Room Measurement Tutorial for Dummies Part 2

[Krunok]

I guess the real question here with miniDSP 2x4HD is if upgrading it to DDRC-24, which costs $200, to be able to use Dirac Live is worthwhile. So, what's the verdict from you who are using Dirac Live?

 

[March Audio]

Yes, 2x4HD has 4096 taps. What do you mean when you said you didn't like what it said the result would be? You couldn't get the curve right playing with those EQ sliders?

I have tried playing with mini dsps 2x4HD FIR filtering and it simply doesnt have adequate taps and resolution. The filter shape goes all over the place without adequate TAPs.

When designing my DSP speakers I did not find any off the shelf DSP solution which was adequately powerful, hence ended up with the XOs being performed on a PC with full Acourate software.

 

[Krunok]

I have tried playing with mini dsps 2x4HD FIR filtering and it simply doesnt have adequate taps and resolution. The filter shape goes all over the place without adequate TAPs.

When designing my DSP speakers I did not find any off the shelf DSP solution which was adequately powerful, hence ended up with the XOs being performed on a PC with full Acourate software.

What are XOs?

 

[March Audio]

I guess the real question here with miniDSP 2x4HD is if upgrading it to DDRC-24, which costs $200, to be able to use Dirac Live is worthwhile. So, what's the verdict from you who are using Dirac Live?

Watch out for the 2x4HD resampling, its poor with lots of low level spuria. There is a thread here somewhere on it.

 

[March Audio]

What are XOs?

Sorry, crossovers

 

[March Audio]

If you are doing more simple room correction or IIR the demands wont be as high but still the DSP power is lacking for FIR.

 

[Krunok]

I see. Do you think Dirac Live would do better with DDRC-24?

Is this that thread?
https://audiosciencereview.com/forum/index.php?threads/measurements-of-minidsp-2x4hd.2459/

 

[Krunok]

Sorry, crossovers

Ahh, ok. I'm watching The Last ship these days so I translated XO as "executive officer" and that made no sense in this context.

 

[March Audio]

I see. Do you think Dirac Live would do better with DDRC-24?

Is this that thread?
https://audiosciencereview.com/forum/index.php?threads/measurements-of-minidsp-2x4hd.2459/

I have used the PC version of DIRAC and it was probably the best "off the shelf" auto EQ I have found, wees all over audyssey.

From memory it uses a proprietary mixture of FIR and IIR filtering to minimise latency and processing requirements.

edit
https://www.dirac.com/live-home-professional-audio-info
https://static1.squarespace.com/sta...+Engineering+Society,+AES+Sweden+lecture).pdf

 

[Krunok]

I have used the PC version of DIRAC and it was probably the best "off the shelf" auto EQ I have found, wees all over audyssey.

From memory it uses a proprietary mixture of FIR and IIR filtering to minimise latency and processing requirements.

edit
https://www.dirac.com/live-home-professional-audio-info
https://static1.squarespace.com/static/55c8a274e4b09cb562cd3ea0/t/57d159a69de4bbd9d306ac8f/1473337780806/Dirac+Room+Correction+(Audio+Engineering+Society,+AES+Sweden+lecture).pdf

Good to hear! So how do you think DIRAC Live would work with DDRC-24?

 

[March Audio]

Good to hear! So how do you think DIRAC Live would work with DDRC-24?

Oh I think it will work fine, Im sure Dirac wouldnt license it if the hardware wasnt capable, I should probably temper my comments above to make it clear I was talking about full blown FIR filtering and crossovers.

 

[Krunok]

Oh I think it will work fine, Im sure Dirac wouldnt license it if the hardware wasnt capable, I should probably temper my comments above to make it clear I was talking about full blown FIR filtering and crossovers.

From what I have been able to understand they made another firmware for 2x4HD to push it to be Dirac compliant and then they call it DDRC-24. I would also imagine Dirac wouldn't license it if it didn't show to work flawlessly.

I have read your 2x4HD measurement thread and DAC section looks acceptable to me. It looks slightly worse than my Topping D10 but still well in the acceptable range.

 

[RayDunzl]

What do you mean when you said you didn't like what it said the result would be? You couldn't get the curve right playing with those EQ sliders?

The little problem I was chasing at the time is a phase cancellation anomaly caused (presumably) by my irregular room, in the area of 50Hz.

A 50 Hz cycle at 96khz = 1920 samples.

(start mathematically ignorant arm waving here)

I suppose that trying to measure/adjust a 1,920 sample wave using a 1,024 sample container is at the heart of the issue.

---

You can set up rePhase with the 96kHz and 1024 taps, slide its sliders to request a filter, and it will graphically show your request and what it is actually capable of doing given those two restrictions. You can request a sharp correction down low, but it can only come up with a gentle change. At high frrequency it can (those wave samples fit in the container).


RePhase Example with 1024 taps @ 96kHz (2x4HD):

Blue: Requested Frequency Response (solid) and Phase (dotted)
Red: Resulting FIR Filter Frequency Response (solid) and Phase (dotted)

At 10khz, the filter FR and Phase adjustment very closely matches the requested change.
At 2khz the filter FR and Phase adjustment rather closely matches the requested change.
At 300hz, it doesn't get very close,
At 50Hz it can't do anything with it, though a wide area of slightly attenuated FR is seen.

With 6144 Taps @ 48kHz (OpenDRC-DI)

It still can't exactly perform the sharp correction on the 50Hz, but the others are good.

 

[Krunok]

Huh, I'm not sure I'm reading these graphs well. Left y-axis scale is in dB? Why would you want such huge corrections at those 4 points? Are these room modes?

 

[RayDunzl]

Why would you want such huge corrections at those 4 points?

This is just an example that shows how fewer taps results in less FIR filter resolution at lower frequencies.

2x4HD very deficient in low frequency FIR capability
OpenDRC-DI almost good enough

---

Back in the real world, I want to twist the phase in the area of 45 to 55Hz +90 degrees in one channel, and -90 degrees in the other, as an experiment to see what happens to my deep hole in that area.

I can't, because I don't have enough horsepower at those low frequencies with the tools I have at hand.

See this post; https://audiosciencereview.com/forum/index.php?threads/understanding-phase.757/#post-23550

 

[Krunok]

Are you sure you wouldn't have enough taps to cover the correction in 45-55Hz range? That would be odd as that's really a narrow range..

 

[RayDunzl]

Are you sure you wouldn't have enough taps to cover the correction in 45-55Hz range? That would be odd as that's really a narrow range..

The Red lines show the output the FIR filter can produce given the restraints of sample rate and number of taps, given the requested change (Blue lines).

It's a narrow range of frequencies, but the wavelength (in samples) of those frequencies exceeds the available filter length (Like "working memory" in taps), so the hardware is incapable of fully supporting an FIR filter at those frequencies (my guesstimation).

For low frequency adjustments with limited DSP power, IIR (Infinite Impulse Response) filters are used. They (somehow) feed back on themselves as the stream progresses to calculate the next output sample.

FIR analyzes and adjusts the next output sample by looking at a "chunk" of samples, limited by the number of taps (sample holders). Fewer taps = less ability to work with lower frequencies.

There are experts here that can explain properly, but this is my meager understanding.

 

[dc655321]

IIRC, FIR filter length is inversely proportional to the transition-band width and the depth of the stop-band.
i.e. narrower, deeper filters require more taps.

 

[Krunok]

The Red lines show the output the FIR filter can produce given the restraints of sample rate and number of taps, given the requested change (Blue lines).

It's a narrow range of frequencies, but the wavelength (in samples) of those frequencies exceeds the available filter length (Like "working memory" in taps), so the hardware is incapable of fully supporting an FIR filter at those frequencies (my guesstimation).

For low frequency adjustments with limited DSP power, IIR (Infinite Impulse Response) filters are used. They (somehow) feed back on themselves as the stream progresses to calculate the next output sample.

FIR analyzes and adjusts the next output sample by looking at a "chunk" of samples, limited by the number of taps (sample holders). Fewer taps = less ability to work with lower frequencies.

There are experts here that can explain properly, but this is my meager understanding.

Sure, everything you said is fine but I was actually asking if you're interested in correcting only 45-55Hz range why didn't you limit that range in the "Range" tab (lower right corner of your graph)?

 

[Krunok]

IIRC, FIR filter length is inversely proportional to the transition-band width and the depth of the stop-band.
i.e. narrower, deeper filters require more taps.

IIRC, that is so.