jakethompson1 wrote on 2023-09-04, 07:43:

[…]
Show full quote

1. It appears safe to use this as an EXT BATT header as-is by removing the jumper on pins 2-3 and instead connecting a 3xAA battery pack across pins 1 and 4, is that correct?
2. I can't read any identification on the diodes, but appears a typical diode in this circuit would drop the voltage by about 0.7V?
3. Is this a NiCd charging circuit with a constant current of 1000*4.3V/150Ω = 29 mA? Isn't that high if these batteries were only 60 mAh?
4. The only purpose of D3 and D4 is to drop voltage? Or is there some chance of power being back-fed to the primary battery even when the connection between pins 2 and 3 is broken, isolating it from the NiCd charging circuit?
5. If I have trouble with boards providing EXT BATT maintaining settings, but not advancing the clock when a 3V CR2032 instead of a 4.5V 3xAA is used, is it safe to remove D1 and D2 and then bypass one or both of D3 and D4?

1) Yes, JP1 obviously is external battery header/connector.
2) Yes, V-drop depending of the type of diode.
3) ---
4) Both.
5) If you want to exchange rechargeable Ni-Cd battery for CR2032, just remove D1(, bypass D2) and put a jumper on JP1 2-3.

What you've outlined in the pic is a fairly common topology for a 4 pin ext battery header. Usually the documentation will describe the jumper on the middle two pins as "enabling the on board battery" or something vs. "clear CMOS" but the operation is the same.

Remove jumper, connect 3xAA or AAA pack to header (use lithium primaries!), done.

To address the other points:
NiCd batteries are incredibly tolerant to "abusive" charging which is why they were so popular back in the day.
Diode forward voltage drop for a standard silicon signal diode is about 0.7V but it's slightly less at extremely low currents. It's not linear.

Note that the schematic shows you can easily replace the NiCd battery with a CR2032 battery holder directly (no need for a pcb module with holder + diode or anything like that) and lift the leg of D1 (I prefer the side closer to 5V) to disable the charging circuit (diode and resistor can be swapped in that path).
In other words there no real need to have a dangling external battery and 3V should be enough to keep the settings which in my opinion is the most important part.

This is my preferred solution to convert motherboards to CR2032 leveraging the existing spot for a battery, the result looks stock and the modification to disable charging is inconspicuous and reversible

3) NiCd chemistry can take 1C charging current. This would be half so safe, and it would still need 2h (plus some extra due to charging not being 100% efficient) to fully charge the battery, not that unusual. Though usually, with daily use, you would not need that long for full recharge. This is a simple circuit, one of the reasons these batteries tend to leak eventually.
4) You can connect CR2032 between pins 3 and 4, that way you will not need to modify the mobo in any way.

Deunan wrote on 2023-12-06, 13:55:

4) You can connect CR2032 between pins 3 and 4, that way you will not need to modify the mobo in any way.

That would make the clock always powered by the battery, even when the system was powered on.

jakethompson1 wrote on 2023-09-04, 07:43:

I have a board that does not have a documented EXT BATT header. But I believe the following CLR CMOS jumper doubles as an EXT BATT header.

Yes, this is correct.

jakethompson1 wrote on 2023-09-04, 07:43:

It appears safe to use this as an EXT BATT header as-is by removing the jumper on pins 2-3 and instead connecting a 3xAA battery pack across pins 1 and 4, is that correct?

The only purpose of D3 and D4 is to drop voltage? Or is there some chance of power being back-fed to the primary battery even when the connection between pins 2 and 3 is broken, isolating it from the NiCd charging circuit?

I haven't traced pin 3 but I can only assume it goes to the appropriate RTC power pin on the chipset (this board doesn't use a discrete RTC).

The two diodes in series are definitely to drop the voltage. For all other purposes, a single diode would do as well. These diodes are not related to the NiCd charging circuit, as that circuit connects only to pin 2, which is left unconnected when you use an external battery. Your diagram does not include another vital component: There will be one or two diodes between +5V and pin 3 of the EXT BATT connector, to power the RTC from the AT power supply while the computer is turned on. The point of the D3/D4 (except for dropping voltage) is to prevent charging the external battery via this path, just as there is D2 which prevents charging the NiCD battery bypassing the 150 Ohm resistor. The "extra drop" with the two diodes most likely adjust from the 4.5V nominal AA battery voltage to the 3.6V NiCd battery pack voltage. It might also be there so you can connect a 4*AA battery pack or 2*CR123 (6V nominal voltage) without exceeding 5V on the output.

Some boards are specified to 4.5V on that connector, others are specified to take 6V.

jakethompson1 wrote on 2023-09-04, 07:43:

I can't read any identification on the diodes, but appears a typical diode in this circuit would drop the voltage by about 0.7V?

Yeah. Unless proven otherwise, just assume those diodes are 1N4148 or diodes with nearly identical characteristics.

jakethompson1 wrote on 2023-09-04, 07:43:

Is this a NiCd charging circuit with a constant current of 1000*4.3V/150Ω = 29 mA? Isn't that high if these batteries were only 60 mAh?

That would be extremely high. The barrel batteries are not optimized for fast charging, and charging them at C/2 without any kind of charge control would kill them within months of operating the computer, making the clock to fail within the warranty period. Your calculation is wrong. While you are correct that the voltage at the node between D1 and the 150 Ohm resistor will be about 4.3V (at low currents, better assume 0.6V, so make this 4.4V), the other end of the resistor is not at ground. During charging of a NiCd barrel, you can assume a per-cell voltage of 1.3V to 1.35V. The battery of 3 cells thus will be around 4V, and the voltage across the resistor will be 4.4V - 4.0V ~ 0.4V instead of the 4.3V you assumed. So the charge current is around 2.6mA, which is just below C/20. This in fact is the recommended "continous trickle charging circuit" for NiCd cells, so the charging circuit is mostly fine.

jakethompson1 wrote on 2023-09-04, 07:43:

[*]If I have trouble with boards providing EXT BATT maintaining settings, but not advancing the clock when a 3V CR2032 instead of a 4.5V 3xAA is used, is it safe to remove D1 and D2 and then bypass one or both of D3 and D4?[/list]

No need to remove D1 and D2. Pin 2 is a dead end when you remove the jumper 2-3 on that connector. Short/bypass only one of D3 or D4, so you still get charging prevention, but cut the drop in half might be the easiest solution to accept 3.0V at the EXT BATT connector (battery still connected to 1-4, as intended by the board manufacturer).

jmarsh wrote on 2023-12-07, 02:00:

Deunan wrote on 2023-12-06, 13:55:

4) You can connect CR2032 between pins 3 and 4, that way you will not need to modify the mobo in any way.

That would make the clock always powered by the battery, even when the system was powered on.

True, but that's how a lot of modern RTCs work anyway. The only problem is older chips that have just too much power draw and would drain the small CR2032 cell quickly, but from my experience RTC integrated into chipset is already quite power-efficient.

That being said I prefer to use 3xAA (or AAA for newer mobos) packs, that last longer (esp. AA batteries on old 286 mobos) and doesn't require charging so with good brand cells, and a nice plastic box away from the mobo it's pretty safe to leave those in for years if the machine is not in storage but also very rarely used.

Deunan wrote on 2023-12-07, 11:18:

That being said I prefer to use 3xAA (or AAA for newer mobos) packs, that last longer (esp. AA batteries on old 286 mobos) and doesn't require charging so with good brand cells, and a nice plastic box away from the mobo it's pretty safe to leave those in for years if the machine is not in storage but also very rarely used.

so fvcking trve king, I'm really sick of the cr2023 zealots shoving them everywhere they don't belong

Deunan wrote on 2023-12-07, 11:18:

jmarsh wrote on 2023-12-07, 02:00:

Deunan wrote on 2023-12-06, 13:55:

4) You can connect CR2032 between pins 3 and 4, that way you will not need to modify the mobo in any way.

That would make the clock always powered by the battery, even when the system was powered on.

True, but that's how a lot of modern RTCs work anyway. The only problem is older chips that have just too much power draw and would drain the small CR2032 cell quickly, but from my experience RTC integrated into chipset is already quite power-efficient.

I would not recommend to connect a battery between pins 3 and 4 before verifying that there is no diode from +5V to pin 3. This was common to power the RTC from +5V while the system is on, and would provide charging to a battery at pin 4 without any current limiting resistor(EDIT: corrected by jmarsh - there needs to be some current limiting resistor to avoid a dead short in the "clear CMOS" jumper configuration! Nevertheless, I still consider some "charging current" at pin 3 very likely). Getting a CR2032 charged is a bigger problem than discharging the CR2032 even while the system is powered on. Thus a diode between the CR2032 and pin 3 would be required - like there already is between pins 1 and 3.

Thus, if you want to use a CR2032, I still prefer my suggestion to bridge one of D3 or D4 and connect the battery to pins 1 & 4. If the voltage isn't high enough for the RTC to run, replace the non-shorted silicon diode with a schottky diode. Or just use 3*AA at 1 & 4 as intended anyway (and also recommended by other posters).

I'd imagine if that were the case, the manual would have a loud warning not to power on the system with the jumper in the CLEAR CMOS position.

Deunan wrote on 2023-12-07, 11:18:

jmarsh wrote on 2023-12-07, 02:00:

Deunan wrote on 2023-12-06, 13:55:

4) You can connect CR2032 between pins 3 and 4, that way you will not need to modify the mobo in any way.

That would make the clock always powered by the battery, even when the system was powered on.

True, but that's how a lot of modern RTCs work anyway. The only problem is older chips that have just too much power draw and would drain the small CR2032 cell quickly, but from my experience RTC integrated into chipset is already quite power-efficient.

That being said I prefer to use 3xAA (or AAA for newer mobos) packs, that last longer (esp. AA batteries on old 286 mobos) and doesn't require charging so with good brand cells, and a nice plastic box away from the mobo it's pretty safe to leave those in for years if the machine is not in storage but also very rarely used.

FYI
I just looked at a few of the 3,6v nicd leaking batteries which I still have laying around after having cut or desoldered them from the mainboards, and they all mention 60mAh.
a google for capacity of a CR2032 gives me 210mAh, though that cr2032 is only 3v still 3 times the capacity of those leaky pests. (not to say that those cr2032 can't ever leak. they sometimes do.)

jmarsh wrote on 2023-12-07, 18:19:

I'd imagine if that were the case, the manual would have a loud warning not to power on the system with the jumper in the CLEAR CMOS position.

Oops, you are right. So there needs to be a current limiting resistor in that path, too. My warning "without a current limiting resistor" is clearly wrong. Thanks for the correction.