[Asterisk-Users] T1-PRI deployment questions...

[Steve Underwood]

Steven Critchfield wrote:

>On Thu, 2003-05-29 at 15:46, Steven Critchfield wrote:
>  
>
>>On Thu, 2003-05-29 at 15:06, Charles E. Youse wrote:
>>    
>>
>>>On 29 May 2003, Steven Critchfield wrote:
>>>
>>>      
>>>
>>>>On Thu, 2003-05-29 at 10:44, Charles E. Youse wrote:
>>>>        
>>>>
>>>>>B8ZS is required for PRI.  It's a digital service and can not handle the
>>>>>loss of data required for AMI.
>>>>>          
>>>>>
>>>>I wasn't aware that AMI lost data. AMI just inverts polarity on the line
>>>>for every other 1. B8ZS does the same thing but intentionally introduces
>>>>errors on the line to maintain 1's density. Neither one is lossy.
>>>>
>>>>        
>>>>
>>>AMI is lossy.  When the ones density in the signal is too low, AMI
>>>will insert ones to ensure that the far end does not lose sync.
>>>      
>>>
>>As I understand it, AMI is not lossy but will may cause problems due to
>>not maintaining 1's density. 1's density is used to make sure both sides
>>are synced up properly. B8ZS is AMI except that it introduces bipolar
>>violoations to make sure the line doesn't stay in an off state for too
>>long.
>>
>>So B8ZS intentionally throws errors on the line in a known manner so as
>>to make sure each side is in sync, but AMI does not care if the line
>>goes all 0's for a while.
>>
>>I still haven't been able to dig up any documentation to back up that
>>AMI is lossy, just maybe prone to errors via slips.
>>    
>>
>
>I still can't find any reference to AMI being lossy, and can't find any
>comments that show where a AMI circuit would introduce 1's to maintain
>1's density. After reading a page describing test patterns and why they
>use certain test patterns, it makes sense why AMI might not be usable
>for a PRI though. 
>http://www.electrodata.com/testpat.htm
>In a PRI, since the signalling is in the D channel, and the consecutive
>B channels could be completely clear, you could run into times with more
>than 15 consecutive zeros. Although I need to do more looking at how D4
>or ESF lays on top of a T1 signal. Anyways, with more than 15
>consectuives zeros you no longer are within ANSI spec.
>
>On a RBS circuit it would be less likely to fall too far out of spec
>using AMI.
>  
>
You are right, Steve. AMI isn't lossy. It stands for alternate mark 
inversion. It simply forces more transitions into the stream to ensure 
good sync at the receiver. With the way old T1s worked this was good 
enough to ensure sync., as the content of these T1s was always voice, 
and not completely arbitrary. With ISDN, or other data applications, 
which carry completely arbitrary content as well as voice, you really 
need a more robust sync scheme. This caused changes in both E1s and T1s. 
E1s got CRC4 to ensure robust frame sync. T1s got 8BZS to ensure robust 
bit sync.

An ISDN T1 *should* be using 8BZS, but isn't always. Similarly, an ISDN 
E1 *should* be using CRC4 framing, but some countries insist on the 
older flakier framing mechanism for some odd reason.

I think the confusion about lossy T1s probably relates to the robbed bit 
signalling that is generally used on non-ISDN T1s. This does, of course, 
lose a little of the 64kbps channel in a rather PITA way.

Regards,
Steve