Power Over Bus (Canbus)

Thread Starter

ilginsarican

Joined Jul 13, 2017
114
Hi,
We are planning to use Power Over CANBUS.I've attached schematics.
The structure in the schematic is the circuit structure located at each node to communicate.
 
Last edited:

Papabravo

Joined Feb 24, 2006
16,165
I'm not at all surprised that you are having problems. This is perhaps the stupidest design I have ever seen. How do you expect the CAN transceiver to drive all of that capacitance, transceivers, and the distributed capacitance of the cable besides? There is no accounting for transmitter drive levels or receiver thresholds. You are way better off running a separate data pair and power pair and using the network to power just the transceivers. Use freaking DeviceNet cable -- it was designed for this purpose, and can support in excess of 32 nodes at 500 kbits/sec. It betrays an incredible ignorance of how the CAN physical layer is supposed to operate. Oh and lose the slope control it is just about the most useless of features.

Puttining optional termination resistors on a device is a foolish thing to do. It will definitely bite you when you go to install these devices and you can't quickly decide which devices have them and which ones don't when things don't work. Terminations BELONG on the cable system, and nowhere else.

Just as an exercise, tell me what the input impedance of 32 transceivers, all connected in parallel is?
 
Last edited:

Thread Starter

ilginsarican

Joined Jul 13, 2017
114
I'm not at all surprised that you are having problems. This is perhaps the stupidest design I have ever seen. How do you expect the CAN transceiver to drive all of that capacitance, transceivers, and the distributed capacitance of the cable besides? There is no accounting for transmitter drive levels or receiver thresholds. You are way better off running a separate data pair and power pair and using the network to power just the transceivers. Use freaking DeviceNet cable -- it was designed for this purpose, and can support in excess of 32 nodes at 500 kbits/sec. It betrays an incredible ignorance of how the CAN physical layer is supposed to operate. Oh and lose the slope control it is just about the most useless of features.

Puttining optional termination resistors on a device is a foolish thing to do. It will definitely bite you when you go to install these devices and you can't quickly decide which devices have them and which ones don't when things don't work. Terminations BELONG on the cable system, and nowhere else.

Just as an exercise, tell me what the input impedance of 32 transceivers, all connected in parallel is?
Actually I was inspired from this design: https://www.ti.com/tool/TIDA-00527 I think you should look at this.
"You are way better off running a separate data pair and power pair and using the network to power just the transceivers. "
This is not the design I planned.
I think when you examine the sample design(TIDA-00527) you will understand why I tried something like this.
I know the problem I'm talking about is impedance.
I just need different opinions.
Also after I put termination resistor on all nodes, data transfer achieved from all nodes(5nodes), R56 and R60 are not optional.
 

Attachments

Papabravo

Joined Feb 24, 2006
16,165
Actually I was inspired from this design: https://www.ti.com/tool/TIDA-00527 I think you should look at this.
"You are way better off running a separate data pair and power pair and using the network to power just the transceivers. "
This is not the design I planned.
I think when you examine the sample design(TIDA-00527) you will understand why I tried something like this.
I know the problem I'm talking about is impedance.
I just need different opinions.
Also after I put termination resistor on all nodes, data transfer achieved from all nodes(5nodes), R56 and R60 are not optional.
TI says they are doing site maintenance and that info is inaccessible. Why don't you give me the executive summary and we'll go from there. I have no reason to doubt your intentions. I think you just stumbled into the wrong pathway.

BTW -- what bitrate are you using, what crystal are you using, where is the sample point and what is the SJW?
 

Ian0

Joined Aug 7, 2020
2,259
There a very important difference between RS485 and CAN. RS485 has push-pull drivers, CAN doesn't, it drives each line one way only, with a resistor to return it to the centre point, so your peak-to-peak signal is much smaller.
Note that the TI app note requires data with no net Dc, and suggests Manchester coding.
CAN with Manchester coding???
CAN always has net DC , and also could transmit no power when it is in the recessive state.
Really, you’re not on to a winner.
If you've only got two wires and ground, how about single-ended CAN or LINbus instead?
 
Last edited:

Papabravo

Joined Feb 24, 2006
16,165
There a very important difference between RS485 and CAN. RS485 has push-pull drivers, CAN doesn't, it drives each line one way only, with a resistor to return it to the centre point, so your peak-to-peak signal is much smaller.
Note that the TI app note requires data with no net Dc, and suggests Manchester coding.
CAN with Manchester coding???
CAN always has net DC , and also could transmit no power when it is in the recessive state.
Really, you’re not on to a winner.
If you've only got two wires and ground, how about single-ended CAN or LINbus instead?
Like I said the TS needs to study CAN and the reasons for the design decisions in more detail. It seems fairly clear that he is lacking in that regard.
 

Thread Starter

ilginsarican

Joined Jul 13, 2017
114
TI says they are doing site maintenance and that info is inaccessible. Why don't you give me the executive summary and we'll go from there. I have no reason to doubt your intentions. I think you just stumbled into the wrong pathway.

BTW -- what bitrate are you using, what crystal are you using, where is the sample point and what is the SJW?
Hi again,
I attached the documents from TI.
Also I found "PHYSICAL LAYER OF CAN BUS NETWORK OVER DC POWER LINE NETWORK " manuel from Yamar.
 

Attachments

Ian0

Joined Aug 7, 2020
2,259
One is data over power lines, the other is power over data lines.
Is the safety systems in my car relied on data transmitted over the +12V power cables, I think I’d be taking the train.
 

MisterBill2

Joined Jan 23, 2018
8,793
Why are there so many capacitors across the data lines? Certainly that will increase the load on the driver so that communication will not be possible. Power filtering is to be done on the power side of the isolation sevice, NOT on the data side.
There is no reason to expect that putting a lot of capacitance across a data pair will allow data transfer to be possible.
 

Ian0

Joined Aug 7, 2020
2,259
The caps aren't across the datalines. They capacitively couple the data on to the power lines, and inductors isolate the data from the power supply smoothing capacitors. I can see it working for RS485.
 

MisterBill2

Joined Jan 23, 2018
8,793
Then it may be that the inductance of those coils is not adequate for the number of points shunted across the line. The impedance is not infinite and so with enough of them in parallel that may be the problem causing the failure to communicate. But that is just a guess, based on incomplete information. My previous comment was at 4AM and so it may have been a bit off.
 

Ian0

Joined Aug 7, 2020
2,259
I think it's a matter of scale - one master isn't going to power 32 slaves, each of which powers a microcontroller taking 30mA or so, but powering one or two low-power slaves is a different matter.
Also, whether or not the slave transmits is a big factor. It has to drive 40mA (5V/120ohms) when there is no power coming in.
In fact, all the time the master is high-impedance (i.e. waiting for a slave to transmit) there is no power being supplied to the slaves.
CAN makes the situation worse - power is only available during a message when a zero is being transmitted (so, about half the time) and not at all between messages. A typical CAN transceiver can use as much as 2.5mA in quiescent state.
The CAN driver peak output voltage into a double-terminate load is 3V, after that goes through two schottky diodes, there's not even enough voltage to power a CAN transceiver. Even if that problem can be overcome, the terminated line takes 50mA leaving only 30mA to supply external loads. Assuming 50% ones:zeroes, that's down to 15mA, and assuming the POWERED node only transmits 50% of the time, it's now down to 7.5mA, that's three transceivers that don't ever transmit, and no power left over for any other circuitry.
I still think the idea of power-over-CAN isn't going to fly.
 

Thread Starter

ilginsarican

Joined Jul 13, 2017
114
I think it's a matter of scale - one master isn't going to power 32 slaves, each of which powers a microcontroller taking 30mA or so, but powering one or two low-power slaves is a different matter.
Also, whether or not the slave transmits is a big factor. It has to drive 40mA (5V/120ohms) when there is no power coming in.
In fact, all the time the master is high-impedance (i.e. waiting for a slave to transmit) there is no power being supplied to the slaves.
CAN makes the situation worse - power is only available during a message when a zero is being transmitted (so, about half the time) and not at all between messages. A typical CAN transceiver can use as much as 2.5mA in quiescent state.
The CAN driver peak output voltage into a double-terminate load is 3V, after that goes through two schottky diodes, there's not even enough voltage to power a CAN transceiver. Even if that problem can be overcome, the terminated line takes 50mA leaving only 30mA to supply external loads. Assuming 50% ones:zeroes, that's down to 15mA, and assuming the POWERED node only transmits 50% of the time, it's now down to 7.5mA, that's three transceivers that don't ever transmit, and no power left over for any other circuitry.
I still think the idea of power-over-CAN isn't going to fly.
Thanks for your return, we don't use schottky diodes,It will be completely canceled in the next revision.
When we were trying this structure, we increased the termination resistance, then now 9 nodes are communicate. This number was 4 when we first tried it.
 

Ian0

Joined Aug 7, 2020
2,259
But now you have D5-D8, which will drop about half a volt if they are schottkies. So the remaining power supply, if you start with a 3V p/p signal from the CAN driver, is now only 2.5V. Not enough to supply a CAN transceiver.
 

Thread Starter

ilginsarican

Joined Jul 13, 2017
114
But now you have D5-D8, which will drop about half a volt if they are schottkies. So the remaining power supply, if you start with a 3V p/p signal from the CAN driver, is now only 2.5V. Not enough to supply a CAN transceiver.
We had this problem in our first attempt, so we removed the diodes, now it can be thought as if there is 0 ohm instead of diodes.
 
Hi,
I'm Mohammad and I'm working on a same subject,
would you please tell me have you done it or not?
there is an addressable fire alarm system witch works in the same way and we are on way to create it,
so here is how it works:
we have a single loop witch can provides up to 255 detectors on just a "2 wire twisted pair" and I'm sure the protocol is CAN;
the current consumption of every detector when they are on standby mode is down to 20-100 uA and when they are activated they use like 5 mA.(and more interestingly they are non-polarity so the user can install it more easier).
it is done so it's not impossible ;)
so any idea so you can help me( i really would be grateful).
 

Attachments

MisterBill2

Joined Jan 23, 2018
8,793
Hi,
I'm Mohammad and I'm working on a same subject,
would you please tell me have you done it or not?
there is an addressable fire alarm system witch works in the same way and we are on way to create it,
so here is how it works:
we have a single loop witch can provides up to 255 detectors on just a "2 wire twisted pair" and I'm sure the protocol is CAN;
the current consumption of every detector when they are on standby mode is down to 20-100 uA and when they are activated they use like 5 mA.(and more interestingly they are non-polarity so the user can install it more easier).
it is done so it's not impossible ;)
so any idea so you can help me( i really would be grateful).
You are asking to be given a whole lot of valuable information that may actually belong to the TS employer. Certainly the concept of your product is interesting, but what I have seen of the workmanship of alarm system installers it is a poor choice to put everything into one loop, subject to failure due to poor installation. There may be exceptions but I have not seen them yet.
I suggest investigating the methods used for Power Over Ethernet.
 
Top