Posted 24 February 2018

I have been working on Wall-E2, my autonomous wall-following robot, for almost three years now, and it seems like I have been struggling with the battery and charger arrangement for that entire time.   I started out with 4 AA batteries, but quickly moved on to a pair of Sparkfun 2000mAh ‘flat-pack’ cells with Sparkfun chargers, with a relay to switch the batteries from series (RUN) to parallel (CHG) wiring.   This worked, but not very well.   The flat-pack batteries weren’t a good match for motor control, and I kept burning up charger modules as well.   After struggling with this through several iterations, I finally abandoned it entirely in favor of a 7.4V 20C LiPo RC battery and an external charger. This worked much better, but forced me to manually disconnect the battery from the robot and charge it externally – not at all what I wanted.   Later on I made another run at the 2-cell series/parallel switching strategy for charging, this time using Adafruit Powerboost 1000C charge modules, each capable of 1A charge rates. Again this worked (actually quite well), but I recently discovered that it has a fatal flaw – this design imposes significant IR drops on the way from the battery terminals to the motors.

So, I have once again been searching for a solution to the battery/charger problem.   While wandering through the Googleverse the other day, I ran across a mention of the 1/2-cell TP5100-based charger module (about 9:50 from start), available for next to nothing on eBay.

Unfortunately, the available technical information on this module is also next to nothing, and what does exist is all in Chinese.   Still, this module has the potential for vastly simplifying my charger setup, so I thought it was worth the effort to perform a thorough study.

In a previous post, I described an Arduino controlled charge/discharge test setup for testing operation of my 2-cell parallel/series switched setup, so I decided to modify it for evaluating the TP5100 module, as shown below

View of TP5100 module showing RUN & CHG indicator connections

Charger test setup, in discharge mode (note 1.1A discharge current)

Charger test setup, in charge mode (note 1.8A charge current)

TP5100 Module Test Circuit

Using this setup, I was able to cycle the battery between a 7.5V load and the TP5100 charge module.   In order to keep the cycle times down to a dull roar, I set the software to switch to charge when the battery voltage dropped below 7.5V, resulting in the plots shown below.

In this case, the discharge current was about 1.1A, and the observed charge current was about 1.8A.  The TP5100 modules seems to work as advertised – with a 12V 5A power supply and a partially charged battery, it successfully charged my 2-cell LiPo pack terminated the charge at about 8.4V (I’m not sure if it is terminating based on current or voltage).

Over the next couple of days, I performed three complete charge/discharge cycles using this same setup.   Discharge was terminated at 6V, and charge was terminated when the TP5100 ‘complete’ output changed from open-circuit to active-low. As can be seen in these charts, performance was very consistent – almost 6 hours run time  into a 7.5V load, and about 4 hours for a complete recharge.

So here’s what I know now about the TP5100 module

• When the ‘1-cell/2-cell’ jumper selector is shorted to select 2-cell, the output voltage stabilized at 8.4 with a 10-15V DC input (I used a 12V 5A supply for the tests).   Below about 10V, the output voltage falls below 8.4V
• with a partially charged battery stack, output current was about 1.8A at the start, tapering to below 200mA at termination
• There is an onboard Red/Blue LED and solder holes for an external bi-color LED.   The onboard LED states are:
  • RED = Charging
  • Blue = Finished
• Both the onboard and external LED connections are tied to +V via the same 1K current limiting resistor.   This resistor is routed to the center hole of the 3-hole external LED breakout.   The rightmost hole is tied to an open-collector gate that goes LOW upon charge termination, and the leftmost hole is  tied to an open-collector gate that goes LOW upon charge initiation.   In my testing circuit above, these lines are labelled ‘Fin’ and ‘Chg’ respectively and were routed to digital inputs with 20K pullups on the Arduino UNO.
• This is  NOT a balance charger – so there may be differences in cell voltages over time.   If this is a potential issue, then separate cell protection modules like these should be installed.

Here’s an annotated photo showing the pertinent features:

So, it looks like this TP5100 module will work fine for my 2-cell LiPo application, with the addition of an external 2-cell protection module like the one noted above.   Not only will this solve my original IR drop problem, but it is  much smaller and simpler too, as shown in the following size comparison shot.   Oh well, at least I had a lot of fun building up and testing the original charger module ;-).

charger module and TP5100 size comparison

Stay tuned!

Frank

Posted 18 October 2017

Wall-E2’s on-board battery system has gone through some major changes over the last three years or so. Back in October 2015 I implemented an on-board battery pack & charger using two Sparkfun boost chargers and two 2000 mAh LiPo batteries as shown in the following image.

Battery pack showing charging modules and switching relay. The tan capacitor-looking component is actually a re-settable fuse

This was a technical success, but a practical failure; the LiPo cells just didn’t have the oomph to handle the high motor currents, and the single relay design had some problems as well.    So, in December 2015 I replaced the entire on-board battery system with a much higher discharge capacity RC battery, the GForce 2200 battery pack, as shown below

Current Wall-E2 battery pack

However, using this battery required that I physically disconnect the battery and connect it to an off-board charger for recharges, which meant I could never implement my idea of an autonomous robot.

So, about a year ago,  I re-implemented the on-board charging system, but with two significant changes; I used two 2-cell stacks of 3700 mAh 18650-style 3.7V li-ion batteries (specifically the NCR18650B cell) instead of the 2000 mA LiPo flatpacks, and two Adafruit PowerBoost 1000C chargers instead of the Sparkfun units.   I also added a second Axicom DPDT relay to solve the previous overvoltage issue.   The result is shown in the photograph below

Finished charging module connected to two 2-cell 3.7V battery packs

This system worked very well, and with the modifications  discussed in this post from last January (January 2017), I had a complete system for autonomous on-board charging.

This system has worked extremely well, but now I wanted to duplicate it for a planned upgrade to a larger robot platform, and I really wasn’t looking forward to hand-wiring another board. What I needed was a PCB, so I (or anyone else) could fabricate any number of charging modules without the PITA factor of a hand-wired perf-board implementation, so I decided to see If I could make a PCB using the free version of DipTrace.

After fumbling around for a while in DipTrace, I soon realized that in order to do a good job with a PCB design, I needed a component and associated PCB pattern for the Adafruit PowerBoost 1000C, and AFACT, none existed – at least not in a form compatible with DipTrace.   So, after some more fumbling around, I came up with the following model for the PB1K

DipTrace component model for the Adafruit PowerBoost 1000C

Creation of this component was complicated by the fact that I needed ‘flying lead’ connections to the ‘charging’ and ‘finished’ LED drivers on the PB1K module – signals that aren’t exposed to the outside world.   Eventually I hit upon the idea of adding a couple of ‘off-board’ pads (shown above at the lower left corner of the pattern), and assigning them to the ‘Finish’ and ‘Charging’ functions of the component layout. The ‘Chg Pwr’ signal already existed on the breakout header as the ‘Power’ signal, so I didn’t need anything for that.   These modifications allowed me to integrate the PB1K module into the charger schematic and the PCB layout using the normal DipTrace project work flow.

After getting what I thought was a successful layout, I clicked the ‘Order PCB’ button in the PCB editor, and ordered their minimum of two PCBs for $30 + shipping (I actually received 4ea PCBs so it was an even greater deal than normal!).

PCBs as received from Bay Area Circuits – nice packaging!

Four PCBs for the price of two – cool!

Test run to populate PCB with actual components

Size comparison between PCB and hand-wired versions. First iteration of PCB is slightly larger

After inspecting the PCBs, I realized I had screwed up on a couple of items.   The pad pattern for the two-wire terminal blocks wasn’t correct (too small), and at least one of them was reversed – oops!   In addition, a couple of PCB traces weren’t correct. So, it was back to the drawing board (literally) for a ‘version 2’ PCB.   On this iteration I learned a good bit more about DipTrace’s schematic/component/pattern relationship, so I felt like I was getting at least some value out of my screwups.   After removing a superfluous 2-wire terminal block and moving a couple of parts around, I was able to make the V2 PCB smaller than the original hand-wired model.   This time I remembered to use DipTrace’s verification routines, and this allowed me to catch some additional problems that had made it through into V2.   In addition, I used DipTrace’s print facility to print out a 1:1 PCB layout on paper, so I could double-check parts placement with real components, as shown in the following photo

Second iteration PCB printed at 1:1 scale on paper for component fit testing

So, many hours and a couple of iterations later, I have what I believe is a successful PCB implementation, and I sent off a new order to BAC.    As usual, I wound up spending  far more time designing the PCB than I ever would have in making a second hand-wired unit for the new robot, but where’s the fun in  that?

Stay tuned

Frank