Thought about synthesizers channel multiplexing

[jxalex]

I have been in thought quite for a while, which can be, I hope some new project research to complete this with schematics and ICs needed to complete solution...

Why so many synthesizers do not use for multichannel output multichannel DACs, but use one monophonic DAC and then do channel multiplexing? Is there some additional reasoning behind that after costs cutting?

From JV1080 model there are already one DAC per channel, but before that still huge pileon of equipment have the channel multiplexing, even JD990, JD800, Korg Wavestation, etc...

Nowadays it is actually possible to make all them with one dac per channel, in a quite simple way. Has anyone done that? (I have seen just once the Yamaha TX802 model with digital outputs and wordclock input).
I see that the benefits are getting rid of that analog filter, lower noise floor or also to equip with digital output without any conversion (in some synths where the sampling rate is 44.1kHz), and in others it has to be upclocked. (Hint! Direct digital sampling...!).


Still, is there any particular reasons why the Roland and Korg flagships used this channel multiplexing technology besides cost?

[Mooger5]

You can add the D50/550 to the list A single PCM54HP is used for both Upper and Lower, probably because a MUX was cheaper than a second DAC.

[jxalex]

The list does not have to be complete as almost any big manufacturers synth before 1996 was using this trickery.

I think that why they did so and such cost cutting with their flagships. Were those DACs really sooooo expensive?
Well, it is quite slippery edge to to go "wanted analog sound harmonics and modulation distortion".

I think that there are not so many who get wet dreams about having digital output on their Roland D50/550 but it can be great help with those synths which noise floor is relatively much higher compared toother synths in use with setup.

[madtheory]

DACs were very expensive. Cost cutting was the only reason. For example, the budget drum machines that followed the Linn and OBDX all used multiplexing. There must be a way to find chip prices from back then?

[Rasputin]

You can add the D50/550 to the list A single PCM54HP is used for both Upper and Lower, probably because a MUX was cheaper than a second DAC.

This may or may not be off-topic, but does anyone know exactly why the D-550 isn't as noisy as (some revisions of, at least) the D-50? What exactly did Roland change?

[Mooger5]

You can add the D50/550 to the list A single PCM54HP is used for both Upper and Lower, probably because a MUX was cheaper than a second DAC.

This may or may not be off-topic, but does anyone know exactly why the D-550 isn't as noisy as (some revisions of, at least) the D-50? What exactly did Roland change?

Apart the revisions, the only differences I notice are EMI screening in the D50 (the aluminium foil wrapping power supply cables) and RFI filtering in the D550 (the two ferrite toroids surrounding audio and front panel cables).
I experimented with some kitchen foil and the D550 sounded subjectively darker. Removing the toroids, it sounded brighter. Not necessarily noisier, but I perceived a change in the overall tone.

[Mooger5]

More on topic, with all the advancements in speed and integration I wouldn´t be surprised if some if not all of the multichannel DACs were internally muxed. Same as always, just everything in one package.
I can see some high end 5.1 HT receiver using six DACs, but you know, is there any objective advantage? Was ever there any pro or semi pro audio interface with say eight outs each one with its own dac?

[Mooger5]

DACs were very expensive. Cost cutting was the only reason. For example, the budget drum machines that followed the Linn and OBDX all used multiplexing. There must be a way to find chip prices from back then?

There are scans of electronics hobbyist´s magazines archived here http://www.americanradiohistory.com
On one Popular Electronics I found an ad by Digikey selling CDXXXX chips for as little as 1.5 and some A/D, D/A chips unknown to me that sold for as high as 30.

[jxalex]

mooger. for multiplexing it clearly was very good dac which was clearly not 5USD only.
You see,... that one dac output which is used after that for multiplexing 6..8 channels has to be very good, in order to get better quality than separate DACs solution and still th. It has to be with with very high setting time. In JD990 if the sampling rate is 44.1kHz, and it is for EACH output, thus CPU feeds it with 350kHz sampling rate with 16bit (corrections?) ), and it is with 16-bit parallel input. (EDIT: more than that: it is serial ).
However. If just there would be used 8 separate DACs, then it would been sufficient that each is capable of having just 44kHz sampling rate, and this somewhat modest requirement would been cheaper and no need for such complex filtering circuit?

So, how much cost the DACs during 90s, while the synthesizer price itself was still 1500USD and over that, when we talk about things like Roland JD990 ? Really curious, how much cost a DAC which can handle 400kHz conversion rate and with 16..20 bits? ANd how much cost a DAC which is good with 44kHz with 16-bit parallel input both.

THe advantage still exists of having separate dacs as there is no need for complex extra circuits.

More on topic, with all the advancements in speed and integration I wouldn´t be surprised if some if not all of the multichannel DACs were internally muxed. Same as always, just everything in one package.
I can see some high end 5.1 HT receiver using six DACs, but you know, is there any objective advantage? Was ever there any pro or semi pro audio interface with say eight outs each one with its own dac?

with latest craplabs blasters I would not be surprised if they do so and also in common DVD players it is with serial interface.
On older days I do not know if there were ANY "semipro" multichannel digital stuff at all. Their price was already high for usable stereo cards in music production. Even the expensive Cardplus DAL+ card was just one stereo pair and the another option was HDD recorders for multitrack.


EDIT: the JD990 output interfacing to DAC is serial.

[Mooger5]

What about clocking? Would it make more sense If each DAC had its own clock or a single clock for all the DACs?

[jxalex]

Wnat You mean? It really depends on the synthesizer itself.

My one thought as a experiment was like such to make, however I have no idea what are the real results. But moreover I would be interested just as a challenge, however I have no such noisy digital synthesizers in my setup which would be in a real need for higher signal-to-noise ratio.
Still after all things done I would do that trickery with several DACs with my ROland JD990 and JD800.

Also with my Korg M1R and Wavestation A/D I would do it, the most challenge is the parallel to serial conversion in order to add the digital AES/EBU output.
(However, just to get the digital 16bit stereo PCM data would not be at all difficult to get from that synth to me as I have I/O board and can make direct digital recording from Wavestation A/D
then add the digital output perhaps it will give the answer to the eternal question about the direct comparison with VSTs )

[Mooger5]

Wnat You mean? It really depends on the synthesizer itself.


My one thought as a experiment was like such to make, however I have no idea what are the real results. But moreover I would be interested just as a challenge, however I have no such noisy digital synthesizers in my setup which would be in a real need for higher signal-to-noise ratio.
Still after all things done I would do that trickery with several DACs with my ROland JD990 and JD800.

Some crazy idea. That dedicated clocks might provide better channel separation...

Also with my Korg M1R and Wavestation A/D I would do it, the most challenge is the parallel to serial conversion in order to add the digital AES/EBU output.
(However, just to get the digital 16bit stereo PCM data would not be at all difficult to get from that synth to me as I have I/O board and can make direct digital recording from Wavestation A/D
then add the digital output perhaps it will give the answer to the eternal question about the direct comparison with VSTs )

You want to capture the PCM from the EPROMs? With CoolEdit (it´s now Adobe Audition) you can read raw PCM data directly from the .bin files... It´s easy to get the waves from the DW8000. A near future project is to edit or replace them with new single-cycle waves and reprogram the PROM. There´s a blog on how to hack the TR626 for this; it´s where I got the idea from.

Years ago, I built a non oversampling, filterless DAC on Vero board. The main chip was the low cost TDA1543, and had to add a CS8412 to convert S/PDIF to I2S. It´s an easy build, the schematics are around. As the TDA chip has always been a far from excellent performer, I tried several in parallel (up to four) to achieve better linearity.
Now I got a spare PCM54 from a DX. It has parallel inputs but I wonder if it can work in tandem with another PCM54, like the TDAs

[jxalex]

@Mooger.

No you really lost me now the DACs in the JD990 have separate clocks. it cant be done otherwise.

ABout Korg M1 and WS the comparison is about in another thread. The EPROM readout is not enough as its algorithms change the sound too.

[meatballfulton]

As an engineer, I don't understand this fascination with a DAC per voice. What is the perceived benefit? Certainly a DAC per voice becomes totally ridiculous with modern digital synths having 128 voices or more.

[jxalex]

No, no, no, not so extreme

No, not one DAC per voice, but one DAC per output socket, Sir! To make things straight.
So it means that if Korg M1 has 4 outputs on its panel, then it will be 4 DACs, and Roland JD990 which has 8 outputs, gets 8 DACs. Since the current schematic is just as it is now -- multiplexing the one dac between 4 ..8 outputs, depending on the synthesizer architecture.

Still, it is experimental phase and I havent done it, as I have many other projects going on, and such, but I guess that if I would have all the details then I have done it already. The another beauty of this project is that it certainly can exist with the parallel system (since the signal lines are taken from CPU in parallel) and thus, the comparison of the improvement is direct! However I do not know if the difference can be more measured than heard if to do that with the current DACs with over 120dB dynamic as on papers (PCM1792, PCM1795) , but my hope is that JD990 can sound clean as XV5080 module since both are 44.1kHz synthesizer modules.