You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@plc4x.apache.org by Niels Basjes <Ni...@basjes.nl> on 2023/01/07 16:27:35 UTC
Modbus addresses offset by 1?
Hi,
I ran into the effect that all modbus addresses I specify are shifted down
by 1.
This turns out to be code in most of the ModbusTag implementations (not
all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
Why was this done?
If I'm writing software to read modbus addresses and I follow the
manufacturer specified information it will run into unexpected effects
(like I have today).
Also: Now getAddressString() yields something different then what I used
to build the tag with.
--
Best regards / Met vriendelijke groeten,
Niels Basjes
Re: Modbus addresses offset by 1?
Posted by Ben Hutcheson <be...@gmail.com>.
Hi Niels,
I know there was a reason why the offset didn't get applied to the extended
registers, however I can no longer find out why I thought that it shouldn't
have applied.
I would be in favor of adding the offset to the extended register to make
them the same as the others.
Ben
On Sun, Jan 8, 2023 at 11:07 PM Łukasz Dywicki <lu...@code-house.org> wrote:
> Based on my own experience with modbus devices - you might also face
> situation where manufacturer starts addressing with own offset by not
> following official guidance. A fairly common practice I observed is use
> of addresses ie. 0xxxx, 1xxxx, 4xxxx for coil/input/holding which are
> being called "entity numbers". [1]
> Out of tens of modbus devices I interfaced with I had a matching
> combination of manufacturer docs, library approach and actual device
> addresses once or twice. Speaking here about heat pumps, inverters,
> electric meters and PLCs, all are subject of same trouble.
>
> Maybe implementing whole driver is not needed as you can try to use
> optimizer which lives between your program code and driver and can
> re-arrange fields/tags before they reach device?
>
> Best,
> Łukasz
>
> [1] https://en.wikipedia.org/wiki/Modbus#Entity_numbers_and_addresses
>
> On 8.01.2023 12:27, Niels Basjes wrote:
> > Hi,
> >
> > Understood. Keep the offset it is.
> > But why does the ModbusTagExtendedRegister not have this shift? Is that a
> > bug?
> >
> > Also there are a few other problems I found in this area, one is that the
> > getAddressString will yield a different address (which is also
> unparsable).
> > I'll see if I can fix those.
> >
> > Niels
> >
> > On Sat, Jan 7, 2023 at 8:35 PM Christofer Dutz <
> christofer.dutz@c-ware.de>
> > wrote:
> >
> >> Hi all,
> >>
> >> yeah … just wanted to say … this stupid offset is by spec that way.
> >> I would say, that in this case we have two different specs with a common
> >> base.
> >> Perhaps if we had “variants” of drivers, we could make different
> >> assumptions.
> >>
> >> If you get a Modbus TCP connection things are the way, they currently
> are.
> >> If you get a Modbus TCP connection with “SunSpec” variant, then you
> don’t
> >> have the offsets.
> >>
> >> Cause … if we undid the “+1” shift, then other would start complaining
> >> because their specked registers are in different locations.
> >>
> >> I have also heard that for EIP for full support of the PLC features,
> we’ll
> >> have to have Allen Bradley, Schneider Electric variants next to the
> default
> >> ones.
> >> I was told most of the major vendors do all the complex types and
> browsing
> >> differently :-(
> >>
> >> Chris
> >>
> >> From: Niels Basjes <Ni...@basjes.nl>
> >> Date: Saturday, 7. January 2023 at 18:31
> >> To: dev@plc4x.apache.org <de...@plc4x.apache.org>
> >> Subject: Re: Modbus addresses offset by 1?
> >> Ok, the "off by one" problem is ancient.
> >> Looking around the internet I read that in the past when register 123
> was
> >> needed you would have to request for 122 over modbus.
> >>
> >> I find the transition from the logical number to the technical number a
> >> responsibility of the application, not of this library.
> >> Take for example the SunSpec (modbus standard for solar systems)
> >> specification it states:
> >>
> >> *6.1.1 **Modbus Address Location*
> >> *All Modbus device maps MUST be in the holding register address space.*
> >> *The beginning of the device Modbus map MUST be located at one of three
> >> Modbus addresses*
> >> *in the Modbus holding register address space: 0, 40000 (0x9C40) or
> 50000
> >> (0xC350). These*
> >> *Modbus addresses are the full 16-bit, 0-based addresses in the Modbus
> >> protocol messages.*
> >> *The first two Modbus registers at the start address MUST have the
> >> following well-known*
> >> *constant values as a marker: 0x5375, 0x6E53 (hexadecimal values of the
> >> ASCII string SunS).*
> >>
> >>
> >> On my solar inverter if I request 2 registers at 40000 I get the
> mentioned
> >> 0x5375, 0x6E53.
> >> Yet with plc4j/modbus I must specify address 40001 because the library
> >> shifts it by one.
> >> It took me the better part of today to figure that one out.
> >>
> >> Niels
> >>
> >> On Sat, Jan 7, 2023 at 6:13 PM Niclas Hedhman <ni...@hedhman.org>
> wrote:
> >>
> >>>
> >>> I am not sure what you are asking; But Modbus documentation addresses
> >>> are (in my experience) always offset by 1, which I think goes back to
> >>> the 1970s when we weren't yet sure whether numbers started at 0 or at
> 1,
> >>> and number ranges were a scarce resource so we wouldn't sacrifice one
> >>> for consistency... Those were the days.
> >>>
> >>> Niclas
> >>>
> >>> On 2023-01-07 17:27, Niels Basjes wrote:
> >>>> Hi,
> >>>>
> >>>> I ran into the effect that all modbus addresses I specify are shifted
> >>>> down
> >>>> by 1.
> >>>> This turns out to be code in most of the ModbusTag implementations
> (not
> >>>> all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
> >>>>
> >>>> Why was this done?
> >>>> If I'm writing software to read modbus addresses and I follow the
> >>>> manufacturer specified information it will run into unexpected effects
> >>>> (like I have today).
> >>>>
> >>>> Also: Now getAddressString() yields something different then what I
> >>>> used
> >>>> to build the tag with.
> >>>
> >>
> >>
> >> --
> >> Best regards / Met vriendelijke groeten,
> >>
> >> Niels Basjes
> >>
> >
> >
>
Re: Modbus addresses offset by 1?
Posted by Łukasz Dywicki <lu...@code-house.org>.
Based on my own experience with modbus devices - you might also face
situation where manufacturer starts addressing with own offset by not
following official guidance. A fairly common practice I observed is use
of addresses ie. 0xxxx, 1xxxx, 4xxxx for coil/input/holding which are
being called "entity numbers". [1]
Out of tens of modbus devices I interfaced with I had a matching
combination of manufacturer docs, library approach and actual device
addresses once or twice. Speaking here about heat pumps, inverters,
electric meters and PLCs, all are subject of same trouble.
Maybe implementing whole driver is not needed as you can try to use
optimizer which lives between your program code and driver and can
re-arrange fields/tags before they reach device?
Best,
Łukasz
[1] https://en.wikipedia.org/wiki/Modbus#Entity_numbers_and_addresses
On 8.01.2023 12:27, Niels Basjes wrote:
> Hi,
>
> Understood. Keep the offset it is.
> But why does the ModbusTagExtendedRegister not have this shift? Is that a
> bug?
>
> Also there are a few other problems I found in this area, one is that the
> getAddressString will yield a different address (which is also unparsable).
> I'll see if I can fix those.
>
> Niels
>
> On Sat, Jan 7, 2023 at 8:35 PM Christofer Dutz <ch...@c-ware.de>
> wrote:
>
>> Hi all,
>>
>> yeah … just wanted to say … this stupid offset is by spec that way.
>> I would say, that in this case we have two different specs with a common
>> base.
>> Perhaps if we had “variants” of drivers, we could make different
>> assumptions.
>>
>> If you get a Modbus TCP connection things are the way, they currently are.
>> If you get a Modbus TCP connection with “SunSpec” variant, then you don’t
>> have the offsets.
>>
>> Cause … if we undid the “+1” shift, then other would start complaining
>> because their specked registers are in different locations.
>>
>> I have also heard that for EIP for full support of the PLC features, we’ll
>> have to have Allen Bradley, Schneider Electric variants next to the default
>> ones.
>> I was told most of the major vendors do all the complex types and browsing
>> differently :-(
>>
>> Chris
>>
>> From: Niels Basjes <Ni...@basjes.nl>
>> Date: Saturday, 7. January 2023 at 18:31
>> To: dev@plc4x.apache.org <de...@plc4x.apache.org>
>> Subject: Re: Modbus addresses offset by 1?
>> Ok, the "off by one" problem is ancient.
>> Looking around the internet I read that in the past when register 123 was
>> needed you would have to request for 122 over modbus.
>>
>> I find the transition from the logical number to the technical number a
>> responsibility of the application, not of this library.
>> Take for example the SunSpec (modbus standard for solar systems)
>> specification it states:
>>
>> *6.1.1 **Modbus Address Location*
>> *All Modbus device maps MUST be in the holding register address space.*
>> *The beginning of the device Modbus map MUST be located at one of three
>> Modbus addresses*
>> *in the Modbus holding register address space: 0, 40000 (0x9C40) or 50000
>> (0xC350). These*
>> *Modbus addresses are the full 16-bit, 0-based addresses in the Modbus
>> protocol messages.*
>> *The first two Modbus registers at the start address MUST have the
>> following well-known*
>> *constant values as a marker: 0x5375, 0x6E53 (hexadecimal values of the
>> ASCII string SunS).*
>>
>>
>> On my solar inverter if I request 2 registers at 40000 I get the mentioned
>> 0x5375, 0x6E53.
>> Yet with plc4j/modbus I must specify address 40001 because the library
>> shifts it by one.
>> It took me the better part of today to figure that one out.
>>
>> Niels
>>
>> On Sat, Jan 7, 2023 at 6:13 PM Niclas Hedhman <ni...@hedhman.org> wrote:
>>
>>>
>>> I am not sure what you are asking; But Modbus documentation addresses
>>> are (in my experience) always offset by 1, which I think goes back to
>>> the 1970s when we weren't yet sure whether numbers started at 0 or at 1,
>>> and number ranges were a scarce resource so we wouldn't sacrifice one
>>> for consistency... Those were the days.
>>>
>>> Niclas
>>>
>>> On 2023-01-07 17:27, Niels Basjes wrote:
>>>> Hi,
>>>>
>>>> I ran into the effect that all modbus addresses I specify are shifted
>>>> down
>>>> by 1.
>>>> This turns out to be code in most of the ModbusTag implementations (not
>>>> all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
>>>>
>>>> Why was this done?
>>>> If I'm writing software to read modbus addresses and I follow the
>>>> manufacturer specified information it will run into unexpected effects
>>>> (like I have today).
>>>>
>>>> Also: Now getAddressString() yields something different then what I
>>>> used
>>>> to build the tag with.
>>>
>>
>>
>> --
>> Best regards / Met vriendelijke groeten,
>>
>> Niels Basjes
>>
>
>
Re: Modbus addresses offset by 1?
Posted by Niels Basjes <Ni...@basjes.nl>.
Hi,
Understood. Keep the offset it is.
But why does the ModbusTagExtendedRegister not have this shift? Is that a
bug?
Also there are a few other problems I found in this area, one is that the
getAddressString will yield a different address (which is also unparsable).
I'll see if I can fix those.
Niels
On Sat, Jan 7, 2023 at 8:35 PM Christofer Dutz <ch...@c-ware.de>
wrote:
> Hi all,
>
> yeah … just wanted to say … this stupid offset is by spec that way.
> I would say, that in this case we have two different specs with a common
> base.
> Perhaps if we had “variants” of drivers, we could make different
> assumptions.
>
> If you get a Modbus TCP connection things are the way, they currently are.
> If you get a Modbus TCP connection with “SunSpec” variant, then you don’t
> have the offsets.
>
> Cause … if we undid the “+1” shift, then other would start complaining
> because their specked registers are in different locations.
>
> I have also heard that for EIP for full support of the PLC features, we’ll
> have to have Allen Bradley, Schneider Electric variants next to the default
> ones.
> I was told most of the major vendors do all the complex types and browsing
> differently :-(
>
> Chris
>
> From: Niels Basjes <Ni...@basjes.nl>
> Date: Saturday, 7. January 2023 at 18:31
> To: dev@plc4x.apache.org <de...@plc4x.apache.org>
> Subject: Re: Modbus addresses offset by 1?
> Ok, the "off by one" problem is ancient.
> Looking around the internet I read that in the past when register 123 was
> needed you would have to request for 122 over modbus.
>
> I find the transition from the logical number to the technical number a
> responsibility of the application, not of this library.
> Take for example the SunSpec (modbus standard for solar systems)
> specification it states:
>
> *6.1.1 **Modbus Address Location*
> *All Modbus device maps MUST be in the holding register address space.*
> *The beginning of the device Modbus map MUST be located at one of three
> Modbus addresses*
> *in the Modbus holding register address space: 0, 40000 (0x9C40) or 50000
> (0xC350). These*
> *Modbus addresses are the full 16-bit, 0-based addresses in the Modbus
> protocol messages.*
> *The first two Modbus registers at the start address MUST have the
> following well-known*
> *constant values as a marker: 0x5375, 0x6E53 (hexadecimal values of the
> ASCII string SunS).*
>
>
> On my solar inverter if I request 2 registers at 40000 I get the mentioned
> 0x5375, 0x6E53.
> Yet with plc4j/modbus I must specify address 40001 because the library
> shifts it by one.
> It took me the better part of today to figure that one out.
>
> Niels
>
> On Sat, Jan 7, 2023 at 6:13 PM Niclas Hedhman <ni...@hedhman.org> wrote:
>
> >
> > I am not sure what you are asking; But Modbus documentation addresses
> > are (in my experience) always offset by 1, which I think goes back to
> > the 1970s when we weren't yet sure whether numbers started at 0 or at 1,
> > and number ranges were a scarce resource so we wouldn't sacrifice one
> > for consistency... Those were the days.
> >
> > Niclas
> >
> > On 2023-01-07 17:27, Niels Basjes wrote:
> > > Hi,
> > >
> > > I ran into the effect that all modbus addresses I specify are shifted
> > > down
> > > by 1.
> > > This turns out to be code in most of the ModbusTag implementations (not
> > > all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
> > >
> > > Why was this done?
> > > If I'm writing software to read modbus addresses and I follow the
> > > manufacturer specified information it will run into unexpected effects
> > > (like I have today).
> > >
> > > Also: Now getAddressString() yields something different then what I
> > > used
> > > to build the tag with.
> >
>
>
> --
> Best regards / Met vriendelijke groeten,
>
> Niels Basjes
>
--
Best regards / Met vriendelijke groeten,
Niels Basjes
Re: Modbus addresses offset by 1?
Posted by Christofer Dutz <ch...@c-ware.de>.
Hi all,
yeah … just wanted to say … this stupid offset is by spec that way.
I would say, that in this case we have two different specs with a common base.
Perhaps if we had “variants” of drivers, we could make different assumptions.
If you get a Modbus TCP connection things are the way, they currently are.
If you get a Modbus TCP connection with “SunSpec” variant, then you don’t have the offsets.
Cause … if we undid the “+1” shift, then other would start complaining because their specked registers are in different locations.
I have also heard that for EIP for full support of the PLC features, we’ll have to have Allen Bradley, Schneider Electric variants next to the default ones.
I was told most of the major vendors do all the complex types and browsing differently :-(
Chris
From: Niels Basjes <Ni...@basjes.nl>
Date: Saturday, 7. January 2023 at 18:31
To: dev@plc4x.apache.org <de...@plc4x.apache.org>
Subject: Re: Modbus addresses offset by 1?
Ok, the "off by one" problem is ancient.
Looking around the internet I read that in the past when register 123 was
needed you would have to request for 122 over modbus.
I find the transition from the logical number to the technical number a
responsibility of the application, not of this library.
Take for example the SunSpec (modbus standard for solar systems)
specification it states:
*6.1.1 **Modbus Address Location*
*All Modbus device maps MUST be in the holding register address space.*
*The beginning of the device Modbus map MUST be located at one of three
Modbus addresses*
*in the Modbus holding register address space: 0, 40000 (0x9C40) or 50000
(0xC350). These*
*Modbus addresses are the full 16-bit, 0-based addresses in the Modbus
protocol messages.*
*The first two Modbus registers at the start address MUST have the
following well-known*
*constant values as a marker: 0x5375, 0x6E53 (hexadecimal values of the
ASCII string SunS).*
On my solar inverter if I request 2 registers at 40000 I get the mentioned
0x5375, 0x6E53.
Yet with plc4j/modbus I must specify address 40001 because the library
shifts it by one.
It took me the better part of today to figure that one out.
Niels
On Sat, Jan 7, 2023 at 6:13 PM Niclas Hedhman <ni...@hedhman.org> wrote:
>
> I am not sure what you are asking; But Modbus documentation addresses
> are (in my experience) always offset by 1, which I think goes back to
> the 1970s when we weren't yet sure whether numbers started at 0 or at 1,
> and number ranges were a scarce resource so we wouldn't sacrifice one
> for consistency... Those were the days.
>
> Niclas
>
> On 2023-01-07 17:27, Niels Basjes wrote:
> > Hi,
> >
> > I ran into the effect that all modbus addresses I specify are shifted
> > down
> > by 1.
> > This turns out to be code in most of the ModbusTag implementations (not
> > all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
> >
> > Why was this done?
> > If I'm writing software to read modbus addresses and I follow the
> > manufacturer specified information it will run into unexpected effects
> > (like I have today).
> >
> > Also: Now getAddressString() yields something different then what I
> > used
> > to build the tag with.
>
--
Best regards / Met vriendelijke groeten,
Niels Basjes
Re: Modbus addresses offset by 1?
Posted by Niels Basjes <Ni...@basjes.nl>.
Ok, the "off by one" problem is ancient.
Looking around the internet I read that in the past when register 123 was
needed you would have to request for 122 over modbus.
I find the transition from the logical number to the technical number a
responsibility of the application, not of this library.
Take for example the SunSpec (modbus standard for solar systems)
specification it states:
*6.1.1 **Modbus Address Location*
*All Modbus device maps MUST be in the holding register address space.*
*The beginning of the device Modbus map MUST be located at one of three
Modbus addresses*
*in the Modbus holding register address space: 0, 40000 (0x9C40) or 50000
(0xC350). These*
*Modbus addresses are the full 16-bit, 0-based addresses in the Modbus
protocol messages.*
*The first two Modbus registers at the start address MUST have the
following well-known*
*constant values as a marker: 0x5375, 0x6E53 (hexadecimal values of the
ASCII string SunS).*
On my solar inverter if I request 2 registers at 40000 I get the mentioned
0x5375, 0x6E53.
Yet with plc4j/modbus I must specify address 40001 because the library
shifts it by one.
It took me the better part of today to figure that one out.
Niels
On Sat, Jan 7, 2023 at 6:13 PM Niclas Hedhman <ni...@hedhman.org> wrote:
>
> I am not sure what you are asking; But Modbus documentation addresses
> are (in my experience) always offset by 1, which I think goes back to
> the 1970s when we weren't yet sure whether numbers started at 0 or at 1,
> and number ranges were a scarce resource so we wouldn't sacrifice one
> for consistency... Those were the days.
>
> Niclas
>
> On 2023-01-07 17:27, Niels Basjes wrote:
> > Hi,
> >
> > I ran into the effect that all modbus addresses I specify are shifted
> > down
> > by 1.
> > This turns out to be code in most of the ModbusTag implementations (not
> > all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
> >
> > Why was this done?
> > If I'm writing software to read modbus addresses and I follow the
> > manufacturer specified information it will run into unexpected effects
> > (like I have today).
> >
> > Also: Now getAddressString() yields something different then what I
> > used
> > to build the tag with.
>
--
Best regards / Met vriendelijke groeten,
Niels Basjes
Re: Modbus addresses offset by 1?
Posted by Niclas Hedhman <ni...@hedhman.org>.
I am not sure what you are asking; But Modbus documentation addresses
are (in my experience) always offset by 1, which I think goes back to
the 1970s when we weren't yet sure whether numbers started at 0 or at 1,
and number ranges were a scarce resource so we wouldn't sacrifice one
for consistency... Those were the days.
Niclas
On 2023-01-07 17:27, Niels Basjes wrote:
> Hi,
>
> I ran into the effect that all modbus addresses I specify are shifted
> down
> by 1.
> This turns out to be code in most of the ModbusTag implementations (not
> all) which have a PROTOCOL_ADDRESS_OFFSET to shift it by.
>
> Why was this done?
> If I'm writing software to read modbus addresses and I follow the
> manufacturer specified information it will run into unexpected effects
> (like I have today).
>
> Also: Now getAddressString() yields something different then what I
> used
> to build the tag with.