TOPIC:

Hello Winfried,

Hoffmann wrote:

- What is rePhase's capability or your suggestion for Driver Offset alignment?

The current version (0.9.9) includes offset alignement capabilities within the impulse, but you would be better off using the channel delay functionality of the miniSHARC to save those precious taps for your filters.

- Not having the Math Background... Which FIR x-over slopes are possible when using rePhase with miniSHARC at 96 ks/s?

I would really urge you to use the 48kHz version of the plugin.
Doubling the sampling rate leaves you with a quarter of the FIR correction capabilities. That would probably make your 4-way project impractical for the miniSHARC platform.
As for the possible slopes you can try for yourself in rephase by looking at the resulting curve (red) for a given target curve (blue), number of taps and sampling frequency.

Here are a few examples of what you can expect at 96kHz and 48kHz (you can probably double that number of taps on some specific channels).
The taps vs slope vs frequency vs sampling rate relations are linear (ie twice the frequency = twice the slope, twice the number of taps = twice the slope, twice the sampling rate = half the slope, etc.)

Hope this helps

Thomas

Hello Thomas,

thank you soooo much for your instructive and eye opening answers! Well, if I use 48 or my preferred 96 kHz sampling may be a decision I make while implementing (latest when I run out of taps... ;o) )

Thinking about the concept, it appeared to me that, potentially, I could do most if not all Amplitude Optimisation with the available PEQs and Cross overs with the IIRs, IF it was possible to to the phase linearisation basically when the Amplituude stuff is done. I hope I can express myself clearly enough:

- Cross Overs with IIR filters (probably LR4)
- Amplitude Linearisation with PEQ/shelving IIR
- Amplitude Target curve alignment with PEQ/shelving IIR
- Phase Linearisation on basis of acoustical measurement of above together with drivers

All this obviously in some iteration loops. What do you think? Is that a realistic approach?

Thank you for more (mabe again sobering) hints!
Winfried

For this approach the best solution would be to have a stereo convolution engine on the input section, similar to the openDRC plugin, followed by a pure IIR 8 channels crossover, all handled within the same miniSHARC module.
There was some talk about such a plugin in the past, but I don't know if/when it will appear (or not )

As a rule of thumb, in practice you can linearize the phase of a crossover with half the taps you would need to build one channel the amplitude filter with a linear phase.
ie looking at the curves above, you would need "only" 512 taps at 96kHz to linearize the phase of a LR 48dB/oct filter at 500Hz.
And then of course if you can put that linearization filter on the input you can linarize the phase of several filters in the same time (one crossover= two filter, and the upper channels will come for free...)

OK Thmas. Very good suggestions!

Thinking about that a seemingly wild and maybe outrageous idea sprang to mind:

If I understood you correctly, a MiniSHARC board could potentially do the preprocessing / overall FIR/IIR Stereo-Filtering and the 4-way Cross-over (maybe with additional FIR processing) could be done with a second MiniSHARC connected with one I²S channel. I understand two separate USB connections would be needed, thats manageable... But would this approach even make any sense?

The question also arising is one of total latency, though. My FIR Filtering could start at ~ 200 Hz, lowest cross-over 180 Hz, Filtering below that could be IIR only. Anything you can suggest latency wise here?

Thank you for one more advice, I hope I'm not too insisting...

Greetings,
Winfried

In fact I was thinking about a stereo FIR convolution (similar to the openDRC plugin) on the inputs, and then 8 channels of IIR crossover, all this in a single miniSHARC module.
The hardware should definitely be able to do this (with probably fewer taps than the openDRC, but 4096 taps would already be plenty for example), but the plugin does not exist yet.
As for the latency, once again you will have to check what you can get with rephase, with either "middle", "energy", or any arbitrary (is time, distance or percentage) centering.
There is no (good) reason to have two convolutions acting on the same channel (eg on the inputs and then again in the crossover).

What is your project (drivers, intended crossover points, etc.) ?

I made a comment/request along these lines a while back. I think it might still be on their to-do list, but not sure.

www.minidsp.com/forum/opendrc-series-sup...in-minisharc-plugins

Dave.

Hi Dave,

thanks for the link, I have revived that topic. Let's see what comes off it.

Thomas,

I did not mean to hijack or side-track your rePhase Intro Topic with my Project. Maybe it would be better to start a separate "4-Way FIR miniSHARC rePhase project" topic if you agree to keep posting there as well. ... with all the descriptions necessary and the state of it and so on.

Regards,
Winfried

Hello Thomas,

sorry for the delayed answer, got distracted. OK, a few word regarding my Project:

I'm owning a set of 1992 Backes&Müller BM-8, 3 Way fully active Speakers.
The Drivers each have sensor Feedback: 245 mm Woofer, 120 mm Midrange, 38 mm Dome Tweeter, Cross Overs are 180 Hz / 1300 Hz
I'm adding an Air Motion Transformer Tweeter as 4th Way Setting in @ 3500 Hz & driven by a current source poweramp module.

The whole set fits into a new self designed cabinet, where all Drivers are as near to each other as mechanically possible, re-designed 750 VA Powersupply, an own volume for the electronics.

The original analog cross over goes away and I want a 4-way DSP cross over with Driver time alignment and FIR linearisation, ready for compromise use of PEQs if insufficient FIR Taps are available.

Biggest obstacle for usage of miniSHARC is, though, that I want to be able to listen to up to 24/96 Musicfiles without re-/down-sampling...

So, at this point I'm actually not so sure miniSHARC will make it into this project.

Kind Regards and thanks for the valuable advice!
Winfried

Hi Winfried,

Nice speakers!
What current amp module are you using?
Current drive is probably the most overlooked technology when active filtering is used...

Regarding your crossover, building a 4-way minimum-phase IIR crossover is no easy task, and is in fact much more difficult than building a linear-phase FIR crossover from the start, because of the way phase behavior on one channel impacts others.

Here is Linkwitz explanation: www.linkwitzlab.com/frontiers_5.htm#V

You have to include all the successive HP filters of the lower channels in the upper ones, even including the natural box one, to recreate their phase shift and get a proper phase-coherent summation.

Hello Thomas,

thanks for your feedback and your interest in my project! To me this seems starting a drift off-topic, but it's your thread so I continue...

I agree with your assessment and advice very much! And because of all the conceptual hurdles in miniSHARC to make my activation work, I've actually started to consider alternative solutions...

First of all the Amp-Module is the original one from B&M "just" extended by a fourth poweramp and modified by cutting out the analog x-over and extending the inputs to the Poweramp/Feedback-Control to the outside (Lundahl-transformer-coupled) in order to connect an external four way crossover/convolver with 8 DACs.

So far I have run three-way DEQX controlled set-up where the DSP does the FIR cross-over/convolving and IIR/PEQ Fine Tuning & Room mode damping.

Well, the fourth or AMT Tweeter way lead me to think about a non-DEQX Solution. And the good experience with a miniDSP 2x4 Implementation for a friend lead me to consider miniSHARC. Although, in the meant time, I'm having second thoughts because of the limitations popped up: Low number of FIR Taps, only one Sampling Rate supported, meaning resampling involved, a LOT of hand work getting the x-overs and Filters tuned (i.e. much more than I'm used to).

The Alternative under consideration is: Use a DAC with on-board DSP supporting to implement a 180 Hz 2-Way (hopefully FIR 48...96 dB/Oct., where rePHASE may come in...) x-over and use this on the DEQX Woofer internal I²S connection. This way I retain the DEQX with all its crunching power, multi sample rate concept, measuring/alignment/calibration workflow I'm used to, ..., and get my 4-Way System. Thoughts...

This is where I am now.
Regards,
Winfried