Replacing an old 3.6V backup battery

[cincin]

I'm working on an old synthesizer that has a 3.6V GE DataSentry backup battery in it, to keep RAM from blanking out when it's off. I hear these are notorious for leakage that ends up corroding the underlying PCB.

I've replaced a similar battery in a Korg PolySix before, it was a rechargeable 3.6V battery, and the generally agreed upon replacement was to put a 3V CR2032 coin battery holder in its place, along with a diode to prevent the circuit from trying to charge the CR2032. In that case 3V worked fine in place if the 3.6V.

I'm having a harder time finding the prescribed solution for the synth that's currently on the bench: a Kurzweil K250.

I'm attaching the schematics for that board, with the battery highlighted in green. It looks to me like there's a charging mechanism in place on this one too.

My questions are:

- Will a 3V CR2032 coin battery work in this synth even though it's 3V, not 3.6V? I imagine this may depend on the ram chips in place. They are "JAPAN 8284HM6264LP-15", datasheet is here:
https://www.digchip.com/datasheets/parts/datasheet/740/HM6264LP-15-pdf.php
The standby requirement are given in a way that I'm not sure I understand, picture of that attached as well.

- Could I use a rechargeable LR2032, which are 3.6V, in this scenario? Would the charging circuit work at charging an LR2032?

- If the 3V CR2032 would work, but not the 3.6V LR2032, can I just place a diode between Q2 and the battery (but not between Q2 and CR5) to prevent the 5V rail from touching the battery and attempt to charge it?

Thanks!

[mariush]

You could just use 2  CR2032 in series  followed by 3-4 diodes (1n400x or something that has high forward voltage even at low currents)

2 x 3.2v = 6.4v  - ( 4 x ~ 0.7v) = 6.4 - 2.8 = 3.6v ... make it only 3 and you'll have a wider operating range.  The diode will also block the circuit from charging the battery.   But you can probably disable charging the battery by cutting or breaking the trace between battery and Q2.

Doubt the actual voltage is really critical, because there's a diode CR5 there which will drop another 0.5v .. 0.6v so your actual voltage going to chips will be more like 3v .. 3.3v

[CaptDon]

It would be good if you had a DMM and could look at the current drain on a 3 volt coin cell after powering down the unit. A CR2032 has around 225mah of energy and hopefully your total drain is 10ua or so then you can calculate the time until a replacement battery will be required. I think those 6264 rams are good down to 2.0vdc guarenteed. Your diode on the collector of Q2 could work with an LIR2032. The charge current could be a tad high and may need the value of the Q2 emitter resistor increased. Be aware, the rechargeable LIR2032 I think only has 95mah of stored energy.

[wasedadoc]

3 NiMH button cells in series.

The original battery was probably 3 NiCad cells in series.

[tszaboo]

Terrible ideas all around.
Get a LiSOCl2 battery, use that. That's also the original chemistry, but a 1000mAh would last forever + 10 years in place of the original 150mAh one

[cincin]

Thank you all for the input! However I am still unclear about the way to proceed

I'd like to keep it as clean and "drop-in" looking as possible. So if I can avoid multiple coin batteries and/or multiple diodes that would be better.

I do have a DMM and could try measuring the draw on the original battery, I'm not sure wether it's dead or not. Likely is. I'd rather not buy and then install a coin holder just to test the draw though.

I like the idea of the LiSOCl2 but they are not rechargeable (right?) so I would still need to include a diode between Q2 and the battery? And it's not a cheap solution, $12 for the battery and $8 for the holder, but that would look tight and last a long time.

Any more tips or ideas? Other alternatives?
Is there a 3.6V rechargeable battery that would get properly recharged by the existing circuit?

Thanks!

[tszaboo]

Wait, the Datasentry is rechargeable? Never mind, I'm an arrogant idiot then.
If you still want to use that, removing Q2 would probably do the trick.

[shapirus]

so I would still need to include a diode between Q2 and the battery?

If the schematic is correct (i.e., matches the actual circuit) then you just need to cut the trace going from the collector of Q2 to the battery. The CR5 diode will do the rest.

Removing Q2, as advised above, may be a better option since it'll allow to roll back to the original state, if it will ever be desired, whereas fixing a cut trace, while possible, may be a more difficult job and the result may not look too nice.

[wasedadoc]

As I wrote above. https://uk.rs-online.com/web/p/button-rechargeable-batteries/0525792

This is a generic type, available with several brand names.

This

"There were some for the S-100 buss that used these GE "DataSentry" batteries which were 3.6v NiCd blocks " is stated in a post at https://forum.vcfed.org/index.php?threads/history-of-the-cmos-battery.15244/

[shapirus]

I can't quite understand how charge voltage and/or current limiting works in the circuit made of Q4, Q3, and Q2, if there is any. But apparently there is, since all those parts must be there for a reason, and it's not just a single resistor.

A simple explanation, anyone? :)

[mariush]

Another option could be to use 2 LITHIUM AA batteries in a regular 2-cell holder... freshly charged they'll be 1.8v and they'll go down to 1.4v as they discharge.

See for example https://www.digikey.com/en/products/detail/energizer-battery-company/LN91/18449229   or look on amazon for the "ultimate lithium" version or lithium based cells from other brands.

[langwadt]

I can't quite understand how charge voltage and/or current limiting works in the circuit made of Q4, Q3, and Q2, if there is any. But apparently there is, since all those parts must be there for a reason, and it's not just a single resistor.

A simple explanation, anyone?

when Q3 is on R30,R33 and CR3 sets the voltage on the base of Q2, the current though Q2 will be so that across R29 will be roughly that voltage minus the Vbe of Q2

[wasedadoc]

I can't quite understand how charge voltage and/or current limiting works in the circuit made of Q4, Q3, and Q2, if there is any. But apparently there is, since all those parts must be there for a reason, and it's not just a single resistor.

A simple explanation, anyone?

When the unit is powered up and 5 Volt supply is present Q3 is on.  That turns on Q4 (which powers the RAM) and turns on Q2 which lets current limited by R29 flow into the battery.  When the +5v rail is not there all three transistors are off and the battery via the diode keeps the RAM alive.

R33, CR4 and R30 form a potential divider which sets Q2 base voltage at around 3.6 Volts (I have not done the exact calculation.) Q2 collector can never rise above that so limiting the voltage the battery is charged to.  In other words the charging current becomes zero when that voltage is reached.

[Jwillis]

There tends to be a lot of confusion on the designation of button batteries. CR2032 is an IEC number and has no reference to the exact chemistry of a battery other than Lithium. C is the designation for lithium, R means it's round, 20mm diameter and 3.2 mm thick. Many companies have taken it upon themselves to try to break the confusion between rechargeable and non-rechargeable.
But it's probably best to consider all lithium batteries with CR designation as non-rechargeable.
LR and LiR are both rechargeable 3.6-3.7 V  and would work in your particular case.

[shapirus]

When the unit is powered up and 5 Volt supply is present Q3 is on.  That turns on Q4 (which powers the RAM) and turns on Q2 which lets current limited by R29 flow into the battery.  When the +5v rail is not there all three transistors are off and the battery via the diode keeps the RAM alive.

I figured that much, but it raises questions:

- what is the purpose of Q3 when Q4 will be off anyway when there's no power at +5V?
- what is the purpose of Q4 when RAM could be powered directly from +5V?
- what is the purpose of Q2 when the battery could be powered via R29 alone, followed by a diode (to prevent current flowing into the PSU)?
- is it ok to allow the battery to charge to 5 volts, when voltage drop across R29 becomes negligible and Q2 is turned on?
- R29 is only 51 Ohm, which, if Q2 is fully on, and the battery is at 3.6V, will result in charging current (I'm ignoring the RAM consumption here) = (5V-3.6V)/51R = 27mA, is this current not excessive for a NiCd (must have been this type originally)?

R33, CR4 and R30 form a potential divider which sets Q2 base voltage at around 3.6 Volts (I have not done the exact calculation.) Q2 collector can never rise above that so limiting the voltage the battery is charged to.  In other words the charging current becomes zero when that voltage is reached.

Ah okay, this later edit answers most of it, then .

[amyk]

Why replace a rechargeable battery with a non-rechargeable one? I'd probably go with 3.6v NiMH if NiCd isn't available.

[CaptDon]

One of the original poster's concerns was that of corrosion from leaking NiCad cells damaging the P.C.B. On some of my expensive equipment I have run wires and moved the battery to a harmless location. On one of my rack mount computers I moved the CMOS battery to the outside of the cabinet so I can do a hot swap without un-racking and opening the unit.

[cincin]

Well, I wouldn't mind using a rechargeable battery as long as the existing circuit works at charging it and keeping it heathy, and that it doesn't leak. I don't know much at all about battery charging, so I may be concerned about a non-issue.

I'd also like to keep the mod tight and pretty, and as original looking as possible, so moving the battery to a new location is probably not ideal. And AA cells are just too big to be a neat replacement.

So which kind of rechargeable battery then?

- wasedadoc and amyk proposed a 3.6V NiMH Button Rechargeable Battery, 160mAh, which looks a lot like the leaky battery I removed from the Korg PolySix.

- Jwillis suggests an LR2032 or LiR2032 which are also 3.6V and rechargeable.

Would either battery work equally well?
Would the charging circuit work well with either of these?

[shapirus]

Would either battery work equally well?
Would the charging circuit work well with either of these?

With regards to this, I would rather stick with the original chemistry. Well, actually, with NiMH, since the original was likely a NiCd, which is now (almost?) obsolete, but NiMH should be a near drop-in replacement.

I'm not sure if that's so, but the *2032 ones sound like they're Lithium-based, and if it's true, then that's a different chemistry, which requires that you:

1) read and thoroughly understand their datasheets;
2) perform measurements of max voltage and max current that the charging circuit outputs and make sure that they don't fall outside of the ranges allowed for the respective batteries.

The second one would be useful to do in any case, if you ask me: whatever type the new battery will be, it's always a good idea to make sure that its in-circuit operating conditions will be right.

[cincin]

Ok, I've been reading on batteries and charging, and it sounds like:
- All NiCD batteries will eventually leak
- NiCD charging circuits are not suitable for either NiMH or LiR batteries. (the original DataSentry is a NiCD)

SO...

I think the easiest / cleanest option may just be a non-rechargeable option with a Q2 disconnected to prevent charging.

Which brings me back to my original question, sort of: How do I figure out if a CR2032 3V would be enough to keep the ram alive in this synth? (other than actually doing the mod to see)

In the Korg PolySix, a 3V battery worked fine at keeping ram that was originally kept alive by a 3.6V NiCD. But the RAM in the current synth may not be the same.

The standby current info in the synth's RAM is cryptic to my newbie eyes, so I asked ChatGPT to clarify things for me. I gave it the data from the datasheet and it said:

Isb (Standby Power Supply Current):

Symbol: Isb
Minimum: Not specified
Typical: 1 mA
Maximum: 3 mA
Unit: mA
Test conditions: CS1̅ = Vih, CS2 = Vil
This specification indicates that under the specified conditions (CS1̅ = Vih and CS2 = Vil), the typical standby power supply current is 1 mA, with a maximum of 3 mA.

Isb1 (Standby Power Supply Current with Additional Conditions):

Symbol: Isb1
Minimum: Not specified
Typical: 2 mA
Maximum: 50 mA
Unit: mA
Test conditions: CS1̅ ≥ Vcc - 0.2V, CS2 ≥ Vcc - 0.2V, or 0V ≤ CS2 ≤ 0.2V, 0V ≤ Vin

This specification provides additional conditions for standby power supply current. It suggests that under the specified conditions (CS1̅ ≥ Vcc - 0.2V, CS2 ≥ Vcc - 0.2V, or 0V ≤ CS2 ≤ 0.2V, and 0V ≤ Vin), the typical standby power supply current is 2 mA, with a maximum of 50 mA.

Now, let's relate this information to the DC Characteristics provided:

Vcc = 5V ± 10% (which means Vcc can be in the range of 4.5V to 5.5V).
Vss = 0V
When the chip is in standby mode, it is drawing the standby power supply current, and these values are given based on different conditions (CS1̅, CS2, and Vin). CS1̅ and CS2 are likely control signals, and Vih and Vil are voltage levels for logical high and low, respectively.

To summarize, the HM6264LP-15 RAM chip's standby power supply current can be in the range of 1 mA to 3 mA under certain conditions and 2 mA to 50 mA under additional conditions, depending on the states of control signals and voltage levels.

Tres cool. But I'm not sure how the control signals are used in this synth.

I'm also unsure of how Voltage and Amperage plays out here. The battery is directly connected to the RAM chips except for the diode CR5. A CR2032 can typically provide 0.19mA at 2.9V. ChatGPT seems to be saying I'd need 1mA *minimum*, more like 3mA, to keep one RAM chip alive, and there are 8 chips. So a CR2032 wouldn't cut it then? I'd need at least something capable of providing 8mA, possibly even 24mA?

I can't seem to find the capacity of the original DataSentry battery, @tszaboo suggested it was something like 150mAh... I asked ChatGPT directly if a CR2032 could keep 8 of those RAM chips alive, and it said yes, for 9.7 hours. Which isn't anywhere near enough.

Does that make any sense? (ChatGPT is questionable at times)
Should I consider a CR2 then? (CR2450 don't seem much beefier)

[cincin]

Yeah I don't think CHatGPT's estimates are realistic. It said:

8 x HM6264LP-15 RAM chips using a CR2 battery, we need to consider the total standby current and the capacity of the CR2 battery.

Let's assume a worst-case scenario where each RAM chip draws 3 mA (maximum specified standby current), and we have 8 chips:

Total Standby Current = 8 chips × 3mA/chip = 24mA

CR2 batteries typically have a capacity in the range of 800 to 1000 milliampere-hours (mAh). Let's assume a capacity of 900 mAh for this calculation.

Battery Life (hours) = Battery Capacity (mAh) / Total Standby Current (mA)


Battery Life = 900mAh / 24mA

Battery Life ≈ 37.5 hours

It makes it sound like it would take a HUGE battery to keep this RAM alive for any sort of functional amount of time.

So maybe a LiSOCl2 after all? (35mA max recommended continuous current)

[shapirus]

So I was wrong assuming that NiMH could be charged with circuits designed for NiCd. I stand corrected.

Now, for the RAM chips. The datasheet says that a typical standby current is 15 μW. Let's make a rough estimation based on a first found datasheet for a 2032: https://data.energizer.com/pdfs/cr2032.pdf.

Typical capacity seems to be about 600 mW*h.
Then, 0.6 W*h / 15e-6 W = 40000 h = 1666.6(6) days. Divide this by the number of RAM chips.

That's probably higher than the real number will be, because the useful voltage range is a question. The RAM IC datasheet says that the minimum recommended supply voltage is 4.5V, and not 3.6V, let alone the CR2032's 3V. The obvious answer here is try and see. An interesting fact is that the datasheet says "Battery backup operation capability" but doesn't elaborate further on this: is it implied by the low standby power***?

The estimated standby time sort of suggests that a rechargeable battery may actually be desirable. And if a circuit modification is inevitable, then a very basic linear regulator with voltage and current limit plus a lithium (LiFePO4 for better safety I'd say) battery would be an easy and reasonable way to go, as long as the ram chips will be fine with its usable voltage range (~2.9..3.3V minus the diode voltage drop). If not, then a LiIon battery (3.7V nominal voltage, charge up to 4.2V -- almost perfect for a linear reg with a 5V supply). But then you might want to implement a battery overdischarge protection...

Or just a beefy non-rechargeable battery. Three D cells will last a long time.


***it's on page 12: "Low V_CC Data Retention Characteristics"

[amyk]

So I was wrong assuming that NiMH could be charged with circuits designed for NiCd. I stand corrected.

If you want rapid charging, there are differences. But trickle charging with constant current, which is what your application does, works fine for both NiCd and NiMH with the only difference being that NiMH should be trickle-charged at ~half the current of NiCd, which can easily be overcome by putting in twice the capacity of NiMH as NiCd.

[wasedadoc]

Was the original battery marked as 150mAh?

If not, was it marked and what was it?

If yes, then the original designer was satisfied that 150mAh was sufficient.  In which case the 3.6V NiMH I linked to is an adequate replacement in that regard.

The original charging circuit topology is OK for charging a 3.6v NiMH type battery.  And it is very likely that the existing component values are both safe and adequate for a NiMH battery of the same or higher capacity than the original NiCd.  If you want to be absolutely sure of that install one and measure:

1.  The voltage across the battery when charging is complete.  That is, when the voltage does not rise any further.

2.  The voltage across the 56 Ohm resistor when fully charged.  If that voltage goes to zero then there is no risk of excessive trickle charging.  If it does not eventually go to zero then the final voltage divided by 56 gives the trickle charge current. Both NiCd and NiMH can tolerate continuous trickle charge, typically a few percent of the rated capacity number.  Consult datasheets.

3.  The voltage across the 56 Ohm resistor when the battery is being charged from its low voltage state.  That divided by 56 gives the maximum charging current the battery is being subjected to.  Check that against datasheet.

Should any of the above three readings and calculations raise concern it is trivial to adjust one or two of the resistor values in the charging circuit.

[cincin]

I like your estimates better that ChaptGPT's, shapirus! If I go with a 3.6V 1200mAh Li-SoCl2 that should give me twice that, or 3333.2 days divided by 8 = 416.65 days of off time. Which is... acceptable. Having to replace a $12 battery every year or so is a bit of a pain. But what can I do.

This page here, say:

The differences in trickle charge current and the need for more sensitive full-charge detection render the original NiCd charger unsuitable for NiMH batteries. A NiMH in a NiCd charger would overheat, but a NiCd in a NiMH charger functions well.

It's the "more sensitive full charge detection" that led me to believe that would be a bad idea. A rechargeable solution would definitely be better, but I don't want to do anything questionable that may result in the battery over heating and damaging the synth. The Kurzweil K250 is a bit of a historical piece.

Was the original battery marked as 150mAh?

The battery doesn't say on the case itself, and I couldn't find a datasheet for it. I'm going with 150mA from tszaboo's post earlier on the this thread:

Get a LiSOCl2 battery, use that. That's also the original chemistry, but a 1000mAh would last forever + 10 years in place of the original 150mAh one