TOPIC: DDRC-88BM - how many Dirac filter taps?

Hi,

how many filter taps does the DDRC-88BM uses for Dirac for each channel?

Thanks.
Regards,
Ben

This kind of question is usually met with the response that everything about the internal workings of Dirac Live is proprietary. However it's no secret that Dirac Live uses a combination of IIR and FIR (their "mixed phase" filtering).

Why do you ask (just curious)?

Hi,

i want to compare to other devices using Dirac.... so it's interesting to see how many filter taps the device used for each channel.

Thanks.
Regards,
Ben

The eight-channel MiniDSP Dirac DDRC-88D (or full rack width DDRC-88A variant) is built around the same underlying hardware and SHARC processor (Analog Devices ADSP21369) as already found inside the eight channel MiniDSP OpenDRC-DA8.

Both machines use the same exact DSP chip - both running at 333MHz processor clock, and 32 bit floating point math.

The OpenDRC-D8 version runs the miniSHARC 4x8 plug-in and allows up to max 2048 taps on any individual channel, providing the total for 8 channels does not exceed 9600 taps total (at 48kHz sample rate) or 3400 taps total (at 96kHz sample rate) thus allowing you typically to allocate resources using more taps for your woofers and fewer for tweeters, etc.

Both Dirac DDRC-88 (D or A) machines can run the DDRC-88BM bass management plug-in or the factory eight-channel Dirac Live proprietary firmware, but their max DSP performance in terms of taps is still governed by the hardware, so should be basically identical spec to the eight channel OpenDRC version.

Thanks. How can i decide where to use the taps? Or will it only use some at Subwoofer channel to get the maximum on the other channels?

When i use full 7 channels + Sub.... how will it be used by Dirac (decision for which taps at which channel)?

The Dirac Live software will automatically allocate available taps between the 8 channels in use, as it sees fit, to give the best overall result / compromise based on the transfer functions it's trying to achive - based on the average responses from your measurements. As John Reekie already said, Dirac's algorithms use a hybrid of IIR and FIR filtering to generate the closest possible correction curves from the available DSP resources - it's not trying to do the whole thing within FIR realm only.
Obviously nobody here on this forum knows what your speakers / room / measurement results look like.. so nobody can tell you EXACTLY how those resources will get divided up or what the results will be.

If you have a specific application where you want or NEED to be in total control and customize your taps / transfer functions / correction curves, then rather than buying an eight-channel "Dirac Live" based system, you'd be better off buying four MiniDSP OpenDRC 2x2 boxes instead which can run 6144 taps per channel each, at 48kHz, equivalent to 128 milliseconds convolution window duration, and are currently the most powerful hardware convolution processors on the market for this type of serious pro audio work with professional digital I/O, and are 100% customizable with any 32 bit impulse response .bin file you create and load into it via USB.
In theory, with OpenDRC boxes you could run identical - or even superior - correction curves to whatever the Dirac system auto-generated, but you would have to "do the homework" yourself after taking the mic measurements, to mathematically average the spatial location responses and generate the necessary FIR correction filters, etc. using something like rePhase software.
If you know what you're doing and make a real nice job of it, you can get far superior / more sophisticated results than Dirac method because you can take many more than just 9 spatial measurements, and you can manually customize it all however you want. The Dirac software is just really a time-saving, labour-saving shortcut for typical hi-fi enthusiasts who want good results quickly without too much arduous setup effort on their part.

The main commercial rival to Dirac systems are the DEQX which can also do much of the number-crunching automatically for you in their proprietary software too. DEQX runs 4096 taps per channel, on 6 channels, but costs around £4000-ish. It also runs 96kHz sample rate which effectively halves the convolution window duration(!) so DEQX offers the equivalent of a 42.67ms max window duration (equivalent to what you'd get from 2048 taps running at 48kHz, whereas the OpenDRC offers 6144 taps at 48kHz giving a 128ms window duration, which makes a lot of difference to woofer correction accuracy. You can easily seeMiniDSP offers far better value when you compare the cost of OpenDRC stereo boxes running 6144 taps per channel for $325 / £245 stereo. Buying three stereo boxes would thus cost only $975 / £735 for 6 channels and clearly outperform the DEQX with the higher number of FIR taps and much longer convolution window durations that OpenDRC offers.

Alternatively, if you just want the maximum number of taps / longest convolution window you can find from any piece of hardware, then look for a Logidy EPSI (guitar effects pedal with jack sockets only) which can run 65,536 samples at 44.1kHz, equivalent to 1486 milliseconds of duration from any 24 bit WAV impulse file you load into it via SD card.

But remember the golden rule - any FIR correction solution, whatever the cost or spec, is only as good as the raw mic measurement data you give it.