Adding voice feedback to an electronics product? (voice chip)

I have an electronics product application where I need voice feedback
to feature selections.
I am considering using a ISD2560 from Windbond's chipcorder series
http://www.winbond-usa.com/mambo/content/view/36/140/
They seem quite popular (based on web search). However, unless I use
their consulting services for them to pre-burn a large number of chips,
I have to manually speak each voice clip into a small electret
microphone for each unit. This is time consuming with questionable
quality from such a recording interface.
Are there any other voice chip alternatives? Ideally a voice chip that
interfaces to an external ROM or Flash chip. Of course some type of
software would need be provided that would allow me to compile voice
clips on my PC to compatable ROM files.
I am also incorporating an PIC18Fxxxx microprocessor into this product.
Perhaps someone has been able to use a PIC processor to emulate a voice
chip. I would appreciate any advice.
Thanks

 

You should be able to make a setup to record sound into the winbond
chips automatically, without using their setup.

But depending on how much flash you have, how much work you want to do,
and what voice quality you want you should also be able to store your
sounds digitally in flash and play them back. You may need to use an
external ADC.

You could use wav files, which I think are 16 bit with selectable
sampling rates. You should be able to sample at 8kHz for just voice,
which would imply about 16kB/sec of sound. That's about 64 seconds/MB,
which isn't too bad but isn't much to write home about. You'd need at
least a 12-bit DAC, and carefully conditioned sound (but you'll want
that anyway). This would be simple to implement in the PIC software;
you'd just be shuttling bits around.

You could use mu-law or a-law encoding, which will give you telephone
quality and cut your memory requirements in half. You can still get
mu-law codecs (at least I think you can), so you'd only have to shuttle
8 bits/sample.

You could use ADPCM (do a web search) at just about any level of
compression you were willing to tolerate, with up to 8 times smaller
files than mu-law (IIRC, it's been a while). This would load the PIC
much heavier, but I _think_ it'd be up to the task as long as you're not
asking it to do much else.

Modern cell phone style compression where you have a model of the vocal
tract is out unless you use something other than a PIC, and I suspect
it's beyond what you want to tackle anyway.

 

I read in sci.electronics.design that wrote (in
<>) about 'Adding
voice feedback to an electronics product? (voice chip)', on Fri, 7 Oct
2005:

You can record a whole bunch at once using any microphone you want and
an audio mixer or even an amplifier with a microphone input. Connect the
chips in parallel to the output of the mixer or amplifier thorough a
simple two-resistor attenuating pad. You need only a few millivolts for
recording by your mixer or amplifier will produce much more.

You can get quite good electret mics at low prices.

 

Regarding Tim's comments

Motorola SPS (now Freescale) was the usual source for ADPCM codecs, but
they seem to be getting out of the business, which is quite a blow as
everyone seemed to use their parts.
I recently went through this as I am just bringing some new products
into production that include voice (and voice prompt feedback).

Voice codecs using mu-law / A-law / 2s complement are available from a
number of sources, including TI. Typical cost is about $2 - $3 in 1k
qtys.

There are software packages around that permit you to use a PC to
record sound to virtually any format you wish, including mu/A law and
2s complement formats.

Driving a codec is fairly simple using a PIC that's fast enough. The
key is to meet the bit timing requirements (typically 64kb/s minimum
for mu/A law encoded data) which should not be beyond a PIC.

As noted, these formats are all 8k frame rate. The codecs can be set to
generate the frame pulse so you can sync a processor up easily.

To deal with .wav files you would need to use a device capable of AC97
(such as the Philips 14000 series or the Wolfson WM9712 for example)
which requires a separate AC97 controller - that would probably be
beyond what you need, and probably beyond your budget.

For what it's worth, I ended up using the TI TWL1103T-Q1 (because I
also have to run those prompts across a bluetooth link) for the codec.

Just my $0.02

Cheers

PeteS

 

Also their text to speech chips send the message as asci text
http://www.winbond.com/e-winbondhtm/partner/b_2_a_5.htm

Use in
http://www.robot-electronics.co.uk/shop/Speech_Synthesizer_SP032006.htm
http://www.acroname.com/robotics/info/examples/sp03-1/sp03-1.html

Alex

 

You could always try: http://www.romanblack.com/picsound.htm

I've messed around with it enough to produce sounds, but never tried to
do voice.

You'd need an external eeprom, since I doubt you'd get to hold much
speech in the 18fxxxx, but if you have 4 or 5 pins open it should work.

Mark

 

Uh ?

Mask ROM and 2.4 secs playback ? Not much good for anything I'd have thought.

Graham

 

Take a look at

http://www.elan-europe.com/products/media/device/index.asp?Subline=Multimedia

With the development environment for these EM55xxx chips, you can
convert wav files and program them into the chip.

Volume and playback frequency are also adjustable.


Cheers Markus

 

The very first product my company ever did was a natural
voice speach playback board for LLNL. It was a Multibus
form factor board with an Oki decoder, a homebrew 3-4 watt
audio amp, 24 sockets for 27C256 eproms, and a couple
dozen TTL chips. They used the boards in their security
system to prompt users as they passed through access points.

I think it would hold 2 or 3 minutes of speech, addressable
at 1/2 second intervals.

We also had to design a capture board for a PC and write
the capture software. My next-door neighbor did the
narration.

I've been looking at doing an updated version using MP3
files and possibly an ARM chip doing software decoding.
The current batch of MP3 decoder chips are all too
expensive for hardware decoding. I'd want to have the
files storeable on a compactflash with a FAT filesystem
so they'd be trivial to transfer from a PC to the play-
back board.

 

In our design we use the EM55M450 Module with OTP. It's o.k. for our 12
sec gong.

Markus

 

ISD had once the ISD-t360

Nice chip i used in a project with DTMF en/de coding and other PSTN stuff
(tone detection for call progress ...)

External flash up to 64Mb for up to 60 min of sound (8000samples / s )

Winbond took over the ISD business and stopped the production of this nice
product , shame on Winbond ! )

I never found a replacement chip that had all of this in a single chip
solution.