Author Topic: Very wide range isolated DC-DC converter  (Read 3426 times)

0 Members and 1 Guest are viewing this topic.

Offline forrestcTopic starter

  • Supporter
  • ****
  • Posts: 720
  • Country: us
Very wide range isolated DC-DC converter
« on: January 18, 2019, 08:06:22 am »
A while back I asked a similar question (same project), and I don't think I explained what I was hoping to find very well.   I don't seem to be able to find the old topic to re-open, so I'm starting a new one...

I have a design which is used in an environement where there are multiple, redundant power sources, all ground referenced, and each can be either + or - in respect to ground.  I don't really have control over the power sources - I have to live with whatever is available.   There might be a +12VDC source and a -48VDC source as an example.   I need to pull power from whatever happens to be available and handle situations where one power supply dissapears during operation.  In most cases I'm needing 3.3V at about 500ma.

Currently my power supply architecture looks like this:



This turns out to be rather expensive, as the negative to positive supply costs about $10 in parts.   I'd like to figure out some way to make this less expensive.

In an ideal world, I'd find something inexpensive which would allow an architecture like:



Unfortunately that wide of power supply range doesn't commonly exist.   I have found exactly one part which might do this, but it seems to be unobtainable at least through distribution.

I've also tried to come up with some sort of low cost switching method which would allow me to just use a single 10-60Vin 3.3V out isolated supply which switches between the positive and negative rails, but so far I haven't been able to come up with something I felt was reliable and was relatively low cost.    A schematic of what I'm trying to describe is shown:



I know I could probably rig some sort of relay up to do the rail selection, but I hate electromechanical devices (they fail).  I haven't seem to be able to come up with an electronic circuit that I'm happy with - specifically because of apparent failure modes.

Right now, I'm primarily looking for options that I might not have considered because I didn't know about them..   A couple additional notes:

1) I show 2 voltage inputs, I can have up to 3 or 4 at times.

2) I'm ok with an output of like +12V (or other similar intermediate voltage), as regulators from 12V down to 3.3V are cheap and plentiful.  In fact, this might be preferable to 3.3V out directly.
« Last Edit: January 18, 2019, 08:22:29 am by forrestc »
 

Offline aheid

  • Regular Contributor
  • *
  • Posts: 245
  • Country: no
Re: Very wide range isolated DC-DC converter
« Reply #1 on: January 18, 2019, 05:09:35 pm »
Now, I'm just a noob so this might be dumb, in which case I apologize.

From what I can see the NCP1031[1] can handle 8V up to 200V. Now I didn't dig into it to see if you could realistically make it work over such a large range, so there's that...

Quote
The NCP103x can be configured in any single-ended topology such as forward or flyback. [...] The internal error amplifier allows the NCP103x family to be easily configured for secondary or primary side regulation operation in isolated and non-isolated configurations.

Maybe it could work for your preferred solution?

edit: forgot to add that at least Farnell stocks it, for about $1.5 a pop.

[1]: https://www.onsemi.com/PowerSolutions/product.do?id=NCP1031
 
The following users thanked this post: forrestc

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15794
  • Country: fr
Re: Very wide range isolated DC-DC converter
« Reply #2 on: January 18, 2019, 07:01:21 pm »
LTC3894: https://www.analog.com/en/products/ltc3894.html
Available at Digikey, Mouser, Arrow (and maybe others).

Not sure I really got your "isolation" requirements here. There is a common ground...? Unless the ground on the right side is not the same as the ground on the left side of your schematics, in which case you should definitely use different ground symbols in order to avoid confusion.

That said, if you need isolation, you can then add a DC-DC isolated converter after the LTC3894 which would down-convert to say 5V. A 5V->3.3V @500mA is easy to find and relatively low cost, such as the Murata MEJ2S0503SC: https://power.murata.com/en/mej2s0503sc.html

If your min input voltage is 10V and it goes through diodes, you obviously can't get a 12V output with a buck converter design only.
But you can always use a 5V-input, 12V-output isolated DC-DC converter for instance after the LTC3894. The RECOM RKZE-0512S would be a good option, it's relatively cheap (around $5). It's a 2W output max; since you first required 3.3V @500mA max, which is 1.65W, this should be ok. Should you still want 3.3V, you can always add a 3.3V buck converter which should you get at least 500mA max output current. Limiting the isolated DC-DC converter to 2W looks like a good idea to keep costs down. Anything higher than that will cost significantly more (often > $10).

Of course you could always design a flyback converter from scratch instead of using the above topology. Would probably lead to a much lower unit cost in the end, but I would advise against that option unless you have already designed flyback converters. Specifying the transformer itself requires skills, and then you'd have to have it custom-made. Eek. You'd have to sell a lot of devices in the end to make that a profitable option compared to the above IMO.

« Last Edit: January 18, 2019, 08:06:22 pm by SiliconWizard »
 
The following users thanked this post: forrestc

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Very wide range isolated DC-DC converter
« Reply #3 on: January 18, 2019, 07:20:39 pm »
A load dump protection circuit would probably be alright.  You're only drawing a couple watts output, so you can afford to drop a few more watts at the highest input voltage.  That greatly reduces the burden on the converter, so an 80V, or even 60 or 40V rated part can be used.

Also consider a controller instead of a regulator: you add an external transistor of whatever rating, and you supply a little power to the controller (say 5-40V of a few mA, easily done with a linear reg).  You can get very high ratings this way, probably limited more by switching loss and control stability than by voltage as such.

You could maybe also arrange some kind of mux to pick one good supply, in such a way that it stays redundant, cutting over seamlessly.  You can't use diodes to, for example, disable the -48V supply while +48 is active, but an active circuit could.  This would limit the converter's liability to only 60V, instead of 120V.

It's not immediately obvious what solution(s) would work here -- it's a nonlinear function, so transitioning from one state to another can be tricky.  If a digital solution is used (comparators and switches), it will be prone to oscillation under some conditions (even with hysteresis), and secondary problems like break-before-make and dead time and propagation delay.  A continuous (analog) solution would be nice, and is probably possible, it would just require some thought to synthesize.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 
The following users thanked this post: forrestc

Offline Gribo

  • Frequent Contributor
  • **
  • Posts: 649
  • Country: ca
Re: Very wide range isolated DC-DC converter
« Reply #4 on: January 18, 2019, 07:45:39 pm »
If you can raise the lower limit to 18V, there are 4:1 Input range DC to DC converters from many sources (Traco, Meanwell, etc).I have managed to find an 8:1 Input range converter, however, it is 300$  :scared:
I am available for freelance work.
 
The following users thanked this post: forrestc

Offline forrestcTopic starter

  • Supporter
  • ****
  • Posts: 720
  • Country: us
Re: Very wide range isolated DC-DC converter
« Reply #5 on: January 19, 2019, 12:48:21 am »
Not sure I really got your "isolation" requirements here. There is a common ground...? Unless the ground on the right side is not the same as the ground on the left side of your schematics, in which case you should definitely use different ground symbols in order to avoid confusion.

The isolation in the schematics are to permit the negative side of the input of the DC-DC converter to be either at 'ground'  or 'negative rail', whichever is lower, while having the negative output of the DC-DC converter be at ground. I notice I missed a wire on the negative output of the DC-DC converter, tying it to ground, which would have made this all make a lot more sense.

The concept being that the isolated DC-DC converter permits the output ground to be different than the negative input of the DC-DC converter.

If your min input voltage is 10V and it goes through diodes, you obviously can't get a 12V output with a buck converter design only.

That is true.   In the current design the range is actually +10-60V on the positive side, and -18-60V on the negative.   The neg to positive switches it to 12.

I'm not particulary hung up on 12V.  I guess the idea would be better described as to use a higher-voltage rail (or some intermediate voltage), which then is further regulated.


That said, if you need isolation, you can then add a DC-DC isolated converter after the LTC3894 which would down-convert to say 5V.

And that suggestion is why I sometimes ask for suggestions.   What a simple & obvious potential solution. 

I have found that often it's the 'oh duh, I don't know why I didn't consider that' suggestions which break these types of problems loose.   Apparently at some point I thought it might be nice to do the isolation in the first step, and then never considered anything else.   

I will say that it would be easier if the first block provided a ground referenced output, but even if it doesn't that just means I have to use small isolated blocks for my other rails.   
 

Offline forrestcTopic starter

  • Supporter
  • ****
  • Posts: 720
  • Country: us
Re: Very wide range isolated DC-DC converter
« Reply #6 on: January 19, 2019, 02:12:58 am »
A load dump protection circuit would probably be alright.  You're only drawing a couple watts output, so you can afford to drop a few more watts at the highest input voltage.  That greatly reduces the burden on the converter, so an 80V, or even 60 or 40V rated part can be used.

Although this is a perfectly reasonable solution, Unfortunately this particular design happens to be used in off-grid/solar sites on a fairly regular basis where every watt counts.   I am going to investigate this as an option though, since doing something similar might make some sense...  i.e. "pre-regulate" at the highest voltages, even if it's a simple 2:1 DC-DC converter.

Also consider a controller instead of a regulator: you add an external transistor of whatever rating, and you supply a little power to the controller (say 5-40V of a few mA, easily done with a linear reg).  You can get very high ratings this way, probably limited more by switching loss and control stability than by voltage as such.

I looked at several controller designs, I think the problem was that I was looking for isolated designs which turn into a rather large circuit, and are fairly few between.  I'm going to go back and look some more.

You could maybe also arrange some kind of mux to pick one good supply, in such a way that it stays redundant, cutting over seamlessly.  You can't use diodes to, for example, disable the -48V supply while +48 is active, but an active circuit could.  This would limit the converter's liability to only 60V, instead of 120V.

It's not immediately obvious what solution(s) would work here -- it's a nonlinear function, so transitioning from one state to another can be tricky.  If a digital solution is used (comparators and switches), it will be prone to oscillation under some conditions (even with hysteresis), and secondary problems like break-before-make and dead time and propagation delay.  A continuous (analog) solution would be nice, and is probably possible, it would just require some thought to synthesize.

I think you've summarized very well why I haven't ended up down that particular solution.   I still have some hope that some magic combination of parts will yield a simple, easy to design mux/switcher, but so far no luck.   
 

Offline forrestcTopic starter

  • Supporter
  • ****
  • Posts: 720
  • Country: us
Re: Very wide range isolated DC-DC converter
« Reply #7 on: January 19, 2019, 02:14:59 am »
From what I can see the NCP1031[1] can handle 8V up to 200V. Now I didn't dig into it to see if you could realistically make it work over such a large range, so there's that...

I somehow wasn't aware of this part.  I'm going to dig and see if the 200V is usable or if it has to be derated.   I've found in many switchers the internal switch is rated at a certain voltage, but due to the topology for say a 100V regulator you actually need a 200V or more switch to be able to handle all of the transient voltages involved. 

 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11332
  • Country: us
  • $
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Very wide range isolated DC-DC converter
« Reply #9 on: January 19, 2019, 04:55:43 am »
Although this is a perfectly reasonable solution, Unfortunately this particular design happens to be used in off-grid/solar sites on a fairly regular basis where every watt counts.   I am going to investigate this as an option though, since doing something similar might make some sense...  i.e. "pre-regulate" at the highest voltages, even if it's a simple 2:1 DC-DC converter.

Holy hell, how many of these things are going to be deployed?  Ten thousand on a site?  You'd think they could muster some cleaner power if there's that substantial of an installation going in!

If this is one-offs, bwahahahah, no, you literally waste more in as many breaths talking about it as saving that one piddly watt. :)

I don't care if they think everything has to save a flea-fart, reality is their problem, not mine!

Or to put it in pure economic terms: you can spend another $10k tuning out another 10 percentage points of efficiency, or you can save another buck (or whatever) on the sale price, and potentially sell more to more customers as a result.

This is very much in the class of problems where it's more economical to spend that budget difference on marketing, rather than engineering.  (And I say this as an engineer! :o )

Don't get me wrong, I'm all for efficiency.  Everything must be done within reason, and demanding that already such a heroic device must be perfect, is outside of that.

:)


Quote
I looked at several controller designs, I think the problem was that I was looking for isolated designs which turn into a rather large circuit, and are fairly few between.  I'm going to go back and look some more.

Another option is two FWBs, one for each input with respect to ground, so each one delivers max 60V (or whatever).  Independent converters or isolators then drive a common rail, or, something (maybe some of that changeover logic, signaled by optoisolators?).

A basic logic function here would simply be using two independent error amps, in parallel.  Normally this would be undesirable: you're effectively wiring two regulated power supplies in parallel, which inevitably will not share current.  But you don't need to share here, you just need one or the other (or both, if you get lucky) to supply enough power.  So, the setpoints could be staggered, so when the output voltage is low, both converters run at full throttle; at the low setpoint, the first converter throttles down, then at the high setpoint, the second one throttles down as well.  If the higher-set supply is unavailable, the output voltage falls slightly, but still well within nominal range.

Compensation might be awkward, because a transient change in the output voltage (period) will tend to activate both error amps; this may not be a bad thing, because recovery will be faster in that case.  Anyway, the transconductance can only double (with two equal error amps acting in parallel), which is easily manageable in a control loop design.  Transient response would be a little weak when just one supply is operating, but I think it's completely manageable.

Anyway, the split supply design allows half input voltage, which is helpful.

This doesn't help necessarily with a floating-ground situation, since the effective (DC) input resistance of a converter is negative.  It will latch to one side or the other.  But the converter sitting at ~0V will run out of power and turn off, stop carrying current, go open circuit, then charge up and go again, oscillating. 

Probably it would be fine to set the "high side" controller to sense the input voltage midpoint, so it can deliver power (even if not commanded to by the voltage error amp?) and stabilize the input voltages.  The increased output current would act to throttle down the one active channel (which would then have to be the "low side" controller), so they naturally share load, it's not a runaway situation.

Yeah, in fact the high side doesn't need an output voltage error amp at all, it just follows along with what it sees at the input.  That'd be fine.

Yeah, that's not a bad method. :)  A bit complicated to describe, easier to draw out in block-diagram form, and possible to prove good behavior in block-diagram form as well.

Or it can be combined with the linear pre-regulator method, or instead of a load balancing control, just put in a shunt regulator there, effectively to bypass the one inactive converter while the other draws load current through both supplies.  That would work too.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline forrestcTopic starter

  • Supporter
  • ****
  • Posts: 720
  • Country: us
Re: Very wide range isolated DC-DC converter
« Reply #10 on: January 19, 2019, 07:49:49 am »
Holy hell, how many of these things are going to be deployed?  Ten thousand on a site?  You'd think they could muster some cleaner power if there's that substantial of an installation going in!

This is much more of a concern for smaller sites...   At a site where the entire power consumption at the site can be as low as 20W, an additional watt (or two) can be a big thing since you've just increased your power needs by 5% or 10%.

At a solar site in northern climates with 1 hour of insolation (not all that uncommon), a single watt 24x7 increases the solar panel needs by 24W.   Plus most off-grid sites like this like to engineer for 2 weeks of batteries in the event of no sun or a solar system failure.   So that single watt increases your battery needs by 336 watt hours.   This doesn't sound like a lot but in these small systems it can add a couple hundred dollars to the cost of the site, and when there are dozens or hundreds of sites where these are deployed it can add significant cost.

The 2-3 battery rails comes from the fact is that it is common for these sites to end up with 2-3 separate systems:   A separate +24V system and a -48V system in combination is really common for historical reasons (gear that came out of telecommunications often runs at -48VDC, gear that came from data communications/industry/telemetry/scada often runs at +24V (or +48V).   Some people will build one battery system, and use DC-DC converters to get the other rails, but others tend to find that building two separate systems works better for them.     

One of our distinguishing factors is the lower power consumption of our solution.   Many of my competitors have chosen to go down the 'it's only a few watts' path, and as a result, I'm picking up a lot of sales from those people where every watt really does matter.   I do a lot of work to be able to put "under 1 watt average power consumption" on the datasheet, because that makes it easier for the customer to feel like they can add our solution to an already tight power budget.



« Last Edit: January 19, 2019, 08:37:37 am by forrestc »
 

Online mk_

  • Regular Contributor
  • *
  • Posts: 243
  • Country: at
Re: Very wide range isolated DC-DC converter
« Reply #11 on: January 19, 2019, 08:06:58 am »
200v is no problem for lt

https://www.analog.com/media/en/technical-documentation/data-sheets/7860f.pdf

abs. max. Ratings say 60V. Tose 200V are only possible with some tricky GND-shifting which is nothing I would use for a (isolated) DCDC-Converter.
 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11332
  • Country: us
  • $
 

Offline jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: Very wide range isolated DC-DC converter
« Reply #13 on: January 19, 2019, 08:24:10 pm »
I hate to be a party pooper, but this looks like an ill-defined and challenging space.

If people want minimum power consumption (ie max efficiency) and (presumably) moderate cost then you should simply make a small number of simpler products to fit the space, rather than trying to build One Converter to Rule Them All.

If you want to persist with the high efficiency + super wide range approach, then the best result might be found using a non-traditional converter. For example:
- polarity can be inverted from -48V to + 48V using a switched capacitor converter at fixed duty cycle and frequency. Sure the output won’t be regulated, but the subsequent buck can fix that.
- multilevel converters can offer a more reasonable way to work with high conversion ratios and still have reasonable duty cycles (eg a Flying Capacitor converter)
 

Offline coppercone2

  • Super Contributor
  • ***
  • Posts: 11332
  • Country: us
  • $
Re: Very wide range isolated DC-DC converter
« Reply #14 on: January 19, 2019, 10:04:43 pm »
this spec has military written all over it IMO
 

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 4004
  • Country: nl
Re: Very wide range isolated DC-DC converter
« Reply #15 on: January 19, 2019, 11:25:22 pm »
Have you considered a switching pre-regulator?

With that you can get your wide input voltage down to a somwhat stable voltage, and then use a standard  isolated DC-DC converter to get to the 3V3 output.

I might have missed it, but how much current do you need at the 3V3 output?

Or else go tell your customor that if they want a non standard power supply with a very wide input range and other strict demands (efficiency, size) that it is not going to be cheap.
 

Offline forrestcTopic starter

  • Supporter
  • ****
  • Posts: 720
  • Country: us
Re: Very wide range isolated DC-DC converter
« Reply #16 on: January 20, 2019, 01:38:09 am »
I hate to be a party pooper, but this looks like an ill-defined and challenging space.

Oh you're not telling me anything I don't know.  This is what an industry which has grown up over 20+ years and no power standards looks like.  It used to be much worse, with the end equipment needing all sorts of random voltages (one would need no more than 12V, one would need no less than 12V, another would need 18V nominal).   Fortunately it's consolidated around 12, 24 and 48V lead acid battery voltage ranges which make things somewhat easier.

If you want to persist with the high efficiency + super wide range approach, then the best result might be found using a non-traditional converter. For example:
- polarity can be inverted from -48V to + 48V using a switched capacitor converter at fixed duty cycle and frequency. Sure the output won’t be regulated, but the subsequent buck can fix that.
- multilevel converters can offer a more reasonable way to work with high conversion ratios and still have reasonable duty cycles (eg a Flying Capacitor converter)

This is largely what I do today (see schematic 1 I posted).   I have a +9-48VDC range DC-DC converter design I've been using for years.   When we needed to add the -48VDC I added a neg to positive inverter using a linear technologies part.   It happens to output +12V but there's no reason it couldn't be pretty much any other voltage (or unregulated).

I'd love to find a simpler/lower cost design for the negative to positive supply, which should be doable since the negative rail is much tighter range than the positive side - it really only needs to support something like -36 to -60V, which is the normal range for a lead acid battery array.

 

Offline Doctorandus_P

  • Super Contributor
  • ***
  • Posts: 4004
  • Country: nl
Re: Very wide range isolated DC-DC converter
« Reply #17 on: January 20, 2019, 01:46:37 am »
Have u thought about using a single diode instead of a full bridge?

Why make 120V out of it if the max input is + / - 60V?

What kind of frequencies do your input voltages have?
A viable method is to put a fet on the input and simply turn the fet off if your buffer elco is charged to a certain voltage.
This can greatly reduce the input voltage range required, and as noted earlier:

Isolated DC-DC converters with an 1:4 input range are readily available off the shelf products.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf