(Almost) finished my HA announcements!


Senior Member
Well, it only took 14 months of having the stuff in my parts closet, but I finally have zoned HA announcements including doorbell ring, CallerID, and TTS! Not completely done, but 80% of the way there.

I created 3 zones of HA announcement speakers using the Elk (intentionally not using my A/V speakers). Right now there's only the CQCServer as the source (ie, not the Elk M1 output1), but that does:
1) Doorbell ring (different per door) (ripped out regular doorchime, i'm now Elk-only)
2) Custom WAV playback on callerID. I recorded 4 of our friends so far, and play back their greeting. This is actually the AWESOME bit, as I used strictly CQC for it, meaning I can sell off the Elk-930 Doorbell detector, Elk-120 4channel Recordable board and Elk-129 computer sound interface. And, since I'm using CQC and WAV files, I have infinite file capability.
3) Custom TTS (text to speech). I'll probably get a custom voice, I really want to try out one of the nuance ones as I think my kids would get a kick out of a british female accent, but it's $90 and there's no demo. I know Kate is a tried&true voice, but it just seems so bland.

Some techie details for those of you who care:
- Doorbells are rigged up directly to Elk zones using CAT5, so they're very simple open/close circuits.
- I'm using an Elk-800 amp which can handle 2-16Ohms
- For each zone, I used a relay on the M1XOVR relay expander
- I mounted 3 of the 32Ohm Elk speakers in parallel so I get a 10Ohm resistance on that leg
- I mounted 1 Elk 8 Ohm External speaker and 1 Elk 8Ohm internal speaker on two seperate legs.
- I *will* be added either a 32Ohm single gang speaker or an 8Ohm regular speaker on leg#4, haven't quite decided which form factor I want.

I'd never run all of them together, but even if I did that's still 2.1->2.6Ohms total, which is above the minimum.

If you want to calculate it yourself, the formula for parallel resistance is
1/x = (1/R1) + (1/R2) + (1/R3) + ...
or, to put it simpler
X (resistance) = 1 / ( (1/R1) + (1/R2) + (1/R3) + ... )

Here's the revised overview screen. I know all the HA speakers don't look like that, but the ones my wife can see do, and one icon is simpler than 3.

Final piece is to step down the M1 output to line level and use the $50 mixer i got off eBay to blend those two, and sell off the Elk-930, Elk-120, and Elk-129. If I get decent $$, I may take a gamble on the Nuance voice.

Just get the Neospeech voices, they are SAPI compliant, works with most apps (including CQC, this is what I use), and they are very cheap. Martin sells them.
Curious what exactly your doorbell ring announcement sound like? Did you custom record or use the seeded lady's voice?
I use AT&T's Natural Voices. They are about $300 each though. Here is a link to Wizzard Software's stire that sells them.



I like AT&T voices but not at $300 !! I get them from NextUp.com for around $35. Even if you have to purchase their $20 TextAloud application you are still going to save money going this route.

Humm, looks like the $300 are for SDK versions. At $35 I may buy a couple more voices. Thanks Mike.