is there a switching regulator equivalent to an LM337?

[robert.rozee]

i have a mains transformer in a piece of test equipment producing a rectified output (lets call it V_RAW) ranging between around 15v and 30v DC. from that single DC supply, i need to create two more rails: one 5v above 0v/ground and the other 5 volts below the rectified 15-30v (V_RAW). now for any low-current application this would normally be simple: just use a LM317 fed from V_RAW to generate the 5v rail, and an LM337 fed from 0v/ground to generate the V_RAW-5v rail. BUT, in this application i need to be able to draw around 800mA from each of the new rails (relative to the adjacent V_RAW or 0v/ground) and so need to use buck converters to keep within the mains transformer power rating.

the 5v rail is no problem, a LM2596 module will do a perfectly fine job. but, finding a solution for the V_RAW-5v is proving a tad more problematic.

simple question: is it possible to configure an LM2596 or LM2577 to take a negative input and convert it to a lesser negative voltage? ie, the equivalent of a switch-mode LM337?


much appreciate any suggestions...


cheers,
rob   :-)

[Siwastaja]

This sounds weird enough so let me confirm:

You want a regulator which tracks the input voltage and produces output voltage which is always 5V less than the input voltage, e.g. from 30V you want 25V, and from 15V you want 10V. Are you sure this is what the test equipment wants? If yes, what kind of response time should it have, i.e., maybe it needs to follow 100/120Hz ripple present in its input, yes?

[Andy Chee]

I’ve wondered the same thing myself, but resigned to myself to alternative means of generating dual positive/negative rails.

Instead, a flyback configuration comes to mind. The output windings of the flyback transformer can give 5V and -5V.

However in my suggested topology, only the 5V is used as feedback, which may mean the -5V isn’t as tightly regulated (which may or may not be a problem).

[robert.rozee]

This sounds weird enough so let me confirm: You want a regulator which tracks the input voltage and produces output voltage which is always 5V less than the input voltage, e.g. from 30V you want 25V, and from 15V you want 10V

yes, that is more-or-less it. imagine an LM337 (an adjustable negative linear regulator) with the input pin tied to 0v, and the adjust pin connected to the middle of a voltage divider strung between V_RAW and the output pin. an 800mA (6 ohm) load is then connected between V_RAW and the output pin of our LM337.

now take out the LM337 and replace it with a switch-mode regulator where the +ve in and out are common, and the -ve in and out have the inductor inline.


cheers,
rob   :-)

[Siwastaja]

I’ve wondered the same thing myself, but resigned to myself to alternative means of generating dual positive/negative rails.

There is no negative rail in what the OP suggests. OP just describes using a "negative" regulator in a weird way to perform a "high-power voltage subtraction", but every circuit node is above GND.

yes, that is more-or-less it.

Out of curiosity (because your description sounds like an X-Y problem, and solving the actual problem is usually more important to everyone), what kind of test equipment would use this kind of unregulated, tracking, "rectified input minus something" power supply, what is it used for? I would understand if it was a measurement signal, i.e. loaded only by some mA, but really high power enough to warrant an SMPS? What does it do?

[robert.rozee]

Out of curiosity (because your description sounds like an X-Y problem, and solving the actual problem is usually more important to everyone), what kind of test equipment would use this kind of unregulated, tracking, "rectified input minus something" power supply, what is it used for?

firstly, the mains transformer in the equipment has a single primary winding that can handle an input of anything from 90v up to 264v AC. hence the very wide range of the unregulated V_RAW. next, the equipment includes rechargeable batteries for use when mains is not available. these batteries are permanently connected directly across the V_RAW supply via a series diode. ie, when/if V_RAW disappears the batteries take over immediately.

the original batteries consisted of four lead-acid cells in series, delivering 8 volts. i am working on replacing these with eight Li-Ion cells arranged as 4 paralleled strings, each string having 2 cells in series. it is easiest to place the series diodes (one for each string) between the two cells, and then have eight independent charging circuits - four for the 'high side' cells, and four for the 'low side' cells.


cheers,
rob   :-)

[Siwastaja]

Oh, so it was an X-Y problem; that's why I asked! You can't safely substitute with li-ion cells, because unlike lead-acid, they absolutely need active current limiting circuitry (and much more; accurate voltage regulation instead of your suggested un-regulation; temperature qualification, initial charge voltage qualification...)

You need a lithium ion charger circuit; plenty of integrated parts which do exactly that. They even have separate sections on part distributor websites.

[robert.rozee]

You need a lithium ion charger circuit; plenty of integrated parts which do exactly that

i've got that part of the problem mostly covered. the cells will be severely de-rated, receiving a maximum charge voltage of 4.0v and a 200mA (C/10) maximum charge current. this will be backed with a small micro monitoring things and shutting down charging of individual cells when the charge current for that cell drops below 50% (100mA). even with these limitations, the design will still provide considerably better performance than the original lead-acid solution.

interestingly, there are few (if any) integrated parts that have any flexibility in reducing the maximum charge voltage. manufacturers tend to instead go for maximum stored energy and minimum charging time at the expense of cell service life. they also seem to prefer relying upon closely-matched cells welded together rather than using diodes (with the associated small voltage drop) to combine multiple strings of cells.


cheers,
rob   :-)

[Andy Chee]

How about a flyback wired like this?

(excuse the crudity of the drawing)

I think I might have the phasing dot or the diode the wrong way, but hopefully you get where I'm going.  Basically subtracting 5V from Vraw.

[Andy Chee]

This buck configuration might work as well

[Siwastaja]

i've got that part of the problem mostly covered. the cells will be severely de-rated, receiving a maximum charge voltage of 4.0v and a 200mA (C/10) maximum charge current. this will be backed with a small micro monitoring things and shutting down charging of individual cells when the charge current for that cell drops below 50% (100mA). even with these limitations, the design will still provide considerably better performance than the original lead-acid solution.

interestingly, there are few (if any) integrated parts that have any flexibility in reducing the maximum charge voltage. manufacturers tend to instead go for maximum stored energy and minimum charging time at the expense of cell service life. they also seem to prefer relying upon closely-matched cells welded together rather than using diodes (with the associated small voltage drop) to combine multiple strings of cells.

As a battery system engineer (too), I can say that series-then-parallel approach, is almost always the wrong one, and if you have to add diodes and special charging circuits it sounds pretty bad, and if you dynamically connect/disconnect cells/strings, then... $deity bless you.

Besides, I still completely fail to see how battery charging requires an anti-regulated supply with accurate-zero PSRR. If you roll your own, you still want a regulated supply, e.g 4.0V/cell, no?

This seems like a very classic X-Y problem, i.e. "hey guys I have this totally unorthodox and weird way of doing a normal thing and I think this is exactly what is needed, I just need a little bit help with X to make it work".

Just don't do it. Look for advice how to do it normally.

[robert.rozee]

How about a flyback wired like this?
[...]
This buck configuration might work as well

i was hoping to avoid a transformer - if going down that path it would be easier to just build a 1:1 isolated DC-DC converter that was run from the 5v buck regulator... i may well end up doing this!


cheers,
rob   :-)

[ArdWar]

i was hoping to avoid a transformer - if going down that path it would be easier to just build a 1:1 isolated DC-DC converter that was run from the 5v buck regulator... i may well end up doing this!

Might as well go all the way with transformers/coupled inductor. You're planning to use "eight independent charging circuits", you at least need 4 transformers already unless you design your own charge control.

[robert.rozee]

As a battery system engineer (too), I can say that series-then-parallel approach, is almost always the wrong one, and if you have to add diodes and special charging circuits it sounds pretty bad, and if you dynamically connect/disconnect cells/strings, then... $deity bless you.
[...]
Just don't do it. Look for advice how to do it normally.

there are a number of other constraints: interface to the existing equipment is via a 10-way cable, and the equipment can not be modified in any way beyond connecting to that cable. charging may only occur when the equipment is switched off on the front panel (the main transformer is permanently powered) to keep within power budgets. the Li-Ion cells must have a service life of 10+ years - if a single cell goes short/leaky it can not adversely affect other cells in adjacent strings. and internal heating during charging is a big no-no.

as i have said, the Li-Ion cells are deliberately being severely de-rated, and a few isolating diodes are of no concern. the charging circuits are extremely simple, so there is no problem creating 8 of them - i have looked at various single-chip charging solutions, but none are suitable. the only stumbling block at the moment is that there appears to be no switching regulator equivalent to an LM337. this surprises me, to say the least. at the moment it is looking like a 1:1 transformer isolated DC-DC converted may be the most practical solution.


cheers,
rob   :-)

[robert.rozee]

Might as well go all the way with transformers/coupled inductor. You're planning to use "eight independent charging circuits", you at least need 4 transformers already unless you design your own charge control.

just one should suffice! i actually have a bag of 5v/200mA modules and was considering the possibility of using them, but it seemed like overkill. i am indeed designing my own charge control: each cell with be protected against over-discharge by a DW01A based module, CC-CV charging will be provided by a TL431 and 2-3 transistors, and charge termination will be controlled by a microcontroller monitoring the current flow into each cell.


cheers,
rob   :-)

[robert.rozee]

This buck configuration might work as well [...]

i think this might be the way to go:



(taken from https://www.ti.com/lit/ds/symlink/lm2577.pdf)

except with the secondary isolated and feedback obtained via a TL431+optocoupler. i've got a few boost modules laying around that would be suitable to repurpose. i think a buck converter wouldn't be suitable, as the feedback would be inverted (referring to your sketch: as V+ increases, V- would decrease, and vice versa).


cheers,
rob   :-)

[PCB.Wiz]

there is also this
https://www.eetimes.com/practical-implementation-of-negative-input-negative-output-step-down-switching-converters/

Shows how to use a level shifter

Addit : a better rated version of the eetimes example and the LM2577 looks to be

lcsc XL6019E1   XLSEMI boost converterType 60V 5V~40V 5A TO-263-5 DC-DC 100+ US$0.4057

[Andy Chee]

i think this might be the way to go:

 i think a buck converter wouldn't be suitable, as the feedback would be inverted (referring to your sketch: as V+ increases, V- would decrease, and vice versa).


cheers,
rob   :-)

Both my sketches only have single feedback from the +5V output, which you can isolate if you wish.

You’re correct, a ground referenced feedback would indeed be inverted, but a floating and isolated feedback can be “polarity corrected”.

Then you could sum both feedback signals (diode OR?) back to the controller.

But I would test the single feedback version first, and check whether regulation remains in spec on the non-feedback rails.

[Zero999]

How about a flyback wired like this?

(excuse the crudity of the drawing)

I think I might have the phasing dot or the diode the wrong way, but hopefully you get where I'm going.  Basically subtracting 5V from Vraw.

(Attachment Link)

This buck configuration might work as well

(Attachment Link)

The diodes on the negative side are the wrong polarity. It should work, if the polarity is reversed. Note there needs to be more current drawn from the positive, than the negative side, for the transformer hacks to work properly.

[langwadt]

quite a few LED drivers that are basically a buck turned upside down, i.e. switch and diode swapped and output is referenced to the input instead of ground

[PCB.Wiz]

quite a few LED drivers that are basically a buck turned upside down, i.e. switch and diode swapped and output is referenced to the input instead of ground

Yes, it is worth checking LED drivers for candidates.
Things to watch for :
 * Ones with high-side current sense, tend to be hysteretic control, which means they expect ~13% ripple on the sense pin.
 * Tolerances of LED drivers is not usually as tight as DC-DC parts.

I did find lcsc have a XL6013 LED driver, which seems to be the XL6019 die I linked above, with minor tweaks : SO8 package, 220mV low side sense, same 3% tolerance spec, same 110mOhm FET
You can use matched pair PNP parts to level shift as per #16 (eg DMMT3906W match 2mV VBE and 2% HFE )