DMAS Design [SPDIF-Optical]

[Ferather]

GaN Systems Extends its Lead in Class D Audio | GaN HEMTs Outperform MOSFETs | Amazon Search: GaN USB Charger.
An upgrade for optical pulses, and PowerDAC's, both apply to the DMAS sound system.

Note: The total THD, THD+N for the main DMAS audio output should be extremely low (SoC mostly).
Conversion to power (main THD point) is done by each speaker (PowerDAC).

----

 

[Ferather]

Here is a SFX digital filter (I got the idea from ferrite beads), it attenuates frequencies beyond the set value, which is set to 22kHz. Human hearing range is 20Hz-20kHz so 22kHz is above.
Digital sample rate divided by two, equals the maximum frequency it can represent (without aliasing), 44.1 = 22.5kHz, 48 = 24kHz audio frequency.

In an ADC situation (analogue to digital), if a DMAS was to support analogue, an analogue filter would do a similar job.
Oversampling allows the higher frequencies to be captured, then removed (down sampled).


Technically speaking, in 32 bit float terms, the attenuation should amount to 'silent aliasing' if any occurred.

----

 

[Ferather]

Interesting SPDIF item: https://www.matrix-digi.com/product/20/X-SPDIF_2 | Not a DMAS but similar in the basic sense.

----

My RTN method uses nil start-stop (off state) and packets, it does have a clock, however its a bit clock.

 

[Ferather]

Here is something else similar to my DMAS design, except the DMAS speakers have their own individual P-DAC, it does focus on lossless audio and bitperfect audio data.
Personally I feel the unit is potentially over priced, considering it doesn't seem to do multichannel. It can be updated via firmware updates.

It uses TOSLink and advertises < 0.001% THD (lower than, distortion), 2 Ohms balanced (see the specs sheet).

https://ecdesigns.nl/product/powerdac-sx/

-----

Not as flexible as the DMAS, as the programmable SFX, MFX and EFX processing points can be used in various ways, including SFX lookahead and MFX or EFX application.
AVR these days also use 32 bit (not float unfortunately) digital processing, but then a 24 bit DAC, in better cases a 32 bit DAC, but still an amp.

The DMAS (Digitally Managed Audio System) is fully digital right up to, into, the speaker, which is essentially a digital speaker.

----

DMAS note, since there is no DAC or analogue amp, the number of channels is limited by input, processing, power supply + physical combo modules (per speaker).


----

Reiteration of [Digital] Class-D, which is what a P-DAC essentially is:



----

Expensive headphones plugged into random DAC's will produce random quality, built in converter will produce constant quality.

What is True Sound? The concept explained | Stuff

Your guide to the philosophy behind ear-pleasingly accurate audio - True Sound

www.stuff.tv

 

[Ferather]

Looks like Linsoul have ownership to the E1DA PowerDAC, its also sold as Linsoul E1DA PowerDAC on Amazon, although I only found v2 (out of stock).

 

[Ferather]

MSC8256FS.pdf (nxp.com) | TMS320C6678ACYP4 Texas Instruments

 

[Ferather]

TMS320C6678ACYP Texas Instruments (more specs to see, slightly cheaper).

 

[Ferather]

Please note that the connection between the DMAS main unit and DMAS speaker is optical, an optical core with 2-3 additional copper wires for power (PCM-optical, power-copper).
In addition none of my designs or ideas are copyrighted so feel free to use them, its an open design. Might as well call the cable a DMAS cable.

Also note that SPDIF is only as good as the hardware either side of it, and whatever the current IEC standards are.

----

Also, if the intention is 2 channel output (stereo), then it should work out better to accept 15+ channels (PCM), then process each channel for spatial output via stereo.

 

[Ferather]

https://ecdesigns.nl/2024/07/11/no-pre-amplifiers-in-the-signal-path-at-all/ (11 July 2024) - We are not far away from the DMAS, and making analogue redundant.

Not sure about the 18 bit part, its true in terms of volume (dB) but not so true in terms of fidelity or wave accuracy from code.

32-Bit Float Recording: What It Is and How to Use It - zZounds Music Blog
Low Level Signals: 32-bit Float versus 24-bit - Sound Devices

In terms of the DMAS main unit, most certainly 32 bit float, both processing and because its also the amp (+- volume, PCM).
If I switch Potplayer between 24 bit and 32 bit float, there is a 'noticeable' difference for me, maybe 1-2%.

Note the use of TOSLink optical, seems to use a converter, makes me wonder if its similar to RTN over optical.


CPU/RAM (Device) > SPDIF (RTN) > TOSLink (125) -- TOSLink (125) > SPDIF (RTN) > CPU/RAM (DMAS).

Its possible to produce a PCIe or USB DMAS optical out (RTN) for PC's and other devices.

====

The PowerDAC in the DMAS speaker is more simple, its code to wave in full power, no pre-amp no amp, simple code to wave out, single channel.
32 bit float will still be accepted as input, the speaker unit its self will attenuate the PCM to whatever volume or power limits.

Theoretically there is no bitrate or channel limits, all set by the restrictions of the hardware used, and output count.
How the speaker manufacturer decides to use PCM and a power supply is down to them.


It would also be possible for a speaker manufacturer to alter PCM to better suit the parts and speaker (to match the original PCM).

----

DMAS - Digitally Managed Audio System (Smart, SoC).

 

[Ferather]

Here are examples of RTN transmission, the first example is variable bitrate, variable bits, using a constant bitrate output, in a packet.
The second is 1 channel, 1 sample, 32 bits in a packet, as a simple example. Packet header = clock data, bit interval.



Basic example: 100% power (bright, 1), 50% power (dim, 0), 0% power (off, -), return to nil (-), 3 states (PAM3).

There is no clock being transmitted constantly, instead its calculated from the data bitrate.
Audio is clocked by samples per second (bits per second in transmission).


Note, it may be possible to use 2 colours and off for 3 states.

----

102 (C) x 24 (B) = 2,448, + 101 (nil) = 2,549, + 20 packet bits = 2,569, x 48000 (S) = 123.312 Mbit/s.
25 (C) x 24 (B) = 600, + 24 (nil) = 624, + 20 packet bits = 644, x 192000 (S) = 123.648 Mbit/s.

----

125,000,000 / 1000 = 125,000 bits per 1ms | 125,000 / 1000 = 125 bits per 1us | 125 / 1000 = [ 0.125@1ns, x 8], 1 bit per 8ns.

====

The default mode for the DMAS would be 32 bit float 'true sound', else other modes (DTS, Dolby, user, so on).

 

[Ferather]

Technically, it should be possible to use lumens to produce an analogue wave, for power conversion of a digital input (speaker receiver).
A basic example would be 1000 lumen, with 500 as 0, every 10 lumen is a height up or down from 500 (PAM 100).

The potential would be an optical to power converter (Optical PowerDAC), no PWM.

If not lumen then colour spectrum will also work. ECDesigns bit switch might work well here also, 100 volt steps (not 100v in power).
You could have a total voltage limit of 10v, 10 / 100 = 0.1v a step (+5,-5), or say 100v, 100/100 = 1v a step, (+50,-50).

 

[Ferather]

Programmable voltage bit switches would allow for a 24 bit (or other) PCM or PAM input to be detected and 24 (or other) bit switches used, where the calculation is still max / bit.
In a 10v max situation (+5,-5), that would be 10 / 24 = ~0.417v a step (+,-), no power clipping, and can support any bit height up to 100 bits +.

Amplification can take place at 3 main locations, PCM, PAM to Speaker, or the final power stage.
A single speaker should limit the total power based on max or supply.

It would be possible for a DMAS speaker to use its own, or another power supply if built to do so, or if more power is needed.

 

[Ferather]

It should be possible to convert PCM into 100 bit (in this example, could be more) as part of the PAM 100 process. The speaker will still use 100 bit as its set max.
This allows, for example, 24 bits to be represented over 100 bits (more volt steps), and no change to volume (10v is 10v).

How the volume is controlled would be different. 100 bits = max volume.

An alternative is to make 100 bits mandatory, and send volume commands to each speaker (or globally) to change voltage over 100 bits (100% v, 50% v, so on).

----

I will label the product simply as 'PAM Speakers', where PAM [X] specifies the bits, for example 'PAM 100 Speakers' (Digital).

 

[Ferather]

It would also be interesting if the reverse was achievable, a 'PAM 100 Microphone', 100+ bits.

----

IET Digital Library: 84 GBd (168 Gbit/s) PAM-4, 3.7 Vpp PowerDAC (2015, PAM4 encoded).

 

[Ferather]

PAM 120 speakers would be backwardly compatible with a PAM 100 output, however only 100 bit switches would be used (still full volume).
Below shows the size (per speaker), of PCM 100 @ 48k, PAM 100 would be the same, using lumen or colour.

PAM 100 'encoded' would be much faster than the 168 Gbit/s PowerDAC linked above.
The DMAS main unit, would need to be set, if above the speaker.



====

Updated design image, includes RTN transmission, and PAM X speaker output. The reverse produces an optical microphone.
RTN: 25 (C) x 100 (B) = 2500, + 24 (nil) = 2524, + 20 packet bits = 2544 x 48000 (S) = 122.112 Mbit/s.



Voltage > PAM > PCM | Code | PCM > PAM > Voltage.

====

The PAM module and power supply can switch off if there is no audio input, and switched on when there is.
There could be a 2ms audio buffer per channel if additional time is needed to power on.

====

 

[Ferather]

Digital > PAM > Analogue | Analogue > PAM > Digital | Digital > PAM > Digital | Analogue > PAM > Analogue.

All you need is bit to PAM and PAM to bit switch (voltage, voltage steps).

 

[Ferather]

RGB Optical, RGB Optical Interface (Infinite 1 bit):

====

If we used colour and colour to bit-map (code, switch), the total number of possible positions would go up to 16.7 million or more, PAM 16.7M+.
In the bit to bit-switch sense (voltage, voltage steps), optical analogue, that would be 16.7 million bits, and PCM 16.7 million.

If we include lumen (colour and lumen sensor), we can use lumen to multiply the total number of positions.




----

1 bit time can represent [X] number of positions, so 1 sample 100 channels would equal 100 bits.

100 (C) x 48000 (S) = 4,800,000 / 1,000,000 = 4.8 Mbits/s.

----

Sample interval and sample rate is an interesting one, reducing the interval will increase total fidelity (if enough), but at the same time add frequency outside of hearing range.
If we calculated 48k as the base of 2 bits (2 intervals), between time is the [bit time, interval], sample rate is number of bits in a horizontal line, per second.

48000 /1000 = 48, /1000 = 0.048, x [20.833~ us] = 1 (1 interval). If we increased to 96k, 96000 /1000 = 96, /1000 = 0.096, x [10.416~ us] = 1.

Total bits would look like [48: 1__1] [96: 1_1_1], as you can see we just went from 2 bits, to 3 bits in [20.833~ us] time. This is not enough to notice much in fidelity.
To get similar to 24 bits in [20.833~ us] time, 1152 sample rate is needed (per channel, per second) (868.055~ nano second interval).

24 x 48 =1152, at the same time, there should be a digital filter attenuating past 20-22k to 0, as we can not hear it.


100 (C) x 1152000 (S) = 115,200,000 / 1,000,000 = 115.2 Mbits/s. Position, positions per second.



----

Principle of PAM: (1) original signal, (2) PAM signal, (a) amplitude of signal, (b) time.



Return to nil (RTN), off state (no bits).

 

[Ferather]

In the digital binary sense, code could be compressed (lossless), it could represent 4 main numbers (R:255, G:255, B:255), and lumen (X) - (255,255,255,X).
Similar to motherboards and GPU's and LED programming (Aura, other), as show in the screenshot posted below.

In the transmission sense (Speaker), you can not compress 1 bit to less than 1 bit, you can only increase-decrease the rate of 1 bit.

Increasing the number of LED's used would change lumen, without changing the colour code per LED.





Google search: RGB LED 16.8 Million

====

A basic military use could be a 500 rocket system controlled by colour-lumen pulses (multi-bit, encoded). Technically usable in some EM situations.
Not sure how far certain colours can be transmitted, so there will be a limit, but it should be further than copper.

----

We might also be able to include ultraviolet and infrared since sensors can detect it.

 

[Ferather]

Please note, none of my design is copyrighted, so feel free to just take all of it and make a DMAS.
Welcome to the optical age. For copper and volt transmission, its game over.

Even 0.0001 volts x 16.8 million is unsafe, and expensive (watts).
If you took it lower, EMI-RFI will be a major issue.

----

You could class the PCM portion of the DMAS as advanced PCM, still code, but colour-lumen.

DMAS (main): 100 channels, 100 positions, 1152k sample rate, X watts per channel.
PAM Speaker: 100 positions, 1152k sample rate, X watts max consumption.


Example main unit and speaker, note the speaker can have its own supply.

----

 

[Ferather]

Note that 24 bits and 100 bits at the same rate, use the same total bitrate, this is due to the infinite 1 bit (colour-lumen pulse = position).
The total number of positions can be defined (for example 16.8 million), but ultimately has no specific maximum.

For audio, the total transmission bitrate is increased by channels and sample rate.

----

Main Unit: 100 (C) x 1152000 (S) = 115,200,000 / 1,000,000 = 115.2 Mbits/s
Speaker: 1 (C) x 1152000 (S) = 1,152,000 / 1,000,000 = 1.152 Mbits/s

 

[Ferather]

Note that binary its self is also a limiting factor, lets say 1 height position required 10 x 1 and/or 0 (10 bits), RGB Optical would produce 10 positions in the same time.
With enough positions and sample intervals (positions per second), there may be no need for any demodulation at the end voltage stage.

A copper wire can only be 1 volt spec at a measured time, the speaker can only have 1 physical position at a measured time.
You can consider speakers as digital, but high rate. Bit, and bit time (position, positions per second).

Technically, this would be a digital to voltage transcoder, almost.

----

PAM: 25 channels, 4608k sample rate: 25 (C) x 4608000 (S) = 115,200,000 / 1,000,000 = 115.2 Mbits/s

Traditional PCM (100 bit) would need: 11.52 Gbits/s to do the above, no RGB.

----



Pulse Code Modulation and Demodulation

 

[Ferather]

Here is a way to put 48k into 'cant I tell if its above' perspective, if the interval was fast enough, why not simply PAM to speaker at that rate?
The answer is, it would not be fast enough, or sound correct if it was sent as modulated voltage, else why the DAC?

RGB Optical allows us to spend the maximum [X] bitrate, all on channels and sample rate.



====

Yes its possible to add more samples after capture (lets say a file already at 48k), however they are pseudo samples, not original samples.

====

The volt step calculation is highest +, added to lowest -, for example +12, added to -12 | 12 + 12 = 24 available voltage.
We can set 100 colours (100 bits), for this example, that would 24 / 100 = 0.24v a step (+ or -).

A microphone could be using 6 available voltage (+3,-3), but still produce 100 positions, 6 / 100 = 0.06v a step.
The resulting colour-lumen pulse would not differ, the same 100 colours regardless.



RTN (off, nil bit) can be used to represent 0 power, other.

----



DMAS = Digitally Managed Audio System (RGB Optical).

 

[Ferather]

Based on 125 Mbits/s TOSLink modules:

PAM (Main Unit): 25 channels, 4608k sample rate: 25 (C) x 4608000 (S) = 115,200,000 / 1,000,000 = 115.2 Mbits/s.
PAM (Speaker): 1 channel, 4608k sample rate: 1 (C) x 4608000 (S) = 4,608,000 / 1,000,000 = 4.608 Mbits/s.

Since a speaker is 1 channel (mono), always 1 bit transmission (so ignored), we can do higher rates.

PAM (Speaker): 46080k sample rate: 46080000 (S) = 46,080,000 / 1,000,000 = 46.08 Mbits/s.
PAM (Speaker): 92160k sample rate: 92160000 (S) = 92,160,000 / 1,000,000 = 92.16 Mbits/s.


The RGB Optical input for the DMAS will need more bitrate than 125 Mbits/s.

====

Digital Class-D uses 48bit to 1 bit then 48k as roughly 200Mhz (M not K), so ~200 Mbits/s, per speaker.

====

Now I am wonder if RGB Optical with infinite bit position counts as a way to do Qubit transmission.

 

[Ferather]

I was looking at other non-binary digital methods yesterday, when I came across Qubits, today I looked at Optical Qubits.
I also wonder if infinite bit (position) counts as superposition, at least until set or measured?

What are Photonic Qubits ?

Photonic Quantum Computing uses light particles as qubits, offering resilience to noise, long-distance coherence, but faces challenges.

www.quera.com

In RGB Optical, there is no 1 or 0, simply a colour-lumen pulse.

----

Some alternative examples, colour green = x volts, or 'k' on a keyboard, or fire rocket 34, other. Still no 1 or 0.

 

[Ferather]

It seems 16 bit PCM audio is 96dB, and 24 bit is 144dB, which is 6dB a bit (96 /16 = 6 | 144 /24 = 6), it should be that 1528dB (32 bit float) is equal to 255 bits (rounded).
In RGB Optical we don't specify dB or volume, but instead max-min position (voltage), its resolution based (positions, positions per second).

256 separate colour-lumen pulses + 256 bit switches, total voltage (+-) / 256, so say 24 /256 = ~0.094 volts a bit switch.

The good news is, due to the infinite bit specification, the bit rate required (compared to 100 positions), will not change, and is not calculated.

PAM 100: 25 channels, 4608k sample rate: 25 (C) x 4608000 (S) = 115,200,000 / 1,000,000 = 115.2 Mbits/s.
PAM 256: 25 channels, 4608k sample rate: 25 (C) x 4608000 (S) = 115,200,000 / 1,000,000 = 115.2 Mbits/s.

Imagine the bit in superposition, not set, or not filled, once filled it has a value, but will always be 1 bit, regardless of total number of available positions (values).

No 1 or 0, however, this does not mean that each colour-lumen pulse can not be transcoded, stored as binary, because it can.
You can also get nil bits, as in no value, but still produce an outcome, for example 0 volts, or a 0 sample.

255,255,255,X where X is lumen, has potential to map basic 3D, 4D.