DIYHiFi.org

I am long tired of going around and so I am finally posting this task.

This is well, well over my head, and so I will appreciate any indication of feasible routes to go.

The thing is since year 2000 there have been coming out different digital mixers, some quite nice; 100% DSP equalization and processing, but almost none has a complete digital path from input to output.
First we had UK's Redsound Infader model, then it was Tascam's X-9, both almost did it... Rapidly Denon saw the opportunity and (Shame) packed two analog models X-400 and X-800 with dig inputs and outputs and marketed them as "Digital".
Numark came out with a pletora of Digital Mixers, but again No dig inputs, this time they did something OK and thus, they could advertise 100 dB S/N. At 2005 the supposedly industry standard, Pioneer, responded with a superb mixer 8 channels, "the first one with a non-DAC/ADC path", but so bulky and priced as to be unobtainable for many of us. A couple of months ago a more reasonable 100% dig 4 channels mixer was released... again expensive($1400) and superb sounding but still with unsufficient EQ knee.

Therefore facing the perspective to wait a couple of more years for the price to come down as usual, is not even an alternative because this late model doesn't provide enough the EQ as we use to.
Frankly, at 2006 I didnt expected to be in this scenary...

Hence, now I've decided to restart the long visits to the places I have already been amused by the impressive results and where I've learnt enough, to put a ground to my hopes... and so the music I so much care would -one day- be proyected as my HD-280 headphones and a few in-house systems sometimes reveal, and even more. Why not?

Ok, that is the general background. Much more could be said but I'd rather would get on topic:

Best modification cadidate is Tascam's X-9, It has the best Equalization available...nothing compares to this one; a mix of fully configurable parametric and graphic, all done in DSP, much on the fly, with 4 channels, flexible routing, plus effects, etc.

Two of the buses have SPDIF inputs(without AES/EBU choice) and also has a coaxial output. The problem is that as soon as these dig inputs get inside, they go into a DAC utilised to accomodate the incoming volume of the signal(important because of the different masterizations) and after that again gets converted thru an ADC to access the actual mixer with its various proceses from then on in nice digital form, 24 Bit.

As the mixer is, before the signal reaches the power amplifiers it demands to go thru 3 analog/digital convertions¡¡¡ Add to that the fact that despite all the economy of analog bits and pots, X-9 present specs show a 84 dB S/N
No wonder the solely TRIM allows for(unnecessarily) (-12/+48) dB

/////////////Ok, With this landscape to move into... after eagerly sarching around, I understand that what I need to locate is:

A pair of good ICs that inside have PGA programmable Gain amplifier stage with +/-12 dB, and being as simple as possible, so they draw little power, hopefully multiplexed and so that are able to receive SPDIF and AES plus housing 2 ADC inputs -if possible-, so the ICs also control the mixer analog line inputs and the effects return (also analog). The control section has to be connected somehow(...) to the present Trim knob that actually controls the op amps.

As you see, that is my most optimistic wish list as I (am almost blind in this) am able to see it.
Then it all would need some PCB built... with good power, conditioning, shielding, probably heatsinked, etc,etc and fitted into the machine(luckily there is space).

I am still trying visiting sites like the cirrus one but you have to read each of the specifications to see what each IC has, I haven't been able to do it the other way so far, but I have seen several interesting chips...

And so, am I erring too far in this, is there any other route? or already built circuits?, adaptable kits? or such?.

Thanks, I would value any comment that helps to steer me into the solution.
I know there is, somewhere.

Thanks.

The block diagram appears at the last page of the manual downable from:
http://tascamforums.com/index.php?showtopic=5529

_________________
WCarlos,GMoroder,JMJarreO2,BAndersson,DParsons,WOrbit,EatStatic,Youth,InfinityP/JReactor,TEclipse,Kox/Flux,MMMorris,PeteNamlook,TSuzuki,JMunro,JSidhu,SPosford,Shpongle,FHolyzewsky,PMartin,KKylven,GusTill,(Pierce)...Thanks

What about the input metering? It is analogue and you would loose it if you go directly to the digital input of the DSP section.

Yes, I saw that too.

Actually this is a minor drawback, because the way it is designed, as it is now, the channel meters pick the signal before the EQ and so it only serves to compare the differences in input, instead of its much better placement after the EQ section, where the user could monitor the effects in output of the powerfull EQ, before he decides to send the signal to the master...

After some "intentively" watchs to the Block diagram, I came to think that the meters could possibly be made to pick their signals after the other DAC that is used at the CUE section, ...though it would probably need to be assigned on/off at the same time one assigns a particular channel/master to be sent later the headphones, and hence it will only light that channel as opposed to the situation now, that all meters light up when receive a signal.

Thanks for your input, it really makes me feel better, that despite all the mistakes and holes that my initial post surely holds(as i've already tried to correct a few but it was after the 60 min limit), the actual task could be of some interest to others.
So by your words may I interpret, that given the right conditions you think it could be done? I am so interested into this as to go from gathering the general info, to listen and learn about its implications, or difficulties, to find the particular suggested candidates as components.

Any other hint will be greatly appreciated.

_________________
WCarlos,GMoroder,JMJarreO2,BAndersson,DParsons,WOrbit,EatStatic,Youth,InfinityP/JReactor,TEclipse,Kox/Flux,MMMorris,PeteNamlook,TSuzuki,JMunro,JSidhu,SPosford,Shpongle,FHolyzewsky,PMartin,KKylven,GusTill,(Pierce)...Thanks

You might find what you are after here. You can disregard the functions you do not need.

For being short answers they do open doors rfbw ¡¡.

I'd still need to read on (and on and on) the various app notes, datasheets and related papers. So far the best candidate looks like it is TAS3103A...

However as the info settles together, I am also realizing quite a better picture about DSP digital amplification, which I understand still needs resampling, though looks much better than Digital/analog convertion, if given the choice.

If Ti products here are indication of what has been recently happening in the audio processors industry, its no wonder that back at 2000/1 the DJ Digital mixers opted for the analoge gain stage at their inputs...
Around that time, I exchanged emails with the Redsound people and they said that the best choice would be to bypass the Trim stage altogether.

Took me a long road to understand a little of what has happened...

The first two chips of this Ti family were only released back at March/April 2001(judging from the datasheet date).
.::TAS3002 works with 16-24 bits and 44-48KHz data, has Data processing at 32 bits and DACs at 24 bits.
.::TAS3004 works with the same resolution and bits, and has the same DSP capacities except it also allows for a stereo analog output, where 3002 mixes both channels into a subwoofer monoaural out.

These ICs main job is to equalize, loudness and vol control in the digital domain, the EQ may be bypassed, plus the DAC out could? be used for a preEQ metering.

TAS3103 came out at Feb2004, looks like a revamped version of the previous chips. Processes at 48 bit and uses Infinite IIR filtering(16), since Tas3002/4 units doesnt say, I assume the later use FIR instead or less filters. Hence all indicates this new "3100" line sounds better.
Plus support 8-98 KHz sample rates, 4 inputs, 3 outputs serial, VU metering output, soft mutes/unmutte,etc.
It looks much more developed, though its controlling is done with specific software and might be complex to do/program???. It does talk of it as being fully configurable and a (PC) GUI-based software development package that is available...
TAS3103A released at Feb2006 looks in the datasheet as the perfect candidate of the four. it has plenty of inf support and evaluation boards,
same capacities as TAS3103, with two years of experience...

However the difference between both is in the errata info ¡¡¡
After reading the both TAS3103 and TAS3103A erratas PDF. I just can't believe what I've read there.

Actually it is not just that the best and newest IC of the line(TAS3103A) has possible gross malfunctioning, but learning that ALL of the released Ti chips in this line of product have consistent errors.

I did know that digital equipment sometimes(usually because of weird electric feed)behaves strangely and has to be reset to get it well functioning again. I also did know that comercial software has to be patched and upgraded often because of bugs, as I know about jitter into digital sound as a manegeable error between certain boundaries, also the notion that ICs have a % of errors increasing with temperature...
But i didnt expected to find that chips like this could be rendered unuseful while at work and hence force the device to be reset: I just hope that "a device reset" means a power off/on of the whole machine.

///////////////Ok, that learnt, I will concentrate here on the latest/best one from Ti, to find out if -yet- this choice is too risky or not... as only a few months after its release have passed and there is already a problem found. What else could come later? though the two years of experience might count in its favor. Maybe this addition needs a dedicated on/off switch, just in case, so the rest of the mixer keeps on functioning while reseting the chip .

TAS3103A

The errata PDF mentions that a change in the MCLK, if it is restarted, stopped or the frequency changes, the chip goes into indeterminate state.
As system impact it says that I2S communication stops(control stops functioning?) and the device might not operate correctly
Since the chip has many operations it may be the case that the volume is not compromised...

As a workaround it only talks about reseting the device when any of that change in the MCLK occurs.

And so before getting into the actual implementation, questions arise...(and please anyone correct me here -even- at formulating the questions, I do have good intentions and try to put up a good effort, but no grounds to back them): If the clock is taken from the source(has master/slave), when the CD player stops(because the user wants to change the CD, -which happens plenty of times with a mixer-)then the MCLK stops in the sense the errata mentions? Is the case that are there alternatives to set the master clock in a more stable configuration? I guess the change in freq, might not be a problem since the gear to connect to the mixer's added-input-board, would be plugged-on from begining to end of the session.

It does look like a Ferrari though.

_________________
WCarlos,GMoroder,JMJarreO2,BAndersson,DParsons,WOrbit,EatStatic,Youth,InfinityP/JReactor,TEclipse,Kox/Flux,MMMorris,PeteNamlook,TSuzuki,JMunro,JSidhu,SPosford,Shpongle,FHolyzewsky,PMartin,KKylven,GusTill,(Pierce)...Thanks

Two points.The newer digital receivers allow for a separate MCLK signal to takeover in the absence of an audio signal from which to derive MCLK and a device like a digital mixer often has an external reference signal and so long as that is present MCLK will be uninterrupted.
There is another option. Since all you need is volume control, a simpler bespoke approach is possible. At this point, I would reach for a CPLD or a FPGA and others a DSP chip. As far as DSP chips go, such a task may well be within the capabilities of a dsPIC but there are others better qualified to comment on that. You might want to try putting out a call to gmarsh. I get the impression he knows these dsPIC things inside out.