MobyGamer wrote:

The speaker's filtering was an intentional design choice for rounding the harshness off of speech signals and speech compression (it is, after all, called the Speech Thing). I don't like using the amplified speaker very much because of the filtering.

Also, the Covox arrived on the market in 1986. Back then most people still had 8088s, often at 4.77 MHz. So you couldn't reach very high sampling rates anyway, with such slow machines. And low sampling rates would also require a lot of filtering.

OK, I added the 100k voltage divider, the supposed 11th pin

thanks for the great normalized comparison !

this is the 100k resistor ladder with added 100k voltage divider :
new recording no youtube : https://youtu.be/LffGptGEq9E
also uploaded the flac file to wetransfer : https://we.tl/42IVkob1Kj

Yeah the speaker smooths thing out, I will test once more with a small cap added .

Great Hierophant wrote:

With the 10-pin package, the two pins on the end seem to go to the audio jack. If that is so, one pin is used for the audio signal and one pin is used for ground. Using your multimeter in continuity mode, you can determine which pin is which by seeing which pin goes to the tip (signal) and which goes to the sleeve (ground). Ideally you should be putting one end of your multimeter on the signal out and the other end on your data pin.

I have confirmed that, on my 10-pin version, the 9th pin of the resistor pack goes to output and the 10th pin goes to ground/shield. Also, despite using a stereo plug, only the tip is wired up, so it only outputs out of the "left" channel.

Jepael wrote:

Between output and most significant data bit (D7, LPT pin #9), there should be 2R.
Between ground and least significant data bit (D0, LPT pin #2), there should be 2R+2R.

Between output and most significant data bit (D7, LPT pin #9), I measure 198.8 kohms.
Between ground and least significant data bit (D0, LPT pin #2), I measure 354.4 kohms.

Based on the 2R measurement of 198.8 kohms, I believe we have confirmed the original Speech Thing uses a 100k resistor ladder. 😀

dreamblaster wrote:

But I still have a spare CVX-2 I can send to MobyGamer, will await for the measurement conclusion first, maybe I can test more first here, because I am running out of CVX-2 prototypes.
Cheerz

I am happy to test whatever you send me, on a single system, against the original 10-pin Speech Thing, and make those recordings available. I'll respond to your PM.

(Edit: I responded to the PM, let me know if you didn't get it)

dreamblaster wrote:

Yeah the speaker smooths thing out, I will test once more with a small cap added .

Keep in mind the speaker smoothing things out was not always desirable. It helps with low output rates but it hurts high output rates. Because the filter was in the speaker, in my opinion it should not be in the adapter.

I suppose, if you were to put a filter in the adapter, it would be best to make it switchable so that people can turn it on or off based on their preferences.

yes I will make it switchable
I will do a quick test now with 100k divider + 500pF

MobyGamer wrote:

dreamblaster wrote:

Yeah the speaker smooths thing out, I will test once more with a small cap added .

Keep in mind the speaker smoothing things out was not always desirable. It helps with low output rates but it hurts high output rates. Because the filter was in the speaker, in my opinion it should not be in the adapter.

I suppose, if you were to put a filter in the adapter, it would be best to make it switchable so that people can turn it on or off based on their preferences.

I agree, the filter should be optional. In most cases, the non-filtered sound is probably preferred over the filtered sound.

This reminds me of the Amiga by the way... It has a low-pass filter built in. It was on by default, causing things to sound a bit muted. You could programmatically turn it off. Various games/demos did this, which sounded much better, at least, when good quality samples were used.
Originally the Amiga 1000 only had 256kB of memory. Not a whole not for high quality samples. The Amiga 500 and 2000 increased that to 512k by default... better, but not great. In the early 90s, 512k RAM expansions became common, later even required. At this point you could spend hundreds of kB on quality samples. So you can really hear a difference between early and later MODs. The early ones may sound better with the filter on, the later ones are generally better with the filter off.
In the end it's all user-preference of course.

okay made a recording with 500pF cap (low pass filter), 100k resistor network + voltage divider. (my pc still has the timing issues, so ignore that)
youtube : https://youtu.be/2yw1srjPvKQ
wetransfer flac file : https://we.tl/vNAm30676h

how does it sound ?
it is less harsh, but also a less bright?
should I try a more subtle filtering ?
or would you like more filtering ?

we're getting there I think

will send my CVX2 to mobygamer for the ultimate comparison.

And will design a cvx3 with selectable filtering

Hi everyone, this is a very interesting thread!
I've never seen the internals of a real covox plug before (but pictures of it on Great Hierophant's blog and on the web some pics the Covox speaker itself).
Anyway, I feel the urge to contribute something to this discussion and so I searched for my Covox clone.
It's a commercially made clone, not something I made myself.
I got it some time ago and it apparently was made in the 80s or early 90s in Germany by a company named RWElektronik
(judging by the price tag of DM27 and the obsolete 4-digit postal code).

It's not the real thing of course, but nevertheless it was made in the good ol' "covox days"..
So I thought it could answer some questions, as to how these things were made then.
Because of this, I took several close-up shots of the circuit and the package for you..
Speaking of the package, it also came with a leaflet and a 5.25" mini floppy.
I haven't ckecked the content yet, but I assume it contains some Public Domain stuff (ModPlay Pro, maybe ?)

The leaflet itself is rather bleak, I think. It contains some kind of clip art (hand drawn with a mouse, I guess).
It's interesting, though, that the drawing shows an old-fashioned stereo/ghetto blaster/boom box.
I haven't thought about using this, I often used my Covox clone with a self-powered pair of speakers or
a selfmade amplifier (the simple types using an LM386 or a single PNP transistor).
Maybe that's something we should also take into consideration here ?
I'm not sure, but perhaps these oldstyle devices performed more low-pass filtering or had a "stronger" sound
(thinking of the Sega MedaDrive, Model 1) ?

I'm ashamed to admit it, but I've never had a real Ghetto Blaster from that time..
OK, I had a tape recorder w/ radio and several cheap miniature ghetto blasters, but not the real thing.
That's why I wonder what kind of equipment people back in the Amiga days used.
I imagine that they used common stuff from before the multi media era, so no self-powered computer speakers,
but rather ghetto blasters, hifi stereos or radios with a phono jack (we once had a tube radio with such an input).

Anyway, I'm getting off-topic again.. 😅
Back to that plug again. It does contain several miniature resistors (SMD, I think) on a printed circuit board: 9x 2002 and 9x 1002.
It also has a cap labeled "104" on the output, which I think is 100 nF.
The casing is black and it also comes with a shielded cable (aprox. 2 metres; about 78inches/6.5 feet) and an RCA connector.
- No capacitor was hidden inside that connector in case you're wondering. 😀

No idea if this was of any interest, but I thought I should tell you about it.
(I just hope my post wasn't so annoying and didn't sound arrogant, whatsoever. My sister often complains about such things whenever I try to talk to her.. 😢)

http://imgur.com/a/2ZYyp

Edit: Success! Managed to read that floppy. Content can be extracted with 7zip.

Jo22 wrote:

That's why I wonder what kind of equipment people back in the Amiga days used.

Most people used their 1084S monitor with built-in amp and speakers (or they connected to the TV, which obviously also had an amp and speakers built in).
The Amiga had two cinch outputs, so it was also easy to conect to the home stereo system. That's what I did sometimes.
But because of the 1084S there was never any need for 'small' powered speakers, like the desktop speakers for PCs, or ghettoblasters or such.

Anyway, judging from your pictures, the clone seems to use 100k and 200k ohm resistors?
It looks cleanly made anyway. Would you say it sounds about as high-quality as the real thing?

Scali wrote:

Anyway, judging from your pictures, the clone seems to use 100k and 200k ohm resistors?

no no --> 1002 == 10k Ohm and 2002 = 20kOhm. ---> So it rather uses values in the range of other clones.

Anyway, Jo22, thanks for sharing, most interesting !, curious to hear crystal dream on it !
Would you remember what year you bought this, when this was produced ?

dreamblaster wrote:

will send my CVX2 to mobygamer for the ultimate comparison.

I will test with Crystal Dream, but I think I should also test with a short (10-second) real-world 44.1KHz audio sample that anyone can run on any system to compare with their own devices. I can try to come up with a short sample that has both loud and quiet parts, and both low and high frequencies, but if anyone already knows of such a sample (preferably musical in nature), let me know over PM.

dreamblaster wrote:

Scali wrote:

Anyway, judging from your pictures, the clone seems to use 100k and 200k ohm resistors?

no no --> 1002 == 10k Ohm and 2002 = 20kOhm. ---> So it rather uses values in the range of other clones.

Ah you're right, good point! So they're an order of magnitude off compared to the real Covox. That would probably result in a much louder signal (that may be the point, given that it has a cinch output only, so a line-out, and no speakers included).
But indeed, is that at the cost of accuracy?

A few ideas:
filter caps selectable for 44k, 22k, 11k sampling rates as well as off
output level potentiometer (log) instead of just a fixed divider as the 100k R2R ladder will have significant loading effects with most pre-amps

Scali wrote:

Ah you're right, good point! So they're an order of magnitude off compared to the real Covox. That would probably result in a much louder signal (that may be the point, given that it has a cinch output only, so a line-out, and no speakers included).
But indeed, is that at the cost of accuracy?

Yes I think so. The covox output is really quiet but more accurate
quality over quantity !

Here are my first recordings with the CVX-1 and CVX-2.

Methodology first.

Playback system: Thinkpad 755CX, P75, 40MB RAM; booted from a clean DOS 6.22 disk, with only himem.sys loaded and DOS=HIGH (otherwise the demo doesn't finish properly, with an "out of memory" error at about 2/3 of its duration; the differecne is 622kB conventional memory with himem and dos=high vs. 560kB without)
Recording system: Eurocom Monster laptop, i7 CPU, 8GB RAM; using on-board VIA audio; the laptop has only a microphone input, not a line input. The mic recording volume is set at about 10% to reduce distortion as much as possible.
The 2 systems were connected using a ground loop isolator.
Recording software: used VLC to capture, saved as FLAC 128kbit, 44100Hz.
The demo is played on LPT1 at 44KHz.

CVX-1 with 4.7nF and 100uF caps (default config): https://soundcloud.com/user-827772994/cvx1-5nf-100uf
CVX-2 with 4.7nF and 100uF caps (default config): https://soundcloud.com/user-827772994/cvx2-5nf-100uf

Both recordings were made with the kits I received yesterday from Serge (thanks, Serge) and assembled this morning.

For some reason the CVX-2 is a lot noisier than the CVX-1; it's not the soldering (I verified all the solder joints) and I don't think it's the sockets I used for the caps either (checked those too, removed/reinserted/wiggled caps etc).

stamasd wrote:

For some reason the CVX-2 is a lot noisier than the CVX-1; it's not the soldering (I verified all the solder joints) and I don't think it's the sockets I used for the caps either (checked those too, removed/reinserted/wiggled caps etc).

Indeed. The CVX-1 is very clean-sounding, where the CVX-2 is noisy in eg the start of the second part (the Axel F song with the synth string sounds).
We've heard that on dreamblaster's own recording as well.

that is strange, both stamasd's CVX-1 and CVX-2 have basically the same schematic.
the CVX-1 has 15k and 7.5k 1% resistors.
the CVX-2 has all 7.5k 0.1% resistors. Maybe some 'random' inaccuracy in the DAC (within the lowest bit limit) helps hiding the noise ?

I will order components to complete some CVX-1 boards with 1% and 0.1% 200k/100k resistors (like the covox).
--> curious how that will sound.
Will also order more different cap values.

I've listened to my CVX-2 again, and I also have a very clean sound, not the noise in the second part.
I am currently running it on my 286-20, so I can't really get more than 30 kHz out of it, but it sounds really good.
I'll see if I can make a recording of this tomorrow. Then I'll also dig out my 486, and do a recording on that (I can do both 30 kHz and 44 kHz, that also allows us to compare directly with the 286 to see if the LPT itself may have any effect on it).

I have measured all of the resistors used in the CVX-2, all cam from the same strip of SMD parts. Most of them measure 7.49kohm, with about 25% of them measuring 7.50kohm. There are no other values among them. I did not measure the 1% resistors used on the CVX-1.

could you measure on CVX-1 too, should have larger diffences between the 15k resistors and the 7.5k resistors (multiply by 2 to compare).
covox has 5% resistors, it seems, so they may have even larger differences (though I expect within a pack,they're closely matched)