I HATE Batteries! - Converting Wireless to Wired (a cheapskate’s guide)

[ImorallySourcedElectrons]

13 hours ago, manikyath said:

actually.. LTT merch idea?

The economic viability of such things is quite questionable if you actually wish to sell it as a company. You got to compete against C2032 batteries and get all the regulatory BS, meanwhile C2032s are dirty cheap.

 

7 hours ago, Key2300 said:

Does your company not have an electronics guy?

They're hiring folks for exactly this.

 

7 hours ago, Key2300 said:

So in general voltage dividers can only be used for constant voltages if there is pretty much zero load on them like with NTC/PTC temperature sensors or if you need a reference voltage for an OP-amp.

Nah, I've used plenty of voltage dividers already where you'd probably throw in a linear regulator. You can use voltage dividers for this purpose if the load is roughly constant, you usually got 10% leeway on supply voltage anyway. And the gate breakdown voltage on a lot of 3.3V CMOS technologies is surprisingly high. The main issue is usually start-up, a lot of digital components don't like slowly ramping voltages because their reset hold circuit is usually designed improperly. Fixable, but annoying.

 

7 hours ago, Key2300 said:

Another thing that immediatly caught my eye was the lack of capacitors in all solutions. You have a very nice video about power supplies where you explain their usefulness when it comes to supplying a constant voltage. And the datasheet of the MIC5504 recommends 2x 1uF capacitors, because like every datasheet for an IC it has a typpical application section that tells customers that wanna use their part how to best implement it. They also only cost about 0,09$ a pop. 

You don't actually need those in a lot of instances for low current applications due to the inherent properties of such loads, putting too much capacitance there can actually cause oscillations. But that goes deep into systems theory territory, TI application note 1148 kind of goes through why you need those capacitors there and which properties the load should have to ensure regulator stability.

 

7 hours ago, Key2300 said:

But then again, for the MIC, why not use a stripboard? They cost like 3 bucks and you can cut them into several smaller ones to use them in several devices and they would've made everything SO MUCH easier! Tbh, I can't imagine that you didn't have one lying around somewhere.

You are clearly not a member of the dead-bug club. I've seen and done worse hackjobs myself, so I ain't commenting on it.

[manikyath]

1 minute ago, ImorallySourcedElectrons said:

The economic viability of such things is quite questionable if you actually wish to sell it as a company. You got to compete against C2032 batteries and get all the regulatory BS, meanwhile C2032s are dirty cheap.

i'd argue false, because the goal is not to be more economical than CR2032, but to not have to replace them anymore. it could cost 10x the cost of a cr2032, because the argument was never about the cost of the batteries in the first place, it's about the annoyance.

 

and if anyone can market an expensive gadget to solve a cheap problem, i'd expect LTT to be the ones

 

also, i'd have to research just how much regulatory BS you actually have to deal with (as long as it's not directly mains powered) but i assume there's plenty of loopholes to get something out the door cheap and functional.

[ImorallySourcedElectrons]

2 minutes ago, manikyath said:

i'd argue false, because the goal is not to be more economical than CR2032, but to not have to replace them anymore. it could cost 10x the cost of a cr2032, because the argument was never about the cost of the batteries in the first place, it's about the annoyance.

True, but how much is that annoyance worth to you?

 

3 minutes ago, manikyath said:

also, i'd have to research just how much regulatory BS you actually have to deal with (as long as it's not directly mains powered) but i assume there's plenty of loopholes to get something out the door cheap and functional.

The issue with these sort of things is EMC directives. Most of the others (e.g., CE) you can self-certify.

[manikyath]

1 hour ago, ImorallySourcedElectrons said:

True, but how much is that annoyance worth to you?

enough to spend the last hour doing this:

i had to substitute some components out because i didnt quite have everything on hand i needed, but essentially:

- standard Qi chargepad

- a hand wound 'coil' from some enameled copper wire that i had laying around from some other wireless charging project.

- diode, and capacitor for crude smoothing

- a relay playing the role of a coil for inrush current smoothing

- two red LEDs that happened to be around 3V voltage drop in total, with a resistor to make sure i dont smash too much current trough them.. but this should pretty much just be a linear voltage regulator.

- a supercap

- an LED to put some load on the circuit.

 

the annoyance of putting batteries in the 3 devices i have that still require button cells is enough for me to actually build solutions.

[ImorallySourcedElectrons]

1 minute ago, manikyath said:

enough to spend the last hour doing this:

i had to substitute some components out because i didnt quite have everything on hand i needed, but essentially:

- standard Qi chargepad

- a hand wound 'coil' from some enameled copper wire that i had laying around from some other wireless charging project.

- diode, and capacitor for crude smoothing

- a relay playing the role of a coil for inrush current smoothing

- two red LEDs that happened to be around 3V voltage drop in total, with a resistor to make sure i dont smash too much current trough them.. but this should pretty much just be a linear voltage regulator.

- a supercap

- an LED to put some load on the circuit.

 

the annoyance of putting batteries in the 3 devices i have that still require button cells is enough for me to actually build solutions.

I must admit I just solder wires to the terminals and use cheap lab supplies.

[Key2300]

3 hours ago, ImorallySourcedElectrons said:

Nah, I've used plenty of voltage dividers already where you'd probably throw in a linear regulator. You can use voltage dividers for this purpose if the load is roughly constant, you usually got 10% leeway on supply voltage anyway. And the gate breakdown voltage on a lot of 3.3V CMOS technologies is surprisingly high. The main issue is usually start-up, a lot of digital components don't like slowly ramping voltages because their reset hold circuit is usually designed improperly. Fixable, but annoying.

Okay yes, they can also be used for constant loads because then the load always influences the voltage coming out of the divider in the same way. But I'd just use a linear regulator immediatly instead of fucking around and finding out because they also cost nothing and it just saves time instead of experimenting.

 

3 hours ago, ImorallySourcedElectrons said:

You don't actually need those in a lot of instances for low current applications due to the inherent properties of such loads, putting too much capacitance there can actually cause oscillations. But that goes deep into systems theory territory, TI application note 1148 kind of goes through why you need those capacitors there and which properties the load should have to ensure regulator stability.

Yes you don't always need them but so far following an ICs datasheet has never failed me and 18 cents for some reassurance doesn't hurt.

 

3 hours ago, ImorallySourcedElectrons said:

You are clearly not a member of the dead-bug club. I've seen and done worse hackjobs myself, so I ain't commenting on it.

Apparently not, no. I mean I'd say that I do hackjobs all the time but I guess I try to apply a minimum standard to them. Using a stripboard for the linear regulator and co. would've been a hackjob in my opinion. The fancy way would've been to order a custom pcb to solder the components on to that has the perfect size to press against the battery contacts.

[ImorallySourcedElectrons]

12 minutes ago, Key2300 said:

Okay yes, they can also be used for constant loads because then the load always influences the voltage coming out of the divider in the same way. But I'd just use a linear regulator immediatly instead of fucking around and finding out because they also cost nothing and it just saves time instead of experimenting.

There are plenty of reasons to avoid linear regulators in some applications. A resistor is simple, reliable, cheap, and doesn't take much volume. It's even better if you can limit it to a single series resistor, but often you got to add the second resistor to get the supply impedance into the right range. It's especially nice if it means you can just grab an off-the-shelf supply, add a resistor to the wiring harness, and just ditch the PCB altogether.

 

12 minutes ago, Key2300 said:

Yes you don't always need them but so far following an ICs datasheet has never failed me and 18 cents for some reassurance doesn't hurt.

It's failed me plenty of times, I learned the hard way that you should always calculate the stability of voltage regulators in any circuit that's pushed into production or continuous use. That was my first €60k screw up.

 

26 minutes ago, Key2300 said:

Apparently not, no. I mean I'd say that I do hackjobs all the time but I guess I try to apply a minimum standard to them. Using a stripboard for the linear regulator and co. would've been a hackjob in my opinion. The fancy way would've been to order a custom pcb to solder the components on to that has the perfect size to press against the battery contacts.

I'm going to guess you mostly do digital work? For analog (RF and high precision mostly) we often resort to dead-bug hovering over a copper clad board, it's easily the best method to get proper performance without having to first design everything on a computer.

[Key2300]

6 hours ago, ImorallySourcedElectrons said:

There are plenty of reasons to avoid linear regulators in some applications. A resistor is simple, reliable, cheap, and doesn't take much volume. It's even better if you can limit it to a single series resistor, but often you got to add the second resistor to get the supply impedance into the right range. It's especially nice if it means you can just grab an off-the-shelf supply, add a resistor to the wiring harness, and just ditch the PCB altogether.

To be honest, you probably have more experience than me, but so far linear regulators and/or switching regulators both have always been less of a hassle for me than trying to achieve a voltage that is necessary to power soemthing with resistors.

 

6 hours ago, ImorallySourcedElectrons said:

It's failed me plenty of times, I learned the hard way that you should always calculate the stability of voltage regulators in any circuit that's pushed into production or continuous use. That was my first €60k screw up.

Here I have to ask how it happened that it was pushed into production. I don't mean to be an ass with that question, I definitely work on a much smaller scale than you do so I'm just trying to gain insight on how those things can happen. When I design a new pcb we order 5 of it and test them rigurously. If there is a design error we notice it (so far at least we always noticed it) and it's back to the drawing board. We wouldn't put anything to use that didn't pass all short and long time tests. And I at least thought that all companies work like that so I'm really curious how that slipped past.

 

6 hours ago, ImorallySourcedElectrons said:

I'm going to guess you mostly do digital work? For analog (RF and high precision mostly) we often resort to dead-bug hovering over a copper clad board, it's easily the best method to get proper performance without having to first design everything on a computer.

By digital, do you mean that the PCBs I design mostly work in a digital sense with 1s and 0s? High and Low? Because in that case, yes. There are some analog ones but most have an MCU on them that takes inputs (analog and digital) and handles outputs. Not a whole lot of analog tech on them.

If by digital you mean the designing and testing process then I'd say not exactly mostly. I design both the circuits and the layout digitally, assisted with simulations and calculations but once the pcb reaches us we solder all components on and test it mostly analog to check if everything is doing what it's supposed to do. The only digital part there is if the MCU is seeing expected values with expected inputs.

[ImorallySourcedElectrons]

11 hours ago, Key2300 said:

To be honest, you probably have more experience than me, but so far linear regulators and/or switching regulators both have always been less of a hassle for me than trying to achieve a voltage that is necessary to power soemthing with resistors.

I think it really depends on the industry you're in, I've mostly used the resistors in cheap disposable sensor tags where volume was a concern (and you can simply print those with carbon paste), and high-reliability devices where the failure mode was a concern. A resistor backed up by a super-zener had a safer failure mode.

11 hours ago, Key2300 said:

Here I have to ask how it happened that it was pushed into production. I don't mean to be an ass with that question, I definitely work on a much smaller scale than you do so I'm just trying to gain insight on how those things can happen. When I design a new pcb we order 5 of it and test them rigurously. If there is a design error we notice it (so far at least we always noticed it) and it's back to the drawing board. We wouldn't put anything to use that didn't pass all short and long time tests. And I at least thought that all companies work like that so I'm really curious how that slipped past.

High speed temperature measurement system for plastics manufacturing monitoring. Combine that with incompetent management pushing unrealistic deadlines, young me not having enough of a spine to tell them to fuck off and give me the time to do my job properly, and you know the rest.

 

As to why it was happening, we had a single lot of test devices from a single manufacturer, purchasing substituting those for a second source without letting us know, and the final devices were running in scorching hot factory halls. Turned out that it was borderline stable in the original configuration, our lab was always a nice 20°C, but sitting in a 50°C ambient environment pushed it over the edge by significantly increasing the current draw through the regulator while simultaneously significantly increasing the losses in the output filter. This caused some issues with the feedback to the adjust terminal and made the stability marginal at best. Intermittent oscillations on power supply rails are tricky to catch though, but luckily we found it quite quickly.

 

11 hours ago, Key2300 said:

If by digital you mean the designing and testing process then I'd say not exactly mostly. I design both the circuits and the layout digitally, assisted with simulations and calculations but once the pcb reaches us we solder all components on and test it mostly analog to check if everything is doing what it's supposed to do. The only digital part there is if the MCU is seeing expected values with expected inputs.

For many analog designs, SPICE and its brethren don't really cut it, say something like injection into an oscillator to capture a carrier is notoriously unreliable in SPICE (to give the stereotype example). So physical breadboarding is nice as a sanity check. Stripboard is fine up to a few 100 MHz, but beyond that it becomes tricky, and that's when dead bug over copper truly shines. Same when dealing with precision circuits, if you're trying to stop nA and pA leakage currents, deadbug on PTFE standoffs is your best bet. But these are analog design issues that most folks don't run into, so they're not as familiar with this prototyping technique. So I can't say there's anything wrong with someone doing this, because we genuinely use it quite frequently.

[BioSehnsucht]

Last year this exact issue with replacing batteries on my Ecobee room sensors bugged me enough to finally do my first PCBWay project to make a CR2032 sized 5V to 3V shim that can be wired up to USB or whatever 5V source. I only just today got around to actually soldering one up (turns out itty bitty SMD parts are practically impossible to hand solder with regular iron, I finally got around to ordering a MHP30 SMD hot plate and some proper solder paste).

 

Only to realize after testing it "worked" (regulated voltage on the "battery terminals") that I must have been half asleep when I ordered the SMD voltage regulators, because I bought 1.5V ones instead of 3V. D'oh! But it makes 1.5V just fine even if that's useless. Ordered up some 3V ones and maybe in 6 months I'll get around to it again ...

 

Also it was a very tight fit into the room sensor, even if nominally a CR2032 is 3.2mm and thus you'd naively think that stacking two 1.6mm PCBs would be 3.2mm, that's forgetting to account for the actual conductive surfaces/traces/etc as the nominal thickness is the PCB itself, so the whole assembly is just a hair too big to fit nicely (though it can be forced in).

 

Doing it this way, versus the hack job LTT did, of course should mean that it actually works since it's a using a voltage regulator... instead of the problems they had with their "simple" approach.

 

If I get around to it I might polish it up a bit and toss the design on Github or something, so others can order their own. I'm definitely not about to go into business selling these.

 

Here's my ugly first time doing SMD with the hot plate, and what the two PCB layers look like bare (everything interesting happens on the bottom PCB, the top PCB just exists to make it 3.2mm thick and provide the top/side positive connection to emulate the physical form of the CR2032, it's connected to the big J3 pad via a castellated edge, the soldered up one you can't really see that well because of how I oriented the wires). Uses an ADP121 series voltage regulator (TSOT-23-5 package) and a pair of 1uF caps (0805 package), so they're itty bitty and a super pain to work with, especially when you start off with zero SMD related equipment and try at first to do it with a regular iron. My iron soldered failures are ugly. The third 0805 pad pair are because I laid out the PCB to work with both the ADP121 style regular and another one, though I don't remember which it was off hand.. basically only one of the Cout1/Cout2 caps gets populated depending on where the output pin is on the regulator.

[manikyath]

On 2/11/2023 at 4:51 AM, BioSehnsucht said:

Last year this exact issue with replacing batteries on my Ecobee room sensors bugged me enough to finally do my first PCBWay project to make a CR2032 sized 5V to 3V shim that can be wired up to USB or whatever 5V source. I only just today got around to actually soldering one up (turns out itty bitty SMD parts are practically impossible to hand solder with regular iron, I finally got around to ordering a MHP30 SMD hot plate and some proper solder paste).

 

Only to realize after testing it "worked" (regulated voltage on the "battery terminals") that I must have been half asleep when I ordered the SMD voltage regulators, because I bought 1.5V ones instead of 3V. D'oh! But it makes 1.5V just fine even if that's useless. Ordered up some 3V ones and maybe in 6 months I'll get around to it again ...

 

Also it was a very tight fit into the room sensor, even if nominally a CR2032 is 3.2mm and thus you'd naively think that stacking two 1.6mm PCBs would be 3.2mm, that's forgetting to account for the actual conductive surfaces/traces/etc as the nominal thickness is the PCB itself, so the whole assembly is just a hair too big to fit nicely (though it can be forced in).

 

Doing it this way, versus the hack job LTT did, of course should mean that it actually works since it's a using a voltage regulator... instead of the problems they had with their "simple" approach.

 

If I get around to it I might polish it up a bit and toss the design on Github or something, so others can order their own. I'm definitely not about to go into business selling these.

 

Here's my ugly first time doing SMD with the hot plate, and what the two PCB layers look like bare (everything interesting happens on the bottom PCB, the top PCB just exists to make it 3.2mm thick and provide the top/side positive connection to emulate the physical form of the CR2032, it's connected to the big J3 pad via a castellated edge, the soldered up one you can't really see that well because of how I oriented the wires). Uses an ADP121 series voltage regulator (TSOT-23-5 package) and a pair of 1uF caps (0805 package), so they're itty bitty and a super pain to work with, especially when you start off with zero SMD related equipment and try at first to do it with a regular iron. My iron soldered failures are ugly. The third 0805 pad pair are because I laid out the PCB to work with both the ADP121 style regular and another one, though I don't remember which it was off hand.. basically only one of the Cout1/Cout2 caps gets populated depending on where the output pin is on the regulator.

since you seem to have boatloads more design possibilities than me (handsoldering SMD components together..)

do something with this if you like:

it's just a 'VL1220' button cell battery with coil and very crude charging circuit squeezed into a CR2032 format.

i'm sure that if it were put together by someone competent that it would make for a great product.

[BioSehnsucht]

1 hour ago, manikyath said:

since you seem to have boatloads more design possibilities than me (handsoldering SMD components together..)

do something with this if you like:

it's just a 'VL1220' button cell battery with coil and very crude charging circuit squeezed into a CR2032 format.

i'm sure that if it were put together by someone competent that it would make for a great product.

Interesting. Seems these are pretty simple devices (the cell), to the point that I can't find any kind of purpose-built charge controller IC (if they required anything fancy, probably one would exist). Looking at https://industrial.panasonic.com/cdbs/www-data/pdf/AAF4000/AAF4000COL6.pdf that in addition to a charging circuit using a voltage regulator to provide a specified voltage, they also give a charging circuit example from 5V that isn't even using any kind of voltage regulator, it's instead just diodes and resistors, so I guess they're pretty confident that it doesn't need any kind of complicated charge cycle, and in the case of using a regulator there's nothing specified, so it must be assumed to be a constant voltage charge.

 

Trying to package it into a faux CR2032 package will be harder obviously though, compared to my external 5VDC input design (ignoring that obviously the wires for my design are exiting a CR2032 and thus it doesn't "fit" in that sense to begin with), since so much space is taken up by the cell. Might be doable, but will be a bit more fiddly to solder up with the cell.

Is that coil off the shelf or did you wrap it yourself? Is it just a coil straight into the regulator? I thought there was more complexity with some kind of signaling for these things. Some brief googling makes it sound like the back channel signaling might be optional though?

 

I wonder if you could etch a PCB coil directly onto the board that small which would work...

[manikyath]

7 minutes ago, BioSehnsucht said:

Interesting. Seems these are pretty simple devices (the cell), to the point that I can't find any kind of purpose-built charge controller IC (if they required anything fancy, probably one would exist). Looking at https://industrial.panasonic.com/cdbs/www-data/pdf/AAF4000/AAF4000COL6.pdf that in addition to a charging circuit using a voltage regulator to provide a specified voltage, they also give a charging circuit example from 5V that isn't even using any kind of voltage regulator, it's instead just diodes and resistors, so I guess they're pretty confident that it doesn't need any kind of complicated charge cycle, and in the case of using a regulator there's nothing specified, so it must be assumed to be a constant voltage charge.

 

Trying to package it into a faux CR2032 package will be harder obviously though, compared to my external 5VDC input design (ignoring that obviously the wires for my design are exiting a CR2032 and thus it doesn't "fit" in that sense to begin with), since so much space is taken up by the cell. Might be doable, but will be a bit more fiddly to solder up with the cell.

Is that coil off the shelf or did you wrap it yourself? Is it just a coil straight into the regulator? I thought there was more complexity with some kind of signaling for these things. Some brief googling makes it sound like the back channel signaling might be optional though?

 

I wonder if you could etch a PCB coil directly onto the board that small which would work...

the way to charge them is to simply apply a voltage (below 3.5 volts) trough a resistor to limit the current. that's all there's to it. it's almost as simple as charging a NiMH.

 

the coil is some very thin enameled copper wire that i just wound '30-ish times' around the base of my flux pen, which happened to be the correct size, and then some glue to mostly set it in form.

 

past that, the circuit is just this: diode => 3.3v linear regulator => resistor => diode (to stop backfeeding) => LR1220

i'm pondering if diode => zener diode to ground (to make 3.5-ish volts) => diode => LR1220 might be enough, given that such a fine coil cant really pass much current anyways.

 

the biggest problem i see with etching the coil is how many turns you could put in a certain size. if doing it at all it should be a flatflex around the outside, just to keep the coil as big as possible (with such a small coil it becomes tricky to receive power from a standard Qi charger)

 

all in all.. especially if the circuit can be slimmed down more (if not zener diode, use a SOT23 linear regulator instead of a SOT223 one..) it should be feasible to assemble it into a CR2032 package without having to do as much leg trimming as i had to do.

[BioSehnsucht]

I did some preliminary sketching up in freecad and kicad trying to figure out how to cram it all together. This was with laying it out assuming that the coil can't be occluded by the negative CR2032 terminal or the battery on the bottom, so you end up with a fair amount of "wasted" space from needing to not have overlap of those things...

 

With those assumptions, it's hard to fit the coil, the battery, the SMT components all in the CR2032 package. Even if you switch from regular FR4 PCB (1.6mm thick) to flex PCB (~paper thin), if you end up with the battery and the SMT components stack on top of each other, the battery is 2mm thick and the SMT components can be up to 1mm or so, leaving you .2mm for top and bottom casing, if you can't squeeze the SMT components around the VL1220 in the remaining space in the horizontal plane... gets pretty tricky real quick to squeeze components in.

 

Probably doable, but will require some real precision engineering.

[BioSehnsucht]

Having done some more researching into Qi charging, I have concerns about the longevity of any Qi charged cells, due to potential heating from inductive charging effects.

 

Apparently, to protect things from inductive heating (as well as to improve efficiency), Qi coils normally backed by ferrite sheet which provides a magnetic shield. However, I'm not sure how to do that and also still have a negative (or positive - whichever side is intended to face the Qi charger) terminal for the "CR2032".

 

Otherwise, I think everything else is solvable. As both a learning experience with designing 3D printed parts (we got a MK3s late last year, have mostly just printed things from thingiverse / printables so far) and to further looking into the practicality of this idea, I printed up a set of parts for forming the positive part of the CR2032 "can". It worked surprisingly well - I had expected to need more force (a screw-clamp, vice, or press) but I managed to form the shell out of a small piece of ~0.25mm sheet metal (nominally it is intended for use as roof flashing around a chimney or similar object) with just the press of a regular Irwin Quickgrip clamp. So a DIY metal housing is entirely doable, in place of simply forming it out of two layers of PCB, and I think you could just barely wedge the coil, VL1220, and components with a flex PCB into the can... but I think you would probably slowly cook things to death while charging over time.

[Mysticode]

I have the newer ecobee sensors that take cr2477 1000mah batteries. 

 

What works the best today to get a wired power hack to work? I'm sick of these batteries!