0

Powerboost 1000C damage protection?
Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.

Powerboost 1000C damage protection?

by paynterf on Fri Dec 08, 2017 9:48 pm

Hi,

I am using two of your PowerBoost 1000C units in a dual-cell LiPo battery charger, along the lines of the two-cell charger tutorial at https://learn.adafruit.com/multi-cell-lipo-charging/simple-balance-charger. However, instead of one charger charging both batteries in parallel, my circuit uses two PB1000C's. In RUN mode, the chargers are disconnected from the batteries, and the batteries are connected in series to the load. in CHARGE mode, the batteries are disconnected from the load, and one cell is connected to each charger. In the RUN mode, the power to both chargers is disconnected, but since this is accomplished by physically disconnecting the input power from both chargers (the robot backs off the charging station), there is a short gap (5-10 seconds typically) between the time the batteries are disconnected and the time charger power is removed. See http://fpaynter.com/2016/12/wall-e2-charging-station-design-part-v/ for the schematic and physical layout.

Your documentation mentions that the chargers should never be operated without a battery connected, and that doing so will damage the charger. I believe I read somewhere that this is due to the large current (voltage?) spikes produced by the charger's magnetics, which are damped out by the battery. So, this leads me to a couple of questions

    Is the damage instantaneous? It doesn't seem to be so, as I have been using the units for some time now, without problems (I think)
    Is there any other way to suppress the voltage/current excursions, like maybe a 5V zener across the battery terminals, or a strategically placed MOV?
    If a battery really IS required at all times, what is the minimum capacity battery that will do the trick? The reason I'm asking this is because I may be able to attach small 'sacrificial' cells permanently to the PB1000C - just for charger protection - while using the larger cells to run the load.

Any other thoughts or suggestions would be appreciated

Frank

paynterf
 
Posts: 37
Joined: Sat Mar 07, 2015 6:41 pm

Re: Powerboost 1000C damage protection?

by adafruit_support_mike on Fri Dec 15, 2017 5:44 am

First, we strongly suggest against doing what you're doing. LiPos have no clear line between 'charging normally' and 'about to explode', and putting random cells in series or parallel imposes stresses that often lead to catastrophic failure.

The balance-charger tutorial does several things to minimize those stresses, and is kinda for people who already know how to do the appropriate monitoring.

Regarding your questions about the PowerBoost 1000C design in general:

paynterf wrote:Is the damage instantaneous? It doesn't seem to be so, as I have been using the units for some time now, without problems (I think)

Failure isn't 100% guaranteed the instant you disconnect a battery, but a current spike at just the wrong time will destroy the board in milliseconds. The most common failure mode doesn't seem to be cumulative, just probabilistic.

paynterf wrote:Is there any other way to suppress the voltage/current excursions, like maybe a 5V zener across the battery terminals, or a strategically placed MOV?

Anything that suppresses the surge current will make the boost converter stop working, or at least dramatically reduce its efficiency.

The surge current is what charges the inductor's magnetic field with energy that turns into 5v output. For the PowerBoost 1000C, the field charges with an average of 150mA @ 3.7v for 60% of the PWM cycle, then discharges an average of 100mA @ 5v for 40% of the duty cycle. Those averages have to obey the inductor's exponential charge/discharge curve though, and both parts of the cycle go from maximum power to zero power.. zero current to maximum current for the charging phase, maximum voltage to zero voltage (relative to the output level) for the discharge phase.

Anything that lowers the peaks reduces the amout of energy the boost converter can use.

paynterf wrote:If a battery really IS required at all times, what is the minimum capacity battery that will do the trick?

You need enough to supply the surge current to the inductor at the at the output power of the load trying to operate from the 5v output. As a rule of thumb, multiply the output current by six. The PowerBoost's average input current is 150% of its average output current, and in boost converter design it's usually best to estimate the maximum surge current at four times the nominal value.

adafruit_support_mike
 
Posts: 54016
Joined: Thu Feb 11, 2010 2:51 pm

Re: Powerboost 1000C damage protection?

by paynterf on Fri Dec 15, 2017 12:59 pm

adafruit_support_mike wrote:First, we strongly suggest against doing what you're doing. LiPos have no clear line between 'charging normally' and 'about to explode', and putting random cells in series or parallel imposes stresses that often lead to catastrophic failure.

The balance-charger tutorial does several things to minimize those stresses, and is kinda for people who already know how to do the appropriate monitoring.


While I certainly agree strongly with your concern about screwing around with high-energy LiPo cells, I happen to have a MSEE, PhDEE, and 50+ years as a working EE with significant experience with battery charging and management. The batteries are all 18650's, and only two cells are charged in parallel (2 cells per PB1K, 2ea PB1Ks). They are switched to series to run the load, and the stack is never allowed to discharge below 3V/cell. The load itself is about 0.25C or less, so I'm pretty sure I'm well within the 'reasonably safe' operating zone ;-).

I'm still a bit confused about the 'Anything that suppresses the surge current will make the boost converter stop working, or at least dramatically reduce its efficiency' statement. Isn't that exactly what the battery connected to the terminal already does? IOW, if having a battery connected to the battery terminals protects the PB1K by suppressing the voltage spikes associated with current discontinuities through the inductor, how would the PB1K know if there was a battery connected or a 5V zener? Both would limit the voltage at the battery terminals. Granted, the efficiency would go down by the proportion of the energy going through the zener and not the battery - is that what you were saying?

Regarding the minimum capacity battery requirement: I'm not trying to power ANYTHING from the 5V output; in fact, I plan to disable the boost section of the circuit entirely by tying the ENable line to GND. All I'm trying to do is prevent the PB1K from dying during the time between the point at which I disconnect the batteries from the PB1K's to the point where USB power is removed from the PB1K's.

Does disabling the boost section (and thereby eliminating the current through L1) reduce/eliminate the damage potential, or am I missing something else?

TIA,

Frank

paynterf
 
Posts: 37
Joined: Sat Mar 07, 2015 6:41 pm

Re: Powerboost 1000C damage protection?

by adafruit_support_mike on Sat Dec 16, 2017 6:28 am

As long as you're familiar with the care and feeding of LiPos, go for it. Most of our customer base consists of people new to electronics, who don't know what's dangerous and are exposed to the collective wisdom of the internet.

paynterf wrote:I'm still a bit confused about the 'Anything that suppresses the surge current will make the boost converter stop working, or at least dramatically reduce its efficiency' statement. Isn't that exactly what the battery connected to the terminal already does?

Not exactly.

The MCP73871 LiPo charger has a load-sharing feature that routes external power directly to the load first, uses any excess current to charge the LiPo second, and draws current from the LiPo to cover any deficit if the external power source can't supply the whole load third. The chip wasn't designed to be the front-end to a boost converter, so there are operating conditions that have been shown to kill the internal switching network.

One condition known to cause damage involves trying to source all the power to the boost converter through the connection to the external 5v power source. Apparently the switches in that section aren't strong enough to handle the full load of surge current into the inductor. The switches to the LiPo seem to be stronger, and can handle the full load from the booster.

The idea that reducing the surge current would lower the converter efficiency is just basic switcher theory: the boost converter approximates a constant-power device over the long term (500mW out requires at least 500mW in), and is a constant-energy device within a single cycle of the switching process. The TPS61090 boost converter charges the inductor's magnetic field by closing a mosfet that shorts the inductor between the MCP73871's output and GND, and the energy stored in the field is the standard integral of Lidi during the time the switch is closed. Then the TPS61090 opens the switch to collapse the magnetic field, converting some of the stored energy to EMF while the switch is open. The amount of energy converted to voltage is proportional to the switch's duty cycle taken as a ratio.. for a 3.7v to 5v conversion, the switch stays closed about 60% of the time and open about 40% of the time.

The inductor's response to the switch is the standard exponential increase in current, and most of the energy used for the boost process arrives just before the switch opens, at the point of maximum current through the inductor. Dropping the peak current reduces the amount of energy in the field available for conversion to EMF, and that will lower the boosted output power. Boost converters like the biggest discontinuities they can get.

Any external components you add to the PowerBoost would have to live upstream of the MCP73871, and if you want to avoid killing the switch to external power, the components would have to be connected in parallel with the LiPo. That means they'd also be subject to the LiPo charging process, so you have to be careful about mixing cell charge/discharge curves.

adafruit_support_mike
 
Posts: 54016
Joined: Thu Feb 11, 2010 2:51 pm

Re: Powerboost 1000C damage protection?

by paynterf on Sat Dec 16, 2017 11:51 am

Thanks for the very informative reply.

So, if I understand correctly, the condition that is most likely to cause damage is with VUSB present, no battery connected, and a significant load on the +5V boosted output line, causing all booster inductor current to be sourced through the VUSB switch chain?

So, in the case of no load on the +5V boosted output line other than the purple LED, can one say that the switching ratio will be nearly 0% (almost always OFF)? with (hopefully) a lower chance for switch failure in the above case (no battery present, VUSB only)?

What about with the booster disabled via the ENable line? In that case, the +5V load should be identically zero (not even the LED load), and one would assume the booster's internal FET switch would remain permanently open (no current through the inductor), thereby eliminating that particular damage scenario?

Thanks for putting up with my incessant questions. In my experience, it is extremely rare (and refreshing!) to come across someone with such deep and detailed knowledge of a product as you obviously have. It certainly gives me a good feeling about other Adafruit products as well ;-).

Regards,

Frank

paynterf
 
Posts: 37
Joined: Sat Mar 07, 2015 6:41 pm

Re: Powerboost 1000C damage protection?

by adafruit_support_mike on Tue Dec 19, 2017 4:37 am

paynterf wrote:So, if I understand correctly, the condition that is most likely to cause damage is with VUSB present, no battery connected, and a significant load on the +5V boosted output line, causing all booster inductor current to be sourced through the VUSB switch chain?

Evidence from the field says the MCP73871 can blow without any load on the 5v output. Apparently charging the 100uF output capacitor from 0v to 5v creates enough initial load to cause problems.

It doesn't happen 100% of the time, and having a load on the 5v output does increase the chance of a failure, but we strongly suggest plugging in a LiPo before connecting a 5v upstream supply.

paynterf wrote:What about with the booster disabled via the ENable line? In that case, the +5V load should be identically zero (not even the LED load), and one would assume the booster's internal FET switch would remain permanently open (no current through the inductor), thereby eliminating that particular damage scenario?

Hmm.. interesting. I don't reccall seeing that combination before.

Your reasoning is good: with the TPS61090 shut down, the load on the MCP73871 should be negligible. In that condition, you should probably be able to detach and reattach a LiPo safely while the 5v supply stays connected. That's a lot closer to the MCP73871's intended use.

paynterf wrote:Thanks for putting up with my incessant questions. In my experience, it is extremely rare (and refreshing!) to come across someone with such deep and detailed knowledge of a product as you obviously have. It certainly gives me a good feeling about other Adafruit products as well ;-).

We're Makers.. our world view is built around understanding technology and modifying it to do what we want.

That's doubly true of our own boards. We enjoy helping people figure out our designs, and providing feedback when someone wants to make changes. It's an ongoing test of how well we know our own hardware, and we learn a lot ourselves along the way. That's how we think the game should be played. ;-)

adafruit_support_mike
 
Posts: 54016
Joined: Thu Feb 11, 2010 2:51 pm

Please be positive and constructive with your questions and comments.