Speech ROM and Other Components

[Coeus]

They're documented at https://en.wikipedia.org/wiki/User:Lord ... DULE_STUFF

Thanks. So the US ROM that had a different word set is described there as the industrial vocabulary which matches my description. This one does seem to contain an index in the style used by Acorn. The next three seem to lack this index but I have been able to index them by simply starting at address zero and letting VSP parse until it gets a STOP frame then pick up the new address. In the following the addresses are in hex. A ? means I could not make out what the word was so maybe someone else can have a listen to fill in the blanks.

VM61003 (Time and Weather Demo)

Code: Select all

         0 Zero
        4D One
        90 Two
        C7 Three
       113 Four
       15D Five
       1B0 Six
       203 Seven
       261 Eight
       2A4 Nine
       336 Ten
       389 Eleven
       414 Twelve
       45D Thirteen
       4E6 Fourteen
       584 Fifteen
       60B Sixteen
       698 Seventeen
       73B Eighteen
       7AB Nineteen
       845 Twenty
       891 Thirty
       8E9 Forty
       956 Fifty
       9C1 Sixty
       A2D Seventy
       AB0 Eighty
       B02 Ninety
       B6C Hundred
       BC3 Thousand
       C72 Million
       CEC Half
       D4F Negative
       DE6 Point
       E33 Affirmative
       EE7 Timed
       F69 Divide
       FD9 by
      1047 Hours
      10AC Plus
      10F1 Minus
      116D Equals
      11DB Red
      1254 Yellow
      12A6 Green
      1325 White
      136B the
      139E the
      13BF Answer
      142A Is
      1482 start
      14E4 Stop
      153D Ready
      157D Temperature
      1610 Time
      167B AM
      172D PM
      17DF O'clock
      1847 Degrees
      18FE Celsius
      19AC Fahrenheit
      1A4B Error
      1AA7 ?
      1B2F ?
      1B74 hour
      1BDB minutes
      1C51 Adjusting to
      1D06 Just
      1D5E East
      1DC4 West
      1E30 South
      1E91 North
      1EFE North-East
      1FB7 North-West
      2061 South-East
      210D South-West
      21C1 Hail
      222B Tornado
      22C2 Wind
      2340 Weather
      23BD Showers
      2448 Sleep
      24AF Smoke
      2520 Storm
      25AC Snow
      2627 Sand
      26B6 TELEX
      2718 Rain
      2788 Ice
      27E6 Feet
      282E Fog
      289D Heavy
      28DF Land
      294F in
      298F On
      29CB Variable
      2A6A Obscured
      2B09 Broken
      2B68 aloft
      2BDF Ceiling
      2C48 Partially
      2CC2 Visibility
      2D7A Drizzle
      2DDD Indicated
      2E74 Severe
      2ED7 Moderate
      2F59 Greenwich
      2FCA Mean
      3043 List
      30C0 Estimated
      314D Decreasing
      31F0 Moving
      325C Pressure
      32C9 Alternate
      334F Clear
      33AA Current
      33F4 Blowing
      3470 At
      34AA Turbulence
      353B Thinly
      3599 Increasing
      3634 More than
      36DB Less than
      3763 Freezing
      37EA Air
      382A Below
      38A6 And
      38ED ?
      3927 Haze
      39A5 ?
      3A0D Scattered
      3A8E Increasing to
      3B6A in
      3BB7 low
      3BE7 of
      3C2E ?
      3C37 Thunderstorms
      3CFB Overcast
      3DC7 Unlimited
      3E7B Low
      3EF4 Percent
      3F46 Over
      3F91 ?

VM61004 (Military Demo)

Code: Select all

         0 alpha
        64 bravo
        BC charlie
       11E delta
       171 echo
       1B3 foxtrot
       248 ought
       280 hotel
       2F6 india
       356 juliet
       3BC kilo
       409 lima
       44C mike
       48E november
       52A oscar
       587 papa
       5C9 quebec
       610 romeo
       670 sierra
       706 tango
       77D uniform
       7FB victor
       86E whisky
       8C7 x-ray
       93B zulu
       984 affirmative
       A38 negative
       ACF mayday
       B74 warning
       C07 icing
       C78 danger
       D0B emergency
       DD2 roger
       E50 hertz
       E8D ?
       E96 security
       F13 target
       F6E vectors
       FE6 light
      103D front
      1098 way
      10DA glide
      115E open
      11C2 lights
      1221 on
      1287 gun dish
      130A fireman
      13B9 select
      1420 filed
      14A5 MIG
      1511 alert
      1579 zone
      15FD terminal
      167E cancel
      16EB radios
      178E speed
      1800 knots
      1868 expect
      18D6 action
      193C radio
      19AA power
      1A07 cat
      1A5F information
      1B09 true
      1B52 pressure
      1BBF check
      1BF6 decrease
      1C63 advise
      1CED you
      1D27 ?
      1D7E error
      1DDA ?
      1E31 wrong
      1E91 no
      1EEA immediately
      1FA3 final
      200F please
      205A start
      20B6 course
      2102 RADAR
      2181 plan
      21EE target
      225B windows
      22F3 watch
      2336 use
      2381 turn
      23CE traffic
      2435 tank
      248A slow
      24F9 release
      256D jeep
      25A8 ignition
      2639 degree
      26A4 drain
      26F9 repair
      277C cancel
      2806 verify
      28A0 ready
      28FA use
      2976 out
      29BF other
      2A25 oil
      2A78 off
      2AC6 near
      2B26 rate
      2B93 mixture
      2C16 much
      2C6E rate
      2CD0 is
      2D28 caution
      2D97 below
      2E13 cylinder
      2E90 contact
      2F0A and
      2F51 fuel
      2FA8 for
      2FE2 sequence
      3068 side
      30C3 slow
      3119 to
      3148 rich
      31B6 pumps
      320C level
      3273 lower
      32E5 rear
      3364 old
      33C9 long
      3435 level
      34AE left
      350F right
      356F open
      35C9 spray
      3659 ?
      36E2 evacuate
      379A failure
      3807 service
      3881 abort
      3902 identify
      3999 too low
      3A2F centre
      3AA9 area
      3B09 base
      3B45 control
      3BBB measured
      3C41 the
      3C74 the
      3C95 vacuum
      3D13 ?

VM61005 (Aviation Demo)

Code: Select all

         0 APU
        CC FSS
       1A8 ILS
       276 IFR
       34B VFR
       420 VHE
       52F landing gear
       60C level off
       6CE under-carriage
       791 spoilers
       83F air brakes
       8DB aerial
       960 engine
       9E4 flame out
       A70 flap load
       B36 inbound
       C08 ignite
       C7A no turn
       D47 flaps
       DB3 evacuation
       E8F doors
       F13 cabin
       F7F arrival
      1018 acknowledge
      10C2 raise
      1156 list
      11C2 approach
      123F departure
      12C5 clearance delivery
      13BB fuel
      142B idle
      1495 increase
      150D up
      154E taxi
      15BB as
      15F6 converging
      1690 above
      16F6 brake
      174E calm
      17C6 braking
      1837 call
      189E cross-wind
      1938 cyrstals
      19A6 cylinder
      1A32 cycle
      1A99 down
      1AF3 ground
      1B4B full
      1B9A new
      1C19 leg
      1C8B maintain
      1D37 landing
      1DC2 lean
      1E2E set
      1E75 short
      1ECB wake
      1F28 <silence>
      1F31 ?
      1F86 ?
      1FF9 RVRs
      20ED squawking
      2161 magnetos
      220E stabiliser
      22C7 ?
      2344 selcow
      23DF vortex
      247F VOR
      256A marker
      25DD heading
      263E ETA
      26F4 downwind
      2789 ceiling
      27F2 mid-point
      288B cygnet
      2906 roll-out
      29A4 ?
      29FD aircraft
      2A89 altimeter
      2B20 runway
      2BAB auto-pilot
      2C47 localiser
      2CFE cow
      2D5D in-flight
      2DF6 over sea
      2EA7 departure
      2F3B lunch
      2F9F tar
      3006 height watch
      30AC stall
      312B touchdown
      31BF squawk
      321D elevation
      32D0 climb
      3342 bank
      33A3 accelerated
      344E trin
      34A0 ?
      3546 slope
      359F niner
      361E galley
      3675 freedom
      36F0 flight
      3749 degree
      37B4 airport
      3820 airspeed
      38AE clearance
      3920 altitude
      39BB radio
      3A53 remarked
      3AEA refuelling
      3B9C outer
      3BF9 middle
      3C5F inner
      3CAC instruments
      3D55 flight
      3DB0 approaches
      3E55 here
      3EA8 boost
      3EE7 telephone
      3F79 ?
      3FEE alpha

The following program would be useful to check individual words.

Code: Select all

   10REM > ByAddr
   20INPUT"ROM Number",R%
   30R%=R%*16384
   40REPEAT
   50INPUT"Address",A$
   60A%=EVAL("&"+A$)
   70PROCsetaddr(R% OR A%)
   80PROCsend(&50)
   90UNTILFALSE
  100DEFPROCsend(Y%)
  110LOCAL A%
  120A%=&9F
  130CALL &FFF4
  140ENDPROC
  150DEFPROCsetaddr(A%)
  160FORI%=0 TO 4
  170PROCsend(&40 OR (A% AND &F))
  180A%=A%DIV16
  190NEXT
  200ENDPROC

[arg]

Thanks. So the US ROM that had a different word set is described there as the industrial vocabulary which matches my description.

The "Industrial vocabulary" ROM was the one that TI supplied as the standard companion to the speech chip - as a demo and potentially for product use if it the application didn't need a dedicated ROM. Acorn used it for developing the OS support for speech, in advance of the arrival of the Kenneth Kendal ROM.

The chart seems to have the dates the wrong way round: the industrial vocabulary ROM had definitely arrived at Acorn very early in 1982 so it must have been produced in 1981 or earlier; conversely the KK ROM didn't arrive until late in '82.

[arg]

As a further note on the speech upgrade I wonder if a mistake was made in constructing the Phrase ROM. There is a command, Read and Branch which, as far as I can tell, reads an address from the ROM and then updates the read pointer to the address just read. That would seem to be an ideal thing to use when wanting to speak a word by index number.

The OS 1.10 release notes (which I've just posted over here) imply that it used to be done that way but got changed between 1.00 and 1.10 due to an "anomaly" in the TMS5220.

[Dave C]

A very interesting thread about the history of the speech upgrade. I am curious as to what the original ashtray fitting and connections looked like.
Dave C

[HHH]

A very interesting thread about the history of the speech upgrade. I am curious as to what the original ashtray fitting and connections looked like.
Dave C

Somewhat belatedly looking at the date on your post but this:

Further to the subject of the speech extension take a look at this:

viewtopic.php?p=374271#p374271

[Dave C]

Thanks,
That looks neat. What was to plug into them?

[Pernod]

Thanks,
That looks neat. What was to plug into them?

Nothing official, but maybe this.

[HHH]

That board connects to the header in place of the second narrower ribbon on the left of they keyboard, I'm not sure anything was ever developed to connect to the blue slot in the ashtray?

[Pernod]

That board connects to the header in place of the second narrower ribbon on the left of they keyboard, I'm not sure anything was ever developed to connect to the blue slot in the ashtray?

Ah, so connects to PL14 and sits internally, good to know.

[HHH]

Hi, Yes,

Gave it a try and the roms on board didn't seem to do much, but when I inserted a TMS6100 into it in the first position and shifted the others up a slot, it did as it is supposed to do. (With a TMS5220 in it's normal position of course) I think I've seen a post somewhere else that mentions the Vnnnnnn rom contents aren't in a format that the Beeb understands.

H

[Pernod]

Gave it a try and the roms on board didn't seem to do much

How did you try using them?

The usual Kenneth Kendal is accessed with SOUND -1, and I believe the Industrial one is -16, so presumably the others are something between -1 and -15.

[Dave C]

Thanks for all the great replies

[HHH]

Gave it a try and the roms on board didn't seem to do much

How did you try using them?

The usual Kenneth Kendal is accessed with SOUND -1, and I believe the Industrial one is -15, so presumably the others are something between -1 and -15.

Was a while ago now, can't remember, will give it another go using your info above, thanks.

[Pernod]

Anyone know what determines whether a speech PHROM responds to SOUND -1 or SOUND -15 ?

I know the TMS5220 can address upto 16 x TMS6100 but what determines which is selected to respond? Is it something that can be identified from a PHROM dump, or something hardwired into the TMS6100?

[tricky]

I think it is in the patent and I'm 99% sure that all the ROMs can be wired in parallel suggesting that it would be in the rom chip but I don't know about the data.
Sorry, I've said nothing

[Pernod]

I think it is in the patent and I'm 99% sure that all the ROMs can be wired in parallel suggesting that it would be in the rom chip but I don't know about the data.
Sorry, I've said nothing

You did prompt me to re-read the TMS6100 datasheet, which says:

The internal chip select can be programmed as one of 16 values. The value for this code appears in the gate placement deck as "BANK". The value specified here is a hexadecimal number between 0 and F. This corresponds to a decimal number between 0 and 15.

Sounds like it's programmed into the TMS6100, and not known from the PHROM data.

So that would mean PHROM_A is bank 15, and PHROM_US is bank 0. Wonder which banks the other VM6100x PHROMs respond to.

[tricky]

I have a valve memory that sprow might list it with his 6100 emulator but it was a long time ago.

[BeebMaster]

It does seem like the PhROM number is programmed into the serial ROM chip separately to the data; I can't see anything (at least in the Acorn PhROM format, but I believe that's specific to Acorn, and in itself unique, because there was only one Acorn PhROM).

What do we think, or know, about how the chips were made? Was the 16K of data sent to Texas for programming into the chips and they allocated a PhROM number at the same time? 15 for PhROM A is a very convenient number, as it used for selection in the low nibble of the 16-bit speech command, making the whole 16 bit number &FFFF, more commonly written in the context of the speech system as -1. This made it very convenient to use the command code &FFFF or -1 to say a Kenneth Kendall word; I don't think it would have been by chance that Acorn's PhROM was given number 15.

[arg]

What do we think, or know, about how the chips were made? Was the 16K of data sent to Texas for programming into the chips and they allocated a PhROM number at the same time? 15 for PhROM A is a very convenient number, as it used for selection in the low nibble of the 16-bit speech command, making the whole 16 bit number &FFFF, more commonly written in the context of the speech system as -1. This made it very convenient to use the command code &FFFF or -1 to say a Kenneth Kendall word; I don't think it would have been by chance that Acorn's PhROM was given number 15.

Normal procedure for things like the OS ROM was to send EPROMs in the post to the manufacturer (multiple identical sets to guard against errors). I think they then sent us back something so we could know that it had been interpreted correctly. Floppies weren't universal enough to be used as an interchange format. The changes from OS1.1 to OS1.2 were sent as edit instructions by Telex shortly after sending the 1.1 EPROMs; I can't remember how we ensured that the edits had been done correctly, but as I say I think something came back to us for checking in the normal case.

For the speech ROMs, TI supplied (loaned, I think) an emulator that took multiple EPROMs for the speech data - probably 2732s, though I remember the emulator being rather large. Maybe it emulated more than one ROM at once? I'm guessing that the ROM number was set in jumpers on the emulator, though I don't remember that detail. Again, I would expect that the EPROMs would have been tested in the emulator and sent to TI along with paperwork for the other masked settings like the ROM number, but I don't know that for sure.

[Pernod]

For the speech ROMs, TI supplied (loaned, I think) an emulator that took multiple EPROMs for the speech data - probably 2732s, though I remember the emulator being rather large. Maybe it emulated more than one ROM at once? I'm guessing that the ROM number was set in jumpers on the emulator, though I don't remember that detail. Again, I would expect that the EPROMs would have been tested in the emulator and sent to TI along with paperwork for the other masked settings like the ROM number, but I don't know that for sure.

Here's the PHROM Emulator, with dipswitch as expected:
https://www.computinghistory.org.uk/det ... tor-Board/

And the PHROM Extender: (connects to PL14 Ext ROM Connector)

[Pernod]

How did you try using them?

The usual Kenneth Kendal is accessed with SOUND -1, and I believe the Industrial one is -16, so presumably the others are something between -1 and -16.

Was a while ago now, can't remember, will give it another go using your info above, thanks.

Let us know which banks they respond to, so I can get them mapped correctly in my emulation.

Considering they're all on the same board I'd assume there are no conflicts, so I'd guess they respond to -16, -15, -14, -13.

[regregex]

What do we think, or know, about how the chips were made? Was the 16K of data sent to Texas for programming into the chips and they allocated a PhROM number at the same time? 15 for PhROM A is a very convenient number[...] I don't think it would have been by chance that Acorn's PhROM was given number 15.

TI wouldn't have been the one assigning bank numbers; the PHROM bus was given over to the customer who could assign the 16 banks as it pleased. There was not going to be a third-party market of PHROMs competing for banks, except one the customer creates (as Acorn intended to do) in which case the customer (as vendor of its own system) would broker bank numbers for use by its clients; since in normal builds the customer would be sourcing all the speech that each system would play in its lifetime.

Does anyone know what algorithm was used to compute the checksums written on the EPROMs in the emulator board? The usual suspects aren't producing a match, at least if those ROMs contain the final PHROM contents.

[HHH]

So using the test prog from Coeus:

Entering Rom Number 1 gives output from VM61003 (Time and Weather Demo)
Entering Rom Number 2 gives output from VM61004 (Military Demo)
Entering Rom Number 3 gives output from VM61005 (Aviation Demo)

In each case the words stack up with the references in the table listed words are delivered in a very much American accent.

In the add on board the socket nearest the connector is empty and the roms are then inserted VM61002 to VW61005 in ascending order in the slots, just in case this has some bearing on how they are referenced?

Is there any more info on VM61002 as I'm not sure I have got any output from this one or what rom number it presently falls under. Ideally a list of words and addresses so I can look for something unique to it.

Cheers

H

[Pernod]

Does anyone know what algorithm was used to compute the checksums written on the EPROMs in the emulator board? The usual suspects aren't producing a match, at least if those ROMs contain the final PHROM contents.

I suspect you'll need to reverse the bits of each byte, so that 'ACORN' is readable in the first few bytes. I thought I had a dump with reversed bits but can't find to confirm.

[Pernod]

Is there any more info on VM61002 as I'm not sure I have got any output from this one or what rom number it presently falls under. Ideally a list of words and addresses so I can look for something unique to it.

It should be bank 0, and is indexed so can be accessed with SOUND -16, x, 0, 0.

[BeebMaster]

This is my own dump of Phrase ROM A which I did recently if it helps. It's in the "right order" (as the comic said the to the maestro).

PhROMA.zip

(15.51 KiB) Downloaded 27 times

Which is VM61002, is it the one from the US Beeb? If so, there's definitely a word-list for that I can dig out.

[BeebMaster]

Looks like it is:

https://beebmaster.co.uk/BM10/USBeebSpeech.html

In which case, these should be the words:

Code: Select all

US PHRASE ROM WORDS	
	
WORD	NUMBER
-	217
-	218
-	219
-	220
-	222
-	223
-	224
-	226
-	228
-	230
-	231
-	232
-	233
-	236
-	237
-	238
-	239
-	240
-	241
-	242
-	243
-	244
-	245
-	246
-	247
-	248
-C-	227
-ER	229
-G-	215
-ILO ?	234
-R-	235
-S-	249
-U-	225
??	207
??	208
??	209
??	212
??	213
(SPACE)	214
A	39
ABORT	73
ABOUT	79
ADJUST	148
ALERT	147
ALL	112
ALPHA	46
AND	173
AREA	139
AT	110
AUTOMATIC	70
B	20
B-	221
BETWEEN	149
BRAVO	49
BREAK	187
BUTTON	151
C	27
CALIBRATE	189
CALL	74
CANCEL	113
CAUTION	176
CH- ??	211
CHANGE	63
CHARLIE	51
CHECK	178
CHEST	128
CIRCUIT	140
CLOCK	152
COMPLETE	102
CONNECT	141
CONTROL	108
CRANE	190
CYCLE	75
D	19
D	33
DANGER	61
DAYS ?	114
DEGREES	179
DELTA	48
DEVICE	153
DIRECTION	191
DISPLAY	76
DOOR	115
E	40
EAST	154
ECHO	47
EIGHT	3
ELECTRICIAN	109
ELEVEN	18
ENTER	192
EQUAL	77
EXIT	116
F	21
FAIL	155
FALSE	129
FARAD	174
FAST	78
FEET	193
FIF-	14
FIRE	100
FIVE	7
FLOW	117
FOUR	2
FOXTROT	50
FREQUENCY	156
FROM	194
G	28
GALLONS	130
GATE	157
GAUGE	118
GET	195
GOLF	52
GREEN	119
GROW	80
H	34
HENRY	53
HIGH	158
HOLD	196
HOURS	188
HUNDRED	10
I	41
INCH	81
INDIA	92
INSPECTOR	120
INTRUDER	159
IS	146
J	22
JULIET	131
K	29
KILO	169
L	35
LEFT	197
LIGHT	177
LIKE	93
LIMA	54
LINE	67
LOW	82
M	42
MACHINE	106
MANUAL	121
MEASURE	160
METRE	60
MICRO	137
MIL-	198
MILI-	175
MINUS	64
MINUTES	150
MOTOR	83
MOVE	122
N	23
NINE	8
NORTH	161
NOT	65
NOVEMBER	132
NUMBER	184
O	30
OFF	68
OFF	199
OHMS ?	138
ON	107
ONE	6
OPEN	84
OPERATOR	91
OSCAR	170
OUT	185
OVER	123
P	36
P-	216
PAPA	55
PASS	162
PAST	200
PER CENT	85
PICO	99
PLUS	124
POINT	186
POSITION	163
POWER	101
PRESS	201
PRESSURE	62
PROBE	86
PULL	125
PURSE	136
PUSH	164
Q	43
QUEBEC	94
R	24
RANGE	202
READY	87
RED	111
REPAIR	103
REPEAT	126
RIGHT	165
ROMEO	133
S	31
SAFE	203
SECONDS	142
SERVICE	180
SET	88
SEVEN	17
SHUT	127
SIERRA	171
SIX	12
SLOW	166
SMOKE	72
SOUTH	204
SPEED	89
START	66
STOP	105
SWITCH	181
T	37
TANGO	56
TEMPERATURE	104
TEN	13
TH- (THE)	58
THIR-	9
THOUSAND	15
THREE	16
TIME	69
TIMER	144
TOOL	167
TRAY ?	205
TWELVE	4
TWENTY	5
TWO	11
U	44
UNDER	90
UNIFORM	95
UNIT	143
UP	145
V	25
VALUE	183
VALVE	182
VICTOR	134
VIGOUR	98
W	32
WAIT	71
WATTS	59
WEST	168
WH- ??	210
WHISKY	172
X	38
X ??	97
X-RAY	57
Y	45
YANKEE	96
YELLOW	206
Z	26
ZERO	1
ZULU	135

[HHH]

Thanks BM that word list works with SOUND -16,n,0,0

It doesn't seem to work selecting 0 in Coeus test prog, I just get random noise sometimes with an odd snippet of a word in the mix.

Reminds me a bit of Bumble Bee in Transformer's speech.

[CHRISJJ]

speech samples from Kenneth Kendal, the BBC newsreader.

Coo. I worked on that. I well remember Paul Bond sending me to Texas Instruments in Bedford to sort out the muddled recordings.