Best way to regulate down from 24V and drive 24V rail-to-rail?

[LaurentR]

I am trying to reverse-engineer the protocol of the Nanoleaf Aurora so that I can drive the tiles myself. See teardown:
https://www.eevblog.com/forum/projects/nanoleaf-aurora-teardown/msg1158410/

They use a 24V supply, which is reasonable given the max power consumption. However, their main signaling is also 24V rail-to-rail. To start experimenting using an Arduino-type device, I would need to:
* Regulate down from 24V to 5V. That's a bit more than I am used to. I am expecting no more than 0.5-1A needed. I assume a linear regulator wouldn't work (power dissipation). Is there any packaged switcher available that has a reasonably small footprint and could eventually be used on a perfboard?
* Drive 24V rail-to-rail. Same thing here. I have done <5V conversion using bufferes or level shifters but I have no experience with driving 24V rail-to-rail. Any opinion on how I should do that (from a 3.3V or 5V source)?

Thanks in advance.

Laurent

[james_s]

For the voltage regulator one of these will do the trick:
http://www.ebay.com/itm/New-Mini-5pcs-LM2596s-3A-DC-to-DC-Buck-Converter-Power-Supply-Step-Down-Module-/192091378431?

Lots of places sell the identical unit, I've seen them as low as 99c each shipped, I've used a bunch of them and they work great.

[Ian.M]

You can get 30V rated MOSFET driver chips.  They'll drive rail-to-rail (- a Vbe drop or two) from a TTL logic level control signal.  However you'll need to 'tame' them with a series resistor on the output to match your cable impedance and you *MUST* have good enough decoupling right at the driver chip.   Alternatively, if the protocol is relatively slow, use a rail-to-rail >+/-15V supply OPAMP fed between +24V and 0V, bias one input at half the 5V logic supply and apply the control signal to the other input via a resistor.   Check the slew rate is fast enough to reproduce the control signal. Another approach is a ULN2803 or similar with bootstrapped active pullups.  Ask me about that one if you need  a lot of output lines.

If you take James's suggestion for an EBAY buck module, you should feed the Arduino power in at 7.5V via a polyfuse with a beefy 9.1V Zener (or 'boosted' Zener + power transistor) as a clamp.  That way,if the buck module fails it doesn't take out all your control logic.   Additional 5V loads in excess of the areduino 5V supply capability can be fed from the 7.5V rail by LM7805 regulators - with only 2.5V across them, they should be good for about 300mA each without heatsinking.   

[LaurentR]

For the voltage regulator one of these will do the trick:
http://www.ebay.com/itm/New-Mini-5pcs-LM2596s-3A-DC-to-DC-Buck-Converter-Power-Supply-Step-Down-Module-/192091378431?

Lots of places sell the identical unit, I've seen them as low as 99c each shipped, I've used a bunch of them and they work great.

Thanks! Looks good. I'll procure a few.

[LaurentR]

You can get 30V rated MOSFET driver chips.  They'll drive rail-to-rail (- a Vbe drop or two) from a TTL logic level control signal.  However you'll need to 'tame' them with a series resistor on the output to match your cable impedance and you *MUST* have good enough decoupling right at the driver chip.   Alternatively, if the protocol is relatively slow, use a rail-to-rail >+/-15V supply OPAMP fed between +24V and 0V, bias one input at half the 5V logic supply and apply the control signal to the other input via a resistor.   Check the slew rate is fast enough to reproduce the control signal. Another approach is a ULN2803 or similar with bootstrapped active pullups.  Ask me about that one if you need  a lot of output lines.

Thanks a lot. I haven't looked at the slew rate (will do tonight). The protocol runs serial @120kb/s so I assume a comfortable slew rate will be easier than a high drive MOSFET with line impedance adaptation (it goes through lots of board, connectors, forks...). Do you have any pointer for an opamp that would work?

[LaurentR]

Based on the posts above and looking at the teardown pics, I see a LM393 (I think) IC on the board. It's a comparator with up to a 36V supply, so I assume that's how they do it.

[Ian.M]

That's a touch under 60V/us slew rate, assuming two edges per bit and a risetime of not more than 5% of the bit period.  That's pretty scary for an OPAMP.  To give you a point of reference, a 741 is about 0.7V/us, and many modern OPAMP are an order of magnitude better but you need two orders of magnitude with enough 'beef' to drive the cable capacitance so IMHO its MOSFET driver territory.  You can get fast enough OPAMPs - at a price, but their current capability and voltage ratings are likely to be limited.   

I was about to say a fast 30V comparator is another option, but the LM393 you mentioned is open collector so would need a pullup circuit - probably active to meet that edge rate spec.

[james_s]

I suppose a crowbar wouldn't hurt, but I've yet to have one of these buck modules fail, I did cook a similar one once by running it right at the max rated output current but this application is not going to do that. Linear regulators can fail shorted too.

[LaurentR]

That's a touch under 60V/us slew rate, assuming two edges per bit and a risetime of not more than 5% of the bit period.  That's pretty scary for an OPAMP.  To give you a point of reference, a 741 is about 0.7V/us, and many modern OPAMP are an order of magnitude better but you need two orders of magnitude with enough 'beef' to drive the cable capacitance so IMHO its MOSFET driver territory.  You can get fast enough OPAMPs - at a price, but their current capability and voltage ratings are likely to be limited.   

I was about to say a fast 30V comparator is another option, but the LM393 you mentioned is open collector so would need a pullup circuit - probably active to meet that edge rate spec.

Turns out the LM393 is used to read back that serial signal for a reason I haven't quite figured out.

The 24V serial signal is driven by what appears to by a 2N7002K mosfet, itself driven (why?) by a 2n2222-type NPN driven @ 3.3V by the main IC.
The output has a 4.7k pullup to the 24V rail. No source termination I could find but there is a diode to the 24V rail (clamp to avoid ringing?).

Measured edge rates: rising 28V/us, falling 60V/us