HElp with constant current limiting a SMPS

[Hiemal]

I've been reading/researching a bit, and noticed there's a few "methods" to making/adding a constant current mode to a power supply.

That being said, I know there's high side, and low side; high side current sense has the advantage of not breaking the ground connection between whatever it is you're powering. Downside, more complex. Usually requires a fancy high side current sense amp.

Low side is "easy" to do, but requires breaking the ground. Usually see this done with a single op amp and some resistors.

I'm working on making an adjustable SG3525-based half bridge supply, using a transformer I pulled from a server power supply (and series'd all the output windings, to get hopefully around 48v). I'm using an opto-isolator, with a TL431 + a dropping transistor to get around the 36v max limit of the TL431.

So, a few questions.

1. What exactly does "breaking the ground" mean for you, if you're trying to make an adjustable power supply?

2. What method of constant current control is "best" for stability, when it comes down to getting that signal back to the PWM controller? I'm seeing methods where they use a differential amp, get a signal, and then use a comparator to check if the current level is above or below whatever you have set; if it's above, then the comparator outputs a high, which goes to the feedback pin causing the output to decrease. Something like this; https://www.smbaker.com/wordpress/wp-content/uploads/2015/10/lm2576_schematic_3-1024x758.jpg

Is this okay to do? I would think it would cause oscillations since you're on/offing the FB pin...

Compared to, say, taking that same signal from the shunt resistor and amplifying it until it starts sinking or sourcing current from the feedback pin. Usually with a diode in series with this.

That, or I'm misunderstanding something entirely, and there's a better methodology that's out there that I don't fully grasp.

Edit: Oh, one more bit; if high side current sensing, and you're using a fixed gain amplifier, is it "okay" to use a resistive divider to knock the gain down to something more manageable? A lot of the amplifiers I'm seeing are more expensive in more "useful" gain configurations, so it'd be a relatively simple way to save money in that regard.

Thanks for the help/explanations. 

[TimNJ]

There's no need to do use "bang-bang" control for the current. You can set up your 431 voltage control loop as normal, and then use a single op-amp for the CC error amplifier. Then place an "or-ing" diode at the cathode of the opto LED, facing into the CC op-amp output. When the constant current condition is met, the CC loop will take over. Generally this works okay.

Alternatively, you can use a part specifically made for this task, such as TSM1052. A more elegant solution I think.

https://www.st.com/resource/en/datasheet/tsm1052.pdf

[MosherIV]

A1 The 0V output from the psu will not be absolute. If the 0V is ground referenced, it will not be at true ground but offset by the voltage drop of the low side shunt.

A2 I am no expert.
Do what TimNJ suggests.
FYI
The difference between Constant Voltage and Constant Current  is that
CV allows the current to vary according to the load, upto the current capability of the psu.
CC varies the voltage to maintain the current through the load, upto the max voltage of the psu.

Ohm law must be obeyed!

[David Hess]

That being said, I know there's high side, and low side; high side current sense has the advantage of not breaking the ground connection between whatever it is you're powering. Downside, more complex. Usually requires a fancy high side current sense amp.

Low side is "easy" to do, but requires breaking the ground. Usually see this done with a single op amp and some resistors.

Nothing prevents using the common circuit topology for low side current control on the high side.  Honestly, if a separate difference or high side current sense amplifier is needed, then you are doing it wrong.  See my last answer below.

1. What exactly does "breaking the ground" mean for you, if you're trying to make an adjustable power supply?

It means breaking the connection between input and output grounds such that there is a potential difference between them.  This does not apply to a regulator which provides a floating output because the ground can be broken on either side with no effect.  Old power supply designs sometimes take advantage of this by using a floating input so that current sensing can be implemented on the low side without breaking any ground.

2. What method of constant current control is "best" for stability, when it comes down to getting that signal back to the PWM controller? I'm seeing methods where they use a differential amp, get a signal, and then use a comparator to check if the current level is above or below whatever you have set; if it's above, then the comparator outputs a high, which goes to the feedback pin causing the output to decrease.

The first step is to not use a topology which requires a difference amplifier, which either adds phase lag or limits precision.

Edit: Oh, one more bit; if high side current sensing, and you're using a fixed gain amplifier, is it "okay" to use a resistive divider to knock the gain down to something more manageable? A lot of the amplifiers I'm seeing are more expensive in more "useful" gain configurations, so it'd be a relatively simple way to save money in that regard.

There is nothing wrong with operating the current sensing amplifier at low gain, except for using a current sensing amplifier at all; it is not required.  The current shunt can directly feed the error amplifier.