PROFINET vs PROFIBUS: the gateway spec that decides whether your HMI still talks

The call always sounds the same. One DP drive on the line has gone obsolete, the only replacement anybody still sells has a PROFINET port, and nobody wants to touch the other forty nodes sitting happily on RS-485. So the plan becomes a gateway. One box, in by Friday. The IO does come across, every time. What tends to quit is the HMI screen that used to read that drive's parameters, because PROFINET vs PROFIBUS is not only a speed comparison. It is a question of what a given box is built to carry over. Here is what separates the two, and the one line on a gateway spec that decides how much of your system survives.

What actually changes between the two networks

Same organization wrote both standards, which is why so much carries over. Application profiles are shared, so PROFIsafe, PROFIdrive, PROFIenergy and the process automation profile behave the same either side. Device descriptions are GSD files in both cases: ASCII on PROFIBUS, XML on PROFINET.

Below that layer they are not the same animal.

	PROFIBUS DP	PROFINET
Physical layer	RS-485	Ethernet
Top speed	12 Mbit/s	100 Mbit/s or 1 Gbit/s
Cyclic telegram	244 bytes	1440 bytes
Addresses per network	126	limited by the controller, not the protocol
Traffic model	master / slave	provider / consumer
Motion axes	32	more than 150
Device description	GSD, ASCII	GSDML, XML
Typical connector	9-pin D-sub	RJ45

PROFINET vs PROFIBUS, the spec differences that change a design.

Two rows there do more damage than the rest. The 244-byte telegram bites on drives and smart instruments: if your DP telegram was already packed, PROFINET gives you room at 1440 bytes cyclic. And 126 addresses sounds like plenty right up until somebody hangs a fifth remote rack off the same segment.

The traffic model gets underrated, though. Master/slave means a DP node speaks when it is spoken to, so every node you add stretches the poll cycle whether it had anything to report or not. Provider/consumer on switched full-duplex Ethernet lets nodes talk when they have data, and nothing collides. Most of the real speed difference lives there, not in the wire rating.

The topology difference behind the PROFINET vs PROFIBUS numbers: a DP trunk polls in turn, a PROFINET segment switches.

You do not have to move the whole line

PROFIBUS is not going anywhere soon. New PROFINET node counts overtook new PROFIBUS node counts back in 2016, and the installed base then was still north of 56 million PROFIBUS nodes worldwide. On a plant that already has a network there are two honest paths: rip-and-replace, which gets you everything at once and charges you production time for it, or step-wise migration, which upgrades what needs upgrading and puts a proxy where the two networks meet.

PROFINET helps here in a way most protocols do not. The proxy is written into its own specification rather than left to whoever wants to build a converter, and defined proxies exist for PROFIBUS DP and PA, AS-i, IO-Link, Modbus and HART. From the controller's side a proxy is just another PROFINET device, and the DP slaves behind it appear as PROFINET IO devices. On the PROFIBUS side, that same box is the DP master.

Worth saying plainly, because PROFINET vs PROFIBUS usually gets framed as new versus old: in a hazardous area PROFIBUS PA is still the correct answer, and a proxy is how you get it onto an Ethernet backbone. Power and signal on one pair, intrinsically safe, and a field device catalog Ethernet cannot match.

Not every PROFINET vs PROFIBUS gateway carries the same traffic

Two tiers of box get sold for this job and the spec sheets look close enough that they get treated as interchangeable. They are not, and this is the part that costs people a weekend.

A plain protocol converter moves IO data, and that is the whole job description. PROFIBUS master on one port, PROFINET IO device on the other, cyclic data mapped between them. HMS specifies its Anybus X-gateway PROFIBUS Master to PROFINET-IO Device exactly that way, and now lists that model as legacy, worth knowing if you are speccing new rather than matching a panel.

A full PROFINET IO proxy does the mapping and keeps going. Siemens' IE/PB Link PN IO is the reference example: it also routes S7 connections across the boundary, which is what lets an HMI or a programming laptop on the Ethernet side reach something living on PROFIBUS.

Traffic	Plain IO protocol converter	Full PROFINET IO proxy (IE/PB Link PN IO)
Cyclic IO data	yes	yes, up to 2048 bytes in and 2048 bytes out
DP slaves behind one unit	model dependent	up to 64
PG/OP and programming access to the DP side	not offered	yes
HMI and S7 connections across the boundary	not offered	yes, via S7 routing
Field device parameter assignment	not offered	yes, data record routing, 244 bytes per record
Non-cyclic connection budget	not applicable	32 S7/HMI, 32 data record, 48 total in any mix
DP slave versions supported	model dependent	DP V0 and DP V1
PROFIsafe over the bridge	model dependent	supported

What each class of PROFINET vs PROFIBUS box carries across.

Read that connection budget before committing an architecture to one box. Forty-eight in any combination is the number that matters, not the individual 32s, and every S7 connection also occupies a TCP/IP connection on the Ethernet side. Two SCADA clients, a historian and a few engineering stations, and 32 arrives faster than it sounds.

The IO always crosses. HMI and programming traffic only crosses when the box does S7 routing.

What catches people on the PROFIBUS side of the bridge

Mode is a commitment, not a setting. The IE/PB Link runs either as a PROFINET IO proxy or as a plain gateway, and moving between them takes a clear and reset that wipes the stored configuration.

Its two PROFINET ports do support MRP, but Siemens only allows the link the client role. If somebody drew it into a ring as redundancy manager, that is not happening.

Do not take the shortest update time the software offers you, either. STEP 7 proposes the fastest value the network can carry, and on a proxy also handling S7 connections, HMI traffic and data record routing, that value is the wrong one. Siemens says take the next step up. Diagnostic and alarm frames off the DP side draw from the same budget.

Supported slaves are DP V0 and DP V1. Run your real slave list against that before the purchase order, not after the box is on the rail.

One more, on the wire. Pin 6 on the DP D-sub carries +5 V to feed the bus terminating resistors, not to supply external devices, and people do try. Get the cable shield contacted to functional ground while you are in that connector: the shield does real work on RS-485, and a floating one turns into intermittent faults you will chase for a week.

Swapping a DP rack head for a PROFINET one and keeping the I/O modules behind it is often cheaper than bridging it.

What I check before ordering anything

Every PROFINET vs PROFIBUS job I have priced comes down to three questions, in this order.

1.    Is the bridge worth building at all? Where the controller is on the replacement list anyway, an S7-300 CPU with PROFINET and PROFIBUS DP integrated gives you both networks off one module: no gateway in the middle, no separate connection budget, nothing extra to spare out.

2.    Does anything on the Ethernet side need the DP side for something other than cyclic IO? If it does, you need a proxy with S7 routing rather than a protocol converter, and that gets decided on the purchase order.

3.    How many non-cyclic connections do you actually need? IO is the easy part and everybody sizes it first. Programming access, HMI screens and parameter tools are what run the budget dry.

The order these questions get asked in decides whether you buy one box or three.

Questions that come up on PROFINET vs PROFIBUS jobs

Is PROFINET just Ethernet?

Physically, yes: standard Ethernet, standard RJ45, standard switches. The difference is what runs on top. PROFINET IO carries cyclic data on its own real-time channel with a configured update cycle, not on plain TCP/IP. A PROFINET device will sit on an office switch and answer a ping. Holding its configured update time there is another matter.

Is PROFIBUS RS-485 or RS-232?

RS-485, on a 9-pin D-sub. Two data lines plus ground and shield, never RS-232. If somebody hands you a DB9 serial cable out of a drawer, it is not a DP cable and the pinout does not match.

Can I put a single PROFINET device onto an existing PROFIBUS network?

That is the harder direction. Nearly everything on the market runs master on the PROFIBUS side and IO device on the PROFINET side, so DP gear can hang off a PROFINET controller. What you are describing needs the mirror image of that, a PROFINET IO controller paired with a PROFIBUS DP slave, and there is far less choice. Check which side of the box is master and which is slave before ordering, because the model names do not always make it obvious.

Which is actually faster, PROFINET vs PROFIBUS?

PROFINET, by more than the wire ratings suggest. The gap closes back up when a lot of non-cyclic traffic goes through one proxy, because the bridge is then the slowest thing in the path.

Where to start if you are pricing this out

The PROFINET vs PROFIBUS decision usually comes down to whether the CPU is staying. If it is, the proxy is the buy, and S7 routing is the first line to check on the spec sheet. If the CPU is already on the list, look at an S7-300 with both interfaces on board: the CPU 317-2 PN/DP and the CPU 315-2 PN/DP each carry PROFINET and PROFIBUS DP on one module, and we keep both as new-surplus in the original box, along with the ET 200M IM 153-4 PN IO for converting a DP rack head. Want a second set of eyes on a slave list before the order goes out? Ask. Cheaper than a return.