LTE Cat1 Speed Comparison – From the Workshop

When designers are evaluating cellular connectivity, bandwidth is always a discussion point because it is more constrained than traditional Ethernet and Wifi speeds. Typically the cellular technology dictates the bandwidth, but the realized bit rates are oftentimes much lower than the theoretical maximums due to uncontrollable factors like network congestion, poor RF environments, etc. However, there are factors that designers can control which affect the realized bandwidth.

We set out to compare bandwidth performance between two 4G LTE CAT1 Verizon embedded cellular modems, specifically NimbeLink’s Skywire vs. Digi International’s XBee®* Cellular. The intent was to capture empirical data to help customers choose which product is the right fit for their application. Not all applications care about bandwidth, but this set of experiments was set up to compare design trade-offs to understand how they impact bandwidth results in real applications.

Both of these modems are in the same technology class, 4G LTE Category 1, which states a maximum download rate of 10Mbps and upload rate of 5Mbps between the cellular radio and the network. The first bottleneck that designers can control is the physical connection between the cellular modem and the product’s MCU. Most cellular radios provide the option for USB or UART connectivity. USB High-Speed 2.0 standard is a 480Mbps interface which has excess capacity to handle the 10Mbps and 5Mbps cellular data paths as well as extraneous command and control information. UART, on the other hand, is a much slower serial interface — the cellular modules in this comparison are capable of a maximum 921,600bps (0.72Mbps with 8N1 framing) which is insufficient to keep up with the 10Mbps/5Mbps download/upload rates of the modem.

The NimbeLink Skywire supports both USB and UART communication paths so designers can choose the physical interface appropriate for their application. The Digi International XBee®* only supports UART so the available cellular bandwidth is constrained.

The second bottleneck is protocol selection. Most cellular modems support multiple protocols like PPP, MBIM, QMI, NCM, CDC-ECM each of which has pros and cons to their use. The two compared here are PPP vs. CDC-ECM. PPP protocol has been around for a while and is fairly universal across modems and operating systems. The benefit is that it is very widely used, making it the most flexible for interoperability with devices. The downside is that PPP has more overhead associated with each packet, so the usable throughput is reduced compared to other protocols. CDC-ECM is the Ethernet over USB protocol. This protocol is lower overhead resulting in higher usable bandwidth for the application.

The NimbeLink Skywire supports both PPP and CDC-ECM protocols, so customers can choose which is appropriate for their application. They should use CDC-ECM for bandwidth critical applications and PPP for maximum compatibility with other operating systems. PPP is the only option on the Digi International XBee.

The test setup was a PC running Ubuntu MATE 16.04.02 with USB connection to the two development kits. The NimbeLink NL-SWDK was used for both USB and UART testing. CDC-ECM was used over USB and PPP was used on the UART. The Digi International XKC-V1T-U was used for UART testing. The XKC-V1T-U was put into bypass mode to allow for PPP connection over the UART. In both CDC-ECM and PPP configurations, the cellular devices provided a direct internet connection to the PC so bandwidth tests could be performed between the PC and public websites. Two Verizon SIM cards were provisioned into the same private network so the cellular side network routing paths are identical. We generated a 5MB file with random content and made it available from an Amazon S3 server so both modems could download via HTTP. The same file was used to test upload speeds to a public FTP server, Speedtest.tele2.net. Two Taoglas TG.30.8113 antennas were attached to each modem and the tests were performed in the same location to minimize any differences in setup. A script was used to automate the testing for 25 cycles each. The script would initialize one modem, run the download and upload tests, then switch to the 2nd modem, run the download and uploads, then switch back to the first modem and repeat.

This worked great in theory, but the Digi International part would stop working after 8-9 cycles. The part was overheating and would stop working, so we did two things: we reduced the file size to 1MB so the modems didn’t have to work so hard, and we added temperature probes to the top of each modem so we could monitor temperature rise during the tests. With the smaller file sizes, the 25 cycles completed without issue.

If your application depends on speed, this chart shows the fastest performance from each part. The NimbeLink Skywire in CDC-ECM mode vs. the Digi International XBee®* in PPP mode, the NimbeLink Skywire is 27 times faster on downloads and 40 times faster on uploads!