 4V, X6S (1206 package) that I could probably make work (still 60 cents each at quantities of 1000), but it just seems excessive doesn't it?
 
Last edited:

ebp

Joined Feb 8, 2018
2,332
I've never even heard of X6S before - and I've looked at an awful lot of capacitor data sheets over the years! Of course each of those characters has long-established meaning, Check the data to verify that it doesn't suffer from excessive negative voltage coefficient of capacitance or temperature coefficient issues that might be a problem.

It really is hard to predict behavior with something like this because of the unspecified behaviors of the converter and the cell. Once the cell has been significantly discharged, the cap is probably supplying a very large fraction of the transient current needed to get the converter running. The negative input resistance characteristic of the converter really works against you.
 
Top