I've only dealt with the smaller end stuff (10kVA 3ph max) and I can't say I've ever seen a transformer after a UPS. The larger UPS did have a transformer internally - I think primarily because it was a simple double conversion type so used the transformer so it could have its own N-PE link for the output circuit. The smaller units (<3kVA) I suspect just degraded to an IT arrangement during grid fail. I guess the fundamental point is that you don't want to be relying on the grid supply's N-PE link during a power fail situation (partly because it might not be reliable and pose a risk grid workers, partly because you may want to work on/disconnect your own mains-side wiring, including N, during maintenance, and of course (under UK law) you can't have PEN conductors in consumer's installations). More modern units tend to be smarter and switch the output between (somewhat cleaned up) direct mains and inverter output dynamically - such system might well have a switchable N-PE link on the output, that closes when mains fails (and the UPS disconnects itself from the supply), but open when the mains supply is used - certainly some domestic "islanding" battery storage systems can work that way during a power cut to keep power on in the home; so the N-PE link issue can solved without needing a transformer.

On the other side of the coin, I suspect IT systems aren't anything like as vulnerable to earth noise as in the past. When I was young (quite a while ago now) earthed referenced data communication (e.g. RS-232) was common, and local differences in earth potential could and did cause problems. Times change though, and recently all the communication I've see has either been differential (e.g. Ethernet or the later versions of SCSI) or even fibre-optic, so that particular problem has certainly reduced. That's not the whole story of course, but is one consideration.

There did seem to be some different approaches in different parts of the world though - the US especially seems particularly fond of throwing transformers into all sort of situations (possibly due to the problems with voltage drop when using a 120V system, so tend to use higher voltages for distribution and then need transformers to reduce it to 120V) whereas this side of the pond we'd be more likely to start with 230/400V system and stick with that throughout. Likewise some parts of the world allow PEN conductors within installations, which can side-step the N-PE link problem (but can introduce other problems of course). I think one well known UPS manufacturer used to leave the supply N connected the the load during power fail (relying on the supply N-PE link) but even they've fallen in line with international standards and isolate supply N as well as L from the output during a power fail. 

I have been told that the larger systems might do things differently though - maybe some of the "big boys" here can enlighten us.

   - Andy.

I've only dealt with the smaller end stuff (10kVA 3ph max) and I can't say I've ever seen a transformer after a UPS. The larger UPS did have a transformer internally - I think primarily because it was a simple double conversion type so used the transformer so it could have its own N-PE link for the output circuit. The smaller units (<3kVA) I suspect just degraded to an IT arrangement during grid fail. I guess the fundamental point is that you don't want to be relying on the grid supply's N-PE link during a power fail situation (partly because it might not be reliable and pose a risk grid workers, partly because you may want to work on/disconnect your own mains-side wiring, including N, during maintenance, and of course (under UK law) you can't have PEN conductors in consumer's installations). More modern units tend to be smarter and switch the output between (somewhat cleaned up) direct mains and inverter output dynamically - such system might well have a switchable N-PE link on the output, that closes when mains fails (and the UPS disconnects itself from the supply), but open when the mains supply is used - certainly some domestic "islanding" battery storage systems can work that way during a power cut to keep power on in the home; so the N-PE link issue can solved without needing a transformer.

On the other side of the coin, I suspect IT systems aren't anything like as vulnerable to earth noise as in the past. When I was young (quite a while ago now) earthed referenced data communication (e.g. RS-232) was common, and local differences in earth potential could and did cause problems. Times change though, and recently all the communication I've see has either been differential (e.g. Ethernet or the later versions of SCSI) or even fibre-optic, so that particular problem has certainly reduced. That's not the whole story of course, but is one consideration.

There did seem to be some different approaches in different parts of the world though - the US especially seems particularly fond of throwing transformers into all sort of situations (possibly due to the problems with voltage drop when using a 120V system, so tend to use higher voltages for distribution and then need transformers to reduce it to 120V) whereas this side of the pond we'd be more likely to start with 230/400V system and stick with that throughout. Likewise some parts of the world allow PEN conductors within installations, which can side-step the N-PE link problem (but can introduce other problems of course). I think one well known UPS manufacturer used to leave the supply N connected the the load during power fail (relying on the supply N-PE link) but even they've fallen in line with international standards and isolate supply N as well as L from the output during a power fail. 

I have been told that the larger systems might do things differently though - maybe some of the "big boys" here can enlighten us.

   - Andy.

Thanks for your insights! I’d like to add that with recent larger modular UPS systems, are designed without isolation transformers, there’s often only a single neutral provided. This setup can introduce more challenges when using two different supply sources with an ATS (Automatic Transfer Switch) at the input, as switching between supplies can lead to instability on the neutral. This situation becomes even more critical without an isolation transformer to manage the neutral’s stability and potential noise.

In fact, using an open-transition (break-before-make) ATS in such cases can worsen these issues, as the neutral can briefly lose its connection, potentially creating noise and transients that sensitive IT loads struggle to tolerate. To mitigate this, a closed-transition (make-before-break) ATS becomes almost necessary to maintain continuity, reducing transients and ensuring a smoother transfer between sources.

It would be interesting to hear if others have found specific configurations or solutions for stabilizing neutral under these conditions, especially in high-sensitivity data center environments.