Author: Gough Lui
Evaluation Type: Power Supplies
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: Unique 802.3bt PoE to USB-C PD adapter with no comparable options at this time.
What were the biggest problems encountered?: No significant issues encountered.
Microchip PoE to USB-C Power & Data Adapter (PD-USB-DP60) RoadTest Review
By Gough Lui – June-July 2021
When it comes to providing data and power on the one cable, USB and Ethernet both have been very important in the consumer and enterprise space. The advent of USB Power Delivery (PD) has turned USB into more than just power for smaller devices, providing up to 100W which is enough to run most modern laptops. Similarly, in the Power-over-Ethernet (PoE) space, the 802.3bt (2018) standard has pushed powers up to the 100W range as well, providing enough power for sophisticated access control, IP cameras and high-performance wireless access points.
The Microchip PoE to USB-C Power and Data Adapter (PD-USB-DP60) tries to meld the two together, allowing USB-C PD devices to be powered from a PoE connection while also allowing the device to communicate over Ethernet via a USB-to-Ethernet LAN7800 bridge chip. This means USB-C PD devices can benefit from simpler deployment at longer distances up to 100m, especially where there is no remote power outlet, benefiting from established PoE infrastructure capabilities including remote power cycling.
I’ve always been interested in PoE, especially as it becomes more inexpensive and applications relevant to the home are becoming more commonplace. It is with thanks to Microchip and element14 for selecting me as a RoadTester for this product, that I have the perfect excuse to explore PoE and USB-C PD in greater depth and the opportunity to put this product through its paces.
As usual, if you found this review interesting, informative or entertaining, I would appreciate if you would leave a like, rate the document, write a comment or share it with others who may be interested. If you have any questions, feel free to leave a comment and I will do my best to answer them.
When it comes to connectivity and power in the consumer space, USB has undoubtedly had an outsized influence. From its humble 2.5W 5V-only beginnings operating up to a leisurely 12Mbit/s, the advent of USB-C reversible connections and the power delivery (PD) standard has allowed it to run equipment requiring up to 100W, while the USB 3.2 SuperSpeed Gen II standard allows for up to 20Gbit/s of data throughput. Initially providing power only for smaller devices, it’s now not uncommon to find laptops powered entirely through the USB-C connector from either a power supply or a docking station.
Over longer distances, especially in small-business and corporate settings, Power-over-Ethernet has had a similar utility for devices such as IP cameras, VoIP desk phones, wireless access points, access-control devices and more. In this setting, it allows remote devices to be powered from the same cable that provides the data connection, saving on infrastructure deployment costs and allowing for conveniences such as remote power cycling.
Early-on, many vendors implemented “passive” solutions using various voltages (12, 15, 24, 36V) and the unused pairs from 100BASE-TX, which worked but was not completely safe especially if plugged into incompatible equipment or non-PoE equipment. Standards-compliant PoE in the form of 802.3af (2003) standardised the 48V-nominal system with intelligence in the power-sourcing equipment (PSE) to ensure interoperability and safety by detecting the downstream load and sourcing up to 15.4W of power where the right resistance “signature” was detected. Higher-speed access points and higher-end IP cameras required more power and the 802.3at (2009) standard (also informally known as PoE+) pushed the power sourced up to 30W with a more complex “signature” handshake and tighter voltage requirements at the PSE. This is perhaps the most common form of PoE available today in mainstream mid-span injectors and PoE switches as it is cost-effective and sufficient for most loads.
However, higher powers may yet be desirable for more sophisticated loads and the 802.3bt (2018) standard addresses this. Informally known as PoE++ or 4PPoE as its modes can use all four pairs of the cable, the standard can source 60W or 100W depending on the mode. This is made possible through a further increase and tightening of the PSE voltage, but still remaining backward compatible. This is sufficient to run whole access control systems, LED lighting and more sophisticated IP cameras. While equipment that is 802.3af/at can be run on an 802.3bt injector with no problems, equipment that requires 802.3bt PoE may not function fully or at all on an injector providing less power. Cable losses further reduce the power available at the device – in the case of 802.3af, this is 12.95W; 802.3at, this is 25.5W; and 802.3bt this is 51W or 71W.
Microchip’s PoE to USB-C Power and Data Adapter (PD-USB-DP60) tries to meld the two together. It can be thought of as a combination of a USB-to-Ethernet dongle device that lets the USB-connected device access the Ethernet network, as well as a USB-C PD 60W power supply powered by PoE allowing the connected USB-C device to be powered or charge from the PoE supply. The solution is compatible with 802.3af/at/bt PoE with the output power available depending on the capabilities of the injector. It is equally capable of operating as an ordinary USB-to-Ethernet dongle, or just providing power to the end device for operation/charging.
In melding the two together, it provides the capability to converge USB-C PD powered devices with PoE infrastructure. This includes the capability to provide power and data over longer distances (up to Ethernet’s 100m maximum) as USB-C cables are both expensive and limited in length, as well as the ability to take advantage of PoE infrastructure capabilities which could include remotely power-cycling the connected equipment through a PoE switch’s administration panel. Because of PoE’s reach, it is possible to power USB-C PD equipment at a distance where it may be too expensive or impractical to install a power outlet, while the 802.3bt capability also ensures that relatively power-hungry equipment is catered for. This is probably something nice to have on hand, especially for those in data centres where looking for a power connection for a laptop is not always easy and may result in costly mistakes (e.g. bumping a power cable for a rack).
For the avoidance of confusion, it should be noted that this device operates in the USB-C device role and does not serve to host or extend USB-C communications over Ethernet. Instead, the connected device needs to be a host and have the necessary drivers to communicate with the LAN7800 chipset onboard to access the Ethernet network. As a result, if you’re looking for a device to overcome USB-C cable length limits by converting to Ethernet (such as the Icron Raven 3104/Pro), this is not an equivalent device.
While there are a few PoE to USB options, they are all 802.3af/at type devices with a “standard” USB-A output of about 10W which can be rather limiting. A bit of research shows that 802.3bt-based devices are relatively “thin” on the ground. In fact, this device is the only device that supports 802.3bt PoE and is also USB-C PD capable. Given that it is the first, launched only in March this year, makes it a very unique proposition. At the time of publishing, Microchip Direct has the unit listed at USD 91 per unit which is quite reasonable given its uniqueness and capabilities. Newark has it listed for a little more at USD 104, but oddly enough, trying to obtain it locally in Australia through element14 has it listed at AUD 349.07 (or USD 264.86) which is unreasonably expensive. I know we do pay a little more being half-way around the world, but that is a bit too much to swallow.
As 802.3bt PSEs are not all that common, to experience the full power output may require obtaining a midspan injector. Microchip recommend the PD-9601GC Single Port, 802.3bt Type 4, 90W, Indoor PoE Midspan injector which is listed on Microchip Direct and Newark for USD135.59. In the case you intend to use the full 60W capability and don’t already have 802.3bt PSEs, this can represent a significant additional cost (especially compared with 802.3at injectors which can be had for USD30) which makes the solution a little less appealing. If you’re in Australia (as I am), the recommended PSE is not available via local distributors and 802.3bt injectors usually top out at the Type 3 60W variety which would not be sufficient and importing from overseas (along with regulatory compliance issues) would be necessary.
The unit arrived inside a glossy colour-print retail hanger cardboard box. The front illustrates the front side of the adapter, while the rear shows the rear along with an included cable and a list of features. Unlike most of Microchip’s products, this is not an evaluation kit but an actual finished retail product.
The top of the box has the Microchip logo, while the bottom has the product barcode, model number and serial number label. The product is Made in China.
The contents are inside a cardboard tray with a pull-out ribbon, with a further “milk tray” plastic insert. Removing the items in the tray reveals this handy hint that there is more to be found underneath.
The inclusions are a RoHS certificate, a quick-start guide, a storage pouch, the adapter and a short USB-C cable. This is noticeably different to the one illustrated on the rear of the package which appeared somewhat longer.
As is customary with high-end networking equipment, the unit is in a textured black metal shell which feels very solid and has some weight to it. The underside has a product label with more specifications including the manufacturer (Changzhou Wujin Hongguang Radio Co. Ltd.), the PoE input of 47-57VDC at 1.73A and USB-PD output of 5VDC/3A, 9VDC/3A, 15VDC/3A or 20VDC/3A. This product is revision A01.
The front end of the unit has a shielded metal jack with port-vista LEDs, one for power and the other for link/activity. The rear of the unit has a single USB-C connector and a threaded hole above where the screw of a captive USB-C connector can be secured.
Accordingly, the provided cable features a captive USB-C connector in moulded plastic and a metal-bodied non-captive plug with LED for the device end. While the connections are entirely reversible and invertible, this e-marked cable’s captive feature should prevent the end from falling out of the adapter. The moulded plastic end has the Microchip logo on the underside, but the cable is almost frustratingly short at just 23.5cm from plug moulding to plug moulding.
The product itself has all of six pages of documentation spread over three documents comprising the Quick Start Guide, “Sell Sheet” Brochure and Datasheet, much of it redundant. This is no issue, as the product is very much “plug-and-play” in nature and requires little-to-no configuring. The only important thing to note is that where data connectivity is desired and drivers are not pre-installed, they are available from the LAN7800 download page. Drivers cover most operating systems of interest nowadays including Windows Vista through 10 in 32-bit/64-bit x86/ARM flavours, MacOS X, Linux, and UEFI PXE. However, as it is a direct descendent of the LAN7500, itself a descendent of the LAN9500, in many cases there is no need to actually install a driver.
In fact, in my testing with my Windows 10 machine, no driver install was necessary as it was automatically installed when plugged in. The driver that was installed was V.126.96.36.199 which is not the latest, but the version before the current one which was the last to be qualified under WHQL.
A similar experience was found using Ubuntu 20.04.2 LTS, where no driver install was necessary. Selected lines from dmesg illustrate the detection of the device and setup using the lan78xx driver:
[ 0.000000] Linux version 5.8.0-55-generic (buildd@lgw01-amd64-050) (gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #62~20.04.1-Ubuntu SMP Wed Jun 2 08:55:04 UTC 2021 (Ubuntu 5.8.0-55.62~20.04.1-generic 5.8.18) [ 1.957811] usb 2-1: new SuperSpeed Gen 1 USB device number 2 using xhci_hcd [ 1.978612] usb 2-1: New USB device found, idVendor=0424, idProduct=7800, bcdDevice= 3.00 [ 1.978617] usb 2-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3 [ 1.978621] usb 2-1: Product: LAN7800 [ 1.978623] usb 2-1: Manufacturer: Microchip [ 1.978626] usb 2-1: SerialNumber: 210460000404 [ 8.778743] libphy: lan78xx-mdiobus: probed [ 8.815338] usbcore: registered new interface driver lan78xx [ 9.009027] lan78xx 2-1:1.0 enx00055ab00687: renamed from eth0 [ 18.399831] lan78xx 2-1:1.0 enx00055ab00687: No phy led trigger registered for speed(-1)
Further driver information shows that the drive is in-tree and supported by WooJung Huh from Microchip directly:
filename: /lib/modules/5.8.0-55-generic/kernel/drivers/net/usb/lan78xx.ko license: GPL description: LAN78XX USB 3.0 Gigabit Ethernet Devices author: WOOJUNG HUH <firstname.lastname@example.org> srcversion: C289622CADC4AAEC8F85AE1 alias: usb:v0424p7801d*dc*dsc*dp*ic*isc*ip*in* alias: usb:v0424p7850d*dc*dsc*dp*ic*isc*ip*in* alias: usb:v0424p7800d*dc*dsc*dp*ic*isc*ip*in* depends: retpoline: Y intree: Y name: lan78xx vermagic: 5.8.0-55-generic SMP mod_unload modversions
The verbose output from lsusb is as follows:
ID 0424:7800 Microchip Technology, Inc. (formerly SMSC) LAN7800 Device Descriptor: bLength 18 bDescriptorType 1 bcdUSB 3.10 bDeviceClass 255 Vendor Specific Class bDeviceSubClass 0 bDeviceProtocol 255 bMaxPacketSize0 9 idVendor 0x0424 Microchip Technology, Inc. (formerly SMSC) idProduct 0x7800 bcdDevice 3.00 iManufacturer 1 Microchip iProduct 2 LAN7800 iSerial 3 210460000404 bNumConfigurations 1 Configuration Descriptor: bLength 9 bDescriptorType 2 wTotalLength 0x0039 bNumInterfaces 1 bConfigurationValue 1 iConfiguration 0 bmAttributes 0xa0 (Bus Powered) Remote Wakeup MaxPower 896mA Interface Descriptor: bLength 9 bDescriptorType 4 bInterfaceNumber 0 bAlternateSetting 0 bNumEndpoints 3 bInterfaceClass 255 Vendor Specific Class bInterfaceSubClass 0 bInterfaceProtocol 255 iInterface 0 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x81 EP 1 IN bmAttributes 2 Transfer Type Bulk Synch Type None Usage Type Data wMaxPacketSize 0x0400 1x 1024 bytes bInterval 0 bMaxBurst 7 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x02 EP 2 OUT bmAttributes 2 Transfer Type Bulk Synch Type None Usage Type Data wMaxPacketSize 0x0400 1x 1024 bytes bInterval 0 bMaxBurst 6 Endpoint Descriptor: bLength 7 bDescriptorType 5 bEndpointAddress 0x83 EP 3 IN bmAttributes 3 Transfer Type Interrupt Synch Type None Usage Type Data wMaxPacketSize 0x0010 1x 16 bytes bInterval 6 bMaxBurst 0 Binary Object Store Descriptor: bLength 5 bDescriptorType 15 wTotalLength 0x0016 bNumDeviceCaps 2 USB 2.0 Extension Device Capability: bLength 7 bDescriptorType 16 bDevCapabilityType 2 bmAttributes 0x0000010e BESL Link Power Management (LPM) Supported BESL value 256 us SuperSpeed USB Device Capability: bLength 10 bDescriptorType 16 bDevCapabilityType 3 bmAttributes 0x02 Latency Tolerance Messages (LTM) Supported wSpeedsSupported 0x000e Device can operate at Full Speed (12Mbps) Device can operate at High Speed (480Mbps) Device can operate at SuperSpeed (5Gbps) bFunctionalitySupport 1 Lowest fully-functional device speed is Full Speed (12Mbps) bU1DevExitLat 10 micro seconds bU2DevExitLat 1500 micro seconds Device Status: 0x0010 (Bus Powered) Latency Tolerance Messaging (LTM) Enabled
Testing with an early-model Raspberry Pi 4 Model B’s USB-C port which supports USB-OTG was successful, although this required a careful selection of cables and changes to the config.txt configuration. Due to early-models containing a “bug” with the USB-C CC-lines, plugging in the provided cable did not result in any power to the Pi 4. Instead, the use of a USB-C to USB-A OTG adapter plus a USB-A to USB-C cable was necessary to connect the Pi 4 in such a way that the PoE adapter understood it was connected to a load, while the Pi 4 was being provided both power and data circuits. Later revisions which fixed this issue should work without this “workaround”.
By default, the Pi 4’s USB-OTG hardware attached to the USB-C connector does not seem to be active, so the addition of otg_mode=1 is necessary inside config.txt to enable the xHCI controller. The adapter is then detected by the lan78xx driver and is fully operational, albeit at USB 2.0 rates because of the limitations of the Pi 4’s USB-OTG controller.
[ 0.000000] Linux version 5.10.17-v7l+ (dom@buildbot) (arm-linux-gnueabihf-gcc-8 (Ubuntu/Linaro 8.4.0-3ubuntu1) 8.4.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #1414 SMP Fri Apr 30 13:20:47 BST 2021 [ 0.000000] CPU: ARMv7 Processor [410fd083] revision 3 (ARMv7), cr=30c5383d [ 0.000000] CPU: div instructions available: patching division code [ 0.000000] CPU: PIPT / VIPT nonaliasing data cache, PIPT instruction cache [ 0.000000] OF: fdt: Machine model: Raspberry Pi 4 Model B Rev 1.1 [ 1.394777] usbcore: registered new interface driver lan78xx [ 1.984215] usb 3-1: New USB device found, idVendor=0424, idProduct=7800, bcdDevice= 3.00 [ 1.984263] usb 3-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3 [ 1.984296] usb 3-1: Product: LAN7800 [ 1.984324] usb 3-1: Manufacturer: Microchip [ 1.984353] usb 3-1: SerialNumber: 210460000404 [ 2.271146] libphy: lan78xx-mdiobus: probed [ 2.271192] lan78xx 3-1:1.0 (unnamed net_device) (uninitialized): int urb period 64
Use with earlier Raspberry Pi models is not as straightforward, as the micro-USB sockets usually only have the power lines wired up and no data lines, meaning that you can only use the adapter to provide power and not USB-Ethernet data. There is, however, an unsupported and non-standard workaround you can use.
The workaround involves back-feeding the power and data up the Raspberry Pi’s USB-A port by virtue of using an illegal USB-A to USB-A cable attached to the USB-C to USB-A OTG adapter. This configuration is generally not recommended as it bypasses some overcurrent protection. I have tested this configuration and it works with my Raspberry Pi Model A, A+ and B. It does not work with the Raspberry Pi Model B+, 2, 3B or 3B+ as these models have an intervening switch on the USB port power that prevents backfeeding. I have not got any other Raspberry Pi models on-hand to test, so the remainder should be considered untested.
Attaching the adapter to my Xiaomi Redmi Note 8T (Android 8) and POCO M3 (Android 12) devices allowed for them to charge, but no data connectivity was available as it seems their Android kernels were not built with the drivers integrated. Attaching the adapter to a Raspberry Pi Model 3B+ was sufficient to supply power, but Ethernet connectivity is not possible as the board design does not have any data pins connected. However, this illustrates that the use of the adapter for power supply only is possible, noting that the power output is USB-PD compliant so it may not be capable of quick-charging devices using proprietary protocols (such as Qualcomm Quick Charge 2.0/3.0). Devices using other protocols and Apple devices were not tested (as I do not possess any such devices).
The operational LEDs on the front of the device are quite handy. The yellow power LED indicates when power is being received over the PoE interface, allowing quick diagnosis whether the PoE circuit is active. The green link/activity LED does not light up until the USB-Ethernet chip is initialised by the operating system, so it is easy to determine whether the adapter has been correctly detected and drivers installed just at a glance.
The bundled USB-C to USB-C cable also has a trick up its sleeve with a green LED indicator that lights whenever power is being delivered to the connected device. This makes it easy to see if the USB-C connection was successful.
Tests consisted of running iperf3 with the default settings five times in the forward direction and five times in the reverse direction. The test peer was my main desktop machine comprising an AMD Ryzen 7 1700 CPU, 64GB DDR4 RAM at 2800MHz, 1TB Samsung 960 EVO NVMe SSD on an Asus X370-PRIME-PRO motherboard running Windows 10 (21H1). The network interface is the onboard Intel I211 Gigabit Network Connection, connected to a Netgear GS724T 24-port Gigabit managed switch. The test device was connected through a 2m cable to the mid-span injector, followed by a 3m AWG24 cable to the device under testing.
Test platforms included a Dell Latitude 7390 laptop with an Intel i7-8650U CPU, 16GB RAM and 512GB SSD. The laptop was running a dual-boot configuration consisting of Windows 10 (RS5 due to corporate requirements) and Ubuntu 20.04.2 LTS. This laptop was chosen as it has a USB-C port capable of charging the laptop and performing USB 3.0 connectivity at the same time. The use of a USB-A to USB-C (2.0) capable cable was used to determine USB 2.0 throughput under Ubuntu as well. It also had an onboard Intel PRO/1000 I219-LM Network Connection which could provide a direct throughput comparison. Tests were also conducted using a Raspberry Pi 4 using the configuration mentioned in the previous section.
The results are summarised in the graphs below:
The benchmark tests using the Intel I219-LM established a relatively consistent baseline in the 939-945Mbit/s range across both Windows and Linux operating systems by which we can compare the performance of the Microchip LAN7800 solution.
When connected in USB3.0, the adapter was able to excel in tests in the forward direction, achieving figures that exceeded that of the Intel I219-LM by a couple of megabits per second which is impressive given the overhead that USB usually entails. The reverse direction performance, however, was quite variable and noticeably slower, especially for the tests run under Ubuntu.
When connected in USB2.0, the bus proves to be the bottleneck, with the highest throughput of 364.2Mbit/s recorded in the forward direction under Ubuntu. The same pattern where reverse direction tests had lower performance persists. The Raspberry Pi 4 was able to achieve slightly lower numbers compared to the USB2.0 tests on the Dell Latitude 7390 laptop possibly because of CPU constraints or driver differences.
The performance overall is not bad considering it is a USB3.0 device, however, the performance asymmetry between forward and reverse direction is unusual and could theoretically pose issues for applications requiring the maximum throughput.
My RoadTest proposal indicated my desire to determine the capability of the Microchip PD-USB-DP60 to service various loads by testing the unit synthetically using a load – in this case, my Rohde & Schwarz NGM202 two-quadrant power supply and B&K Precision Model 8600 DC Electronic Load. I also wanted to understand the quality of the output power and attempted ripple and noise measurements using my Rohde & Schwarz RTM3004 oscilloscope.
It should be noted that all performance tests were conducted with a 3m AWG24 stranded Cat.5E cable to maintain consistency. While this length would be considered short compared to the Ethernet standard which requires operation at 100m, assuming that the unit is designed to 802.3bt compliance suggests that there should not be any issues. This is because the standards have relatively tight PSE voltage requirements which push up towards the high-end to compensate for cable losses and usually assume greater cable loss than actually is experienced. Assuming the use of an 802.3bt Type 4 injector, 71W should be available at the load and thus a conversion efficiency of just 85% would be sufficient to deliver 60W to the USB-C connector. This is not a particularly lofty goal for a well-designed modern switching DC-DC converter, thus I have some confidence that it should be able to deliver the required level of power. That being said, if I had 100m of cable, I would have tried it but I don’t and I can’t see a reason to make this investment.
My proposal indicated the unit would be tested with an 802.3at compliant (PoE+) power injector as this seems to be the most economical and popular standard at this time. As a result, I purchased a relatively inexpensive TP-Link TL-POE160S Gigabit PoE+ Injector which is designed for end-span applications.
This relatively nondescript wall-mountable black plastic box is a sealed unit which can add PoE to one downstream device by taking in the data signal through one jack and putting out the data and PoE combined on the other jack. Mains power is provided through a cable connecting to the clover-leaf connector, while an LED indicates when the injector is active. Just like the Microchip adapter, things are very much “plug and play”.
Using the device with regular USB devices including relatively heavy-draw devices like the Raspberry Pi proved to be just fine. This is no surprise, as 5V/3A adds up to 15W which is comfortably within the 802.3at power envelope.
Testing the adapter with the Dell Latitude 7390 showed that even with 802.3at, the adapter is capable of supporting USB-PD but at a limited power level. The laptop was able to charge once powered down, but if running, would not charge and instead slowly depletes the battery. If the laptop was attached to a standard USB (non-PD) charger, it would not have charged at all, so this is a good sign.
Testing with a Lenovo Miix 2 Tablet with its only single USB-OTG port allowed the tablet to charge and communicate successfully via the USB-Ethernet connection. As a low-power Intel Atom-based tablet from 2015, it is good to see that it can be made to both charge and communicate despite being pre-USB-C.
A measurement of the output I-V curve was made by connecting a four-wire USB-A break-out to a USB-C to USB-A OTG adapter and plugging that directly into the Microchip PD-USB-DP60’s output.
Unfortunately, a small voltage drop due to this adapter is unavoidable, but it should still give a good idea of the stability and over-current protection capabilities.
The output voltage starts almost precisely at the maximum USB 5V voltage of 5.25V when under no load and decreases linearly (possibly in part due to adapter resistance) as the current load increases. Over-current protection is tripped at about 3.333A which indicates a functional OCP with sufficient headroom to prevent inadvertent tripping and voltage maintained above the USB 5V minimum of 4.75V. This result is excellent.
Tests of the USB-PD outputs are illustrated in a later section with an 802.3bt power injector.
With a traditional 5V output, the power quality from the adapter is excellent overall. The ripple was measured at 18.925mV peak-to-peak unloaded, measuring 17.791mV under 500mA load.
Increasing the load to 1A, the ripple remains similar at 17.433mV. At 1.5A, this remains consistent at 17.622mV peak-to-peak.
At 2A, the ripple is just 18.018mV peak-to-peak. Finally, loaded at its rated 3A, the ripple reaches 19.417mV peak-to-peak which is well below what is achieved by many wall-power supplies and computer ports.
An FFT of the spectrum shows a narrow(ish) peak which corresponds to the switching frequency of approximately 306.4kHz which is dominant. Next to it, there is also a lower peak, seemingly spread-spectrum in the 354-377kHz range. Other than that, there seem to be some noise-spikes which may come about due to non-linear effects.
Taking away the 20MHz bandwidth limiter, it seems that the output of the adapter does have quite a bit of high frequency noise content, especially near the FM radio band, centred about 105MHz. This is probably not good for powering sensitive radio-frequency equipment.
While my initial RoadTest application only mentioned PoE+, I felt it would be a great disservice to the “first” 802.3bt USB-C PD adapter to only test it under those conditions. As a result, I decided to splurge and import the lowest-cost 802.3bt 90W-capable power injector I could find.
This turns out to be the rather non-descript DSLKIT BTIN5219 802.3bt PoE Injector used for these tests. For those interested, I have done a teardown of this injector.
Testing this configuration with the Dell Latitude 7390 was successful, as it detected an “undersized” 60W adapter connected to it, rather than the (OEM) 65W adapter. This allowed the laptop to be powered and battery to be charged just fine, with network connectivity being maintained via the USB-C connection as well despite using a CPU stress test to attempt to maximise power consumption.
This validates the USB-C PD capability with an actual device and the correct detection of the 802.3bt PoE power injector.
Furthermore, increasing the power supply would not be great without a way to properly synthetically test the unit under load, so I ordered a bunch of USB-C PD “decoy” boards which serve to request various PD voltages so I can test them using my loads. This change to the RoadTest is the main reason it’s taken a little longer than expected to deliver the report as the items had to be shipped from overseas during which Sydney, Australia underwent an emergency COVID-19 lockdown which it is still in as of the publication of this review.
Note that measurements of voltage would be (very slightly) influenced by the resistance in the supplied USB-C cable and decoy board, while the measured current excludes the quiescent consumption of the decoy board itself (which should be in the ~20mA range approximately).
When the USB PD standard voltage of 9V is requested, the output shows an ohmic response, with an open circuit voltage of about 9.18V decreasing to about 8.77V by the rated 3A which is an excellent result. Powered by either the 802.3at or bt injectors, the full 27W output is available with OCP kicking in at 3322mA.
At the next USB PD standard voltage of 15V, the initial voltage is about 15.17V, rising slightly initially and then falling to about 14.78V by the rated 3A which is again an excellent result. In this case, the 802.3at injector cut out at 2057mA delivered to the load (or approximately 30W) although this is likely to be earlier in case of higher cable losses. The full output was available with the 802.3bt injector, reaching OCP at 3320mA.
At the final USB PD standard voltage of 20V, the open circuit voltage was about 20.4V, decreasing to 19.96V at 3A, maintaining its excellent regulation. With the 802.3at injector, the output cut off at 1534mA delivered to the load (approximately 30W) while the 802.3bt allowed it to reach its OCP at 3320mA.
Based on these test results, the Microchip PoE to USB-C Power & Data Adapter consistently delivers excellent voltage regulation, consistent OCP protection and has demonstrated its capability to deliver output at all standard USB-PD voltages to the level indicated in the specifications.
As this is a bit of a bonus and to avoid overloading this review with oscilloscope traces, I decided to only test ripple and noise for USB PD voltages at the full rated load of 3A. This is because most power converters generally exhibit their worst behaviour at full load (even though this one seems to be the exception at zero load when tested at 5V, this is not seen as a highly relevant use case).
At 9V, the ripple measured an average of 153.41mV peak-to-peak which is approximately 1.7%, well below the 5% that is usually expected for IT equipment power supplies.
At 15V, the ripple increased to an average of 197.48mV peak-to-peak (approximately 1.3%) which is an excellent result.
Finally, at 20V, the average ripple reached 257.14mV peak-to-peak (approximately 1.3%) which is again an excellent result. There should be no concerns regarding the quality of the power output on the Microchip PoE to USB-C Power & Data Adapter – it’s cleaner than some mains AC adapters I’ve previously come across.
Another item of interest is just how efficient would a PoE system be at servicing a USB load. Because I formerly RoadTested the Tektronix PA1000 Power Analyzer and B&K Precision Model 8600 DC Electronic Load, it is possible for me to determine the efficiency of the solution based on power delivered to USB versus power consumed from the wall with my own pyvisa script. This depends on the efficiency of the Microchip PoE-USB-C PD Adapter, power injector and losses in 3m of 24AWG Cat5e cabling, but should give us some indication.
To attempt to get the best understanding of the power consumption, I first measured the no-load quiescent of the PoE injectors and then the no-load quiescent with the Microchip PoE-USB-C PD adapter attached.
The TP-Link 802.3at injector idled at 1.1589W with nothing attached, rising to 2.3375W with the Microchip adapter plugged in. This suggests an increase in power consumption of 1.1786W from the wall can be attributed to the idle consumption of the adapter.
My DSLKIT 802.3bt injector idled at 2.4041W with nothing attached, the consumption falling to 2.3375W with the Microchip adapter attached. This may sound counter-intuitive, however, it is something you can sometimes observe in the wild, as the increased load may actually push the switching converter into a more efficient operating point, thus reducing power consumption while delivering more power.
The end-to-end efficiency is depicted in the left graph, while the idle quiescent injector power was subtracted from the measured power on the right to try and eliminate the effect of the injector standby power on the efficiency (while retaining cable losses and adapter quiescent current losses). Please note that this test does not represent a measurement of the Microchip adapter’s efficiency and the results are dependent on the injector and cable losses as well.
The end-to-end efficiency reached about 81% peak, however, this was only attained for higher USB-PD voltage outputs. At the USB standard 5V, system efficiency was only about 65% or 70% peak. Poorer system efficiencies at light loads especially plagued the 802.3bt injector which has significant quiescent current. Subtracting the standby quiescent injector power pushes the peak efficiency to about 83%, with 5V efficiency reaching about 75% peak.
Such results are not unexpected from PoE-based power distribution due to the losses incurred with AC to DC and DC-to-DC conversion steps and intervening cable losses.
This review wouldn’t be complete without a peek under the covers. As this unit seemed like it could be disassembled without destroying it entirely, I ventured to take it apart.
The unit is held together by two different-sized screws which reside underneath the product label.
The presumably unpainted cast aluminium casing as seen from the inside is nice and clean. A single PCB houses the complete solution.
The top side of the board, contrast enhanced, is shown above. The construction can be seen to be quite high-quality with a gold-finish board that is cleanly soldered with only some minor flux residues visible. It is adorned with several test points and LED indications that are not seen once the cover is closed. Solid electrolytic capacitors have been chosen for bulk capacitance except for one capacitor which is an ordinary aluminium electrolytic, although of quality make from Matsushita. The USB-C connector is an Amphenol product. Larger components are supported by white silicone to improve vibration resistance. It seems the board itself had a code of PR-6000-C00.
The board can be seen to be populated with a variety of components, many of which are Microchip-branded, a few Microsemi-branded (now a Microchip company) and On Semiconductor-branded (a good friend of Microchip, it seems). This really speaks volumes about how this product is enabled by Microchip’s integrated-circuit solutions – truly “eating dog food”.
Major components include:
The underside mostly contains smaller semiconductors and passives. Noteworthy components include an unidentified U33 marked NKAN KUG and unidentified TRS3 marked PKGH P06C. The soldering on the underside is acceptable, but if I am to nitpick, there are some SMD components which seem to show signs of either tarnishing and imperfect solder flow up to the top of the component. This doesn’t seem severe and I don’t expect it to be an issue in terms of reliability.
The Microchip PoE to USB-C Power and Data Adapter (PD-USB-DP60) is a unique product, being one of a small niche of products supporting the highest-power 802.3bt PoE standard at this time, and the only to combine the ability to interface to a host via USB-C to provide 60W PD (Power Delivery) charging and data over an integrated LAN7800 SuperSpeed USB 3.1 Gen 1 to Gigabit Ethernet chipset. While it is cutting-edge in terms of its support for 802.3bt, it is backwards compatible with 802.3af/at PoE infrastructure as well, with a corresponding reduction in available output power for the connected device. In the USA, it appears to be priced reasonably at USD104 (via Newark), however in my home country of Australia, the price is an astronomical AUD349.07 (via element14) which makes it a lot less attractive.
The adapter feels like a solid product with a hard-wearing aluminium shell with an e-Marked captive USB-C to USB-C cable and protective pouch included in the colour retail hanging box. A peek under the covers confirms this impression, with good construction quality on a quality gold-finish PCB with an emphasis on using Microchip “in-house” components supported by high-quality capacitors. Testing its power output performance shows that the output voltage is well within expected range under all loading conditions, that the over-current protection acts consistently to maintain safety and that the ripple are well controlled to ensure clean power to connected devices. The radio-frequency noise component of the output is perhaps the only slight concern.
In use, it proves to be a versatile product, allowing for use as a USB to Ethernet dongle (unpowered via PoE), as a USB power source only (attached to a device without OTG Host role) for powering products or charging batteries, or for both USB PD and USB-Ethernet (where the connected host supports communicating with the LAN7800 and has the appropriate drivers). The indicator LEDs make diagnosis of the connection a breeze, however, its functioning is conditional on using the correct USB cables that allow for the source (host) and sink (device) roles to be correctly negotiated. I found its application to be as “plug-and-play” as can be expected, with no need to manually install drivers on most variants of Windows and Linux. Support for other operating systems will require the installation of drivers or the compilation of kernel modules.
As a product, I have found the adapter to be quite useful when checking out network ports in a building. Where ports are connected to a PoE switch, simply hooking the Ethernet cable to the adapter is enough to know if the port is “alive” or not. Plugging it into the laptop allows for easy diagnosis and packet captures. It has also come in handy to charge up a mobile phone where no power points or chargers were available. For those who may be working in data centres or networking, it may be a very useful tool to have in an everyday carry bag.
Perhaps the biggest caveat is that while it is possible to run a full laptop from the unit, this is only possible when attached to an 802.3bt injector sourcing the full ~90W into the cable. Use with the more cost-effective 802.3at (PoE+) injectors results in a power limitation of about 27W to end devices which is insufficient for heavy draw devices such as laptops, but would still be able to fast-charge some mobile phones through USB-PD. This disadvantage may be less of a disadvantage if 802.3bt were to become more popular in the future, but the number of devices that need more than the 30W that PoE+ currently offers is still relatively limited. Because of this, 802.3bt equipment still remains more costly – factoring in the cost of an 802.3bt injector to make full use of this adapter would add another USD135.59 to the solution cost (assuming you can obtain the recommended PD-9601GC injector).