-
Bacnet MS/TP - Master vs. Slave
Can someone help me understand a little bit better MSTP slave devices? I understand they do not particpate in the passing of the token and therefore cannot initiate information requests, only respond to masters request for info from them. From what I gather most are also only 1/8 load transceivers so you can add more than normal on the bus.
What I am looking for is:
1) What other devices on the network do you need to communicate to them - I have seen some references to a Master proxy - what is that?
2) How many devices can you realistically put on one trunk - the max number of masters is 124 but slave devices can exceed that?
3) What is the impact on polling rates/baud rates/performance, etc. when adding slave devices to a MSTP trunk.
4) Is anyone using them extensively? Most of the bacnet devices I come across are masters on the MSTP network.
-
-
JJ- Those are some good questions. Let's see if I can shed some light on them for you.
1) You can communicate with a slave device just as you would a regular master device in as much as you can read and write points. What makes a device a slave device is the fact that it's not in the token ring, nor can it initiate any messages- only respond to them.
Some master devices can be set up to be proxies for slave devices in BACnet as you have mentioned. These controllers respond for the slave device in certain instances, as though they were the device itself. An important function that a proxy would do would be responding to a who-is for a slave device so that front end thinks that the device is present, alive, etc. I personally don’t use this functionality much myself but there are a number of controllers that can do this for you if you require it. (It is my contention that you likely don’t need this feature, just poll for the data directly.)
2) Technically speaking, a trunk can have up to 255 (0 through 254, the number 255 is wild) devices on the trunk. The number of masters is limited by the range that is applicable to masters, so only 128of them (0 through 127) may be considered masters. Please note that this doesn’t mean that a device in this range has to be a master, it can also be a slave assigned an address between 0 and 127…
The question of “realistic” numbers of devices that can be on one trunk is a bit harder to determine. Every manufacturer has a different rule of thumb and in reality it has to do not only with the length of the wire and the quality of the wire but also the overall impedance of each of the devices attached. I have seen some manufacturers suggest that 32 is a good max number, others that suggest 128, etc. Throw repeaters into the mix and this question becomes arguably worthy of an entire thread.
3) The actual act of adding a slave device does not take up any “time” on an MS/TP trunk. It is invisible to the token passing portion of the communications, so it doesn’t effect the time it takes the token to circle the ring. If you are reading or writing a lot of points on the device then those transactions will take time but no more so than they would take asking the same questions of a normal master device.
4) A lot of manufacturers allow their devices to be put in slave mode and there are more yet that make devices that can only be configured this way. If you absolutely require the fastest token ring around, you can always limit the devices getting the token which will decrease the time it takes the token to loop the entire ring. I personally don’t think it makes much difference but I’m sure people could come up with oddball examples where it made all the difference.
In the case of slave-only devices, often this is the only way to make a very limited controller (slow processor, low RAM, etc) exist on the MS/TP trunk. Implementing the entire MS/TP stack is not always an option for say a PIC processor with 4K or RAM…
I hope this helps!
-B
-
thanks, I think that gets me going - I am starting to see more and more of these and just need to understand the differences.
-
Originally Posted by
BACnet
<snip>
also the overall impedance of each of the devices attached. I have seen some manufacturers suggest that 32 is a good max number, others that suggest 128, etc. Throw repeaters into the mix and this question becomes arguably worthy of an entire thread.
<snip>
Good day BACNet,
You are correct here, however, the number of 32 or 128, etc is totally dependent upon which RS-485 transceiver the device manufacturer used. The original and older style RS-485 transceivers had/have an input impedance of 1 unit-load or 12K ohms and was designed to have a maximum of 32 devices (32 unit loads) connected on a bus (RS-485 specification). However, they do make RS-485 transceivers with 1/4 unit load (48K ohm) and 1/8 unit-load (96K ohm) allowing of a maximum of 128 (1/4 unit load devices) or 256 (1/8 unit load devices) respectively. Sadly, each manufacturer uses whichever unit-load device they wish and usually this is dictated by cost.
What happens if you exceed the 32 unit-load spec? If the device just meets the RS-485 drive specification, then you have reduced noise immunity... which results in a problematic bus and/or reduced bandwidth (because the electrical signals may get corrupted by the electrical noise and so will need to be re-transmitted). The bottom line is you take your chances if you exceed more than 32 unit-load devices on your bus.
The moral of the story... the RS-485 transceivers used by the manufacturer will determine the maximum (electrical) number that you can have on the bus (max may also be limited by the communication protocol) and still maintain a high quality (high noise immunity) bus.
Cheers,
Sam
-
Post Likes - 1 Likes, 0 Dislikes
-
Is there any standard happening out there in terms of transceiver chips being used? Are most slave devices 1/8 load and most master devices 1/4 load? Or is it really something that varies from manufacturer to manufacturer and you just need to make a phone call every time you get a new device to ask?
-
Originally Posted by
justjohnson
Is there any standard happening out there in terms of transceiver chips being used? Are most slave devices 1/8 load and most master devices 1/4 load? Or is it really something that varies from manufacturer to manufacturer and you just need to make a phone call every time you get a new device to ask?
Good day JJ,
In regards to the BACNet MS/TP spec itself, I do not recall seeing any particular requirement for the RS-485 transceiver impedance... although perhaps there might be. Perhaps someone who has thorough knowledge on the spec will chime in. I suspect that it is "implied" that a higher impedance transceiver should be used in order to be able to support all of the devices spec'd. So, given this (potential) ambiguity the transceiver's selected will vary from manufacturer to manufacturer and even product to product. Sadly, most products are manufactured with the lowest bill of material cost without regards to consequence... Higher impedance RS-485 transceivers typically cost more (maybe a $1 more) and so are subject to being substituted with the lower impedance (and cost) versions. I have seen some really silly component substitutions and design changes for the simple sake to reduce a bill of material cost by tens of cents for less. Indeed, one must watch the cost of things, but not at the sacrifice of durability, reliability, and quality...
Cheers,
Sam
-
Not sure if this is the latest trend, but the new BACnet spyders are 1/4 load devices.
-
Originally Posted by
s2sam
Higher impedance RS-485 transceivers typically cost more (maybe a $1 more) and so are subject to being substituted with the lower impedance (and cost) versions.
Sam - Are you SURE about that pricing? What total cost are you talking about if they are one dollar more?
-
Originally Posted by
sysint
Sam - Are you SURE about that pricing? What total cost are you talking about if they are one dollar more?
Good day Sysint,
I am fully aware of pricing of these transceivers, as I am actively involved in electronic (and software, firmware, etc) design and manufacturing daily. As for total cost, one has the define this accordingly. My comment about a $1 (or so) more was based upon the RS-485 transceiver itself and did not include any additional costs for additional circuitry that may be included (like transient protection, etc). As with all electronics, purchase volume dictates price. So, buying 1 or 2 devices may not translate to a $1 differential... Some of our higher quality (and low impedance) transceivers can cost $6 in small volumes (<10) but are $2 or so in the volumes we purchase.
Regardless of the cost differential, my point was/is that bill of material cost should not over-ride usability, quality, and reliability. I say this not as just a consumer of these goods, but also as a manufacturer who is truly aware of all of the costs involved and at times annoyed at what short-cuts some manufacturers take.
Cheers,
Sam
-
Hi Sam -- I ask because I'm trying to figure out what Echelon is up to because they are going to offer their new Neuron 5000 for $0.60 in quantity and no credits charge for commissioning.
I'll have to dig. It may be less expensive to run bacnet/LON than MSTP.... and you would get free topology. Further you could probably use LON routing in IP852 and a configuration server. This would make bacnet networks very open and flexible to route with documentation.
-
Originally Posted by
sysint
Hi Sam -- I ask because I'm trying to figure out what Echelon is up to because they are going to offer their new Neuron 5000 for $0.60 in quantity and no credits charge for commissioning.
I'll have to dig. It may be less expensive to run bacnet/LON than MSTP.... and you would get free topology. Further you could probably use LON routing in IP852 and a configuration server. This would make bacnet networks very open and flexible to route with documentation.
Good day Sysint,
I understand. A fairly bullet-proof complete (i.e with transient protection, etc) high impedance RS-485 interface could be built around $6 to $7 or so depending upon volume, etc. I think if you use this number you can determine what the ultimate cost would be.
As for Echelon's Neuron chip price at $0.60... at what volume are they quoting this at? This price is quite inexpensive considering what the chip does. The issue I would have is if one uses their chip in a design and then sometime later it is discovered their chip has some issues (electrical or protocol), then you are hooped... as you would need to potentially change out the chip in all of your existing installs. I say this not to be alarmist (forgive the pun), but because I know of a site where the Neuron chips had some problems and the result was to replace all of the controllers (maybe 50+)... a number of them were behind finished ceilings, etc...the cost was enormous... and I do not know who actually paid for this repair (customer, manufacturer, Echelon).
That being said, it is an interesting and potentially cost effective solution all things considered.
Cheers,
Sam
-
Of topic a little but just for you info Sysint, Siemens Europe does not use BACNet/MSTP for its building controllers but and actually uses BACNet/LON. Its wonderfully easy to install due to the the free topology and I prefer its other BACNet/IP option.
It is limited to 30 devices per LON segment for some reason but maybe thats so they can sell more routers.
-----------------------------------------------------
Don't be a stupid dwarf. It's not big and it's not clever!
-
Originally Posted by
s2sam
Good day Sysint,
I understand. A fairly bullet-proof complete (i.e with transient protection, etc) high impedance RS-485 interface could be built around $6 to $7 or so depending upon volume, etc. I think if you use this number you can determine what the ultimate cost would be.
As for Echelon's Neuron chip price at $0.60... at what volume are they quoting this at? This price is quite inexpensive considering what the chip does. The issue I would have is if one uses their chip in a design and then sometime later it is discovered their chip has some issues (electrical or protocol), then you are hooped... as you would need to potentially change out the chip in all of your existing installs. I say this not to be alarmist (forgive the pun), but because I know of a site where the Neuron chips had some problems and the result was to replace all of the controllers (maybe 50+)... a number of them were behind finished ceilings, etc...the cost was enormous... and I do not know who actually paid for this repair (customer, manufacturer, Echelon).
That being said, it is an interesting and potentially cost effective solution all things considered.
Cheers,
Sam
Sam, I'm saying they offer what looks like a really cheap solution. However, they aren't the only ones able to implement this so I don't think too much issues here to say you are screwed if something isn't right. It's no different than a complete bacnet controller not working really... and I've never heard of a case like you describe, but things can happen to anybody. This gets closer to the D1G model of tossing down bacnet or LON on the same fieldbus wire.
-
Originally Posted by
picnik
Of topic a little but just for you info Sysint, Siemens Europe does not use BACNet/MSTP for its building controllers but and actually uses BACNet/LON. Its wonderfully easy to install due to the the free topology and I prefer its other BACNet/IP option.
It is limited to 30 devices per LON segment for some reason but maybe thats so they can sell more routers.
Picnik... you should know by now I'm fully aware of this already. As long as you brought it up - Somewhere Siemens wrote a nice article (I'd like to find again) showing how much better this solution is compared to MSTP.
-
Originally Posted by
sysint
Sam, I'm saying they offer what looks like a really cheap solution. However, they aren't the only ones able to implement this so I don't think too much issues here to say you are screwed if something isn't right.
It's no different than a complete bacnet controller not working really... and I've never heard of a case like you describe, but things can happen to anybody.
This gets closer to the D1G model of tossing down bacnet or LON on the same fieldbus wire.
Good day Sysint,
Indeed, in regards to my comment about the Neuron chip's possible future issue(s) the risk is small, but nevertheless there. Where this differs from BACNet (assuming one implements their own BACNet stack) is that if there is a protocol issue (i.e. bug) a manufacturer has the ability to sort it out...as the the problem can be corrected in firmware (assuming the problem is not a faulty hardware design). Using a Neuron chip (or a dedicated third-party BACNet chip), the manufacturer is hooped and at the mercy of the Neuron (or BACnet) chip manufacturer to fix the problem. Secondly, if it is shown that the chip itself has issues, then the only solution is to replace the chip... which is not very simple or economical
Now, to be fair there are other single sourced chips in the manufacturer's design... such as the microcontroller or microprocessor, however, because these chips are generic they are sold in much higher volumes (100's of thousands, millions, etc) and so major bugs within them would have been resolved fairly early on in their release.
I make these points as items to be aware of and not meaning to slight any one vendor or protocol.
Cheers,
Sam
-
Well yes I suppose. However, you can implement LON off the chip. - Anyway, it's not like there is one way to do this. However, this one is rather inexpensive.
-
Originally Posted by
s2sam
Good day Sysint,
Regardless of the cost differential, my point was/is that bill of material cost should not over-ride usability, quality, and reliability. I say this not as just a consumer of these goods, but also as a manufacturer who is truly aware of all of the costs involved and at times annoyed at what short-cuts some manufacturers take.
The full-load (32 device) transceivers are more noise-immune than the 1/4 or 1/8 load RS485 transceivers. VFD's often use the full-load transceivers, because of this. A 480VAC VFD has about 600VDC on their power circuitry which is physically located close to the isolated RS485 circuits. In many cases it is not about saving a few pennies, it is often about the quality of communications.
-
Originally Posted by
AbeS187
The full-load (32 device) transceivers are more noise-immune than the 1/4 or 1/8 load RS485 transceivers. VFD's often use the full-load transceivers, because of this. A 480VAC VFD has about 600VDC on their power circuitry which is physically located close to the isolated RS485 circuits. In many cases it is not about saving a few pennies, it is often about the quality of communications.
Good day AbeS187,
I do not know if you can say this across the board for all 1 Unit Load (UL)RS-485 transceivers. By definition a RS-485 transceiver must comply with the minimum requirements which are set out in the electrical specification(s) for the RS-485 (or should I say EIA-485) devices. Some manufacturers exceed these minimums and some offer devices which guarantee these "enhanced" specifications... but not all. Remember if the specs are not "guaranteed" then you cannot assume that these enhanced specs will always be available as these specs can and will change on the respective manufacturing run. For specific design critical specs I always qualify a manufacturer and I will not allow a alternative manufacturer's component substitution unless I have qualified this new vendor's part. Sadly, at times purchasing department personnel see all components as generic and as such feel they can interchange parts (assuming functional/footprint compatibility, etc) based upon pricing... this is indeed not so on a number of components that are deemed critical to a circuit's functionality and reliability.
Another item which may be related to your comment is that the 1 UL devices have a lower input impedance (minimum of 12K ohms by spec) than the 1/4 UL (48K ohms) or 1/8 UL (96K ohms) devices which in itself may offer less susceptibility to noise.
Cheers,
Sam
-
Originally Posted by
s2sam
Good day AbeS187,
Another item which may be related to your comment is that the 1 UL devices have a lower input impedance (minimum of 12K ohms by spec) than the 1/4 UL (48K ohms) or 1/8 UL (96K ohms) devices which in itself may offer less susceptibility to noise.
The lower input impedance results in pulling more current from the bus hence it is generally MORE immune to noise.
-
MORE immune to noise. less susceptibility to noise.
Aren't these two statements the same?