Is HW based E/C essential for all analogue deployments?

I have seen several posts recently that state hardware based echo cancellation is essential for analogue cards in all deployment scenarios. I disagree and would be grateful for other peoples opinions based on their own real world deployment experiences.

Earlier versions of Trixbox shipped with various software based echo cancellation algorithms that provided mixed results. The latest version of Trixbox ships with the OSLEC echo canceller enabled by default, which has received positive feedback from a large number of users.

There are also commercial SW E/C algorithms that are G.168 compliant and support tail lengths up to 128ms / 1024 taps. These can be purchased for around $10 per channel (or are available for free direct from one card manufacturer) and are auto-tuning so there is no need to mess around with things like ztmonitor.

SW based E/C requires CPU resource and therefore is not suitable for servers that need to support large numbers of channels. One commercial SW based E/C provider recommends using a 3.0 GHz Pentium 4 for 8 channels at 1024 taps. Another states that their echo canceller will support up to 16 channels at 1024 taps and provides some benchmark performance results for a few common processors.

My experience of deploying analogue cards that don’t have HW E/C along with the OSLEC echo canceller has been very positive, with customers being very happy with the voice quality. There have been a few occasions when a commercial echo canceller has needed to be used to eliminate echo completely. However, I have never been involved with an install that has had an echo issue that needed H/W based echo cancellation to correct, and we have never had a card returned due to echo issues.

There are scenarios where I agree that software based echo cancellation is not suitable, e.g. for PRI lines where using software based echo cancellation would have an adverse effect on the processor due to the number of channels. Another example where I think software based echo cancellation is not suitable is for pre-configured IP PBX appliances that require plug-and-play card upgrades without needing to install additional licences/software.

When compared with SW based E/C, HW based E/C offers a number of advantages including:

1. No need to install/configure extra software/licences (reduces time/complexity for new installs and upgrades)
2. Eliminates risk of callouts due to echo issues (which reduces customer support costs)
3. Guaranteed to eliminate echo from day 1 onwards (provider customer and installer confidence)

In my view 2+3 are only applicable for open source echo cancellers because in my experience they don’t eliminate echo in all cases, whereas the commercial SW canceller I have used does. However, this opinion is only based on my own experience and I would be interested in hearing if other uses have had experiences where they have tried a commercial SW based E/C algorithm and have been unable to overcome echo issues and have had to use a card that provides HW based E/C.

When compared with HW based E/C, SW based E/C offers just one advantage – cost. If the cards cost the same there would be nothing to discuss. However, cards that do not provide HW based E/C can be purchased for less than half the price of those that do. When considering the total cost of a new phone system install, the extra cost of the HW based E/C may be insignificant in many cases, however, when considering the cost of installing a new card in an existing system, the extra cost becomes more apparent.

In today’s global economic climate more businesses are looking at cutting costs and are counting every $/£/€. Here in the UK hardware costs have risen by about 35% in the last 6 months due to the falling value of the £GBP. Also, in some countries the cost of hardware can be comparable or even more expensive than the cost of installing it, in which case minimising hardware costs where possible very important.

In my view using SW based E/C instead of HW based E/C offers a viable cost effective solution for deployments with low numbers of channels (e.g.

Looking at things from another viewpoint, I know of a number of phone systems that have been running for years with just the standard open source echo cancellers and the customers have never had an echo issue. In my view the extra cost of providing HW based E/C for these customers would not have been justified, as they would be paying extra for a solution to a problem that does not exist.

However, there are other people that based on their own experiences recommend always using HW based E/C, even for 1-2 channels.

If anyone has any thoughts/views on HW vs. SW based E/C based on their own experiences in deploying phones systems then I would be grateful for your feedback.