[Asterisk-Users] Frustrated with echo...

[Rich Adamson]

Steve Jones wrote:
> I thought the whole thing with the hardware echo cancellation is that
> it was basically in liu of the equivilent echo cancellation done in
> software...  The reason to go to the hardware was for high-density
> systems??  For two FXOs, I thought I'd be safe in getting the
> non-echo cancellation cards, but perhaps no, huh?!  :-(

The echo cancellation issue is highly dependent on the exact pstn lines 
that you use.

The * software EC works well in lots of implementations, however it does 
have limits that seem to be directly related to the delay between the 
time data is "sent" verses when the reflected energy (echo) is received. 
The limit seems to be somewhere in the 30 to 35 millisecond range given 
the tests that I've conducted using various s/w tools. In very general 
terms, it seems the longer the pstn copper lines between asterisk and 
the Central Office, the more likely software EC will not be as usable or 
consistent as the hardware EC.

The hardware EC (from digium or sangoma) have wider limits, and those 
limits are different for the digium TDM2400 verses sangoma A200D. The 
difference between the two cards is related to the exact hardware 
chipset used on the two cards. (The two chipsets have very different 
investment/engineering costs.)

It should also be noted that many telephone companies have implemented 
various remote line concentrators (typically seen as relatively small 
steal boxes in the neighborhood) that can also have an impact on echo. 
There are no published guidelines or guesses that would suggest if the 
telco has implemented "A" then you must use EC "B".

Also note that whatever worked for cards and EC in one case is not at 
all indicative of what will work in another case as the pstn line 
construction is "always" different. (E.g., different length of copper 
cable, different gauge of cable, different methods of terminating the 
telco side of the pstn line, different quality of cables, different 
manufacturers and architectures of remote concentrators, etc, etc.)

In all cases, regardless of whether one is using hardware or software 
EC, the efficiency of the EC function is highly dependent on 
transmission levels.  If the transmission levels are set to high, echo 
is going to happen regardless of what card or EC is implemented. "That" 
seems to be an issue that many asterisk newbies (as well as lots of 
asterisk s/w developers) do not seem to understand.

So, if you were going to be selling asterisk boxes throughout your 
region, one might consider having an arsenal of analog products that can 
be selected based on "each" specific implementation. For short pstn 
lines, the TDM400 card with s/w EC seems to work well for lots of folks. 
For longer pstn lines where echo is not properly addressed in s/w, the 
TDM2400 or A200D seems to address the problem.  For long pstn lines and 
those that have somewhat unusual echo problems, the A200D seems to 
address more issues then what the TDM2400 does.

The above does not address analog fax support, which also enters into 
the engineering decision.

The choice is not necessarily one of supporting digium or not; its 
rather an engineering decision to select the product that addresses the 
technical issue, period. Unfortunately, there is no reasonable way for 
you (or anyone else) to know in advance which product is needed to 
address the issue. Anyone that tries to influence you otherwise is 
absolutely full of BS. That's based on 20+ years doing detailed 
engineering work (including pbx & transmission engineering) for a very 
large US telco, AND, been-there-done-that with asterisk over a two to 
three year period.