Upgrading a passive Power over Ethernet splitter with 802.3af compatibility

As seen on Hackaday!

If you haven’t heard of Power over Ethernet, chances are you’ve experienced its usefulness without even knowing about it. Power over Ethernet (PoE for short) does exactly as the name implies: power is sent over the same Ethernet cable normally used for data transfer. This is often used for devices like IP phones and wireless access points (often you see these APs in restaurants and other establishments mounted to the ceiling to provide Wi-Fi access), as it is far easier, cheaper and safer to provide low-voltage power instead of wiring in AC power which requires the help of a licenced electrician.

A (Very Simplified) Background on Power over Ethernet

The actual PoE standards (click here to learn more) IEEE-802.3af (up to 12.95 watts), 802.3at (up to 25.5 watts) and the newest 802.3bt (up to 60-90 watts) standards provide vendor-independent methods for sending and receiving 48-volt DC power over the Ethernet cable without frying the device on the other end if it’s not equipped to receive power. The PSE (power sourcing equipment) manipulates the Ethernet pairs to sense the presence of a PD (powered device), then queries what power level it should provide; after this negotiation phase, the PSE finally sends 48 volts to the PD (usually on pins 1/2 and 3/6, called Alternative or Mode A) and all is merry, thanks to the help of “phantom power“. However, cheaper devices are available which skip this and simply shove DC power over the Ethernet cable with no regard to the safety or well-being of the remote device – this is called “passive PoE”. There are no regulations regarding passive PoE, but they generally send DC power (often 12, 24 or 48 volts) over Ethernet pins 4/5 and 7/8 (called Alternative or Mode B), usually shorting the two pins on each pair for easy power transmission at the expense of being limited to 10/100 Mbps speed.

Many years ago (I’m talking back in high school, over 6 years ago), I bought some cheap PoE equipment – a D-Link DWL-P200 PoE injector and splitter kit – assuming it was compatible with 802.3af due to its use of 48 volts… it wasn’t. Since I bought this on a trip to the US and I live up in (the arguably nicer 🙂 ) Canada, I couldn’t be bothered attempting to return it to the Fry’s that I bought it from; it served some use powering a wireless router for a few years before I ditched it in favour of a ZyXEL GS1920HP-48HP 802.3at-compatible switch and Ubiquiti UAP-AC-PRO access point. It then sat in my junk bin for a while before I took it back out and conjured up a solution to make the splitter compatible with the PoE standard; this way I could tap into my existing 802.3at-compatible infrastructure I wired into my house (or perhaps use it to siphon a couple watts in other places 🙂 ).

Note I am using the word “compatible” and not “compliant” since this definitely does not attempt to comply with all of the electrical specifications contained in the 802.3af/at standards; however, I have tested this on 802.3af and 802.3at Ethernet switches and have had no issues with the upgraded splitter. One significant attribute is that true PoE requires electrical isolation and my splitter certainly does not provide it; for my use this isn’t an issue and even some commercial splitters omit this feature to reduce cost.

Modifying the D-Link DWL-P200 for 802.3at Compatibility

The DWL-P200 is a near-ideal candidate for conversion to 802.3af/at (I’ll call it “active PoE” from now on) since it already uses 48 volts for power – all it really needs is an active PoE-compatible front-end which requires an Ethernet isolation transformer, two diode bridges, a TVS (transient voltage suppression) diode, a 802.3af PD controller chip (and a partridge in a pear tree?). Easy enough, right… right?

Step 1: Prepare the Power Interface

The DWL-P200 splitter does not use a diode bridge on its power input (pins 4/5 are positive and pins 7/8 are negative), but active PoE requires that PDs include diode bridges for polarity-insensitive operation. Additionally, the splitter does not have an isolation transformer normally used for Ethernet; rather it had 10 ohm resistors on pins 1/2 and 3/6 as series coupling between the input and output – these were removed to provide a spot to install the centre-tapped isolation transformer that active PoE requires for Mode A (power on pins 1/2 and 3/6).

After harvesting an Ethernet transformer from a dead MacBook (seriously, dead computers make for great component stores), I scraped away insulation on the differential data pairs and used 40-gauge magnet wire to connect each pair to the transformer, and used 30-gauge Kynar wire for the power lines which are connected to the centre tap of each pair. To affix it, I used a blob of hot glue (which turned out to be pretty useless since this board runs HOT!), and ran the wires off to one of the diode bridges in the front-end I built.

The data output pins (1, 2, 3 and 6) are terminated to an AC-coupled ground using 75 ohm resistors, often referred to as “Bob Smith termination” to help reduce noise.

Step 2: Build the PoE PD Front-End

The actual front-end was built as two separate boards: the first was the power input board; the second was the 802.3af active PoE PD controller, which had its own construction considerations that I’ll address in a bit.

The power input board is pretty simple and was comprised of two Bourns CD-HD201 60-volt Schottky diode bridges and a SMAJ58A 58-volt TVS surge suppression diode to help overcome voltage spikes that can occur when a cable is unplugged due to the inductance in the cable itself. The inputs of the diode bridges were then connected to the centre taps of the Ethernet transformers and the original pins 4/5 and 7/8 on the power/data input of the splitter.

The second board is the PoE PD controller, which is responsible for negotiating with the 802.3af/at PSE controller at the other end of the cable. I used the Texas Instruments TPS2378 PoE PD controller, which was meant for 802.3at Class 4 (25.5 watts maximum) but I’m only using it for 802.3af Class 0 (up to 12.95 watts). The TPS2378 has a heat-sinking “PowerPAD” on the bottom which must be connected to Vss (ground); I used solar cell tabbing wire underneath and created a sort of fin-like arrangement on the unused area of the DipMicro SOIC/TSSOP-to-DIP adapter board (they don’t sponsor me – I just really like their adapter boards!). The external PoE detection and power class signal resistors were soldered to the DIP pads on the adapter to save space.

Step 3: Put it Back Together Again

Once the two boards were assembled and connected to the original FP5001 DC-DC converter‘s input, the boards were nestled inside the original case and some Kapton tape was wrapped around the case since I damaged some of the plastic clips that held it together during disassembly.

Conclusion

With the active PoE upgrades installed, the splitter now works with 802.3af, 802.3at and passive 48 volt PoE power sources. However, the internal construction of the splitter means it only supports 10/100 Mbps Ethernet. Additionally, I find that the board gets very hot under full load (I’ve measured internal temperatures well above 100 degrees C when the case is closed) which negatively impacts its efficiency, but I consider it a fair trade-off considering this was never meant to work on active PoE in the first place.

Advertisements

Completed: Self-discharge test of Kentli PH5 1.5V Li-ion AA (Part 6)

Looking for the teardown or how well the Kentli PH5 battery performs under load? Click the links to learn more.

It’s finally happened – the self-discharge test of the Kentli PH5 Li-ion AA battery has finally come to an end… and it only took almost 3 years!

 

april 29 2018 stats

Kentli PH5 self-discharge test statistics

Self-Discharge Rate

I never anticipated this test would run for so long; although the PH5 did not have a manufacturer-specified self-discharge rate, marketing materials suggested that the batteries had a storage life that was “3-5 times longer than Ni-MH batteries”. Wikipedia states that after one year, normal Ni-MH batteries lose about 50% of their capacity, and low-self-discharge (LSD) Ni-MH batteries lose 15-30%.

Correlating this with the data collected from the Texas Instruments bq27621-G1 fuel gauge, the battery lost 40% of its charge within one year, placing it in between the standard and LSD Ni-MH chemistries. Using Excel’s SLOPE() function, the self-discharge rate was calculated to be 0.10108%/day.

Experimental Improvements

There is some error in State of Charge measurement when using the bq27621 fuel gauge. As it uses the Impedance Track algorithm, open-circuit voltage is used to determine a battery’s state of charge upon gauge initialization. This OCV curve is chemistry-specific, with slightly different formulations requiring different chemistry ID codes. The bq27621 has a fixed Chemistry ID of 0x1202 (LiCoO2/LCO cathode, carbon anode), but experimental data revealed a better-matched Chemistry ID of 0x3107, 0x1224 or 0x0380; the first two chemistries pointed towards a LiMnO4/LMO cathode chemistry which I was somewhat skeptical of, but did not test further.

Using another gauge with a different, programmable Chemistry ID could have led to a straighter SoC curve. This wouldn’t be too difficult to reproduce, as the battery voltage can be fed to the gauge in order to recompute the state of charge. Additionally, the bq27621 has a Terminate Voltage of 3.2 volts (the gauge considers this voltage to be the point in which it reads 0% SoC), which is higher than the battery’s protection voltage of 2.4 volts (granted, there is very little charge difference in this area of the discharge curve).

My test setup was not temperature-controlled; I live in a house without air conditioning and room temperatures can vary from 15 to 35 degrees C (59 to 95 degrees F), depending on the season. However, I doubt that this would have had too much impact on discharge rate, and this would better represent real-life scenarios where a constant temperature may not necessarily be guaranteed.

Finally, this test was performed on a new, uncycled battery. I suspect the discharge rate would be significantly higher on an aged battery that was subject to a lot of charge cycles and day-to-day wear.

Conclusion

This was the longest-running experiment I’ve ever conducted on this blog. The Kentli PH5’s self-discharge rate lasts longer than a standard Ni-MH battery, but a LSD (low-self-discharge) Ni-MH battery would still last longer, albeit with a lower terminal voltage. The battery, when new, should be expected to last almost 3 years without a charge (although there won’t be any charge left by then); it will hold about 60% of its capacity after 1 year of storage.

To download a copy of the self-discharge test data, click here.

Self-discharge test of Kentli PH5 1.5V Li-ion AA (Part 5)

It’s amazing – 894 days (and counting) have elapsed since the start of my long-term experiment, documenting the real-world self-discharge behavior of the Kentli 1.5V Li-ion AA battery… and it’s still ongoing! How have things fared so far?

Surprisingly, even after spending nearly 30 months on the shelf, there is still 12% capacity left. The voltage has dropped from 4.216 to 3.692 volts according to my bq27621 Li-ion fuel gauge; the State of Charge (SoC) has dropped 50% since my last update.

november 28 2017 stats

The linear end date prediction is holding pretty steady, having changed slightly to an estimated 0% charge date somewhere in February 2018.

On that note, I’m impressed by how much attention this little battery has received, even years after my initial review. Every day I see a handful of views checking out the teardown and performance metrics, and there seems to be hardly any sign that this will change anytime soon. To everyone who stops by to check out my blog posts: thank you! 🙂

Self-discharge test of Kentli PH5 1.5V Li-ion AA (Part 4)

“It’s been a long time… How have you been?”

It’s been almost a year since I started my discharge test of the Kentli PH5 Li-ion AA battery, and the battery has lost almost 40% of its capacity due to self-discharge.

The discharge curve has gotten a lot less… linear since the last time I posted a self-discharge update. The battery is down to 62% state-of-charge, and its voltage has dropped down to 3.89 volts. Still, there’s a lot of time left until this battery reaches empty… but when?

I’m no statistician, but doing a linear extrapolation in Excel gives an approximate end date of January 2018, and the SLOPE() function in Excel gives me an average drop of 0.111%/day. Of course, this can easily change over the course of this test, but only time will tell…

HDQ Utility version 0.96 now available!

Whew, I’ve been working on this version for quite a while. With the helpful feedback of many people that have tried my software, I’ve made a large number of improvements to the software; of course, there are plenty of features that aren’t implemented yet, but are being worked on.

More information about how this utility works can be found here.

Download HDQ Utility v0.96 here: https://www.dropbox.com/s/pf0vszgfei7s8ly/HDQ%20Utility%200.96.zip?dl=0

Updates

  • (Major improvement!) Improved HDQ logging functionality (logs are now saved to a separate file instead of being overwritten).
    • Example: “HDQ Log (2015-10-26 at 19.02.50) – HDQ Utility v0.96.txt”
  • Improved HDQ communication (HDQ breaks no longer require the serial port to be opened more than once, and HDQ no-response timeouts are decreased from 0.5 to 0.3 seconds.
  • Reworded certain error messages for clarity.
    • Example: “Communication error: Cannot read byte from address 0x02 (No response from device).” 
  • Renamed file ‘config.txt’ to ‘Config – COM Port.txt’ for clarity.
  • Improved state-of-health warnings by making them non-modal (they do not require the user to dismiss the message).
  • Added more notifications for unidentified and uninitialized batteries. (Uninitialized batteries are determined by a FULL ACCESS security state, with Impedance Track disabled.)
  • Fixed invalid device name and maximum load current readings for v5.02/sn27545-A4 based batteries (e.g. iPhone 6, 6+…).
  • Added time-to-full readings (for firmware older than v2.24).
  • Improved error-checking for device identification (it will display a notice that the tool may need to be restarted).
  • Updated DingoLib UI library to auto-resize window to 0.9x display resolution for improved readability on larger monitors.

To-Do

  • Create a dedicated section on my blog for the HDQ Utility.
  • Create a user’s manual describing the parameters displayed by the program (in particular, the Advanced Battery Information section).
  • Improve data logging functionality by saving logs to a subdirectory instead of the program’s root to decrease file clutter.
  • Improve error-checking for commands (retry reads if one or more bytes are not received from the device).
  • Add error statistics indicating how many communication errors occurred during data collection.
  • Improve support for older (older than v1.25) firmware.
  • Improve support for v5.02/sn27545-A4 devices (make use of advanced commands available in this firmware version).
  • Add support for restarting of data collection without having to re-execute the program.
  • Add Data Flash memory functions to allow for readout of advanced configuration, serial number, lifetime/black-box data, etc.
  • Rewrite this program in something that’s not LabWindows/CVI… also, use of a GUI rather than a non-console text UI.

Self-discharge test of Kentli PH5 1.5V Li-ion AA (Part 3)

Aw what, it’s October already? So much for having another blog post in September…
But anyway, “more months, more data!™”

The voltage of the PH5 has dropped down to 4.093 volts as of today (October 1st, 2015), and its State of Charge is now 93%. There’s just enough data to guess the discharge rate of the PH5: with the currently logged data, the PH5 self discharges at approximately 0.103%/day. At this rate, the cell should last years before finally reaching zero. Looks like this will be a very, very long term test…

(At least that would give me more time to procrastinate write blog posts.)

Self-discharge test of Kentli PH5 1.5V Li-ion AA (Part 2)

After my first self-discharge analysis of the Kentli PH5 Li-ion AA battery, I have collected another month’s worth of data.

The battery’s voltage drop has been surprisingly linear. Although I didn’t get the exact day when the bq27621-G1’s State of Charge readout dropped to 99%, it is quite clear that the state of charge is dropping with a fairly steep curve now. That said, because the battery’s voltage is still far away from the ‘flat region’ of the discharge curve, it is difficult to determine when the battery will discharge itself completely at this time.

Self-discharge test of Kentli PH5 1.5V Li-ion AA (Part 1)

As an extension to my previous performance analysis of Kentli’s PH5 Li-ion AA battery, I fully charged an unused PH5 and left it on my desk to self-discharge. Every now and then, a Texas Instruments bq27621-G1 fuel gauge is hooked up to the Li-ion battery terminals (in the case of the PH5, the recessed ring around the 1.5V terminal) and the bq27621’s default settings are used to measure the voltage and state of charge.

I started this test on June 18th, 2015 and will keep taking occasional measurements until the protection IC in the PH5 shuts down.

Since the 18th, the voltage dropped from 4.216 volts down to 4.192 volts as of July 6, 2015; the bq27621’s State of Charge reading remains at 100% for the time being. The voltage drop has been fairly linear so far, but I expect it to taper off as the battery discharges to the Li-ion cell’s “flat region”, and only after that do I expect the cell’s voltage to decline more rapidly.

(Day 2 of 4) Mini-Ramble: Dallas! TI! Batteries!

Today was the first day of the actual Texas Instruments Battery Management Systems event. To my surprise, a couple hundred of people showed up from TI employees, a lot of customers (representatives from various companies like Bose, Google, and many others), and me as well. 🙂

The first day was a basic but still detailed introduction to the inner workings of Li-Ion technology as well as its limitations, failure modes (the gas coming from a Li-Po [lithium-ion polymer] cell contains carbon monoxide, hydrogen and a bunch of other gases), with this leading towards battery fuel gauges and why just measuring the voltage is not enough to accurately determine how full a battery is.

The day ended with a lab showcasing TI’s new Gauge Development Kit (GDK), which, in layman’s terms, is a “battery lab on a board”. It includes PC communication hardware, an adjustable charger, adjustable load and an on-board fuel gauge (but it’s set to use an external fuel gauge by default). I even got a chance to talk the TI battery management team, and even had a dinner with a few key TI guys including the one who made THE design for the GDK.

Looking inside a (fake) iPhone 5S battery

Considering how popular the iPhone is, there’s always going to be some counterfeits out there. I’ve been out buying various iPhone batteries to build a database of each generation’s characteristics, but one model has eluded me so far: the iPhone 5S. The iPhone 5C’s battery that I bought appears to be genuine (but with its own issues), but none of the iPhone 5S batteries I’ve bought so far (4 of them at the time of writing this blog post) were genuine. All of these fakes look like a genuine battery at first glance, but all of them share a few common traits.

Battery teardown

The fake battery sports the usual iPhone battery information, complete with some dot-matrix printed data and a data-matrix barcode. It’s labeled with a capacity of 1560 mAh and 3.8 volts nominal voltage.

Comparison between real and fake iPhone 5S battery

Comparison between real and fake iPhone 5S battery

The connector itself has two points for soldering the connector to provide durability. However, with the fake batteries, they are not soldered down. The two spots on the ends of the connectors are dark with a small point visible inside it (that point is the reinforcement pin on the connector). If this connector is installed in an iPhone, it will probably not come out without either damaging the battery’s connector, or worse, leave the plastic connector piece inside the phone, requiring tweezers to remove.

Connector lifted off with a hobby knife

Connector lifted off with a hobby knife

iPhone 5S and 5C battery pinout

iPhone 5S and 5C battery pinout

Removing the black protective tape reveals an iPhone 4 battery fuel gauge board. The connector is soldered to this board, with four solder points visible.

iPhone 4 battery PCB with soldered-on flat flex connector

iPhone 4 battery PCB with soldered-on flat flex connector

Pulling out the PCB  reveals another characteristic of these fake batteries: the positive terminal is cut short, with another metal section being clumsily spot-welded to the stub on the cell.

Note how the battery tab is poorly welded to the PCB.

Note how the battery tab is poorly welded to the PCB.

Battery fuel gauge data

The battery fuel gauge requires proper programming to accurately indicate the battery’s charge status. Because of this, each iPhone battery generation has its own specific configuration.

The fake iPhone battery retains the programming for the iPhone 4’s battery, which is a designed capacity of 1420 mAh, using a bq27541 fuel gauge running version 1.25 firmware. The data inside it is often that of a used/recycled battery as well.

This data can be (partially) read out directly from the iPhone with a tool such as iBackupBot, but more data can be read if the battery is read with another tool. I have the EV2400 from Texas Instruments to read this out on a PC, but this data can be read out with a USB-to-TTL serial port, a logic gate (a logic inverter) and a small MOSFET transistor.

I created a small tool that uses this circuit to interface with the fuel gauge and read out its data. Check it out here.

Using my tool, this is the report for one of these fake batteries. Note how it is identified as an iPhone 4 battery. Don’t be fooled by the calculated state of health. It’s not accurate for this battery as the fuel gauge still thinks it’s still inside an iPhone 4 battery pack.


**** START OF HDQ BATTERY LOG REPORT ****
HDQ Gas Gauge Readout Tool version 0.9 by Jason Gin
Date: 9/30/2014
Time: 0:52:24
Serial port: COM26

Battery Identification
========================
DEVICE_TYPE = 0x0541, FW_VERSION = 0x0125, DESIGN_CAPACITY = 1420 mAh
Battery's configuration matches that of a standard iPhone 4 battery.

Basic Battery Information
===========================
Device = bq27541 v.1.25, hardware rev. 0x00B5, data-flash rev. 0x0000
Voltage = 3804 mV
Current = 0 mA
Power = 0 mW
State of charge = 45%
Reported state of health = 0%
Calculated state of health = 99.3%
Cycle count = 14 times
Time to empty = N/A (not discharging)
Temperature = 27.9 °C (80.3 °F) (3009 raw)
Designed capacity = 1420 mAh
Heavy load capacity = 628/1410 mAh
Light load capacity = 673/1455 mAh

Advanced Battery Information
==============================
Capacity discharged = 0 mAh
Depth of discharge at last OCV update = ~778 mAh (8768 raw)
Maximum load current = -200 mA
Impedance Track chemistry ID = 0x0163
Reset count = 11 times

Flags = 0x0180
Flag interpretation:
* Fast charging allowed
* Good OCV measurement taken
* Not discharging

Control Status = 0x6219
Control Status interpretation:
* SEALED security state
* SLEEP power mode
* Constant-power gauging
* Qmax update voltage NOT OK (Or in relax mode)
* Impedance Track enabled

Pack Configuration = 0x8931
Pack Configuration interpretation:
* No-load reserve capacity compensation enabled
* IWAKE, RSNS1, RSNS0 = 0x1
* SLEEP mode enabled
* Remaining Capacity is forced to Full Charge Capacity at end of charge
* Temperature sensor: External thermistor

Device name length = 7 bytes
Device name: bq27541

**** END OF HDQ BATTERY LOG REPORT ****