Supercap stay-alive schematics

[CDFER]

positron96
2 days ago

Hi @CDFER , sorry for a totally unrelated question, but I was reading through the thread and found out your idea about ESU supercap solution and an IC reference. I've been wanting to design a supercap stay-alive device for some time, and would appreciate any knowledge about existing solutions. Do you know what ESU logic for supercap utilization is? I'd assume they are charged by a linear regulator (or a buck converter?), and then stepped up with a boost converter you mentioned. But what's determining when the boost converter switches on (and buck switches off, as I'd think it's a bad idea to keep it on while boost is on too)? Is there a voltage comparator, or a separate MCU on the loco board? Are all of the components on the loco board dedicated to stay-alive functionality? That's quite a lot of components.

[CDFER]

The best start would be this schematic I found some time ago on a German forum: https://www.stummiforum.de/t171549f21-RE-SuperCapLader-im-Eigenbau-Goldcaps-als-Pufferspeicher-5.html

It uses a Zener diode (D4) on the enable pin of the buck converter.

You just need to ensure there is a secondary output diode on the boost converter and set the boost converter output voltage lower than the Zener diode rating on the buck.

ESU has designed a few PCBs for the British company Accurascale, and as far as I can tell, they use a similar setup (but with different buck and boost chips).

In particular, they used the ME2149 boost converter, which is perfect for this application because it can operate down to 0.7V. This means it can utilize more of the energy stored in the supercap (or might even be able to use a single 2.7V supercap rather than two in series).

ME2149.pdf

I have played around with making my own, and it works well provided the supercaps can supply enough current (basically you need 1F or more to get the 1-2A you need).

Alternatively, I have thought about just putting a cheap MCU on the board, like the CH32, to read the voltages and control the EN pins.

Theoretically someone could build make this design even smaller by making a bi-directional buck-boost converter like this aplication schematic but I can't see a practical way to do it in a compact way with the currently avaliable chips but TI has shown how you could do this in slva726.pdf

[positron96]

Wow, thanks for the schematic, this is extremely useful and these things are hard to come by! It indeed looks like a ready-to-build well thought solution. I'll need to read through all the pages of that forum thread though...
So, if I understand correctly. boost converter always works (if en pin is shorted to vin), but since it's configured to output ~10V, it does not boost when input voltage is higher (via its feedback). Buck converter, on the other hand, starts when input voltage is above ~11V (zenner+ Ven_hi)

I was also wondering how these circuits work when CV readout happens, and it seems they simply don't engage the boost converter because the voltage doesn't drop too low (as command station is supposed to provide enough current).

Before I saw a solution with a zenner diode, I also was thinking about an MCU, for space reasons I considered some SOT-23 MCUs among newer ATtiny chips, like an ATtiny10. A comparator was also among my ideas.

I also saw some bidirectional buck-boost converters from TI, TPS61289 in particular, but it requires some circuitry to control direction. It also needs one external mosfet for some reason, but still promising.

Regarding ESU/Accurascale PCB, were you able to identify other components? I see 2 33uH inductors, are those for switching or for motor filtering? I think these are switching inductors, then the motor filters are strangely missing... The IC at the bottom seems to be XLSEMI XL6007, which is a boost converter (a 33uH inductor is then straight from its datasheet). In top right I think I can read a buck converter, ME3148A, rather then an ME2149. ME3148 has 47uH inductor in the datasheet though, but it's still closer to 33uH on the PCB then the 3.3uH from ME2149's datasheet.
There is also something like an LDO in top left corner, completely unreadable to me.

[CDFER]

I'm not too familiar with DCC programming tracks but I suspect the three wire stay alives have the third wire basically just hooked up to the buck-converter en so that the load on DCC controller can be managed (maybe based on voltage drop or just a random delay at startup idk).

The TPS61289 is rather interesting looking (I'm supprised I haven't come across it before) but it does require quite a few supporting components but I'll have to take a deeper look into it at some point. If only someone made a nice all in one chip like the TPS61094 but at a higher voltage.

I'm not sure how I got that mixed up yes it looks like an XL6007 and ME3148A. I think I saw the ME2149 on some pcb but I can't find any photo for it now. But I have used it on a prototype myself and it works quite well.

I have tracked down another photo and I think the 'LDO' in the corner has the markings ALP11 which leads to the BCX53 power PNP transistor and I suspect that in combination with the surrounding components it is forms a current limiting curcuit for the supercaps.

[positron96]

So, I've read all the pages on the forum, and got some ideas.

To summarize, it seems the thread has major 2 versions of the PCB design:

1. First design where charge current is limited by a 10ohm resistor.
2. Second design, "fast charger", that uses the same fixed voltage buck converter, but hacks its feedback to also limit current. (I've seen the same technic in chinese CC/CV modules).

Here is a variation that I've found, I think it's not from the author of the thread, but from the same source as the schematic from your post. Uses different ICs

source

This got me thinking that charging side can be better implemented by not a fixed voltage DC-DC, but something like a CC/CV battery charging IC. This should probably make it simpler compared to second design.

By the way, can you tell the origin of the schematic you supplied in the top? I've looked through posts of original author of the charger PCB, and his PDFs have a different-looking schematic (from Eagle it seems), while this one looks drawn without an EDA.

[CDFER]

I've tracked down the origin of the schematic on the German forum: https://www.stummiforum.de/t171549f21-RE-SuperCapLader-im-Eigenbau-Goldcaps-als-Pufferspeicher-3.html.

I agree that a CC/CV approach for charging is better, but I'm worried that most CC/CV battery charging ICs use an LDO or similar regulator. This would produce a lot of heat given the high input voltage (>16V) even with small currents.

Ideally, I think we would want a buck converter with a built-in configurable current limiter (set with an external resistor). While buck-based battery chargers do exist, finding one that is compact and can handle >16V input is proving difficult. It seems like it's just a matter of trawling through hundreds of components to find a fit!

I also am not quite sure where to set the current limit. We need a low startup current if 10+ locos are starting simultaneously, but we also want a fast recharge after going over a dirty track section. This is where I think adding a micro could be a neat solution, allowing us to implement either a random delay at startup or switch between two constant current settings for charging. But the trouble is getting that to fit in a compact size and I'm not sure the space it would take up would outweigh the benifits that it would add.

[positron96]

It seems like it's just a matter of trawling through hundreds of components to find a fit!

Yeah, I agree. I tried to search a bit on octopart, haven't succeed yet.
To complicate matters, there should in theory be 3 categories of devices that can potentially fit the bill: buck DC-DC converters with current limit, LED drivers (that are inherently current-limited) with additional max voltage limit, CC/CV battery chargers with configurable voltage and current. All of these can come in switch-mode flavor.

We need a low startup current if 10+ locos are starting simultaneously, but we also want a fast recharge after going over a dirty track section.

Honestly I'd be fine with one current set for both cases (I think I saw one of NMRA standards requiring startup current limit of ~100mA). Considering that the current goes through decoder's rectifier that has its own max current, drawing too much to charge quickly can stress the decoder excessively.