On the CVX-1 I have:
7.48 x2
7.47 x3
7.46 x3
14.94 x5
14.95 x2
14.98 x1
14.97 x1

TBH the differences don't seem very significant to me; I'm not sure they explain why the part with the better precision parts is so much noisier. Perhaps some capacitive effects due to the extra traces between resistors? But those would be on the order of picofarads, i.e. several orders of magnitude lower than the filtering caps.

stamasd wrote:

On the CVX-1 I have: 7.48 x2 7.47 x3 7.46 x3 14.94 x5 14.95 x2 14.98 x1 14.97 x1 […]
Show full quote

On the CVX-1 I have:
7.48 x2
7.47 x3
7.46 x3
14.94 x5
14.95 x2
14.98 x1
14.97 x1

TBH the differences don't seem very significant to me; I'm not sure they explain why the part with the better precision parts is so much noisier. Perhaps some capacitive effects due to the extra traces between resistors? But those would be on the order of picofarads, i.e. several orders of magnitude lower than the filtering caps.

Yes the resistor values are fine, it is still a mystery to me too, at the moment.
I will do some more tests on CVX-1 once I get the parts. Ordered 1% tolerance through hole resistors and 0.1% through hole resistors

I ordered some SMD resistors too; those pads look like they will accommodate 0805 parts too, right? 😀
(got a whole bunch in both 0603 and 0805 ranging from 5k to 200k)

Anyway, it will be a while until I get them - they're coming from China.

Regarding the oft-cited 1986 release year of the Covox Speech Thing:

• The applicable patent, US4812847, lists a filing date of Oct. 2, 1987.
• The "Speech Thing" trademark, 1506939, lists a filing date of March 1, 1988.
• Finally, a company profile from September, 1991, re-posted from the Covox BBS, states that, "In late 1987, the IBM PC version of Voice Master made its debut along with the popular "Speech Thing"".

Thanks for clearing that up; looks like the first year the hardware was available was 1987. I came up with the year "1984" based on the copyright notices on the diskette labels and software.

I would be wary of listing patent filings and trademark registrations as gospel for when things were available commercially. I can think of a few products where the patent was finished years before a product was available, and vice versa (patent and/or trademark was finished years after commercially available, sometimes even after the product was discontinued).

I believe 5% resistors will make no audible difference.
Yes, the reproduction of the waveform will not be as 'ideal' as with the 0.1% resistors but the content played through the speech-thing is FAR from pristine itself.
To actually hear the deformation of a waveform one would have to listen to a pure sinus-wave with low noise floor to hear the added harmonics, this is not the conditions the ST provides.

PS.
Isn't there something other than this abominable techno dance Crystal Dream demo????
Maybe some sort of MP3 or WAV player for DOS that can reach the LPT?

I think DSS can do that.

Edit: here

James-F wrote:

Isn't there something other than this abominable techno dance Crystal Dream demo????
Maybe some sort of MP3 or WAV player for DOS that can reach the LPT?

When I make controlled recordings of the samples dreamblaster is sending me, I'll also play back several 10-second audio corpuses (corpii?). I'll make the audio files and player available at that time.

Not that it is conclusive, but we have 360K disks from 1990 and most of the Covox programs have copyright dates of 1988 and later.

MobyGamer wrote:

I would be wary of listing patent filings and trademark registrations as gospel for when things were available commercially.

Not gospel, no, but certainly helpful. I generally look for a "Date of First use in Commerce" statement in the trademark information as well, but that isn't available here.

A "late 1987" release kills some of the Speech Thing significance and mythos for me though, given the availability of other sound/speech products for the PC at that time. It's certainly a clever product, but given a lack of applicability in my setups today (outside of some clone-supporting French titles), my own LPC DAC devices are little more than seldom-used curios.

I think that the Covox, similar to the GUS, was mostly a darling of the demoscene.
In the early 90s, SB cards were very expensive, especially for young talented computer artists who were still in high school. So a cheap Covox (or even better, making your own for just a few bucks) was a good way to get a DAC, which was just what demos needed for their MOD music.
I think the fact that the circuit was included with software like MODPlay and FastTracker had a big hand in the popularity.

stamasd wrote:

I ordered some SMD resistors too; those pads look like they will accommodate 0805 parts too, right? 😀
(got a whole bunch in both 0603 and 0805 ranging from 5k to 200k)

Yes, these are actually 1206 footprints, by hand it is easiest to solder 0805 onto it.
0603 also works, (I did that for the 100k test). (and 1206 should also work)

Cloudschatze wrote:

I generally look for a "Date of First use in Commerce" statement in the trademark information as well, but that isn't available here.

Scratch that. The actual trademark registration certificate for the Speech Thing (rather than just the summary information I'd looked at last night) specifies a first use in commerce date of December 18, 1987. Since that information would have been provided by Covox, this may be about as authoritative a release date for the Speech Thing as we'll ever see,

Here's my three captures.
The 286-20, which could barely scrape by at 36 kHz: https://soundcloud.com/scali/286-36khz
The 486DX2-80, also doing 36 kHz for a direct comparison: https://soundcloud.com/scali/486-36khz
And finally the 486DX2-80 all out at 44 kHz: https://soundcloud.com/scali/486-44khz

Scali wrote:

I think that the Covox, similar to the GUS, was mostly a darling of the demoscene.

No, it had a lot of use outside of the demoscene for computers that couldn't use ISA sound cards, like laptops and PS/2 systems. It was probably the most common device connected to laptops so people could play audio on a laptop in the early 1990s. Around 1993, Media Vision and other companies released other LPT dongles that could do more, such as stereo, and had record capability, and they took over until PCMCIA was popular. But from 1990-1992, I saw many Speech Things on businessmen's laptops, and a few on friend's PS/2 MCA systems for audio in games.

The Media Vision "Dongle" is highly complex compared to your average Covox-a-like tho.
Sadly I don't have any pictures of the internals at hand but I think it basically came down to being a ThunderBoard on a LPT port instead of having an ISA interface.

DigiSpeech DS301 board ... not your average resistor array either.
http://nwserveur.no-ip.org/win3xorg/devices/P … -board-back.jpg
http://nwserveur.no-ip.org/win3xorg/devices/P … board-front.jpg

Scali wrote:

Here's my three captures. The 286-20, which could barely scrape by at 36 kHz: https://soundcloud.com/scali/286-36khz The 486DX2- […]
Show full quote

Here's my three captures.
The 286-20, which could barely scrape by at 36 kHz: https://soundcloud.com/scali/286-36khz
The 486DX2-80, also doing 36 kHz for a direct comparison: https://soundcloud.com/scali/486-36khz
And finally the 486DX2-80 all out at 44 kHz: https://soundcloud.com/scali/486-44khz

No feedback on these captures?
I'll try to stream a 44.1 kHz 8-bit unsigned PCM file through the CVX-2 next.
If all works well with the code (and the hardware, it depends on the HDD controller), I should be able just to rip CD tracks directly, and convert them to the right format, then play through the DAC.
That should give us a good way to build up more reference material.
It will probably require a 486 or Pentium system to pull off the 44.1 kHz though. A quick test at 22.05 kHz on my 286-20 was successful, but 44.1 kHz was a bit too much. Not in the least because IDE controllers and clone BIOSes tend to have dodgy performance that steals a lot of CPU time.

Hi Scali,
THANKS! the recordings are quite close,
after a few listens, I'd now really say the 286 (at 36Khz) is less noisy than the 486 (at 36kHz, and even at 44kHz)!

Have a listen :
compare this part (286) https://soundcloud.com/scali/286-36khz#t=11:52
with this (486-36Khz) https://soundcloud.com/scali/486-36khz#t=4:00
and (486-44Khz) https://soundcloud.com/scali/486-44khz#t=3:57

dreamblaster wrote:

Hi Scali,
THANKS! the recordings are quite close,
after a few listens, I'd now really say the 286 (at 36Khz) is less noisy than the 486 (at 36kHz, and even at 44kHz)!

Yup. I even noticed that in the waveform view when I recorded it. The 286 had pretty much a straight line in the silent parts. The 486 output never quite 'flatlined'.
So that difference is purely caused by the different computer. Possibly the PSU/motherboard design.
Also, the 486 has the printer port on a VLB card. Perhaps that picks up extra interference, because it is connected directly to the high-frequency bus of the CPU.

These were just some brandless generic clones, built from standard Taiwanese parts.
I'll be using some high-end Compaq DeskPro next. I wonder if that improves the quality. MobyGamer used a real IBM 486 I believe? Pretty high-end anyway.

Here's my Pentium Pro streaming an old track of mine at 44.1 kHz via the CVX-2: https://soundcloud.com/scali/cvx-2-alive-441-khz
Original version is here: https://soundcloud.com/scali/mdb-scali-alive-master-128kbps