Problem with Amiga 3000T , Please help !

[Castellen]

Some good suggestions from Zac67 regarding the memory, certainly worth checking all of the ZIPs are the same.  More information here.

The Diode for the CR2032 Battery is preferred as a 1N914 although the 1N4001 should work the both have the same VF Characteristics, although the IN914 is the recommended item according to some of the Amiga battery hack files.

As for the RTC backup power supply; you're far better off using a small signal Schottky diode as opposed to a standard silicon diode (e.g. 1N914, 1N4148, 1N400x, etc).

To look at the maths behind it...
From the RP5C01 datasheet we can see the minimum input voltage is 2.2V at a current of 15µA.  The RTC has a 1.2k Ohm resistor in series with the battery (see schematic).  So it'll develop (1.2k x 15µA) = 18mV, or round to 20mV across it.  So the actual minimum voltage at the RTC becomes (2.2V + 20mV) = 2.22V

If a 1N914 diode was used, we can see it has a forward voltage drop of around 0.55V at 15µA.  So it'll only provide a clock backup until the cell reaches a voltage of (2.22V + 0.55) = 2.77V.  The CR2032 cell is 3V nominal, meaning it only has to discharge by 0.23V to a terminal voltage of 2.77V before it'll need to be replaced, so that's using well under half of the cell's capacity.

If a Schottky diode was used instead, say a common BAT85 for example, we can see the forward voltage drop will be less than 0.24V at 15µA.  So the cell can discharge further to (2.22V + 0.24) = 2.46V before it needs to be replaced.  You still won't get to use the full rated capacity of the cell (230mAHr), but at a guess you'd still get about 60% of that (140mAHr).  At a current draw of 15µA, I'd expect a clock backup time of 389 days or 13 months.  Of course that depends on the exact cell being used, age/condition of the cell, exact current draw of the RTC on standby and the temperature.

So in summary, while the silicon diode will work, you'll get far longer out of the coin cell if you use a small signal Schottky diode.

[Castellen]

Question - is there any harm in removing and bridging that resistor (R179), when replacing the original battery with a diode and 2032 coin? I've done this on my 2000's, by simply installing the diode (1N60 in my case) in place of the resistor (R803). Based on what you've written, it would seem the RTC battery should last quite a bit longer. Any potential issues with this?

The purpose of the resistor in series with the battery is to limit charging current to a safe value when a rechargeable nickel based battery is fitted.  With the lithium cell + diode, the resistor serves no purpose so can be bypassed (short circuit) if you want to, although the gain in battery life in doing so will be miniscule; hardly worth the effort.

As I recall, the 1N60 is a germanium diode so should have forward voltage characteristics comparable to a Schottky diode.  Can't remember the exact specs of them, but put your voltmeter across the diode.  If you see any more than about 0.3V with the computer off, then you can make a worthwhile gain by replacing the 1N60 with a BAT85 or whatever.  

Edit: Just pulled my 2000 from the closet, it's been sitting off for well over a year, and the battery is nearing 3 years old. I was surprised to find that it had only lost 20 minutes. Looks like the RTC on the A2000 is good all the way down to 2.0V, which helps.

The A2000, A500+, A501, etc, use a different RTC device, a Seiko/Epson MSM6242 which does have a lower data retention voltage of 2V as opposed to 2.2V on the Ricoh RP5C01.

Of course it depends on the device and temperature.  While the minimum data retention voltage of 2.0V is the specification, in practice some devices may operate below that.


A little off topic but perhaps worth adding is the fact that the crystal oscillator on the RP5C01 and probably the MSM6242 varies in frequency slightly according to the applied voltage.  I generally calibrate them for f0=32.768kHz at 3.6V (for nickel batteries).  If the voltage drops much below that, the oscillator runs slightly slow.  When the computer is running and the supply voltage is 5V, it runs slightly fast.

And of course as quartz crystals age, they usually drift low in frequency, so that and the lower operating voltage explains your time drift.  Not to mention that those crystals aren't brilliantly accurate over wide temperature ranges.

While you can trim the crystal frequency using the trimming capacitor, you'll need a high impedance input (10M Ohm or better) scope or frequency counter to do it correctly.  Any load on the oscillator circuit from the measuring equipment will change the frequency.