Silly Sound Bastard: yet another Covox Speech Thing clone 

Hehe, true :)

i did

This is more of a CrossFire to me :D

No, just no. That's not how it works. It's two similar devices working in tandem.

@@forevercomputing If SLI stands for Scan Line Interleave so there may be something like - for example - ACI - Audio Channel Interleave ;-)

Bear in mind that this is only 8 bit sound.

@@xys007 Most DOS games used 8-bit samples anyway at 11025 Hz (or lower) sampling frequency, so it's not like there's a big hit in sound quality, which wasn't there to begin with.

It’s a R-2R ladder DAC, just built out of discrete components.

Really for sample rate under 16khz or so, the added benefit of more bits is limited... I would say maybe 10. But a lot of those bits is mostly use for volume or dynamic range that is less of a factor in 90s games

Wait untill ultra silly soundblaster comes.

Yeah, for me it is so impressive, because of it's simplicity in the first place. You wouldn't expect such result from something like this :) It was funny to make, furthermore, I got some practice in EDA, as I wanted and needed for another project I'm working on. And last, but not least, I have an IBM PS/2 machine to restore. It has only MCA slots and I have no sound option for it, so the Sound Bastard will come quite handy there :)

Same here, although nowadays they are on a network share.

I am an E.E i did not think it was possible as well till it was mentioned then i felt a pop in my brain and understanding was unleashed and it makes perfect sense now.

DAC is resistors there is support logic and what not, but DAC itself is extremely accurate resistors and nothing more

Unfortunately Doom will not work with this. The sound blaster emulator works only in real mode games, so all the older ones, but Doom runs in protected mode.

@@necro_ware Ahhhh, makes sense! :)

I have to revise this statement. There is one open source fork of doom now available, which supports DACs on LPT, like Covox. See github.com/viti95/FastDoom

I grew up with ibm computers all the way back from xt and have never heard of this before now

Hehe, thank you very much. This channel exists already since two years, it's just that I mostly made videos for myself until last month..... I will never understand, how YT works :)

Yes, I love this simple device too, it wakes up a lot of nostalgic feelings in me as well.

@@necro_ware 😁

Thank you Jorge! Glad you liked it :) Yeah, I think too, that it sounds the best in the recording, when unfiltered. But as I told, in the passive headphones it was too noisy and 100nF didn't make any difference at all. However, 1uF was just right and in the recording it sounds awful. I guess, it depends very much on the speakers, also if they are active or passive. That's why I made it switchable, just to make such experiments.

Some of the Sound Effects sounded pretty good through my Surround Sound System to be fair! 🙂

Yes, as I told in the video, unfortunately I don't have my anymore, which I built 30 years ago....

or maybe IBM appointed some CEOs in Creative Labs, allegedly, lol, stranger things have happened

Nobody expect at this time IBM PC will be used as a game computer, so no need to offer sound better that a beep to acknowledge user action...

IBM was some serious business. I think even the idea that their computers could be used for entertainment was viewed as sinful, perverted and completely incompatable with the core values of the company. lol

Probably just opted not to include much sound capability. Keep in mind that most people were probably listening to records or maybe cassette tape back then. Also, since these have to be directly driven by software they probably eat a lot of CPU… Not that big a deal on a 486 or later, but the original IBM had an Intel 8088.

It plays only digital sound. That takes a lot of memory, which was very expensive back then. Adlib on the other hand was FM sound playing notes and not digital samples. That needed significantly less memory and CPU power. That's why Adlib was more successful.

Emulate nothing, it *IS* the real audio card! :D

yeah :) i'm an Amiga guy, and did a little bit of research in that regard. thing is: you can think of the speech thing as a reverse parallel port sampler. as you know, when sampling sound on your Amiga's built-in parallel-port, it becomes useless, your sampling software will freeze the Amiga. that's because the parallel-port, as opposed e.g. to the sound outputs, cannot be accessed by a custom chip (with DMA). parallel port it is handled by the CPU, thus high (usable) sampling rates, or playback rates on the speech thing, will pretty much halt the entire system. (the situation might be different on e.g. an A4000 with some parallel-port on a Zorro-card, i don't know.) so, long story short: speech thing and Amiga is super tempting, but doesn't go together too well. (sadly.) :) (PCMCIA port on A600/A1200 and 16-bit hi-fi speech thing is super tempting, too. :-D )

@@amigaalive6266 You have done the research and I have not used my Amigas for a long time (I did have 2 different samplers though). And yes that is true, sampling programs did not multitask (but I think they could) it was just badly non OS written software. And when sampling you did not want any task having more priority because then you could loose bytes. But for playback I think some form of interrupt (playing sounds i.e just putting bytes on the parrallellport lets say a 8 Khz equivalent sample on the parallel port for a game that has some leftover CPU time (adventure games, and stuff like that.. cutscenes on other games) would be doable... And ofc pure sound programs at least for the A1200 it would not be a problem I think even with higher sampling rates on parrallellport? or do you mean accessing the parallel port (in this case puting a byte on the datalines cuts the DMA off completely?) I do know the parallel port data needs to be handled by the CPU and has no DMA. But you can put samples anywhere in memory you can just not play it using DMA and Amiga custom chips. This sounds like something I would loved to do, but my Amiga 500 is a long way from where I live and my A1200 is in bits and both needs a recap. But it should not be impossible for interrupt driven routine to play samples (from either fast or chipmem) thru an simple D/A on the parallel port In my mind at least. I think you need to convince me some more to make this a really bad idea 🙂 The ultimate would also be do (as I said in my original post) a small mixer (analogue mixer) that you could also in the same case as the parallel D/A converter could just put you normal Amiga 4 channel (2 leads) cables and a couple of options or maybe just one... so the parallel port "channel" could be mono mixed into both stereo channels. If you understand how I mean?

@@alexanderwingeskog758 yes, totally :-) i'm not trying to convince you that it's a bad idea. more like the opposite. i just think there might be some serious limitations. but it is certainly possible. i wrote a workbench-compliant sampler program, and i think it kinda worked up to somewhere around 10khz or so. when switching off some stuff (multitasking & display, i think) i got a little more, around 15khz i think. so maybe as a rough estimation if you use the speech thing at maybe 5khz there might be enough bandwidth left to do something else meaningful with the computer.

It should be far superior to the PC speaker though which is afaik similarly driven.

@@jnharton From my experience, it is.

Person who reverse engineered the Disney Sound Source and built a compatible circuit here. The DSS wasn't exactly capable of playing samples on it's own, the buffer is basically implemented to prevent buffer overflows/underruns - if the software supports it. Afterall the buffers are tiny ... 8 byte if I'm not mistaken (has been a while since I worked on it). In fact if you set the playback rate to 7KHz a covox would even be less CPU consuming, as it's a "fire and forget" type device while the DSS needs constant attention to the buffer in most cases - games by Apogee very likely just constantly streamed empty data, even when there was nothing to play back. While the DSS design made sense for very low end systems, it barely made sense in the 486 era. One major benefit of the DSS over the Covox is that it can be autodetected in software.

I guess so, but you'd have no software, which would support it.

You are welcome! As I told all the schematics are open hardware and you can use them if you want. Another user @HeavyD6600 also made a Silly Sound Bastard for himself based on my plans. You can watch some videos about it on his channel. He makes a nice review and even speaks more in detail about the assembly. See: ru-vid.com/show-UC795052gCeYosTs1_fODjkA

Hi there! Thank you very much for your support, but are you sure that wasn't a mistake? Please don't get me wrong, I really appreciate the help, it is really huge, but I don't make this channel professionally and I'm always worried about such big donations.

@@necro_ware no mistake - as said before, I owe this era of hardware my life. I wouldn’t be where I am today so please enjoy it and keep the videos coming :)

yes, in the recording indeed. But since this device has no amplifier, it sounds differently with various devices. With headphones the sound is the best with a strong filter.

@@necro_ware oh okay 👌 I thought it might be something like that.

While it is affected by the load, it's also because the sampling rate is really low. It depends on the output (mixing) rate of the player, and also the quality of the samples themselves -- which in a 4-ch MOD are usually pretty.. terrible. :-) The HF content you do hear isn't actually supposed to be there. The sampling process, when sampling at low 8kHz rates common for MOD samples, actually removes all the HF information, samples the remaining bits, then you're supposed to run it through a DAC with the same low-pass filter as the input. When you don't filter it, there's kind of a "mirror image" of the audio, with all the content BELOW the Nyquist frequency (1/2 sample rate) being mirrored ABOVE the Nyquist frequency. This is called aliasing, and it isn't supposed to be there. But since the sample rate is so low, there's no intended signal in the HF region, which sounds dull. So sometimes that aliasing noise makes it sound, subjectively, a little brighter and more interesting. A little bit, at least. Too much and it quickly gets fatiguing. It's distortion, and like all distortion, it's not _supposed_ to be there, but sometimes it's OK that it is. :-)

Not really. DSS had a buffer clocked at 7kHz, this is just a simple DAC, what Covox basically is.

I think, it should be possible, but you would have to write your own software to use it. I didn't heard about such approach yet. I think, that stereo was more important, than 16-bit samples. However, nobody said, that you can't use 4 DACs on LPT1 to LPT4 and make a stereo 16-bit setup :)

I don't know :D I left the links to the music in the description, you can use them to play the files directly in browser. You could compare the quality to your OpenMPT playback there...

It's actually very easy to make, you don't even need a parallel port controller, since you can simply setup the address per software and decode directly on the bus. I made a prototype of such card, but another tinkerer was faster and he made it even in stereo: www.retrokits.de/index.php/dual-isa-dac-r0-covox-speech-thing-on-isa-bus/

:) haha... in chains

You can do it with one row of resistors, but it's really hard, because they have to be very precise to get the right proportion. Else the sound quality will be very bad. It is much easier to make it with 3 rows of resistors of same value. This way even cheap resistors will sound ok'ish.

Yes, true. I also heard dual mono or something similar, but may be that was in another language. I mix sometimes the terms.

@@necro_ware Dual mono is another term for the same thing. Just putting this out there, I'm in no way trying to be mean or anything. I love the videos. I think it's awesome to see some hardware from my childhood live again. My goal is to help you since it appears English is not your first language

@@thomasvlaskampiii6850 Everything's fine :D I appreciate every constructive help and you are not mean at all. Thank you very much.

Not so simple, DSS needs some active elements, a buffer and a clock.

Yeah, for recording 100nF was a bit too much, but unfortunately I had nothing else between 1nF and 1µF at hand. And as I told, for my passive headphones 1µF was perfectly fine. If you are interested in something similar, but more sophisticated, then watch this ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-wW1th462IKM.html With a transmitter, a DAC IC and an amplifier, the quality will be even better and you will have a much more stable output, I guess, so that the filters will not make such a big difference, when you take some different speakers. However, it is the simplicity of the solution I've chosen, what fascinates me so much. Regarding the Tandy emulation, unfortunately, it needs a 386. May be, there is a way to get it running on a 286 as well. However, since audio over a parallel port can't use any direct memory access (DMA), you will need all the processing power you can get and an 8088 or 286 would be just not enough anyway. I tested it on a 386SX-25 and there it was running ok. In some games you could feel a slowdown, when a sound was played. It's a nice thing for experiments and learning about the tech, but as I told in the video, If your primary aim is to play games, probably an Adlib or SB-Compatible is a better solution. Especially, if you think, that the most games back in the days were written for OPL2/3 and not for digital sound.

LOL :D It just needed an old covox to be unleashed.

They have higher capacity, being just as small.

@@necro_ware Ah, I see! :) By the way I ordered five SSB long version PCBs earlier today. Thank you for making the design freely available!

I don't see a use case yet, but may be you can. As far as I know, at least in Windows 95 there was a Covox Speech Thing driver and may be it survived up to Windows 10? I suppose not, but who knows?! And if you use some old DOS program from the command line in Windows 10, probably it will be able to access LPT and use the Covox Speech Thing (clone) directly. I don't have windows 10 here however, so I can't check it.

@@necro_ware My big project I have been working on has lots of 8 bit Arduinos and one Mega 2560. I just need to generate noise and maybe a little music to mix into a sound board to give it a little robotic sound. I have found that a modified version of Josh Levine code for his video "Build a giant scrolling LED text display for about $15 per foot" worked sort of. It annoyingly speeds up as the loop runs its course. Link to my project: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-NCm6vHOR6uY.html

I guess, I don't understand yet. Could you explain a bit more, what you mean?

@@necro_ware Just a demo of the dual parallel port setup where 25% of the left is mixed into the right, and 25% of the right mixed into the left so it's not as jarring on the ears. Things are rarely hard panned in audio for that reason. I mean if you still have the raw tracks you could just mix it in your audio program of choice. Would be fun to hear.

@@PJBonoVox Ah, ok, I understand now. This has nothing to do with a Covox itself, but with an external mixer, right? I'll try, when I have time, out of curiosity, but in the video it was my intention to show, that the channels are really separated :) However, this experiment would more fall into the category audio, then retro hardware, don't you think?

@@necro_ware No idea :) I was always annoyed with my Amiga and how the stereo separation was 100%. Terrible in headphones. Just would be cool if you built an adapter to allow you to mix the stereo. All good though, still love the content!

Ok, thank you for the feedback. I'll see if I can embed it somehow in another video, in case I get it working. Cheers!

The company? :p

Yes, I know how it is :D I also have 1000s of projects running....

Yeah, I still love this project, even if it's not a real replacement for a sound blaster.

Yeah, definitely, it would be a lot better to have an amp. But just as I told in the video, I wanted to have it passive and as simple as possible. Unfortunately, a parallel port doesn't supply any power, so I'd need an external power supply then, which I didn't want. However, if I'd make it active, beside an amp, I'd also use a bus transceiver and a DAC as IC and not as bunch of resistors. Something like this: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-wW1th462IKM.html

@@necro_ware Every digital signal provides power. ;-) You just have to harvest it. You would need a very low-power op-amp willing to run with a single supply at low voltage. But thanks to mobile electronics, that isn't too hard to source.

@@nickwallette6201 Sure, this was how serial mice did work. They pulled their power directly from the digital signals. But with parallel port you are somewhere around handful of mA, if you pull from all registers. I prefer not to burn my parallel port and I know couple of people, who already did ;) They said, that they will need only 10mA, well, was a bad idea.

@@necro_ware An important point. However, it depends on what you use, I think. I just did a quick search on Digikey for LV audio or general-purpose single-channel op-amps with supply current less than 2mA, and found (just picked one at random) the Rohm BU7485G. Typical current is 1.5mA, max 2.0mA. That ought to be 100% safe to phantom-power from a couple diode-isolated port status pins or something. And there are lower-power options still - down to nA! Of course there are also some that consume tens or low hundreds of mA, without considering output load even, so choose carefully. :-) IMO, a buffered line-level output could be done in a way that should be compatible with any parallel port. After all, driving headphones from the DAC output can consume a few mA as well. ;-) And there are those troubleshooting devices with pass-through ports and LEDs to show activity on the pins.. those would have to be pulling at least 1mA, from any pin at logic high. Anyway, just my 2c. :-)

It's just noise, similar to a dial-in modem :D

I didn't look into the code yet. VSB was made by the same guys as TEMU AFAIK and the source is openly available. I'll have to find some time to look into, however, I guess it is full of implications and hacks, so it is probably not quite easy to understand. If you are interested: github.com/volkertb/temu-vsb

I didn't try it on a XT, but I guess, to some degree, you could. At least, playing music at 8kHz sampling rate could work. However, without DMA, all the work of data transfer has to be done by CPU and for a 8086 it could be just too much. Furthermore, the tools I've shown to emulate Tandy sound and Sound Blaster, need at least 386 CPU. Still, just out of curiosity, I'd give it a try :) You can find all the gerber files to order the PCBs in the repository from the description, but for an experiment, you can solder something yourself, or even use breadboard. Just to see if it works.