fast constant current / constant voltage power supply (IEC62386 / DALI bus)

[martin_nga]

Hi@all,

I'm having some problems in implementing a voltage/current source that maybe some analog guru can give me a hint on where to start  . Searching the net for current/voltage supplies is pointless, you will mostly find LED drivers 

We're currently designing a power supply for IEC62386 DALI lighting bus. The bus normally operates at a voltage of 16V and can supply connected devices up to 250mA directly. Communication on the bus is done by shorting the bus lines. In this case, the power supply needs to operate in constant current mode and supply 250mA.
The specification requires the switch between constant current and constant voltage mode to happen in the region of 10us.

I've started designing my circuit based on this one: https://www.mikrocontroller.net/topic/378103#4414174 (power supply part). This essentially is a standard LM317 power supply with added constant current control loop. The loop works by taking advantage of the transistor base threshold voltage.
I've modified this circuit to use an LT3088 instead of the LM317 as this can operate down to 0V output voltage, so I can reduce current shunt resistance. The LT3088 is also specified to work stable without output capacitance.



This works more or less well inside simulation (see attachment). However, in the real circuit I can't get the current control loop to work reliable:
- in some cases oscillates at 2MHz in constant voltage mode (so much for stable...)
- starts oscillating at ~1MHz in constant current mode
- current value depends heavily on transistor
- current value changes with temperature

I've played around with the circuit for some time, but I can't find a way for this to work properly. The LT3088 datasheet gives some hints, but those either don't change anything or slow down regulation. I also can't change the transistor part (feedback) as this will change the control loop behavior.

Do you have a hint or keywords on where to start looking? I'm open to any approach.

Thanks for your help!

regards,
Martin

Edit: 2nd try on schematic.

[duak]

Have you seen http://ww1.microchip.com/downloads/en/appnotes/atmel-42071-dali-slave-reference_design_application-note_at01244.pdf ?

I would consider changing the design to a constant current source with a voltage clamp to set the maximum output voltage.  I see the problem as designing a stable circuit that has to change its output impedance from high when in constant current mode to low when in constant voltage mode.  Many voltage regulators require an output capacitor for stability that makes the transition from low impedance to high impedance difficult.

You might want to look at the hp 6186C constant current supply for ideas.  Since it is a lab instrument it's probably more precise and complex than is needed to meet your needs.

[Marco]

How about a buck converter with a switch which is normally always on with say a 10 nF output capacitor and 100 uH inductor, with a fast comparator which turns the switch off when current goes over 250 mA and some circuit to force a minimum on/off time of say 100 ns?

So in normal operation it's just a 16V power supply with a 100 uH/10 nF output filter, when you short it out it becomes a semi constant current buck converter.

[martin_nga]

Thanks for your answers.

Have you seen http://ww1.microchip.com/downloads/en/appnotes/atmel-42071-dali-slave-reference_design_application-note_at01244.pdf ?

Yes. However, I can't find anything about power supplies in there. There are also appnotes from Microchip, Motorola, NXP, Renesas and ST. Only the Microchip one contains a simple one transistor power supply.

I would consider changing the design to a constant current source with a voltage clamp to set the maximum output voltage. ...

I had the same thought already. Do you think this will reduce the tendency to oscillate? Keep in mind that this will have up to 300m of AC wire with unknown impedance connected to it... [I left out EMC / over voltage protection stuff in circuit diagram]

How about a buck converter with a switch which is normally always on with say a 10 nF output capacitor and 100 uH inductor, with a fast comparator which turns the switch off when current goes over 250 mA and some circuit to force a minimum on/off time of say 100 ns?

So in normal operation it's just a 16V power supply with a 100 uH/10 nF output filter, when you short it out it becomes a semi constant current buck converter.

We have no experience with designing switching supplies. I have played a bit in LTspice with buck converters, none had a control loop even close to fast and stable enough to meet the specification requirements. So I went the "save" route with linear.
I also have the problem that semi constant current is not acceptable. Voltage can be in range 14..21V, whereas current needs to be as high as we can get, while it mustn't exceed 255mA (with some overshoot on switchover).

[Marco]

I have played a bit in LTspice with buck converters

I'm talking about buck configuration, not any specific buck controller. Much like linear regulators as current sources, abusing existing specialized ICs is unlikely to be the best solution. I was thinking about a circuit something like follows. U1 is not a real device, it's a parametrized comparator with 20 ns delay, 3 ns rise/fall, 2 mv hysteresis and switches the gate from 16 to 6 Volt. To actually implement it you'd use something like a TS3011 and a fast gate driver.

PS. it has no compensation, so it will have a tiny bit of voltage overshoot after the short ends (About 4V in simulation.) It's just the basic concept though ... making it into something practical is your problem.

[duak]

I'm sorry, I gave the wrong link - this should be correct: http://ww1.microchip.com/downloads/en/Appnotes/01465A.pdf  It shows a simple analog current limiter on page 3.

The analog current limiter will have to dissipate a significant amount of power if there is a lot of traffic on the DALI or if there is a short circuit so Marco's idea could be an advantage.  It may have a problem with the value of the series inductor and the load capacitance forming a resonant circuit.  The resonant frequency should be chosen to be well above the bit rate.  Some sort of damping and clamping may also be needed to prevent high voltages from being generated.

[Marco]

There was a certain method to the madness of the values I chose, there's limited room to mess with them. A flyback Schottky to the power supply to clamp the output voltage seems to work pretty well to keep everything relatively smooth.

PS. digikey only has one inductor with a resonance frequency comfortably high enough compared to the frequencies involved ... so that might be a consideration too.

[none]

@martin_nga Hi, did you find a solution? Did you go with Marco's "chopper"?