Design of a power supply for a non certified KNX bus

I'm not sure what you are talking about. Perhaps post a link that has a PS schematic.

(one inductor will be enough... current has to flow out one leg and back the other.)

George H.

Wow, being an open standard, theres not much information on the choke value used. I gather it's a couple of uH from looking at the DIN rail chokes from Gira.

Cheers

Il 05/05/2016 17:38, George Herold ha scritto: > [...]

KNX

specifications are open, but there isn't a direct download link. You can register to MyKnx and make an order for KNX specifications. The price of the product (specifications) is zero. After applying the order, you will be able to download the KNX specifications.

The specifications are very big, because KNX is a very complete standard for building automation and specify almost everything, from phisical medium (DC bus, powerline, RF, IP) to application layer, addressing, fragments, and so on.

I'm interested only in the physical and data-link layer named KNX TP1 (TP stands for twisted pair). TP1 is descripted in Chapter 3_2_2, but here there isn't a good description of power supply unit (with its choke).

Some info about the PSU for TP1 is descripted in Chapter 9_2. I uploaded two images that you can see here:

I'm not an expert of power supply units and transformers and, as I wrote in my original post, I don't need to be KNX certified, because I'm going to develop a proprietary solution inspired from KNX TP1 standard, resuing ready-to-use KNX transceivers (ON Semiconductors makes nice products).

The big problem is the choke. It is a transformers with three windings. What is the mean of the right winding/circuit with BC327 transistor? Is it necessary?

As you can see, there are two windings/inductor: one on positive and one on negative rail (in parallel with a resistor). Are they coupled? Why there are two and not only one on a single rail?

KNX devices are very common in Europe, so I think some transformers manufacturers has already designed and produced a suitable KNX choke, but I couldn't find any.

Any help?

The windings are all coupled and act as a big "differential mode" choke to partly isolate the data signalling from the power supply. My guess is they chose symmetrical chokes in positive and negative lines to improve balance to ground. If your intended version is only on a short cable with just a few devices (as opposed to running round whole buildings) or is free of ground connections then you might get away with a single choke.

I'm still thinking about the clamp circuit (winding 1 -2) but my guess is that it prevents overshoot under some kind of event?

How is the data signalling done? What voltage/current levels and speeds, you mentioned 9600 bps but is that modulated onto some higher frequency carrier?

piglet

One or two chokes is not a problem... the big problem is the transformer (third winding).

I'm not able to understand that section of circuit. How the secondary winding will be able to avoid overshoot on primary?

It is all descripted in KNX TP1 specifications (Chapter 3_2_2).

Something is written in KNX transcervers datasheets, for example here (page 16):

The device that wants to transmit should hear if the bus is idle (CSMA/CA). If it is idle, it can start transmitting the telegram, a sequence of 0s and 1s.

Only 0s can be seen on the bus, because 1s don't change the idle condition of the bus.

The device that wants to transmit 0 puts a suitable load on the bus that draw as much current as need to decrease the voltage on the bus of 6-9V respect the idle value (typical 29Vdc). This step should be maintened for 35us (active period), followed by an equalization pulse. The total bit time is 104us (9600bps).

Of course the voltage step is created thanks to the choke in the power supply.

Here is my attempt at an explanation, which could be all wrong:

The 100R resistor and 220nF capacitor are a snubber. During data signalling current flows through winding 3-4 (and 5-6) and the dot end is more positive, this happens too in winding 1-2 where the dot end becomes more positive and makes the diode conduct and the capacitor charge up. When the signal current ceases the voltage in winding 1-2 reverses and the diode is non-conducting and the transistor is turned on to discharge or reset the capacitor. However the two 47R resistors across the primaries also do plenty some damping and I do not yet understand the need for the extra assymmetric snubber.

My guess is you could probably forget all that in a minimalist version and just shunt a 5mH inductor with a 100R resistor and if you are worried about bus overshoot put a 47V zener across the data bus?

piglet

Surprisingly, it is. Find a reverse engineered schematic here:

You may also read the whole thread, unfortunately it is in German: There seems also a patent involved:

The circuit around the transistor is to limit bus voltage when sending (pulling the bus voltage lower for 6V iirc) stops.

If you want to google for it, try EIB instead of KNX. EIB is the original name of KNX but it was not wisely chosen since it limits the market to Europe (European Installation Bus) :-)

Cheers

Robert

Hi Robert Thanks for your answer.I couldn't download schematic.Could you please upload it somewhere else?

Regards Behdad

Am 25.04.2017 um 22:35 schrieb snipped-for-privacy@gmail.com: ...

Hello Behdad,

try this link:

the other seems to be broken. Does not work for me, too.

Cheers

Robert

Thank you Robert.It is so kind of you.