Voice annunciated test box circuitry

256leds

one

pin

so

and

im

Hi, Mirach. You're definitely heading in the right direction.

I've done things like this before. In this kind of work environment, you don't want to have some kind of contraption no one else can understand in exchange for possibly saving a few bucks. The KISS principle (Keep It Safe and Simple) applies here. The most important thing is reliability, and next that your test fixture will be easy to document and calibrate/test.

If you've got some bucks to spend, nuke the problem by buying a Measurement Computing CIO-DI-192 board ($299 USD, providing 192 digital inputs to a PC through an ISA slot), and a CIO-DI-96 board ($179 USD for 96 inputs). Get some 1K SIPs to use as pullups, and solder them on the boards before installing them in the closest junker PC with sound card. Then wire it up and write a program in VB to scan the lines and output the audio you want to the speakers.

Something which might give you some flexibility in the cost/hassle tradeoff without adding enormous complexity is to get a CIO-DI48 ($119) and a CIO-DO48H($179) and use shottky diodes to make a keyboard reading matrix. You'll have to put the diode matrix in a blivet box outside the PC.

I would recommend against using the 8255-based DIO-48 for this purpose, even though it would be a lot cheaper ($119 USD covers the whole thing

-- you can program one 8255 to be all inputs, one to be all outputs, and you're done) because the I/O are going to be in the "real world". Even with proper ESD precautions, the 82C55 is trivially easy to hang up, and you don't want to make a test fixture that's going to make you look bad because it always messes up. But no matter what route you use, you really should use a PC for this test.

I hope this has been of help. Simple test fixtures like this can really help make you look good. If there's a chance of overrunning outputs, make sure to place a beep sound between outputs to avoid chatter. Remember to spend enough time on the user interface to make it user friendly (and don't forget to place your name under your supervisor's name prominently on the splash screen ;-) ).

Chris

I'm sure you'll need to create a circuit to encode all the input lines into an 8 bit binary number first. From there, you can interconnect it to a PC so that the speech processor, in WinXP can read the decimal equivalent to you.

A program should be running that reads the port for input then clears and writes the decimal equivalent to the screen. The Windows accessibility tools include a audio screen reader that I have only played with at times. (Microsoft Sam sounds like the obsolete SC-01 from Federal Screw Works)

You can buy digital voice recorder chips from windbond

that allow you to record some amount of voice. You can cue the voice at any point. Thus, you could record yourself saying 1,2,3.. and then cause the chip to replay the individual digits on command. You control it using SPI from a microcontroller. They go up to 240s of recording time, so you could conceivably record all 256 numbers, but individual digits are probably easier. You can buy them at digikey or futurlec. I'm sure you can also get pre-recorded voices, but I didn't spend the time searching.

As to encoding the data, you could use individual multiplexer chips. However, you could also get 8 octal buffers with high impedance outputs. Tie each input to a correponding buffer pin, and tie the outputs together into a bus, all bit zeros together, ones together, etc. Then, select the buffer chips one at a time, and read the bits for that buffer in parallel using 8 pins on your microcontroller. Once you have a hit, just program the SPI device to output the proper voice codes. 16 bits on the final device, 8 for input, 8 for output. Plus, of course, the 3 or 4 bits for SPI to control the recorder.

The total part count for this solution would be the voice chip, the microcontroller, the buffers, and such. You could run it off of a 9V battery, or use a set of rechargable batteries, and build a recharger unit for it. Alternately, if it didn't need to be portable, you could run it off of a wall wart. Total cost would be around $30 for parts, I think, including PCB and enclosure, depending on the cost of the plug. If you were to mass-produce these, you could probably make them for $10 each.

I'll work up an estimate for 10 if you want... ;)

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

Meh. Multiplication. You'd need 32 buffers, and 32 pins to select them. You need a pin per input. There are probably better ways to do this...

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

Thanks guys for the great responses . I could actually get by at present with 128 individual numbers. (we currently arnt working on the units with 256 wires). Some of the involved electronics is beyond my scope of knowlege, so im not against paying someone to put a couple of these together if a price could be worked out that we both could live with. I could even send you the empty boxes probably 8"Wide by 6" high by 2" deep. and they would probably be assembled without the plug (we can wire the actual plugs) so basically its making the boards and chips work then I can take it from there. I like the idea of a portable device ,this is the way our current light boxes are and it would be a plus to be portable. I had originally wondered if you might be able to get 128 small cheap 1 or 2 second voice chips that could each be triggered by a voltage when that wire is touched . But I may be dreaming on that one :) . I was originally thinking of these chips replacing the LEDs currently in the boxes. That may be too simplistic of an idea since im not sure whats avialiable. Let me know what you think guys. And again Thank You

The voice chips are fairly expensive, so I think getting a slew of them won't be practical.

I spent some time attempting to redeem myself by coming up with a simple way to input your 256 signals.

One simple way would be to use a 16x16 grid of NPN transistors, arranged in rows and columns. The collectors of all the transistors in a particular row would be wired together. All of the bases in a column would be connected together, in such a way that all the bases in a column could be driven by raising a single pin for that column.

The plug pin you want to identify is then grounded. The others are left to float.

Now, by powering the columns one by one (through a 1k resistor), you can find the grounded plug pin by noticing which row goes to 0, using the column number as the low address byte, and the row number as the high address byte. If no emitter is grounded, the row will be near logic 1 due to the diode action of the collector.

You can walk through the rows by asserting them one at a time, and then reading the result with a microcontroller. Once you find the row that is low, you know the row and column address, and can thus tell the voice chip which digits to say.

The voice chip will be expensive to run in terms of power, particularly if you want volume. However, using an earplug would make it a lower power device, thus allowing for the ability to carry it around.

How big are the plugs?

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

Robert

The plugs are large . The one im looking at is 4"X 1.5" and has 260 pins in it. This is pretty typical of these plugs,just a little different configurations etc. Thanks

Robert I tried to send you a reply to your email but it came back as un-deliverable. Send me another email and an address that i can send to and I'll send you some info.

Thanks

--
You will need a microcontroller with 16 bits of IO and an IRQ input,
seventeen 4 line to 16 line decoders, two 8 bit serial-in parallel-out
shift registers, four 4-bit programmable counters, a 512K Eprom with
the words 'zero', 'one', 'two', 'three', 'four', 'five', 'six',
'seven', 'eight', 'nine' burned into it, an 8-bit parallel in DAC, an
audio amp, a loudspeaker, and a test probe.

and a partridge in a pair tree.

Actually, those winbond SPI voice chips are very cheap, and are addressible for both writing and reading. They have onboard flash too. This cuts the complexity quite a bit. Once the voice is programmed, you can copy the data from chip to chip. This replaces the DAC, eeprom, and all the addressing logic. Just send an SPI command, and it says the digit. I think the speaker and amp isn't required, if they use headphones. The winbond chips will drive them directly, I believe.

One could also use one of those phoneme chips (remember "macintalk"?). They sound horrible though, unless you take great care in building the phoneme string and timing the output.

The horror is going to be the 256 wires to the plug. Hopefully there is a reasonable header connection that can be built.

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

--
Got a part number and a price?

The ISD400x looks promising. It's $7 at futurlec

I'm sure you can get it far cheaper in volume, although fururlec tends to have alarmingly good prices for certain things. I'm guessing one could sample a few as well. It's available it in PDIP, which is nice for quick prototypes.

I reread the datasheet last night. It does have an internal amplifier that can be used for a headset. The datasheet has a schematic for a small system using it.

I was hoping they have a version with a prerecorded 'voice model' doing digits, but, sadly, they don't appear to.

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

Where is the DAC? Are you planning on using the EEPROM to bitbang a filter to get the right sounds? I've seen tricks like that for the PIC.

Also, you don't include the tools for creating the sounds and burning them into the eeprom. He only wants a few of them. The winbond chip has a facility to do it itself.

You could probably clock the counters using an output from the uC. The PIC chips, have the option of outputting the instruction clock on one of their pins. However, I'm guessing that you would need to divide down the output somewhat.

One advantage of the SPI interface for the winbond chip is that it's probably less than a day of work to get working for both record and send, and has a facility to cue at any point for either recording or playback. I noticed you are using a counter to clock the eeprom. How are you going to start at particular digit values? I suppose you could preset the counters using microprocessor pins, but that means you need extra uC pins and traces for that.

Maybe outputting a couple of high address pins from the microprocessor would work. That way, you could divide the address space up into 16 chunks, and easily clock through any of them. All you would need is a reset line for the counters, and a clock source.

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

--
2ea HC164 @ 0.50........ $1.00
4ea HC163 @ 0.50........ $2.00
1ea 27C256.............. $1.50
8ea 1% 1/8W R........... $0.25
1ea LM386............... $0.75
                         -----
Total, Digi-Key......... $5.50 

Approximately...

+------+ | Q7|----[128R]--+ | | | | Q6|----[64R]---+ | | | | Q5|----[32R]---+ | | | | Q4|----[16R]---+ | | | | Q3|----[8R]----+ | | | | Q2|----[4R]----+ | | | | Q1|----[2R]----+ | | | | Q0|----[R]-----+---|+\ +------+ | >--+-- 27C256 +---|-/ | | | +--[R1]---+ | [R2] | GND

--
The tools for creating the sounds are a microphone, a PC sound card,
and whatever editing software came with the card or with Windows.  I
already have an EPROM programmer but if the OP doesn't, oh, well...
He'll have to work out all the addresses and do a lot of bit-banging
on the SPI interface to set up the addresses to do the recording, so
he'll have to, essentially, build a programmer and program each chip
in real time unless he buys a Winbond programmer.

Your resistors are set up for Q0 outputting MSB. Do Windows sound programs save bits this way? I've seen gang programmers can be set up to reverse the order of bits.

Do you know what kind of fidelity you'll get with that setup? Not that it matters much for digit output, but you want it to be intelligible.

Or, he could use a simple onboard programmer, and perhaps a simple uC program to copy from one device to another using the analog ports. I don't think you can gang-program the winbond chip, unfortunately, and it doesn't look like the SPI interface allows programming the flash. It really is designed to be a voice recorder playback toy.

You could save the shift registers by using 4 bits to set the high address bits, dividing your eeprom up into 16 equal sections. Program the digits at those sections, set those bits, and perform a reset when you want to start playing a section. It's a couple more uC pins, and it's less flexible, obviously, but it saves the shift registers. That's a buck!

You can get the ISD4002 winbond chips from digikey for $5.08 each if you buy 25.

You can get the 20 second verion (which is probably enough for 10 digits) for $3.27, although it may be more difficult to cue the output, since it doesn't have an SPI interface.

There are also dedicated sound output chips that use external eeprom, similar to your scheme, but they seem designed for higher end systems.

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

This last bit is in error. I was looking at the european digikey catalog by mistake (google took me there.) Those prices are in EUROs.

--
Regards,
   Robert Monsen

"Your Highness, I have no need of this hypothesis."
     - Pierre Laplace (1749-1827), to Napoleon,
        on why his works on celestial mechanics make no mention of God.

--
It's an EPROM, so the bits can be weighted at the discretion of the
designer.

--
An R-2R ladder would require 17 resistors, so as a discrete
substitution it would be more than twice as expensive.  Eight 5%
resistors (plenty good for this application) cost less than a quarter,
lots cheaper than a DAC-08.

Plus, take a look at

http://www.analog.com/UploadedFiles/Application_Notes/72999933AB6.pdf