SMP does not power off when the calculator is off

[vladkorotnev]

When the calculator is off, the module still remains powered on the VCC line.
Probably the Vcc line is just connected directly to the computer's battery input.

Proposed solutions:

• Find if it is possible to turn it off in software
• Rewrite using interrupts and sleep mode to lower the power consumption (preferred method)

@piotr433 Can you please chime in on how the Vcc line of the SMP socket is wired?

Thanks!

[piotr433]

I confirm that the supply voltage is present on the SMP connector even when the calculator is turned off.

Probably the Vcc line is just connected directly to the computer's battery input.

No, not directly, but through a voltage regulator.

Find if it is possible to turn it off in software

I'm afraid it's not possible.

Rewrite using interrupts and sleep mode to lower the power consumption (preferred method)

Yes, the microprocessor should be put in sleep mode after a few seconds of bus inactivity, and waken up by a CLOCK signal.

Measurement results supplied by fellow @spark-sys:

MK-90 turned off:
no load, Vcc = 3.5V
R = 10 kohm, Vcc = 3.5V, I = 0.35 mA
R = 1 kohm, Vcc = 3.1V, I = 3.1 mA
R = 470 ohm, Vcc = 2.8V, I = 5.96 mA
R = 330 ohm, Vcc = 2.6V, I = 7.9 mA
R = 300 ohm, Vcc = 2.2V, I = 7.7 mA
R = 220 ohm, Vcc = 2.1V, I = 9.5 mA * MK-90 doesn't turn on under such load

MK-90 turned on:
no load, Vcc = 4.8V
R = 10 kohm, Vcc = 4.7V, I = 0.47 mA
R = 1 kohm, Vcc = 4.5V, I = 4.5 mA
R = 470 ohm, Vcc = 4.5V, I = 9.6 mA
R = 330 ohm, Vcc = 4.5V, I = 13.6 mA
R = 300 ohm, Vcc = 4.5V, I = 15 mA
R = 220 ohm, MK-90 doesn't turn on

[vladkorotnev]

Too bad, that leaves us no capacity to make a WiFi version, at least for now.
But I hope microSD can run with an Arduino with no problem off that :-)

[piotr433]

There's no guarantee that every MK-90 can supply 15mA. My earlier project of this kind failed, because it worked only with some of them. I guess that it drew too much current.
http://web.archive.org/web/20071102215701/http://www.pisi.com.pl/piotr433/mk90rcre.htm

Perhaps your device could be powered from the pin 39 of the rear expansion port? It's directly connected to the positive battery terminal. It's an ugly solution, I know...

[vladkorotnev]

Is the Vbat terminal on the SMP port powered by the MK, btw?
Maybe it can be somehow repurposed to supply power?

[vladkorotnev]

Also I noticed that arduino-based SMPEmu won't work if it's plugged in while the calculator is off.

I suppose this is related to the fact that standby voltage on the Vcc line of the SMP slot is 3.5V, which might be below the default BODLEVEL fuse setting by Arduino IDE.

I will also try removing the extra LEDs from the arduino to reduce current consumption.

However looking at microSD datasheets, most of them use roughly up to 50mA when writing, so it might cause a reboot of the calculator.

Seems like there is a need to somehow hardwire the emulator cartridge into the power circuits of the calculator or the battery slot...

[vladkorotnev]

After removing the power LED, I was able to implement the power saving mechanism in 5c17558

However some issues arised.

1. You still need to plug in the module while the calculator is active (Vcc is at 5v)
2. Sometimes the calculator doesn't deselect the module on time, which keeps it active and prevents it from going to sleep
3. Sleep mode cannot be lower than SLEEP_MODE_ADC. I'm not sure what this is caused by, but the SMP seems to be unreadable with any sleep mode lower than that.

@piotr433 If you could please test in regard of point № 3 and measure the new current consumption of the module, that would be very helpful :-)

[piotr433]

Unfortunately, this test cannot be performed on my present ATMEGA8 board. I have just ordered an ATMEGA328 board and hope to get it within a few days.

[piotr433]

Thoughts without testing - the problem is surely caused by the long startup time of the crystal oscillator (milliseconds range) and the SUT configurations bits setting. Proposed solutions:

1. Selecting the internal RC oscillator (which starts immediately) as a clock source. This choice would affect the Arduino bootloader communication reliability, though. It would be safer to burn the flash through the SPI port.

2. Not using the microcontroller power saving features, but connecting the board through an additional circuit cutting off the power when the SMP supply voltage drops below 4V.

[vladkorotnev]

@piotr433 Will anything bad happen if we feed stabilized +5v into the SMP port? Because that way we could add a stabilizer and a lipo battery and turn the flash cartridge into a memory+battery pack, for even further ease of use of the calculator :-)

[piotr433]

Feeding +5v into the SMP port wouldn't hurt the calculator, but wouldn't power it, either. In order to be sure I have tried this on my MK-90.

[vladkorotnev]

OK, so maybe we could have a supercapacitor or a small battery inside our SMP to keep up with the current spikes of up to 30-50mA that an SD card can do?

[piotr433]

It's a good observation that the increased current demand will last only for short periods of time. I like your idea!

[vladkorotnev]

Do you know the proper calculation formula though? I hardly remember anything apart from the ohm's law and some RLC filters :-(

[piotr433]

What do you want to calculate? The capacitor value?
When a capacitor is charged or discharged with a constant current, then the voltage across it ramps linearly according to the formula:
delta U = current * time / capacitance
Assuming current spike duration of 20 ms, current value of 50 mA and allowed voltage drop of 0.1 V we get:
capacitance = 0.05A * 0.02s / 0.1V = 10 mF

[vladkorotnev]

10uF is easy, these days we can get even more, can we go up to 1s then?

[piotr433]

mF means millifarad = 1000 microfarad
10 mF = 10000 uF
Too high value would increase the loading time after power on. Loading a discharged 10000 uF capacitor with a constant current of 10 mA to 3.3 V would take:
0.01F * 3.3 V / 0.01A = 3.3 seconds

[vladkorotnev]

oh, I'm not sure if small 10mF caps exist (I mean, I have some, but they are about the size of a vodka shot glass :D)

yeah but those 3.3s will be after plugging in the SMP, not when turning on the calculator. It however is interesting if the high inrush current of the empty cap charging might burn the power circuitry of the calculator...

[vladkorotnev]

Oh, it seems they indeed do exist! https://www.digikey.com/product-detail/en/rubycon/16PX10000MEFC16X31.5/1189-1125-ND/3134083

The huge ones I have are for 60V, that's why they are so big it seems.

Probably 16V is a safe margin, as I doubt 6V or 10V ones exist, or if they do, that they are much smaller

[piotr433]

The SMP port seems to be short-proof, but from my experience, when it is directly loaded with a capacitor > 100 uF, the calculator wouldn't turn on. A series current limiting resistor (470 ohm) would be needed.
BTW, when a capacitor is charged from a constant voltage source through a resistor, following formula applies:

The supply voltage required by an SD card is 3.3V, right? A voltage regulator would be needed between the capacitor and the card.

[vladkorotnev]

Yes, I have a few high efficiency switch mode 5-to-3.3v regulators, as well as some SD card slots with built in level converter ICs. Wouldn't the resistor create problems at the time the current needs to be released (e.g. During a card write)? // Ak.R. iOS/Mac/Windows & Web developer Vaporwave/ambient producer, sound/video engineer Genjitsu Gadget Lab Member 001 http://genjit.su

…

Sent from my iPhone

On 19 May 2018, at 01:07, Piotr Piatek ***@***.***> wrote: The SMP port seems to be short-proof, but from my experience, when it is directly loaded with a capacitor > 100 uF, the calculator wouldn't turn on. A series current limiting resistor (470 ohm) would be needed. BTW, if a capacitor is charged from a constant voltage source through a resistor, following formula applies: The supply voltage required by an SD card is 3.3V, right? A voltage regulator would be needed between the capacitor an the card. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.

[piotr433]

Wouldn't the resistor create problems at the time the current needs to be released (e.g. During a card write)?

It would limit the SD card average current, i.e. the duty cycle (how intensively the SD card can be accessed). Some current limiter is necessary, though, otherwise the MK-90 won't turn on. Perhaps a JFET BF245C with the gate connected to source?

BTW, ensure that the capacitor is discharged before connecting it to the calculator.

[vladkorotnev]

JFET BF245C with the gate connected to source?

Could you please provide a schematic of what you mean?

[piotr433]

The advantage of a JFET over a resistor is its nonlinearity. At low drain-source voltage it acts as a resistor of 120 ohm, otherwise like a current source of 16 mA (values measured on my specimen). The capacitor is loaded faster than through an ordinary series resistor of 5V / 16 mA = 312 ohm.
The circuit was tested on my MK-90.

[vladkorotnev]

Good point! Just need to see how well it works with current bursts of 50mA now. Probably that will only happen after I finish my graduation paper though :( // Ak.R. iOS/Mac/Windows & Web developer Vaporwave/ambient producer, sound/video engineer Genjitsu Gadget Lab Member 001 http://genjit.su

…

Sent from my iPhone

On 20 May 2018, at 00:39, Piotr Piatek ***@***.***> wrote: The advantage of a JFET over a resistor is its nonlinearity. At low drain-source voltage it acts as a resistor of 120 ohm, otherwise like a current source of 16 mA (values measured on my specimen). The capacitor is loaded faster than through an ordinary series resistor of 5V / 16 mA = 312 ohm. The circuit was tested on my MK-90. — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.

[piotr433]

I had to modify the AVR program because of following issues:

1. The subroutine recvCommand() cannot handle a SELECT pulse without CLOCK pulses (which appears on the SMP bus when a key is pressed or a beep emitted). It stays then in the WAIT_CLOCK_FALL loop consuming power. A crude fix was applied to the do_get_byte() routine.

2. The sleep / wake up code is overly complicated.

Measured current consumption:

1. No power saving (subroutine sleep_mode() commented out) - 10mA

2. SLEEP_MODE_ADC - 2mA

3. SLEEP_MODE_PWR_DOWN - less than 0.1mA (cannot measure more accurately), works only with an internal RC oscillator 8MHz with minimal startup time (HFuse=0xD9, LFuse=0xC2). This fuse modification disables the Arduino bootloader!

[vladkorotnev]

I guess 2mA is enough, because the convenience of arduino bootloader is just too damn good, and 8MHz is likely too slow to read the SD card and output data at the same time. Thanks for your help! :-) // Ak.R. iOS/Mac/Windows & Web developer Vaporwave/ambient producer, sound/video engineer Genjitsu Gadget Lab Member 001 http://genjit.su

…

Sent from my iPhone

On 30 May 2018, at 23:35, Piotr Piatek ***@***.***> wrote: I had to modify the AVR program because of following issues: The subroutine recvCommand() cannot handle a SELECT pulse without CLOCK pulses (which appears on the SMP bus when a key is pressed or a beep emitted). It stays then in the WAIT_CLOCK_FALL loop consuming power. A crude fix was applied to the do_get_byte() routine. The sleep / wake up code is overly complicated. Measured current consumption: No power saving (subroutine sleep_mode() commented out) - 10mA SLEEP_MODE_ADC - 2mA SLEEP_MODE_PWR_DOWN - less than 0.1mA (cannot measure more accurately), works only with an internal RC oscillator 8MHz with minimal startup time (HFuse=0xD9, LFuse=0xC2). This fuse modification disables the Arduino bootloader! — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.