New Joulescope JS110s sometimes don't start streaming data

:warning: If you purchased a Joulescope between very late April 2021 and May 10, 2021, please read!

If your Joulescope has a manufacturing date from March 2021 through May 2021, it may have been manufactured with an unexpected variation. Fortunately, the variation is acceptable, but your Joulescope may need new firmware to operate reliably.

If you suspect that you are being impacted by this issue, you can update your firmware today. On Windows 10 with the Joulescope UI already installed:

  1. Download the latest 1.3.4 firmware image to your Download directory.
  2. Open a Command Prompt
  3. Type: "c:\Program Files (x86)\Joulescope\joulescope.exe" program upgrade "%USERPROFILE%\Downloads\js110_1_3_4.img"
  4. Wait for about 30 seconds until the sensor-side LED illuminates. Unfortunately, this approach does not display any status on Windows.

Details

The Joulescope JS110 includes the TPS2552 which limits the current to the transformer driver that powers the JS110 sensor when it first turns on. This ensures that the JS110 keeps all your other USB devices happy and powered when it starts up. It’s a great Texas Instruments part that we are happy with and recommend.

The TPS2552 has two variants: constant-current (TPS2552) and latch-off (TPS2552-1). All other JS110s use the constant-current version. Unfortunately, many units in this build use the latch-off version. With the latch-off version, the part permanently disables the output if the overcurrent condition lasts for more than a pre-configured duration.

The Joulescope JS110 sensor-side is designed to hit the configured TPS2552 current limit as it powers on. Fortunately, the latch-off version does not “trip” until between 7.5 milliseconds typical (5 ms min to 10 ms max) duration of the overcurrent limiting condition. The JS110 powers up each analog rail separately, which helps spread out the overcurrent durations. It just so happens, that the JS110 sensor seems to take about 6 or 7 ms to power up the most demanding -16V rail. Most TPS2552-1 parts work fine. However, some combinations of TPS2552-1 and the effective -16V capacitance are not happy and cause the TPS2552-1 to trip and latch-off.

When the TPS2552-1 latches off, it disables power to the JS110 sensor. The JS110 then becomes unresponsive indefinitely, or at least until you unplug it from USB and then plug it back in. Since we designed for the TPS2552 (not TPS2552-1), this behavior is unexpected.

During our most recent manufacturing run, we found an unusual number of units that failed at programming. An initial investigation did not reveal a cause, and we kept building. We successfully built many units, and took partial delivery of fully working units to ship to our customers.

After successfully building many units, we spent yesterday performing detailed failure analysis on these misbehaving units. We found that the firmware was actually programming correctly. Immediately after programming the sensor, it starts running the programmed firmware, and we found that it became unresponsive right when turning on the analog power. Further investigation led us to find the TPS2552-1 installed on these units.

Fortunately, the sensor firmware can modulate the -16V regulator enable pin as it is turning on. This keeps the TPS2552-1 overcurrent duration to 1 millisecond, which is perfectly acceptable for the TPS2552-1. Although the effective capacitance on the -16V rail is high, the current requirements are very low, so modulating the enable pin as turning on works great. In fact, it’s a better approach for all JS110 Joulescopes to keep other USB devices happy.

What are we doing

We are currently in the process of deploying this firmware update to all completed units in inventory. All units shipping on or after May 11, 2021 will have this fix. We will also deploy this fix to all manufacturing work in progress.

We do not know the exact date we started shipping these units, but it was very late April or even early May. If you ordered a unit during this time, we will support you to get the firmware up and running on your new units. If you run into any issues or need support for macOS or Linux, please post below or email support at jetperch.com.

How did this happen?

Building with the right parts is critical. Our contract manufacturer (CM) has processes in place to prevent this exact type of problem. Our CM did a full investigation to understand how this happened, and it is actually three independent failures:

  1. Our CM confirmed that they ordered the correct part TPS2552DBVR directly from the TI store. TI shipped the TPS2552DBVR-1. Now, the ending numbers on parts are usually packaging options, like different tape & reel sizes. In this part, the “-1” indicates different functionality. TPS2552DBVR-1 is a bad part number choice by TI, and it appears that even TI can’t handle it correctly!
  2. Our CM has a receiving process. The receiver diligently noted the “-1” variation. Unfortunately, the receiver also interpretted that “-1” as a packaging variation.
  3. Our CM uses optical inspection. The TPS2552 is labeled “2552” and the TPS2552-1 is labeled “CHX”. Unfortunately, the first few units were built with correct TPS2552 parts from a previous build. When that quantity was used up, the CM switched to a new component reel. The optical inspection caught and flagged the “CHX” change. The operator looked at the TPS2552 datasheet, and saw that “CHX” is a valid indicator for the TPS2552. Since this part had already been cleared by receiving, the operator assumed that this variation was appropriate. The operator updated the optical inspection to accept the CHX and the build proceeded.

Lessons Learned

  1. Suppliers: choose part numbers according to the defacto standard, typically a part family prefix, a specific part, part variations, temperature range, and ending with packaging variations.
  2. CM receiving: Don’t assume that the end of a part number is packaging options.
  3. CM optical inspection operators: If part markings change, raise the issue to the manufacturing engineer to double-check, possibly raising the issue to the design engineer.
  4. Engineers: If you use a part with an unusual part number scheme, communicate that to your CM.