Transformer isolation of SMPS is not just for safety?

[zenerbjt]

We are doing a power supply which runs off a 14S LiPO (42-59V).
Vout is 24V and 130W.
This is under the SELV rating so i assume that we are not required to have transformer isolation?

However, is it true that one other reason for isolation is so as to prevent damage to the load in the case of a shorted FET?. For example, if a primary side transistor in a transformer isolated SMPS fails short, then the Vin doesnt get impressed across the load....whereas eg if a high side buck fet fails short, then vin goes across the load, potentially destroying it. So is this the "other" reason for isolation?.....ie other than safety?

[ahbushnell]

We are doing a power supply which runs off a 14S LiPO (42-59V).
Vout is 24V and 130W.
This is under the SELV rating so i assume that we are not required to have transformer isolation?

However, is it true that one other reason for isolation is so as to prevent damage to the load in the case of a shorted FET?. For example, if a primary side transistor in a transformer isolated SMPS fails short, then the Vin doesnt get impressed across the load....whereas eg if a high side buck fet fails short, then vin goes across the load, potentially destroying it. So is this the "other" reason for isolation?.....ie other than safety?

That's why you do fault analysis.  There are many different cases for using a transformer or not.  If three is a large difference in voltage from input to output then a transformer would have an advantage.  And if a shorted transistor is a problem then you might want to use a transformer. 

[tom66]

There are ways to prevent a FET failure from causing catastrophic over-voltage.

For instance, a beefy TVS on the output or a separate isolation/overvoltage protection FET.  This would not be expensive to add, and almost certainly cheaper than a 130W capable transformer.

[Siwastaja]

Design properly so that the FET doesn't short.

If it shorts, then accept the whole unit dies, and just replace the whole unit, and fix the reason for the fault.

Downtime is most problematic anyway. Having a "product you can fix by just replacing a few components on a PCB" is most often false economy. You replace the whole unit anyway.

The only expection to this would be if the thing after the MOSFET would be some very special technology, BOM worth of many thousands, or poor availability. Then you would add extra protection.

Otherwise, it doesn't make a difference whether a design fault burns part of the circuit or a larger part of the circuit. I wouldn't add any extra complexity, especially not an isolation transformer, just to limit the amount of damage in a situation that shouldn't happen to begin with.

On the other hand, do protect against external sources of errors, like overvoltages in the input lines.

[chris_11]

There are sometimes reasons to break the ground loop i.e. in large 24V industrial control PLC systems. If that is not the case then a simple non isolated hard switched buck will give you 95%+ efficiency in a smaller volume than any transformer solution.

br
Christian

[vk6zgo]

Design properly so that the FET doesn't short.

If it shorts, then accept the whole unit dies, and just replace the whole unit, and fix the reason for the fault.

Downtime is most problematic anyway. Having a "product you can fix by just replacing a few components on a PCB" is most often false economy. You replace the whole unit anyway.

Try this with an installation which is thousands of km away from your base depot, in a country tens of thousands of km away from the country of manufacture.
In many cases, stuff trundles on happily for years, spare units were either not obtained, or were all used in earlier years, but when this device dies, it needs immediate repair, as it means loss of income to your customer.

The only expection to this would be if the thing after the MOSFET would be some very special technology, BOM worth of many thousands, or poor availability. Then you would add extra protection.

It's not the BOM, it's the complete device it is part of, which may be rendered unusable.
"Many thousands of dollars" are more the rule than the exception with Broadcast equipment.

Otherwise, it doesn't make a difference whether a design fault burns part of the circuit or a larger part of the circuit. I wouldn't add any extra complexity, especially not an isolation transformer, just to limit the amount of damage in a situation that shouldn't happen to begin with.

On the other hand, do protect against external sources of errors, like overvoltages in the input lines.

The amount of the circuitry which burns is of vital interest to those who have to repair the equipment, under strict time constraints.

"Board swapping" is a long term Management dream, but is often not possible, nor economic.

[Siwastaja]

Try this with an installation which is thousands of km away from your base depot, in a country tens of thousands of km away from the country of manufacture.
In many cases, stuff trundles on happily for years, spare units were either not obtained, or were all used in earlier years, but when this device dies, it needs immediate repair, as it means loss of income to your customer.

And you think someone there is magically capable of troubleshooting the issue to a blown MOSFET, get a spare MOSFET, replace it, and get it back running? FYI, the MOSFET gate driver (often the DC/DC controller) is very likely blown as well, and because they blow short, number of upstream components can be toast, too.

Finally, using a transformer does not even guarantee safety to the downstream parts as suggested by treez. A crowbar + fuse could be more appropriate.

No, in such cases either complete units, or replacement modules, are used.

In any case, this doesn't make a lot of sense. If the designer fucks up such basic design so it blows MOSFETs randomly, what makes you think it always and only blows that one specific MOSFET which causes no further damage thanks to the transformer isolation? The problem could be after the secondary side of the transformer; actually it's likely there, assuming most of the design work and largest part of the circuit is on that side.

Do you think that adding major complications like transformers make the design more robust? If the designer struggles building a non-isolated buck which does not blow the MOSFETs, how do you think they are able to design a transformer-isolated version of the circuit, which adds the design complexity by an order of magnitude? For example, depending on the topology of course, transformer brings leakage inductances and much higher parasitic ringing energy and voltages into play, increasing the stress to the transistors. So now you are designing snubbers you didn't need with the simple buck.

Adding complexity to add robustness is seldom an answer. Simulation, and prototype verification by measurements is. Adding true redundancy is. Specifying maintenance procedures beforehand, securing the availability of spare modules/units is.

I'm really assuming here we are talking about an integrated product the power supply is part of, where the complete control is on OP's company's shoulders. As hinted earlier, if this is a power supply designed to power expensive or specialized external loads, by all means add extra layers of safety to prevent ever exceeding output voltage spec, even under failure conditions. And for this, again, I suggest a crowbar circuit instead of adding isolation and thinking it solves the problem. After all, shorted high-side MOSFET in a non-isolated buck isn't the only case which can cause overvoltage on the output. Failure of voltage reference or either of the feedback resistors are obvious possibilities as well. Separate output voltage monitor / crowbar protects from all such possibilities.

[vk6zgo]

Try this with an installation which is thousands of km away from your base depot, in a country tens of thousands of km away from the country of manufacture.
In many cases, stuff trundles on happily for years, spare units were either not obtained, or were all used in earlier years, but when this device dies, it needs immediate repair, as it means loss of income to your customer.

And you think someone there is magically capable of troubleshooting the issue to a blown MOSFET, get a spare MOSFET, replace it, and get it back running? FYI, the MOSFET gate driver (often the DC/DC controller) is very likely blown as well, and because they blow short, number of upstream components can be toast, too.

Technicians routinely do stuff like that------ that is what they are employed for.

Finally, using a transformer does not even guarantee safety to the downstream parts as suggested by treez. A crowbar + fuse could be more appropriate.

No, in such cases either complete units, or replacement modules, are used.

Thirty plus years in Electronics repair says otherwise.

[Gyro]

59V isn't SELV.

[Siwastaja]

Sorry vk6zgo but you are completely off-topic and failed to look at the actual question. It is if changing to an isolated topology makes the solution more robust against problems like output overvolting. The answer is, no, I don't think so. (tom66 got it right from the start.)

I'm sure your 30 year experience in electronics repair shows that transformers can fail, and transistors fail because of unsnubbed inductances. Or maybe you just replace parts. I prefer to go for root cause analysis if at all possible, to eliminate the problem forever. But this needs to be done by the designer. So ship a replacement, and ship the failed unit back to manufacturer. It's slow and expensive first, but ultimately leads to ultimate products. It's OK if you think it's idealistic and not realistic.

But there is the classical "repairability bias" like I call it; products that fail end up on your desk, and you are happy whenever they are "repairable", while products that never fail never end up on your desk!

I do repair things for fun, and for learning from the mistakes of others, and don't remember ever seeing dead MOSFET in a simple buck converter. OTOH, dead MOSFETs in power path switches! These things exceed SOA due to inrush all the time. Dead MOSFETs in larger transformer-based converters. Dead inductors and transformers from too much heat and windings shorting together. (Won't even mention electrolytic capacitors.) The giveaway is, products that fail tend to have 1-2 typical weak spots, and once they are fixed in design, these products just stop failing.

I'd suggest simplicity and familiarity is the key, then adding, if necessary, independent safety layers (or independent redundancy, if that is more applicable), instead of changing to more complex, more error-prone basic solution expecting it to be fundamentally more reliable without much proof of such claim.

[Siwastaja]

59V isn't SELV.

Depending on where you live, the limit and the terminology vary, so it may be 75V DC (as in EU AFAIK), or 120V DC, or something else, but chances are high that 59V DC is SELV. By IEC standard as shown in Wikipedia (https://en.wikipedia.org/wiki/Extra-low_voltage), it's <120V DC.

[Gyro]

Ah, you're right - I was remembering the AC figure (SELV traffic controllers). 


P.S. Irrc, SELV requires isolated supply, so battery is fine, with tolerance of a single fault scenario.

[David Hess]

A transformer isolated switching topology is one way to protect against switching transistor shorts but there are at least two other ways:

1. The SEPIC topology isolates DC using a series capacitor and may be used specifically to protect against switching transistor shorts.  It has the disadvantage of requiring the full ripple current to pass through the series capacitor so is not suited for high current applications.

2. An SCR crowbar and fuse can protect against a switching transistor short.

[Siwastaja]

2. An SCR crowbar and fuse can protect against a switching transistor short.

Highly recommended because it protects from all other possible sources of failures causing overvoltage, including control IC, voltage reference or feedback divider failure.

And, I think fuse + SCR crowbar ends up simpler and less intrusive than changing to a more complex switching topology.

[zenerbjt]

Thanks, what if you are just using Vicor DCDC modules?.....would you then say its best to pick isolated vs non-isolated? (from the point of view of load damage due to fet failure).

The non -isolated are  often much cheaper.

[tom66]

I would say the chances of a well designed Vicor module failing high output is quite low.  If you are really genuinely concerned, put an overvoltage protection shunt on the output and use a non-isolated part. An isolated part can still fail high voltage if for instance the feedback optocoupler goes bad (if it uses opto feedback.)

The isolation barrier also reduces efficiency and arguably this additional heat will contribute to a shorter lifespan (across all units) of the product.  Having an isolated ground can also create interesting EMC headaches for certain applications.

So in short, don't just use an isolated part due to concerns about high FET failure.  Consider the whole system and whether it needs isolation, and if it needs OVP, then implement that without isolation if required.

[pwrtrnx]

A fuse and some suitably rated zeners will save the load if the series pass element goes short  ...  else buy some "isolated"  types,  e.g. Sepic or transformer based ...

[f4eru]

Try this with an installation which is thousands of km away from your base depot, in a country tens of thousands of km away from the country of manufacture.

Yes, one reason more to make sure the design is sound so the FET doesn't burn.

[Electrodynamic]

zenerbjt

We are doing a power supply which runs off a 14S LiPO (42-59V).
Vout is 24V and 130W.
This is under the SELV rating so i assume that we are not required to have transformer isolation?

However, is it true that one other reason for isolation is so as to prevent damage to the load in the case of a shorted FET?. For example, if a primary side transistor in a transformer isolated SMPS fails short, then the Vin doesnt get impressed across the load....whereas eg if a high side buck fet fails short, then vin goes across the load, potentially destroying it. So is this the "other" reason for isolation?.....ie other than safety?

I have built countless DC/DC power supplies and unless 100% isolation is required I never use a transformer.
For example in your case a 5$ Arduino nano could monitor input/output voltage through the analog inputs using an opto-isolated voltage divider.  If you want to throw another $3 at it you could also monitor the component temperatures with some 10K thermistors. Then if any variable you choose is outside the preset limits it simply opens a relay shutting it down.

In my opinion it's a lot easier and safer to throw $10 worth of microcomputer monitoring intelligence at it than mess around with other components. In fact I'm not sure why these old unmonitored transformer type power supplies are still legal. 30 years ago we had no choice because we had to use what was available however it's 2020 and we have many options.

Regards

[S. Petrukhin]

For example in your case a 5$ Arduino nano could monitor input/output voltage through the analog inputs using an opto-isolated voltage divider.  If you want to throw another $3 at it you could also monitor the component temperatures with some 10K thermistors. Then if any variable you choose is outside the preset limits it simply opens a relay shutting it down.

In my opinion it's a lot easier and safer to throw $10 worth of microcomputer monitoring intelligence at it than mess around with other components. In fact I'm not sure why these old unmonitored transformer type power supplies are still legal. 30 years ago we had no choice because we had to use what was available however it's 2020 and we have many options.

By placing an MCU or other reasonable device on the high side, you get a potential point of trouble. You need to take care of the possibility of updating the firmware. This is not always convenient and economical.