We’re working on a product that makes use of USB HS, and are having some connection and consistency issues.
Originally, our product was USB FS-based and we had no connectivity issues; with this second iteration the design was adjusted to use USB HS with a new MCU and the same cable. The ergonomics of the cable in particular are important to our application, rendering typical USB cables with more robust shielding unsuitable.
The cable’s specs:
- A 2-meter long run of Tensility 4-strand cable (not twisted-pair, but has a lay length of around 3.5cm)
- Gold-plated USB connectors and a 4-pin Molex Picolock (gold-plated on both male and female ends) on the device side.
- Inductance of the cable is 1.2uH, capacitance is 97pF, resistance is 0.92 ohm, and impedance is ~111 ohms (which we believe to be relatively close to the USB spec of 90 ohms).
On some systems, the USB HS connection will work consistently and flawlessly each time. On others, it will be port-dependent with success rates ranging anywhere from 30-80%.
Issues seem to occur in decreasing frequency with cable lengths around 2m to 1.6m, and are mostly resolved at around 1.5m. Additionally, any of the following seem to improve our device’s performance:
- Shielding the cable in aluminum foil
- Twisting the cable (~10-20 turns in the 2 meter length) axially
- Creating a twisted pair out of D+/D- on a fully stripped cable (with no shielding)
We have access to some rudimentary test equipment, namely an oscilloscope (Siglent 1104XE), multimeter, LCR meter (DE-5000), and a digital logic analyzer if there are any relevant measurements that are missing.
We have many cables ready to go -- enough for the entire foreseeable run of the product -- but production of the boards is done iteratively in smaller batches with no large batches yet produced. With that in mind: are there any board or cable- related mitigations that are worth investigating in our case?
