MDAC First Listen (part 00011110)

[DANOFDANGER]

Iv never liked eq correction, works ok in cars but that's because there's not much choice. Id rather fix the room than compensate with eq.

Bluetooth sounds convenient but unless the audio is high quality profile like aptx i'm not interested, i dont even use it in my car.

Both no from me.

 

[JohnW]

This is sacrilegious.

All that fine detail, dynamics, texture and harmony sacrificed for more impressive measurements.

At least it is optional. It reminds me of seat belt reminders in cars that sound off in rebuke if you unplug your seatbelt a fraction of a second before the car reaches a complete standstill, i.e. a nasty and annoying little gimmick.

If it's what the market wants. . .

Steve,

You don't need to use the DSP, it can be completely bypassed - it would only be for those who would like to try Room EQ or even simple Tone controls etc.

 

[JohnW]

All,

I agree with JohanH. It seems as we're trying to build an Audio God Machine. I see the schedule drifting into the far future.

/Lars

The DSP would be mounted on the Digital board - adding DSP should have no impact on the production date as Dominiks will start designing the PCB once he's completed a current project. As Dominik will take care of the digital board I only need to design the Analogue board, its this Analogue board that requires the greatest design time - Dominik should have the Digital board ready before the final version of the Analogue board is completed.

 

[fusion5]

One of the most often requested feature is some form of "Simple" room correction / Room EQ / Tone controls / equalization.

I discussed it with Dominik at length this evening / Morning and the FPGA is not best suited for DSP type processing - so we propose adding an onboard DSP processor allowing us in the future to offer room EQ and basic room correction which would go some way to help solving room resonant issue etc.

The advantage of having the DSP onboard allows sampling rates upto 192KHz (and maybe 384KHz) as a one box solution – although the room acoustics analysis software and user interface for measuring and analysing the room and loudspeaker responses would be run on a computer – once the optimal settings have been found then these results are downloaded to the MDAC2’s DSP via the USB interface and the MDAC2 can then be operated as a standalone unit with no further need for the computer.

We would make the MDAC2 onboard DSP compatible with PC software from a company like REW’s Room EQ Wizard http://www.roomeqwizard.com/

With the onboard DSP engine we could also offer features such as Tone controls and Headphone Crossfeed support etc – anything that requires DSP.

If there is the general consensus with MDAC2 “owners” to go ahead with the additional DSP then the extra hardware cost (which includes the cost of the DSP IC and hardware development) would be GBP100.

Adding the DSP processor to the hardware design would add less then a weeks delay to the release of the MDAC2 – although the software / features would be enabled after the initial release of the MDAC2 (upon release of the MDAC2 the hardware will be DSP enabled but it will take Dominik a few months to get the software full operational).

As a side note, we would also add Bluetooth audio streaming as it can be added to the design for less then GBP10 – while the Bluetooth input is not selected the module will be completely powered down so there can be no concern about RF issues.

The idea behind adding Bluetooth is as a convenience feature allowing you to stream YouTube / Spotify / Skype type audio from a laptop or Mobile device – its not intended for serious listening – but if you anything like me, then I spend 90% of the time I just having music / news steaming in the background while I work, and Bluetooth is ideal for this.

DSP:
Sophisticated tools for room correction like Acourate or Audiolense create complex FIR filters for both frequency and phase corrections. You'll need different filters for each samplerate and serious processing power for the online convolution.

Although I don't know if your DSP is up to the task. I prefer running this on my PC

No need for simple eq tasks here.

Bluetooth:
I don't need it (yet). On the other hand, why not?

 

[shaolin]

I agree with timelines concern, if it affects MDAC2 release too much, rather to avoid it. But just 1 week (if I got it correctly) to implement DSP on board seems ok. However, I also worry releasing correct firmware might take months for 1 man show Than it's probably better just use existing solutions.
Some add-on card would be great to have it as an option, but might bring even more complications to the design.
And I just hope additional chip doesn't bring (too much) additional EMI/RFI.
Would FPGA stay or replaced by DSP completely?

I know that applying too much correction is wrong. But for those who can't afford change/tune room, lower the peaks in LF might help a lot.
I think when HF is untouched, fine detail, dynamics, texture and harmony is still there.
At the end, it's all about how it's done.

BTW, any updates on MDAC2? Where we are? Impressions?

 

[JohnW]

DSP option, yes please

However, could this possibly be a plug-in board/module option?

The DSP IC itself and local PSU decoupling cost etc about £10 to £15 - so its not really worth designing a separate PCB for it - the connectors and separate PCB assembly would cost more then the DSP itself - then packaging etc...

Asides from the material cost of the DSP & Bluetooth, its the extra hardware design time (a week or so) + Dominiks programming time (1 to 2 months). There already has been cost creep on the digital side due to the requests' - well I'd say almost the insistence that we added Galvanic isolation for the USB which was not factored in the original design requirements.

It would be more cost effective to build two versions of the MDAC2 - the same PCB just different component population.

1. Entry level version without DSP + Bluetooth

2. Advanced version with DSP + Bluetooth

Those who are interested in the DSP + Bluetooth version would be asked to contribute an extra £100 to cover the extra material + development costs.

To be quite honest, I'm quite interested in trying the DSP room / speaker correction as it can be performed in full 56 bit mathematical precision - with this 56 bits being feed directly into the FPGA modulator thus maintaining as much mathematical precision as possible reducing DSP rounding errors.

I'm sure with the DSP power available we could try other ideas in the future, the combination of the DSP + FPGA gives us vast future possibility's

Both versions will have the same analogue stages - so its only a question of extra digital features. I'm hoping that with the extra DSP power we might be able to design more advanced filters then with just the FPGA alone. The FPGA is ideal for the ASRC, Modulator and handling the DSD BitStream - while the DSP is better suited for Digital filtering / EQ etc.

 

[Bigbobby666]

Those who are interested in the DSP + Bluetooth version would be asked to contribute an extra £100 to cover the extra material + development costs.

John, with these additions my brow is starting to get raised - how does one go about getting 'onboard' and where are there details of cost / process?

Cheers
Rob.

 

[JohnW]

I agree with timelines concern, if it affects MDAC2 release too much, rather to avoid it. But just 1 week (if I got it correctly) to implement DSP on board seems ok.

Yes it would take about a week extra PCB design time to add the DSP hardware - Dominiks handling the design of the Digital PCB so it does not impact my design time on the more complex analogue board..

I also worry releasing correct firmware might take months for 1 man show Than it's probably better just use existing solutions.

The idea is to use the software such as REW’s Room EQ Wizard http://www.roomeqwizard.com/ to handle the analysis and user front end - with Dominik only having to write the MDAC2 DSP software to work with the downloaded filter coefficients from REW’s Room EQ Wizard software. As we now only need to worry about programming our on-board DSP with the measured results we are not having to reinvent the wheel, and dramatically shortens Dominiks workload.. The advantage of on board DSP is that we can maintain double precision (56bits) from the DSP to the modulator without needing to truncate the data, and also work with 192KHz input sample rates.

Some add-on card would be great to have it as an option, but might bring even more complications to the design.
And I just hope additional chip doesn't bring (too much) additional EMI/RFI.

As the DSP IC and surrounding components only cost say GBP15 its not worth the extra expense of adding connectors and an extra PCB - just cheaper to have mounted on the Digital mainboard.

The Digital and Analogue sections are now isolated from each other - this will help eliminate any circulating Ground plane noise currents - also the DSP would be operated fully synchronous with the audio sample rate - thus side stepping EMC issues.

Would FPGA stay or replaced by DSP completely?

The power of the FPGA is still required for the Bit intensive tasks such as the ASRC, Modulator and handling the raw DSD Bitstream + bit manipulation... DSP's are not optimised for this type of task.

I know that applying too much correction is wrong. But for those who can't afford change/tune room, lower the peaks in LF might help a lot.

I think when HF is untouched, fine detail, dynamics, texture and harmony is still there. At the end, it's all about how it's done.

I agree, I'm thinking along the lines that the DSP would just cancel room mode resonate peaks - and ONLY working upon the peaks - not trying to boost nulls... The beauty of having the DSP is we can try a few "correction" strategies, nothings fixed in stone...

BTW, any updates on MDAC2? Where we are? Impressions?

I'm currently got my head down designing the MDAC2 Analogue stage PCB - once the analogue stage is completed, I'll issue a PCB which can be fitted and tested on a few owners MDACs for first impressions - the new Analogue stage design & Dual ESS are where I expect to reap the greatest sound quality improvements (Only the new analogue stage will be tested on the MDAC units)...

 

[JohnW]

John, with these additions my brow is starting to get raised - how does one go about getting 'onboard' and where are there details of cost / process?

Cheers
Rob.

Rob,

The first task is to source a donor MDAC unit (Ebays a good bet) - once you have secured a unit then you can pay the first 2 development pledges of GBP100 - via our very crude MDAC2 webpage (which I must update with latest developments).

 

[GWM]

Hi John
The hight of my listening room is exactly half it's width so I have 1 bad mode which is difficult to totally avoid so some modest DSP correction might be a help so am interested as long as the cost doesn't creep up too much.
Geoff

 

[JohnW]

Hi John
The hight of my listening room is exactly half it's width so I have 1 bad mode which is difficult to totally avoid so some modest DSP correction might be a help so am interested as long as the cost doesn't creep up too much.
Geoff

The target cost would be an extra GBP100 for the DSP version.

Cheak out http://www.roomeqwizard.com/ to see it it looks like it would work for you

 

[misterdog]

To be quite honest, I'm quite interested in trying the DSP room / speaker correction as it can be performed in full 56 bit mathematical precision - with this 56 bits being feed directly into the FPGA modulator thus maintaining as much mathematical precision as possible reducing DSP rounding errors.

I'm sure with the DSP power available we could try other ideas in the future, the combination of the DSP + FPGA gives us vast future possibility's

Both versions will have the same analogue stages - so its only a question of extra digital features. I'm hoping that with the extra DSP power we might be able to design more advanced filters then with just the FPGA alone. The FPGA is ideal for the ASRC, Modulator and handling the DSD BitStream - while the DSP is better suited for Digital filtering / EQ etc.

Yes Please

I think that some folk get too fixated in believing that we are creating a perfect reproduction of recordings in our own environment.

Yet if our environment is not acoustically identical to that of the recording
environment it cannot be.

Reality is that we are creating an illusion of 'perfect reproduction' and though I have never experienced DSP from what I have read it is already on my wish list.

 

[Mikas]

It would be more cost effective to build two versions of the MDAC2 - the same PCB just different component population.

1. Entry level version without DSP + Bluetooth

2. Advanced version with DSP + Bluetooth

I would love to have the Bluetooth for its convenience (and I do not believe it will be significantly worst in terms of SQ).
As for the DSP and Room correction, for me it is a bit like cheating and it will produce 'artificial' results (IMHO).
Since all these tricks are in the digital domain, I'm not concerned about loosing any detail or sound quality, but nevertheless I believe the results will always be a bit artificial. It's like going back to the times when (analogue at the time) sound effects appeared with different equalization settings.
I'm not interested at the moment.
That said, I'm interested in this part:

I'm sure with the DSP power available we could try other ideas in the future, the combination of the DSP + FPGA gives us vast future possibility's

So, since there isn't option 1 with Bluetooth, I will sign for option 2.

But I believe this is going to get things a little bit more complicated in the future with two different software upgrade paths...


Michael

 

[misterdog]

I'll issue a PCB which can be fitted and tested on a few owners MDACs for first impressions -

Plenty of volunteers available in Sheffield.

 

[shaolin]

Thank you John for comprehensive reply. Now I'm tempted even more!
Such a precision (56bits) looks very promising. And using REW is a good bet I believe.

DSP:
Sophisticated tools for room correction like Acourate or Audiolense create complex FIR filters for both frequency and phase corrections.

Does REW applies also phase corrections? I'm not familiar with this yet, wondering what are Acourate or Audiolense advantages over REW (if any).

 

[Mikas]

If we're going away from the original concept then I want network streaming capability added as well. For me wired would be ok, but others will want wireless as well. Plus a nice app to control everything.

Cheers,
Johan

+1 to that !!

But unfortunately (I've tried this approach with John before), that's like transforming the MDAC2 more into a little computer, and that will take more time and resources. And also I'm afraid there aren't enough people interested in.


Michael

 

[ChrisPa]

It would be more cost effective to build two versions of the MDAC2 - the same PCB just different component population.

Understood

1. Entry level version without DSP + Bluetooth

2. Advanced version with DSP + Bluetooth

Those who are interested in the DSP + Bluetooth version would be asked to contribute an extra £100 to cover the extra material + development costs.

On the basis of what you've outlined I'd definitely go the DSP option, even if it remains unused for some time

To be quite honest, I'm quite interested in trying the DSP room / speaker correction as it can be performed in full 56 bit mathematical precision - with this 56 bits being feed directly into the FPGA modulator thus maintaining as much mathematical precision as possible reducing DSP rounding errors.

I'm sure with the DSP power available we could try other ideas in the future, the combination of the DSP + FPGA gives us vast future possibility's

Both versions will have the same analogue stages - so its only a question of extra digital features. I'm hoping that with the extra DSP power we might be able to design more advanced filters then with just the FPGA alone. The FPGA is ideal for the ASRC, Modulator and handling the DSD BitStream - while the DSP is better suited for Digital filtering / EQ etc.

Totally agreed. It all sounds excellent

The idea is to use the software such as REW’s Room EQ Wizard http://www.roomeqwizard.com/ to handle the analysis and user front end - with Dominik only having to write the MDAC2 DSP software to work with the downloaded filter coefficients from REW’s Room EQ Wizard software. As we now only need to worry about programming our on-board DSP with the measured results we are not having to reinvent the wheel, and dramatically shortens Dominiks workload.. The advantage of on board DSP is that we can maintain double precision (56bits) from the DSP to the modulator without needing to truncate the data, and also work with 192KHz input sample rates.

<snip>

I agree, I'm thinking along the lines that the DSP would just cancel room mode resonate peaks - and ONLY working upon the peaks - not trying to boost nulls... The beauty of having the DSP is we can try a few "correction" strategies, nothings fixed in stone...

Totally agree with only looking at peaks

REW is a good tool, and using it as a user interface for generating correction seems absolutely right. It may need watching as in its default form it tries to correct pretty well everything including notches.

The programming interface for najda (Najda Under Control) may also work as its output is in the form of a series of paramater files passed over to the DSP unit
http://www.waf-audio.com/products.php?pos=1&sub=0&lang=en
http://www.diyaudio.com/forums/digital-line-level/215379-dsp-xover-project-part-2-a.html

However, all that is for the future. As long as the silicon is on board it can be configured later.

 

[JohanH]

+1 to that !!

But unfortunately (I've tried this approach with John before), that's like transforming the MDAC2 more into a little computer, and that will take more time and resources. And also I'm afraid there aren't enough people interested in.


Michael

Adding DSP feels about the same amount of transformation....

REW requires an additional SPL meter; as I already have a DEQ2496 + mic I'd rather have the digital loop function added.

Cheers,
Johan

 

[JemHayward]

Put me down for the "Advanced with DSP" version! Having used digital correction before, I'd be keen to use it again to control my annoying resonance, though I can (and do) live without it, as my Summits are pretty much flat in my rather odd room, whereas my Quads were all over the place.

(Also wondering what a single driver speaker with digital correction would sound like... a modern version of the Baxandall speaker that was featured in Wireless World in the 70's. A friend built a pair, for their day, they were pretty amazing.)

 

[ChrisPa]

This is sacrilegious.

All that fine detail, dynamics, texture and harmony sacrificed for more impressive measurements.

At least it is optional. It reminds me of seat belt reminders in cars that sound off in rebuke if you unplug your seatbelt a fraction of a second before the car reaches a complete standstill, i.e. a nasty and annoying little gimmick.

If it's what the market wants. . .

That depends on your religion

Off topic, but...

Every item in the reproduction and recording chain applies eq to the sound we listen to. Within our hobby we mostly try to sort the eq out by box and cable swapping. This isn't always the most sensible way of achieving the desired result

Digital eq has been applied whilst recording, often many times over. If nothing else then the anti-aliasing filter on the ADC input, and the resampling process of going from whatever the native sample rate was to 44.1KHz

In particular, speakers are equalised to flatten the response, whether through driver, cabinet, acoustic loading or electrical design, and through the use of crossovers.

I know that the biggest limitation for my system is the room. Some of this can/should be sorted by room treatment - in particular, reflections. However, for simple resonant peaks, and for many other gentle corrections, corresponding eq. at a high resolution is a perfectly valid way to correct.

Adding DSP feels about the same amount of transformation....

Primarily, it's only 1 chip on the pcb. This is a world away from a computer + network + user interface. In particular the 'nice app to control everything'

Does REW applies also phase corrections? I'm not familiar with this yet, wondering what are Acourate or Audiolense advantages over REW (if any).

Where the anomaly is is a natural resonant peak, a simple corresponding filter inherently applies the corresponding phase correction. Hence the reason for only treating peaks.

Otherwise phase corrections tend to be of greatest relevance in the crossover.
John hasn't proposed this for use as an active crossover
Yet