DC-DC Modules: What am I missing here?!?

[IDEngineer]

We're working on a project that requires part of its circuitry to be galvanically isolated. Easy answer: A DC-DC isolation module as made by Murata, Recom, TI, CUI, and many others. They come in various flavors of input voltages (3.3, 5, 12, 15, and 24 are common) and output voltages (same choices) and you can often have mismatched input and output voltages so you can obtain different supply rails than your existing power system. All well and good.

...except for two things. First, these devices are completely unregulated. I'm not talking about "they're unregulated until you pull some minimum current", I'm talking COMPLETELY unregulated. The curves in the spec sheets show a basically linear relationship between output current and output voltage - the very definition of an unregulated source. OK, that sucks, but at least if you know that going in you can add a downstream voltage regulator.

...except for the second thing: Those output voltage options I mentioned above! If you have an unregulated supply, you're likely to use a voltage regulator. Voltage regulators need some overhead to operate. Even LDO regulators need almost a volt, and traditional regulators need 1.5-2V to be safe. But instead of rating these DC-DC modules with that in mind (say, 7V output voltage so you can safely run a 5V regulator), their nominal output voltages are right at the traditional rails of 3.3, 5, 12, etc. Leaving no headroom to run a regulator.

If they wanted to produce a supply device at traditional voltages, they could have included actual regulation. But they didn't. Conversely, if they wanted to keep them simple and not have regulation, fine - they could have increased the nominal output voltages to make it easier to use a downstream regulator. But they didn't. Instead, they took the worst of both worlds - the least optimal of every option - and built that.

Getting back to the title: I must be the one missing something here, because virtually all of the DC-DC module manufacturers build them this way. So what is the typical way to use these? I've had precious few projects where the supply rails could just vary unpredictably all over the place with no ill effects. At the moment our plan is to go to the next highest secondary voltage (12V) and use a downstream 5V regulator, but that dissipates 7V across the regulator and dramatically reduces the output current rating for a given DC-DC module package size and cost. And it could all have been so easily avoided.

What am I missing here?

[GreggD]

I have used many of these. First off they require at least a 10% minimum load. This keeps them closer to there specified output voltage but they can malfunction without that load. There output, say 5V is close enough to drive RS485 ics "75176".

[0xdeadbeef]

There are actually also types with regulation. Still, most need 10% load and might create twice the specified voltage or so when no valid load is connected.
When I used them and wanted to avoid this, I added some shunt regulator to draw a bit of current until the voltage was somewhat in spec:

[IDEngineer]

I have used many of these. First off they require at least a 10% minimum load. This keeps them closer to there specified output voltage but they can malfunction without that load.

"Closer to their specified output voltage" is exactly the problem. 

There output, say 5V is close enough to drive RS485 ics "75176".

OK, yes, that application probably doesn't care about voltage regulation. But most projects sort of rely on stable supply rails. I guess these modules were not intended for such applications, though I'd think they'd be intentionally designing themselves OUT of the majority of possible design wins. Weird.

[IDEngineer]

I added some shunt regulator to draw a bit of current until the voltage was somewhat in spec

"somewhat in spec" 

Again, I have no problem with them being unregulated - IF they'd chosen voltages that left headroom for a separate regulator. It's like they suboptimized them on purpose.

[lem_ix]

Depending on your requirements this might be relevant. Allows for more flexibility on your design.

http://www.ti.com/product/sn6501-q1

Probs similar solutions from other manufacturers.

[H.O]

I understand your point but it's not like there aren't any regulated DC/DC converter modules available if that's what you need. I don't think you mentioned what sort of power you're looking for but Traco have various models from 1W and up. Murata MEF1 seems to be another option in the 1W range and their NCS3 is an isolated, regulated 3W job and CUI Inc has a whole bunch of them as well.
 

[IDEngineer]

I understand your point but it's not like there aren't any regulated DC/DC converter modules available if that's what you need. I don't think you mentioned what sort of power you're looking for but Traco have various models from 1W and up. Murata MEF1 seems to be another option in the 1W range and their NCS3 is an isolated, regulated 3W job and CUI Inc has a whole bunch of them as well.

Total current requirement is around 10mA at 5V (e.g. 50mW). We've been looking at the 1W units because they're plentiful, not because we need the power.

Yes, there are internally regulated units but we are trying to stick with what passes for an industry standard here in terms of packages and pinouts, so we have multiple sources. When you get into the internally regulated units things start varying a bit more.

And even then... since regulated units ARE available, why have the UNregulated units work at the target output voltages instead of leaving some headroom? If you already don't care about your voltage going all over the place, why not center that variation a couple of volts higher so your modules can be used in a huge number of additional situations?

[chris_leyson]

I agree regulation isn't too good with these small DC-DC converters. However, there is another issue to watch out for and that's input voltage risetime, if your input supply comes up too slowly then some of them won't startup and just sit there getting hot until they cook themselves. It's in the small print in some of the data sheets. Input supply needs a fairly fast risetime for them to startup properly.

[0xdeadbeef]

Total current requirement is around 10mA at 5V (e.g. 50mW). We've been looking at the 1W units because they're plentiful, not because we need the power.
Yes, there are internally regulated units but we are trying to stick with what passes for an industry standard here in terms of packages and pinouts, so we have multiple sources. When you get into the internally regulated units things start varying a bit more.
And even then... since regulated units ARE available, why have the UNregulated units work at the target output voltages instead of leaving some headroom? If you already don't care about your voltage going all over the place, why not center that variation a couple of volts higher so your modules can be used in a huge number of additional situations?

Well, it doesn't really help you to complain. I was in the same boat, but I don't see what's so terrible about the shunt regulator approach I suggested.
Don't get upset too much about that "somewhat in spec" phrase of mine. For the actual use case, I wanted to get 5V on the isolated side of a 5V to 5V isolated DC/DC.
The DCDC converters I used had a +-5% output voltage accuracy - which was OK for me. It was just that this accuracy was only guaranteed at >=10% load.
Since I didn't want to waste energy by adding a dummy load resistor, I came up with that shunt regulator thing which just drains current if needed and also doesn't limit the output current.
The shunt regulator has a better accuracy than the +-5% of the DC/DC but I didn't need it, didn't waste to waste energy and therefore set it to just 5.25V instead of 5.1V or whatever.
This is what I meant with "somewhat in spec" (for the USB voltage spec I wanted to meet). Actually with the shunt regulator, the DC/DC operates fully in its +/-5% spec.

[IDEngineer]

The DCDC converters I used had a +-5% output voltage accuracy - which was OK for me. It was just that this accuracy was only guaranteed at >=10% load. Since I didn't want to waste energy by adding a dummy load resistor, I came up with that shunt regulator thing which just drains current if needed and also doesn't limit the output current.

Yes, I agree your shunt circuit is elegant and clever. What frustrates me is that is should be unnecessary!

With regards to +/-5% accuracy above 10% load... I don't know which manufacturer and model you were using, but every spec sheet I've reviewed has a straight line inversely relating current to voltage. That means there is no voltage regulation at all, unless you HOLD your current consumption at a constant level. Your shunt circuit may inadvertently have a higher duty cycle than you intended!    If their spec sheets can be believed, that is. Perhaps a more accurate expression would have been "output voltage repeatability" (above 10% load), since based on those curves they don't seem to provide any sort of actual regulation at all.

You're right, complaining doesn't improve things. But it's frustrating when their design choices don't seem to make sense, and in this case actually get in the way of using their parts. That's why I presumed up front I must be missing something, that there was some good reason they emit unregulated voltages so close to standard target supply rails.

[Someone]

The DCDC converters I used had a +-5% output voltage accuracy - which was OK for me. It was just that this accuracy was only guaranteed at >=10% load. Since I didn't want to waste energy by adding a dummy load resistor, I came up with that shunt regulator thing which just drains current if needed and also doesn't limit the output current.

Yes, I agree your shunt circuit is elegant and clever. What frustrates me is that is should be unnecessary!

There are many many designs that don't need regulation on the power rails, these parts meet that market even if it doesn't match your requirements. If a reference voltage is needed then add a reference.

Or as suggested above there are also well regulated isolated DC-DC supplies available from multiple sources in common through hole SIP "modules".

[IDEngineer]

They are designed to run off regulated input bus.

Yes, but their output voltage varies with output CURRENT (read: load), no matter how well regulated the input power may be. This is straight from their spec sheets.

[HackedFridgeMagnet]

They are designed to run off regulated input bus.

Yes, but their output voltage varies with output CURRENT (read: load), no matter how well regulated the input power may be. This is straight from their spec sheets.

What am I missing here?

At least a couple of links to part numbers or spec sheets.

[David Hess]

They are intended for applications where the load regulation is not critical.  This includes most analog circuits and many higher voltage digital circuits and includes applications where a floating bias supply is needed.

There are ways to regulate the secondary side output voltage from only the primary side and compensate for resistive and rectifier losses but I rarely see this done.  It is too bad they do not give access to the primary side signals to do this but it would mean extra pins that most users would ignore.

Coilcraft and others make small high frequency transformers intended for isolated power supply applications so you could roll your own.  Gate drive transformers can also work for this.

Nobody runs precision analog or RF magic with those noisy, unregulated module without cleaning the noise up first.

Often the only thing which is necessary is an RLC filter to reduce residual noise and this might be needed to isolate one circuit from another anyway.  Clever circuits or at least not the wrong circuits can have considerable power supply rejection.

A soft switched inverter using one of the transformers I mentioned can have low output noise and this fits well with primary side regulation.

[IDEngineer]

They are intended for applications where the load regulation is not critical.  This includes most analog circuits...

Unfortunately, analog circuits that involve A/D's or D/A's - including MCU's that incorporate such devices - can be sensitive to (a lack of) regulation. 

Coilcraft and others make small high frequency transformers intended for isolated power supply applications so you could roll your own.

We've checked into that, but it's hard to beat the price, second sourcing, and footprint of those little modules. So our present plan is to move to a 5V in, 12V out isolated module and use a downstream 5V regulator. We don't need much current so the reduction in a 1W package from 200mA out to 83mA out isn't a problem, and low current means low dissipation from the regulator despite the 7V drop across it.

In our specific application, we can get away with going to the next highest secondary voltage. But it's still a shame that doing so is necessary. If, as everyone says, these things are intended for unregulated applications where the absolute voltage doesn't matter all that much then it wouldn't hurt to have the output voltage be (say) 8V instead of 5V... which would allow existing insensitive applications to work just fine AND allow downstream regulation if needed without such a huge loss of output current.

[tszaboo]

Look, the standard ones are good enough to run a RS485 or CAN transciever off it, even a micro. Just place a Zener on the output, and it will limit the voltage, and provide the 10% ish load.
If you need regulation, you can still use a ie. 12S09 instead the 12S05 and place a simple 7805 afterwards. You can literally run precision analog components on these, if you clean up the power supply, and they cost almost nothing. There was this one-off circuit, where the input voltage was 24V, and I've just used one of these to drop it down to 5V, because I couldn't be bothered to design an SMPS to do it, and linear would have been silly.

[mariush]

If the power requirements are low, maybe you could just drop in a tiny 5v LDO ... there's plenty out there which have a voltage drop as low as 30mV to 100mV ... so you have 5..8v in, 4.95v out or something like that.

ex
lp3985 ( 6v in,  100mV drop at 150ma out )  : http://uk.farnell.com/texas-instruments/lp3985im5-5-0-nopb/ldo-reg-0-2vdo-0-15a-5v-5sot23/dp/2436533
lp2980 , lp2981, lp2985  (up to 16v in, 120 to 300 mV at 50 .. 150mA depending on ic )
mcp1711 (6v in 150ma)

etc
 

[Damianos]

I have a cargo-container full of magnetic material. I want to transport it with my scooter but the manufacturer has not available any accessory for that purpose! The suggestions are to use a truck, a ship, a train... Why they don't help me to use my scooter that is cheap, economic in fuel consumption and environmentally friendly?

Sorry if the above is seeming offensive, but:

They are intended for applications where the load regulation is not critical.  This includes most analog circuits...

Unfortunately, analog circuits that involve A/D's or D/A's - including MCU's that incorporate such devices - can be sensitive to (a lack of) regulation. 

Coilcraft and others make small high frequency transformers intended for isolated power supply applications so you could roll your own.

We've checked into that, but it's hard to beat the price, second sourcing, and footprint of those little modules. So our present plan is to move to a 5V in, 12V out isolated module and use a downstream 5V regulator. We don't need much current so the reduction in a 1W package from 200mA out to 83mA out isn't a problem, and low current means low dissipation from the regulator despite the 7V drop across it.

In our specific application, we can get away with going to the next highest secondary voltage. But it's still a shame that doing so is necessary. If, as everyone says, these things are intended for unregulated applications where the absolute voltage doesn't matter all that much then it wouldn't hurt to have the output voltage be (say) 8V instead of 5V... which would allow existing insensitive applications to work just fine AND allow downstream regulation if needed without such a huge loss of output current.

The question is: why they don't change the specifications of million devices, to do my work, even if I have selected an improper device?
The next step will be something like: why they don't make the desk-lamp's power supplies stable enough and noiseless, so as to use them as voltage references in 40-bit ADCs and to have the last digit stable?

To answer the question of the title and the initial post: you have to realize that these devices are meeting specifications that were established years ago. Is it the time to change them?
If you insist to use them in your application, I think that you have to follow one of the proposals that are given to you. Otherwise you have to select a proper device.

[IDEngineer]

The question is: why they don't change the specifications of million devices...

My warmest thanks for your helpful and supportive comment.

Nowhere did I suggest they should change the specs. I was simply wondering why they didn't spec them to be more flexible from the start. And my last post before yours specifically outlined our present plan to accommodate their existing specs.

Engineering is an art of compromise. Everything is a tradeoff. My most recent post described how we intend to balance the specs to make these parts work, with both their advantages and disadvantages. How your post then improved things for this community isn't clear to me, but I'll go back and read it again in case I missed some useful nugget of technical insight.

Thanks again.

[David Hess]

They are intended for applications where the load regulation is not critical.  This includes most analog circuits...

Unfortunately, analog circuits that involve A/D's or D/A's - including MCU's that incorporate such devices - can be sensitive to (a lack of) regulation. 

ADCs and DACs have power supply rejection as well.  Except in ratiometric application, using the power supply as the reference is asking for trouble.

One thing I am always surprised about is that designs rarely use a simple operational amplifier, discrete transistor, and voltage divider to make precision remote sense references or regulators as needed.

If, as everyone says, these things are intended for unregulated applications where the absolute voltage doesn't matter all that much then it wouldn't hurt to have the output voltage be (say) 8V instead of 5V... which would allow existing insensitive applications to work just fine AND allow downstream regulation if needed without such a huge loss of output current.

I noticed the same thing looking through Coilcraft's selection of suitable transformers however there are a lot of analog parts now which cannot tolerate voltages higher than the absolute maximum of their parent digital process; they work fine at lower voltages but not higher ones.

Regulators are so inexpensive now unlike in the past that secondary side regulation is probably the way to go but that conflicts with the above so they cannot use the same floating output inverters.

[IDEngineer]

I noticed the same thing looking through Coilcraft's selection of suitable transformers however there are a lot of analog parts now which cannot tolerate voltages higher than the absolute maximum of their parent digital process; they work fine at lower voltages but not higher ones.

Agreed, so we have a conundrum. Absolute output voltage and stability don't matter - except when they do.    With modern analog parts some form of regulation is required, since based on the spec sheet curves the output can vary from -6% to +10% based on load.

And lest we forget: "Lower than 10% minimum loading will result in an increase in output voltage, which may rise to typically double the specified output voltage if the output load falls to less than 5%." While other respondents herein have posted helpful shunt circuits to overcome this, handling such variation is quite literally why voltage regulators exist.

Hence our plan to have a downstream regulator... cheap, low parts count, and excellent regulation performance. Of course, almost anything is better than -6% to +100%!

[IDEngineer]

One thing I am always surprised about is that designs rarely use a simple operational amplifier, discrete transistor, and voltage divider to make precision remote sense references or regulators as needed.

I concur but will offer one reason I've seen: Because no spare opamp section was available, and there was no real estate to add another package and its attendent discretes.

BTW, I resolved this in one design by using an internal bandgap reference on the MCU. Measuring that known voltage with the A/D yielded a value whose difference from the predicted value gave a correction factor. That correction factor could then be applied to other measurements to improve accuracy. It worked amazingly well. I maintained a filtered average of the bandgap conversion value to further remove noise and normal regulation variability, and did the correction in floating point with proper rounding (had lots of spare CPU cycles available). On paper it was troubling not to have a formal hardware reference, but given the constraints of the design it worked quite nicely.

[Sceptre]

I see a half-dozen 6V 1W isolated modules in stock at both Mouser and Digikey (for example, 1470-2978-5-ND, 580-MTE1S0506MC); if you can't find an LDO that works with that, there are plenty of 9V modules as well (1470-1399-5-ND, 490-PDS1-S5-S9-M).  There's also a 7V module from Bel (DSP1N5S7), but it's pricey.  So, unless I'm missing something, I'd say your issue is sub-optimal search functionality.

[IDEngineer]

I see a half-dozen 6V 1W isolated modules in stock at both Mouser and Digikey (for example, 1470-2978-5-ND, 580-MTE1S0506MC); if you can't find an LDO that works with that, there are plenty of 9V modules as well (1470-1399-5-ND, 490-PDS1-S5-S9-M).  There's also a 7V module from Bel (DSP1N5S7), but it's pricey.

As I said, it's a compromise. All of the parts you mentioned are more expensive than the ones we've been talking about, and don't have as many second sources for the same specs. And as I and others here have noted, regulators are cheap. Hence our plan to go up in secondary voltage and then regulate down.

You can usually find an expensive solution to a one-off problem. But for volume production, the trick is to find the optimal solution that balances cost, multiple sources, PCB real estate, etc. It's not just about whether a part can be found... it's about finding part(s) that work in volume.

Thanks!