RTC options for E20C / E52C?

[willy]

Hi,

I was wondering if there are any options to populate the HYM8563TS RTC on E20C/E52C. I’m really thinking about replacing my aging devices with combinations of these ones which, from now, are the best match for my use case, but given that upon power outage, my machines boot in roughly 20s while the ISP’s box takes 5 minutes, there’s no way to restart at the correct time for a while, and RTC solves this.

I’ve looked at the pads on the E52C, I feel like I could handle soldering the RTC controller itself, but the 0201 resistors are too small to just maintain them in place during soldering, especially the pull-up ones which are very close to each other. Also, finding a flat xtal is not trivial, the ones I’m having are a bit thick. Thus I’d rather pay one extra buck to have them on board!

[Mitchell]

Hi willy

You can populate the HYM8563TS RTC on the E52C while following the RTC related circuitry on the schematic.

What is your confusion?

[willy]

I’m speaking about having it soldered on the board by default. My board doesn’t have it, and the components are too small for me to risk doing it (the chip should be OK but not the 0201 pull-up resistors). I have no problem with the schematic, that’s where I figured where it was supposed to be (U17 IIRC).

[hipboi]

Sure, let’s just solder it by default. It’s not soldered because the network switch usually doesn’t have RTC builtin. But we are smart network device, we should have it.

[willy]

Ah, many thanks for this @hipboi . I’ll be interested in knowing when it’s safe to order new devices with the chip in them. I’ll happily order a new E52C to replace the current one, as well as a few E20C for my low-power proxies.

[willy]

FWIW I finally managed to rebuild such a tiny module by salvaging some components from an old RK3288 HDMI dongle. I’ve added a 0.33F supercap, a diode for the charge, and two 6k8 pull ups, and connected it to the HYM8563’s SCL/SDA pads (which fortunately are directly connected to the bus), and to the 3.3V capacitor behind the SD card’s connector. I had trouble getting it to be recognized by the driver (it said it had corrupted data) but it turns out that I “just” needed to clear the “VL” (voltage low) bit in reg addr 2 using i2cset and rebind the driver. At this point it was properly recognized and at least it didn’t lose the power during the time it took me to better resolder the wires and reassemble everything. 10 years ago I successfully used a 0.47F capacitor on an ALIX board and it would last about 20 days, so I think I’ll have at least as long. I’ll see. I’m adding two photos of the assembly (I used shrink tube to protect the connectors and forgot to take a photo before that).

So the conclusion is:

• it’s possible to connect an I2C slave to those pads provided you add your own pull-up resistors, which should make it possible to use more accessible devices such as DS3231. Still you’ll need a 0.2mm soldering iron tip and a magnifier to avoid creating shorts between the pads.
• there’s quite some room over the SD connector which is about as tall as the RJ45 connectors and larger than the SD connector, suitable for a tiny CR1220 or supercap plus a small board.
• thin enough wires (I used wrapping colored wire) have no problem passing between the PCB and enclosure to place the device on the top despite the chip being at the bottom.
• there are plenty of RTC boards on aliexpress, I’ll see if the DS3231 I ordered can fit once I desolder the connector.

[willy]

For those interested in adding their own RTC themselves, I managed to make another such module, and it’s definitely doable for a somewhat trained individual. I’ve documented all the steps there with accessible and affordable components. Be prepared to spend at least half a day on it, though:

Adding a cheap and simple RTC to Rockchip devices

[willy]

Hi!

I tried to do the same on the E20C and discovered (after soldering) that there’s already an AT24C16 soldered on board, and that the HYM8563 (and all that family) has the good idea of being addressed at 0x51, which is the second address of I2C eeproms.

I’ve looked for a while and found that DS1308U and DS1338U have the same pinout but use the DS1307 driver and are addressed at 0x68, thus out of the EEPROM scope. I’ve ordered 5 of them and will give this a try.

[willy]

Argh! I’m disgusted! I’ve received my DS1338 chips and they were still not detected. After looking closer at the schematic, I noticed that I was confused by the pins naming: I2C0_SDA_M1 vs I2C1_SDA_M0, so no, the RTC and EEPROM are not on the same bus, so they don’t conflict. In fact I2C0 was disabled in the DTS, reason why I didn’t find it suspicious to have the two on what I believed was the same bus.

After enabling I2C0 in the DTS with this:

--- a/arch/arm64/boot/dts/rockchip/rk3528-radxa-e20c.dts
+++ b/arch/arm64/boot/dts/rockchip/rk3528-radxa-e20c.dts
@@ -18,6 +18,7 @@
 
        aliases {
                ethernet0 = &gmac1;
+               i2c0 = &i2c0;
                i2c1 = &i2c1;
                mmc0 = &sdhci;
                mmc1 = &sdmmc;
@@ -198,6 +199,12 @@
        status = "okay";
 };
 
+&i2c0 {
+       pinctrl-names = "default";
+       pinctrl-0 = <&i2c0m1_xfer>;
+       status = "okay";
+};
+
 &i2c1 {
        pinctrl-names = "default";
        pinctrl-0 = <&i2c1m0_xfer>;

Then my chip is properly detected at 0x68:

Also, I found that DS1338, while interesting (and sucking very low power), requires a distinct Vbat pin that matches the /INT pin of the HYM8563. The copper trace passes very close to the battery diode so it should be reasonably easy to derive it from there, but I’d rather reinstall the 8563 to match the DTS and avoid cutting tracks. Also, one nice thing about DS1338 is that it includes its own capacitors for the Xtal, so it could be more usable for hand soldering. I’ll try without, as often the capacitance of the xtal is within tolerance margin and works, and I’ll see, because quite frankly, soldering 0201, even with a microscope, is a pain, they stick to the soldering tip, and the even fine tweezers are too large to hold them while soldering. Hot air is not easy due to all components in the vicinity.

Going back to replacing that chip now…

Edit: HYM8563 now soldered and working like a charm. I only added the 8563, the 32kHz XTAL (on the left), a low dropout diode between 3v3 and Vcc (below the chip), a 100 ohm resistor to charge the supercap without pulling too hard from the diode (below the diode), and the supercap on the other side, it fits fine on the RTC connector’s pads. I also had to solder two jumpers for SDA/SCL. This is difficult to do at this scale (0.3mm wide by 0.6mm long, the chip is 3mm wide to give an idea of the scale) because the solder wire tends to make big balls when fusing, and to stick to the soldering tip. Anyway I was happy not to have to solder any caps nor the pull up diodes. The chip was pre-positionned using hot air and finished with the soldering iron and flux under a microscope (it’s slightly rotated because I couldn’t use the microscope with the hot air). Here are photos of the bottom (chip) and top (supercap):

I wouldn’t do that every day but now I’m happy that the device is hardware-complete, and I’ll soon be able to use it.

Oh and BTW regarding the supercap, I used these ones: https://www.aliexpress.com/item/1005006869819627.html . That’s 0.07F at $0.16/pc which I estimated should last roughly a month. It avoids the lithium issues and even saves the RTC connector. IMHO it could even be smaller (0.033) since such machines are not supposed to deal with long outages while keeping their clock. @hipboi you might be interested in seeing it at scale here to get an idea of how it fits on the board. I think that for such devices, anything lasting at least a week (~0.022F) would be suited, and then only the lowest price matters.