Power sequences

[Chasm]

Posting in Beginners since this should be simple without my mental block.

A device uses different auxiliary voltages.
They have to be activated in sequence on power up and deactivated in reverse order on power down. No Nothing new there. The device is quite dumb, no special requirements other than the order.

How to realize this, cheap?

[tyblu]

What does it need to do, specifically? (What voltages and timing?)

[Chasm]

It's a GLCD, it has no strict timing requirements for the sequence.

Vcc -> Vgl -> Vgh (-> Data -> Backlight)
Vcc 3.3V
Vgl -10V
Vgh 15V
Backlight 20V

[Psi]

Quick and dirty solution.
You could have a chain of relays so one provides power for the next.
Extra contacts on each relay can switch things on in order.
The time between them will be small, just as long as the contacts take to move state.

Ideally you'd want to switch each power connection on with a transistor or fet using different resistor+capacitor changing rates, or a mcu.

[Chasm]

Too much current. =)

The LCD should draw less than 300 micro A on Vgl and Vgh. I want to spec the bost/buck converter quite close and as cheap as possible.
That is the plan, for some reason one implementation uses a LT1172CS8 - very nice but a 1.25A 5€ regulator is a mite excessive.

I can also switch the voltages with the MCU, another 3 I/O is not a problem but I wanted to get rid of more parts.

[EEVblog]

Sometimes the datasheet says that, but just powering them all up at once can sometimes work just fine.
Is it a hard and fast "this thing will blow up if you don't..." or is it just "it's recommended that..."
Have you actually tried it in practice?

Dave.

[Chasm]

Hi Dave,

Note 2: Be sure to apply GND, V_CC, and V_GL, to the LCD first, and then apply V_GH.

They are the pos. and neg. power supply for the scan drivers of the TFT. In almost all other aspects the datasheet is pretty vague.

I don't have the LCD's at the moment.
They where cheap so I guess I could sacrifice one.

If it does not work I would have to change the PCB again, which is more expensive than the LCD in the first place.
Or designing both options into in from the beginning.

[JohnS_AZ]

Everyone knows that you can type a part number into http://www.findchips.com , but you can also type in things like "Power Sequencer".

In the resulting list somewhere you'll see LTC2924.

Then http://www.alldatasheet.com will provide you with the datasheet "LTC2924IGN - Quad Power Supply Sequencer - Linear Technology"

[Balaur]

You can use linear regulators with a "Power Good" output and an enable input. Then you cascade the regulators in the required sequence by tying the power good signal of an early regulator to the enable of the following one.

This will take care of the power-on startup requirement.

For the power down, you may try playing with the values of the filter capacity in such way that the different auxiliary voltages dry up in the required sequence.

Alternatively (and more costly) you can use something like a FSM to drive the enable inputs of the supply regulators or some kind of power switches.

While I understand perfectly the need of a specific start-up sequence, it is less clear for me why you should care for the power down. A sanely designed device will not suffer any negative effect during the power down. You could mention latch-ups or clamp diodes as potential sources for heavy damages, but come on...

Cheers,
Dan

[Chasm]

Everyone knows that you can type a part number into http://www.findchips.com , but you can also type in things like "Power Sequencer".

Thank you, there is the missing search term.
I think I'll try the thoroughly simplistic LM3880MF-1AA, it should be sufficient for this task.

Cost is a driving issue, or at least the exercise in this case, can't use a Power Sequencer that costs more than the GLCD. =)

You can use linear regulators with a "Power Good" output and an enable input. Then you cascade the regulators in the required sequence by tying the power good signal of an early regulator to the enable of the following one.

This will take care of the power-on startup requirement.

That would work, but the cheap DC DC converters rarely have such an output.
Seems like I'll end up with a MC34063, it's hard to get cheaper in single digit quantities and there is no current to speak of. Ah, better add a couple mA load to the scan driver supplies.

While I understand perfectly the need of a specific start-up sequence, it is less clear for me why you should care for the power down. A sanely designed device will not suffer any negative effect during the power down. You could mention latch-ups or clamp diodes as potential sources for heavy damages, but come on...

Probably it is quite excessive to control the power down sequence in this case.
In a NXP eval board for this display they controlled the power on sequence via the MCU and a couple of FETs. The power down sequence is controlled by the position of the FETs, they are between bypass caps and load for the aux. voltages. Should also work in my case with the simplistic sequencer.

[saturation]

A very simple circuit is add an RC time delay to the V+ input of each circuit, each delay greater that the previous one in the series.  This is efficient only if each subcircuit doesn't draw much power, otherwise consider the series resistor wastes energy.  On power up:

1234

but on power down its

1234 too.


When power is applied, the capacitor will appear as a short to ground and ramp up.  By using different RC combos on each circuit, you can either calculate it or just wing it, you can delay power up by a defined amount; low tech and parts count, easy to implement.

However, on power down, the last item to power up is also the last to power down because it has the largest capacitor in the sequence.

To shut down in reverse order you need to design a switching circuit, or quickly and dirty, a multi-pole mechanical switch.  When the circuit is switched off, each pole of the switch should short smaller value resistors to ground to reduce the series resistor in V+ of each subcircuit,  so it discharges the capacitors at a faster rate, choose the values so the time constants for 4,3,2, are reduced for the right sequence.

You can do the above in silicone too and use the AD chip, its more compact and precise.  

All those added caps also debounce mechanical switches for you, and you can add a sufficient large cap across the main power supply to compensate for the switches momentary disconnect, if that is an issue.  

[Zero999]

What's the primary power supply?

A comparator IC such as the LM339 with an RC circuit connected to the inputs and a PNP transistor on each output.

Where are you getting the -10V from? It's possible to switch it from a positive supply but it requires another transistor.

[Chasm]

What's the primary power supply?

A comparator IC such as the LM339 with an RC circuit connected to the inputs and a PNP transistor on each output.

Where are you getting the -10V from? It's possible to switch it from a positive supply but it requires another transistor.

The aux. voltages are generated from a 5V supply.
A comparator is also a nice solution, and even cheaper than the power sequencer.

Another transistor, FET or other chicken food is not a problem.


What is this all about?
I came across some dirt cheap slightly used 3.5" QVGA touch TFTs.
The goal is a breakout board for it, probably with a ARM processor directly on it.

With the TFT at about $4 a pop I want to keep the rest on the cheap side as well. At least in relation.

[Bored@Work]

The goal is a breakout board for it, probably with a ARM processor directly on it.

Use I/O pins from the ARM plus some simple software at the beginning of your program to turn on the TFT power in the desired order. If the I/O pins can't source enough current to drive the TFT, or if the ARM hasn't the right voltage, use some additional FETs on the I/O pins to switch the power.

[Chasm]

Yeah, that would be the obvious thing to do. Three outputs and a few FET.
I want to do it in a discrete way in case I decide to use it as a "stupid" breakout board without populating the MCU.

[Zero999]

The aux. voltages are generated from a 5V supply.
A comparator is also a nice solution, and even cheaper than the power sequencer.

Another transistor, FET or other chicken food is not a problem.

Then you need to additional components: connect the emitters to all of the PNP transistors together and connect the 5V input of each regulator to the collector of the appropriate transistor.

The formulae for calculating the timing can be found on Wikipedia under RC circuit and potential divider. I would remove the supply voltage from all the derived equations as it'll make it easier to calculate.

You can of course get power supplies with enable/disable pins and use RC circuits but the LM399 and PNP transistors are so cheap, they might cost less and are certainly cheaper than an additional microcontroller.

Another option is to use an astable multivibrator (couple of logic gates, Schmitt trigger oscillator, 555 timer etc), a CD4017 and some thyristors which is more complicated but you can have up to 10 outputs using only two ICs. The diodes are only requited if the voltage supplying the load is greater than Vdd. This circuit is also better for higher voltages.

[Chasm]

While a 555 solution would get me into the contest I think I keep it to something a bit more practical. Especially since I'm not into the run for a lifetime supply of LCD's.