Novena PVT2 ECO List

This is a list of all the changes applied to the board from PVT1 to PVT2 (Crowd Supply initial campaign) release. If it's not on this list, it didn't happen.

Each change has the format of issue summary/resolution, and specific change

ECO1: Improve HPD margin
HPD on i.MX6 has a low impedance (10k). This causes marginal HPD performance on some boards. Make R29L stronger.

ECO2: EOL issue U11D
STMPE610 is EOL. Switch to STMPE811, pin-compatible and not EOL.

ECO3: EOL issue audio amplifier
NS4890 is EOL. Replace with compatible NS4890B in two places.

ECO4: Improve kernel panic logging
U10S is too small for good kernel panic logging. Replace with larger capacity I2C EEPROM to store several crashes worth of logs.

ECO5: 47uF 1206 capacitor is wrong AGAIN
The wrong package type (1210) is being ordered for this 1206 part. This was supposed to be corrected on PVT1. Please double-confirm that instruction is understood that this part cannot be substituted with a 1210 footprint.

ECO6: Upgrade accelerometer
Old accelerometer is probably approaching EOL.

MMA8452Q is stocked at 3mm+ levels in the factory and has existing Linux drivers in main-line. A little more expensive than MMA8653FC, but twenty cents is not going to kill this product. So, we go with MMA8452Q.

ECO7: Upgrade FAN53540UCX
FAN53540UCX marked as not recommended for new designs.

FAN53541UCX is a pin-compatible upgrade. Difference between the two is that the enable and mode input no longer need 1k ballast resistors. However, leaving them in place causes no harm and allows for downgrade compatibility in case parts are in short supply.

ECO8: Update i.MX6 P/N
Previously schematic listed the 1.0GHz P/N as the default. Updated to reflect the plan to only produce with the 1.2GHz P/N.

ECO9: Confirm change of D10C
Measurements on diode leakage indicate D10C was not properly updated on BOM at last ECO. Confirm factory received ECO update.

ECO10: U12N is EOL
U12N, the lower synchronous switch of the 5V converter, has gone EOL. Replace with recommended substitute by On Semi.

The recommended substitute NTMFS4C35NT1G is pin-compatible. It has a slightly lower power rating due to a higher Tj to the case, but the RdsON is the same. The converter is actually highly over-spec'd for the measured loads, so this substitution is not a problem. In fact, the substitute may improve efficiency slightly as the gate charge and other parasitics of the new device are lower.

ECO11: Remove AW-NU137 option
AW-NU137 is EOL and also, we're just never going to use it. It was originally included as an option for wifi in devices that needed to be cheap and where it was acceptable to not have PCIe and also binary blobs are acceptable. There has been zero demand for this configuration option.

Also, swap USB_PCI with USB_VID ports on upstream hub for improved routability. This should have no impact on the software.

ECO12: Fix CDP issue on USB port EXT2
USB port EXT2 has 1.5A-capable limits but does not configure itself as CDP. Fix this by swapping Port4 (EXT2 on PVT1) for Port 1 (USB_VID on PVT1). Amber1/Green4 NC (as currently configured) sets only Downstream Port 1 as CDP-capable.

This is purely a wire fix, no BOM-visible ECO required.

ECO13: Improve inductor profile
The ferrite MSS1260 used as the primary power inductor is a bit too tall and large. Use a composite molded core XAL7030 which is a bit more compact.

Pros/cons: XAL7030 has a better rated Irms than MSS1260, but seems mostly due to better heat conduction out of the package, not due to better electrical performance. The DCR of XAL7030 is 19.5mOhm, MSS1260 is 12.60mOhm, so conduction losses are greater. Isat of the XAL7030 is better due to soft-saturating characteristic of the composite core. Overall, I expect there to be very little change in practical circuit performance, perhaps a 1-2% max loss in battery life with this change under Ipeak conditions.

The MSS1260 isn't removed from the PCB layout. Instead, XAL7030 shall be fitted inside the footprint of the MSS1260. This allows for going back to the original inductor in case there is a design issue with the XAL7030.

This change is coordinated onto Senoko BOM as well.

ECO14: Rework front panel connector
For integration with laptop, a unified front panel connector is introduced.

P16D (8-pin FH19SC from Hirose) is removed. All signals from this connector are re-routed to a new P16D, now an HRS FH34SRJ-30S-0.5SH.

This 30-pin header now contains the following features:


 * both USB ports going to the front panel
 * 6x GPIO routed from the FPGA (so as to emulate at least SPI (with WP/HOLD), 3x I2C, or just function as 6x GPIO)
 * TS_ANA analog input
 * power routing for prototyping convenience
 * some CPU GPIO to provide for blinkenlight on the front panel
 * the power switch from Senoko (to turn the system on and off from the front panel)
 * a front panel general purpose switch (to allow for recovery boot and BT keyboard association)
 * speakers (original headers still retained) to allow for cleaner in-case routing of wires

The major downside of this revision is the internal USB headers can no longer be tapped using a simple DIY 8-pin ribbon cable. However, the 8-pin ribbon cable was found to have inadequate performance for stable USB2.0 operation anyways.

U14F was sacrificed. It's a SPI ROM that hangs off the FPGA, which has never been used in any design. The intention of U14F was to store private keys or other local NV storage for the FPGA that was off hard-drive. Instead, the SPI pins are run to the 30-pin header so users can have more flexibility on what SPI device to attach (you can still attach a device similar to U14F, but now it can be anything else as well).

ECO15: DNP digital mic header
Can't use header at the same time as the LCD adapter board, so the part is redundant.

ECO16: DNP backside switch
Backside user switch is not used in this configuration. Remove.

ECO17: USB hub resets
Sometimes the USB hubs don't reset nicely across soft-power cycles. Wire up hard resets to the CPU so during power on the hubs can be properly reset.

DI0_PIN4 / ball P25 is USB_HUB1_RST (most upstream hub) DISP0_DAT6 / ball R23 is USB_HUB2_RST (most downstream hub)

Further validation indicates resetting the hubs has dubious value, so wire these using DNP 0 ohm resistors so this is a stuff option, but not actually implemented in the run

Reset lines also have pull-ups added to them, so in case they float they go to a sane value.

ECO18: Power LED turn-off
Power LED cannot be turned off currently as it monitors the 2.5V line, and during suspend this stays on.

Rewire LED to go to P1.8V_VGEN4, which is currently an unused LDO output from the regulator. Change color to red so it still turns on (dimly) at 1.8V. LED can then be powered off finally by configuring the LDO once the kernel is up and running.

The power LED is important because the LCD screen doesn't turn on for several seconds after hitting the power button, and also it's impossible to distinguish between a messed up bootloader and a power failure without a power LED at boot.

ECO19: Use external RTC
RTC drains a lot higher current than 4uA according to spec sheet.

Introduce an alternate RTC to provide time. Eliminate supercap for LICELL, as we're not using that line anymore.

We use the same RTC as on the Hummingboard.

ECO20: Eliminate redundant CM chokes
L15U and L12U are ineffective and cause yield issues. Remove.

ECO21: Add IRQ line to Senoko
This helps simplify getting the host's attention because it's too hard to write a console driver in kernel space that listens to the UART interface for alerts.

GPT_CLKIN is repurposed to run to the Senoko header. This means you can't use the high-resolution timing feature with UART3 at the same time as the battery board, which is probably an obscure enough combination that maybe only one person might complain about that.

=Post bring-up ECO=

ECO22: swap audio control option
KEY_ROW1 is meant for the front panel button. Swap resistors to gang audio amp control again.

Issued to factory as ECO-0001 against PVT2 rev A on September 30 2014.

ECO23: improve regulator stability on cold power-on
ECO13 changes the inductor. This has introduced some instability on cold power-on. Oscilloscope waveforms show a lot of noise on the 5V line after jamming the power cable in, and the 5V regulator subsequently shuts itself off.

Adding another 22uF cap on the 5V output filtering (C23N DNP->populated) eliminates this noise and the system seems to perform stably across multiple cold power-on events.

Issued to factory as ECO-0002 against PVT2 rev A on October 15 2014.