Switched bulk capacitance

[foxtrot]

I have an application in which I've got a buck converter for which I need to tightly control inrush power consumption, and also need relatively high bulk output capacitance (1mF). I feel like I've seen this before, but has anybody here come up with (presumably) a inrush-controlled means of switching in bulk capacitance? I was thinking a low-side switch on one or more big caps w/configurable inrush control would be good, but there's nothing off the shelf that seems to work nicely at my voltage range (24V). Thoughts, napkin sketches, and links to reference designs would be much appreciated 🙏

[langwadt]

just current limit the buck?

[jbb]

just current limit the buck?

+1

For bonus points, a buck with input current limit can be the most effective thing for charging those caps up fast.

——

If your load equipment doesn’t like a slow voltage rise time, you could add a power switch between the buck converter (incl 1mF cap) and the load.

——

1 mF is a lot of output capacitance. It might be absolutely necessary, but can you get it down somehow? For ripple, maybe a multiphase buck could improve the situation? Or go 0.5V high and use a linear LDO for cleanup?

If you need big cap bank for a pulse discharge, I suggest that you really don’t want to add extra resistance to the cap bank in the form of some extra switch

[twospoons]

If your buck has a soft-start function, just set that up to be long enough to keep the inrush low.

[foxtrot]

If your buck has a soft-start function, just set that up to be long enough to keep the inrush low.

That was my original thought; however, I'm a little worried that a long-enough soft start (maybe half a second?) will cause problems (i.e. the device thinking it has some sort of fault).

just current limit the buck?

Are you thinking basically a current sensing amplifier tied to the buck's feedback pin? I haven't designed something like that before :/

1 mF is a lot of output capacitance. It might be absolutely necessary, but can you get it down somehow? For ripple, maybe a multiphase buck could improve the situation? Or go 0.5V high and use a linear LDO for cleanup?

Yeah; there's a class D amplifier attached to this thing – as well as a few other bucks for various elements in the system. Suggestions are always appreciated.

[mariush]

NTC/PTC resistor in series with the voltage, whichever version has high resistance when cold and low resistance when hot.

A timer or something that would activate a mechanical relay or something equivalent to short out the resistor after a few seconds... this is common in computer power supplies

[jbb]

On soft start: I don’t remember seeing ‘soft start too slow’ in buck controller data sheets. Maybe you could experiment?

On current limiting: theoretically you can put a current shunt and sense amp etc down yourself, but I recommend picking a buck controller with current limiting built in.

Yeah; there's a class D amplifier attached to this thing – as well as a few other bucks for various elements in the system. Suggestions are always appreciated.

Hmm. Perhaps a dumb question: could you throw out the buck converter and run your Class D amp straight off the original power supply?

 Traditional linear audio amplifiers are known for liking great big DC caps because they need to ride though the nulls in the incoming AC power. Have you experimented with different sizes of cap to see what’s necessary for you? And if the sound quality is impaired, do you know how/why?

I haven’t worked with Class D before, but I imagine 2 things to look out for are “the DC rail dips too much for bass”, “the switching noise from the buck makes undesirable tones in the outputs.”

If it’s the latter, have you looked into synchronising the buck converter and Class D amplifier switching frequency? This can (as I understand it) help you place weird frequency mixing products outside the audible range.

[Siwastaja]

Every buck converter in existence, which does not self-destruct, already has this current limiting built-in. This includes basically every buck control IC on the market. You are solving an imaginary problem.

Worst offenders from 1980's do it with "soft start" circuitry (which just makes the voltage reference ramp up slowly) alone, but it's still there; if it wasn't, the inductor just saturates and the MOSFETs go bang on first power-up, with much less capacitance than 1mF. Most sensible modern controllers add some way of more or less approximate current sensing (e.g., Vds drop sensing of integrated mosfet, or connections to explicit current sense shunt resistor), and cycle by cycle limitation.

[David Hess]

Are you thinking basically a current sensing amplifier tied to the buck's feedback pin? I haven't designed something like that before :/

For current mode regulators, it works better to current limit by applying feedback to the regulator's compensation pin which itself controls the peak inductor current.  This is also how soft start should be implemented.

That was my original thought; however, I'm a little worried that a long-enough soft start (maybe half a second?) will cause problems (i.e. the device thinking it has some sort of fault).

It seems odd that such a large output capacitance is required.  Switching regulators are usually more limited by the impedance of the output capacitor rather than the capacitance itself.

[the_cake_is_a_lie]

and also need relatively high bulk output capacitance (1mF)

Yeah I haven't seen more than 220uF bulk capacitance on the output of a switching regulator. I'd question if 1mF were really necessary.

[foxtrot]

You are solving an imaginary problem.

This specific application is powered by an upstream buck that has a very precise power limit. If I exceed this power limit, the upstream buck will shutdown for a period and reset. I have no control over this, so I have to design my buck to play by its rules. This is – at least, in my mind – the reason that I need to soft start: charging that much output capacitance is going to cause me to exceed my power limit. Is this assumption somehow incorrect?

Yeah I haven't seen more than 220uF bulk capacitance on the output of a switching regulator. I'd question if 1mF were really necessary.

Yeah, I would love if it weren't. I may try to get around it by distributing bulk capacitance along the output rail behind inrush-controlled load switches.

On soft start: I don’t remember seeing ‘soft start too slow’ in buck controller data sheets. Maybe you could experiment?

Sadly, this was relayed to me by TI's engineers as their concern. So I guess it's something they just don't consider relevant enough to put in their datasheets. 😂

Hmm. Perhaps a dumb question: could you throw out the buck converter and run your Class D amp straight off the original power supply?

Not a dumb question! I have a wide V_in range that's roughly double my output voltage. The Class D amp operates *very* efficiently and with minimal THD at this output voltage – which is why I chose it as the primary rail voltage. Unfortunately (as you see elsewhere in this post) there's a buck upstream of this one that is *very* tightly power-limited, with a circuit-breaker behavior. If I exceed the power limit, I get shut off. Not my choice, and out of my control.

Traditional linear audio amplifiers are known for liking great big DC caps because they need to ride though the nulls in the incoming AC power. Have you experimented with different sizes of cap to see what’s necessary for you? And if the sound quality is impaired, do you know how/why?

I don't have a lot of room to experiment with this one; the datasheet of my amp does a lot of heavy lifting, since it clearly specifies max ripple, its effect on THD, etc. From the perspective of my device, I've got a DC power source.

I haven’t worked with Class D before, but I imagine 2 things to look out for are “the DC rail dips too much for bass”, “the switching noise from the buck makes undesirable tones in the outputs.”

If it’s the latter, have you looked into synchronising the buck converter and Class D amplifier switching frequency? This can (as I understand it) help you place weird frequency mixing products outside the audible range.

I suspect you're right re: DC rail dipping for too much bass. Luckily, my Class-D will change its equalization on the fly if it detects that its supply rail is browning out, but the overall goal of sound quality is compromised in this case – which is also something to avoid. I'm sizing for 1.5x RMS current.

[Siwastaja]

This specific application is powered by an upstream buck that has a very precise power limit. If I exceed this power limit, the upstream buck will shutdown for a period and reset. I have no control over this, so I have to design my buck to play by its rules.

But you still have "your buck" - so just use a low enough current limit. Practically, maybe this means looking at buck controller IC which actually have inputs for current sense resistor, this way it can be more accurate.

Input current limiting, as mentioned by some others, would do the whole charge a bit faster, but it is a more specialized feature, while output current limiting exists in every sensible buck converter, and even more accurate versions are not that hard to find. Also look for an IC where you can disable the hickup/fault mode so that it can run in current-limited mode for unlimited time.

This is – at least, in my mind – the reason that I need to soft start: charging that much output capacitance is going to cause me to exceed my power limit. Is this assumption somehow incorrect?

It is incorrect, because the charging power is something you can choose, and as I said, it is something every buck converter already controls to some extent. Just use lower current, and if your application requires especially exact regulation, then just find a controller which allows more exact control; those with explicit shunt resistor inputs are better, and true CC-CV parts are even better.

[foxtrot]

But you still have "your buck" - so just use a low enough current limit. Practically, maybe this means looking at buck controller IC which actually have inputs for current sense resistor, this way it can be more accurate.

I've had somewhat of a difficult time finding them.. Perhaps you'd be willing to take a quick look? My constraints are as follow:

V_in = 36-60V
V_out = 21V
V_out_ripple = 200mV p-p
I_out = 3.8A

[mariush]

Maybe have a look at ICs like SIC461 / SIC462


SIC461 (max 10A out) : https://www.digikey.com/en/products/detail/vishay-siliconix/SIC461ED-T1-GE3/7682210

SIC462 (max 6A out) : https://www.digikey.com/en/products/detail/vishay-siliconix/SIC462ED-T1-GE3/6574659

FAN65004B may also work : https://www.digikey.com/en/products/detail/onsemi/FAN65004B/8636353

maybe renesas raa211651 ?  https://www.digikey.com/en/products/detail/renesas-electronics-corporation/RAA2116514GNP-HA0/15196000

[Siwastaja]

I've had somewhat of a difficult time finding them.. Perhaps you'd be willing to take a quick look?

I remember using LM25117 once, this has shunt resistor connections, and datasheet has an example circuit to feedback the current for accurate CC mode if the crude one doesn't suffice (it probably does). It has pins to disable hickup/fault mode, too, so it will happily output against current limit for unlimited time; I used it for CC-CV battery charging. Voltage range does not suffice for you but maybe it helps to know I found the part using TI's WEBENCH which has this "IC FEATURES" filter more capable than those at distributor sites.

[foxtrot]

I've had somewhat of a difficult time finding them.. Perhaps you'd be willing to take a quick look?

I remember using LM25117 once, this has shunt resistor connections, and datasheet has an example circuit to feedback the current for accurate CC mode if the crude one doesn't suffice (it probably does). It has pins to disable hickup/fault mode, too, so it will happily output against current limit for unlimited time; I used it for CC-CV battery charging. Voltage range does not suffice for you but maybe it helps to know I found the part using TI's WEBENCH which has this "IC FEATURES" filter more capable than those at distributor sites.

Yeah; V_IN(max) of the LM25117 is insufficient. The TI portfolio is pretty limited in this regard.

SIC462 (max 6A out) : https://www.digikey.com/en/products/detail/vishay-siliconix/SIC462ED-T1-GE3/6574659

This may be a good option. I'm going to dig into it further! Thanks for the suggestion.