Speaker placement & room acoustics – DSP

[Jagged24]

I found the discussions on the recent thread started by Hifiwolf quite interesting, but think that DSP deserves more attention. As Serge and others point out in the other thread, DSP is not a panacea that will overcome truly awful room acoustics or poor placement, nor will it replace proper acoustic treatment. But if – like most of us on this forum – you don’t have a dedicated listening room, which limits options for speaker placement or room treatment, then DSP is well worth a try IMO.

I’m posting to share my experience in case anyone else wants to have a go.

My ~6m x 3.5m listening room doubles as a living room, so I have no option but to place my speakers (Audio Physic Avanti IIIs) roughly 2m apart along the longer dimension, about ~0.5m from the back wall, with each speaker ~2m from the nearest side wall. My listening position is almost nearfield – about 2m from the speakers, near the opposite wall.

Room EQ using the recently released miniDSP 2x4HD unit (which I got from Keith at Purite Audio) has made a major difference at relatively low cost (~£200). The 2x4HD has analogue and toslink digital in, can work as a USB DAC, and has analogue outputs for up to 2 subs besides the main speakers. Any standard TV remote can be programmed to select inputs, digitally control volume, mute etc.

Measurements with REW freeware (which can be downloaded with a complete set of instructions from http://www.roomeqwizard.com/ ) and a calibrated microphone (UMIK-1, also from Keith at Purite Audio) at the listening position showed this frequency response (1/6 octave smoothed) (image below= before correction, the 3 curves show the responses from both channels + sub below 600Hz and the L or R channels separately from 450-20kHz):
 



It is quite clear that the frequency response is not optimal, particularly between 30-500Hz  – probably a combination of room acoustics, speaker placement and the effects of the passive crossovers within the speakers themselves (the drive units in the Avanti IIIs cross over at 200Hz (6dB/octave), 500Hz (6dB/octave), and 2.2kHz (18dB/octave))

EQ using REW (which automatically generates filters that can be directly uploaded into the miniDSP 2x4HD unit) allowed this equalized response with L+R channels and sub across the 10Hz-20KHz range (also 1/6 octave smoothed) (image below= after correction):



I can clearly hear the improvements after EQ - it really is night and day.

I’ve tried out a few other pairs of speakers before and after DSP done in a similar way, and in all cases, there has been a marked and easily audible improvement.

I haven’t used many of the very flexible capabilities of the MiniDSP 2x4HD unit  - for example, it could be used as an active crossover for active bi-amping, and even allows FIR filters for phase correction.

Also, miniDSP has recently released a firmware upgrade for ~£200 that gives the 2x4 HD unit Dirac live capabilities – so the unit will automatically scan the room response, and calculate the right FIR filters including phase correction and sub integration, provided you plug in a calibrated microphone like the UMIK-1. This will be useful for those who don’t want to mess around with REW themselves.

All in all, a big improvement in sound for an ~£200 outlay.

 

[Gizza]

I found the discussions on the recent thread started by Hifiwolf quite interesting, but think that DSP deserves more attention. As Serge and others point out in the other thread, DSP is not a panacea that will overcome truly awful room acoustics or poor placement, nor will it replace proper acoustic treatment. But if – like most of us on this forum – you don’t have a dedicated listening room, which limits options for speaker placement or room treatment, then DSP is well worth a try IMO.

I’m posting to share my experience in case anyone else wants to have a go.

My ~6m x 3.5m listening room doubles as a living room, so I have no option but to place my speakers (Audio Physic Avanti IIIs) roughly 2m apart along the longer dimension, about ~0.5m from the back wall, with each speaker ~2m from the nearest side wall. My listening position is almost nearfield – about 2m from the speakers, near the opposite wall.

Room EQ using the recently released miniDSP 2x4HD unit (which I got from Keith at Purite Audio) has made a major difference at relatively low cost (~£200). The 2x4HD has analogue and toslink digital in, can work as a USB DAC, and has analogue outputs for up to 2 subs besides the main speakers. Any standard TV remote can be programmed to select inputs, digitally control volume, mute etc.

Measurements with REW freeware (which can be downloaded with a complete set of instructions from http://www.roomeqwizard.com/ ) and a calibrated microphone (UMIK-1, also from Keith at Purite Audio) at the listening position showed this frequency response (1/6 octave smoothed) (image below= before correction, the 3 curves show the responses from both channels + sub below 600Hz and the L or R channels separately from 450-20kHz):
 



It is quite clear that the frequency response is not optimal, particularly between 30-500Hz  – probably a combination of room acoustics, speaker placement and the effects of the passive crossovers within the speakers themselves (the drive units in the Avanti IIIs cross over at 200Hz (6dB/octave), 500Hz (6dB/octave), and 2.2kHz (18dB/octave))

EQ using REW (which automatically generates filters that can be directly uploaded into the miniDSP 2x4HD unit) allowed this equalized response with L+R channels and sub across the 10Hz-20KHz range (also 1/6 octave smoothed) (image below= after correction):



I can clearly hear the improvements after EQ - it really is night and day.

I’ve tried out a few other pairs of speakers before and after DSP done in a similar way, and in all cases, there has been a marked and easily audible improvement.

I haven’t used many of the very flexible capabilities of the MiniDSP 2x4HD unit  - for example, it could be used as an active crossover for active bi-amping, and even allows FIR filters for phase correction.

Also, miniDSP has recently released a firmware upgrade for ~£200 that gives the 2x4 HD unit Dirac live capabilities – so the unit will automatically scan the room response, and calculate the right FIR filters including phase correction and sub integration, provided you plug in a calibrated microphone like the UMIK-1. This will be useful for those who don’t want to mess around with REW themselves.

All in all, a big improvement in sound for an ~£200 outlay.

 

Thanks for sharing your experiences of DSP. I have had my eye on the HD unit myself and after reading of the results you obtained, will definitely have a dabble myself once my holiday is out of the way and funds become available. Repped.

 

[Jagged24]

Thanks for sharing your experiences of DSP. I have had my eye on the HD unit myself and after reading of the results you obtained, will definitely have a dabble myself once my holiday is out of the way and funds become available. Repped.

Cheers Gary. I think you will be pleasantly surprised how much difference DSP can make - I was! Do make sure you install the latest firmware when you get your 2x4 HD unit, as the first version had some niggles. Happy to help out if you need any advice.

 

[h.g.]

I fully agree that DSP will usually bring significant improvements but plots like the pair shown can be a bit misleading. The smoothing in the plot is hiding the shape of the room modes and what has been done to them by the DSP. A plot of the actual frequency response at the listening position will be less flattering and more useful in working out what is going on and what might help. Plotting the phase may show up regions where the DSP is going a bit wrong since the algorithms have problems handling non-minimum phase regions. A plot of the impulse response will show if the algorithms have a problem with pre-ringing or pre-echo. Some interpretation required since some forms of pre-ringing will not be audible but, based on some of the plots in DIY speaker forums, some of the software being used with minidsp systems does not look problem free in this respect. It is all fixable but to assume all the DSP software being used with minidsp-type systems is going to be glitch free is probably unwise. 

 

[Jagged24]

Thanks h.g. - you make some good points. I didn't think it necessary in sharing my experience with DSP on this forum to show the phase response and impulse response plots, but yes, I have checked them.

Regarding octave smoothing, you are quite correct that smoothing may mask certain room modes, but you will also be aware that it is questioned whether the human ear & brain actually 'perceive' unsmoothed frequency response curves, since there is a significant amount of perceptive 'smoothing' introduced by our brains. Many appear to use between 1/6 to 1/20 octave smoothing to approximate what we hear; I don't take a consensus viewpoint on this, but find that 1/6 smoothing works well for me. (I doubt that an unsmoothed frequency response as you suggest is relevant to what we hear.)

Nevertheless, as you will see even in this 1/12 smoothed frequency plot from 10-300 Hz (blue line) coupled to an excess group delay plot (black line) to give some insight into the phase response, the miniDSP 2x4HD doesn't do too badly for a DIY unit costing ~£200! Here the excess group delay plot represents the slope of the difference between the measured and minimum phase (the excess phase) as calculated using REW - except for a small spike around 120Hz, the excess group delay is pretty flat, suggesting that the frequency response is at or near minimum phase.



I should emphasize that although the miniDSP 2x4HD has the potential to use FIR filters to even better EQ with minimun phase, I've only used IIR filters for the EQ. When I have the time and patience, I may get into calculating the FIR taps.

Your point about the impulse response and pre-ringing is also a good one. I don't have the IR plot handy, but it didn't look bad. I certainly don't perceive any intrusive pre-ringing.

Perhaps the most important issue for me - theoretical points about physics and engineering aside, the miniDSP EQd system very clearly sounds better to my ears!

 

[SergeAuckland]

My understanding from old conventional wisdom is that  third octave smoothing corresponds with the brain's sensitivity. I have no evidence for this, just what 'custom and practice'  over the years has done. 

Sharp spikes or cuts in a frequency response carry little energy, so even if the spikes/dips are large, as long as they're not wide, they could well be inaudible, or if audible, not be anything like as disturbing as their amplitude might suggest. You may recall one of the 'audio watermarking' systems used deep narrow nulls in frequency response to code for the origin of the audio. Although it was widely derided as being audio vandalism, it was surprisingly innocuous in practice, if not completely inaudible. 

So, I've used third octave smoothing for my measurements, and third octave filtering for my equalisers as a reasonable compromise between less smoothint /increased equaliser resolution and greater smoothing/ coarser resolution which makes accurate EQ difficult.

S

 

[Jagged24]

So, I've used third octave smoothing for my measurements, and third octave filtering for my equalisers as a reasonable compromise between less smoothint /increased equaliser resolution and greater smoothing/ coarser resolution which makes accurate EQ difficult.
 

As in my posting above, that's the route I've taken too, except that I find 1/6 smoothing works well for me.

My understanding from old conventional wisdom is that  third octave smoothing corresponds with the brain's sensitivity. I have no evidence for this, just what 'custom and practice'  over the years has done.

There's evidence that the degree of perceptual smoothing varies across the frequency range of hearing, with the ear able to 'perceive' more variation (down to 1/20 smoothing)  at lower frequencies. That's why I cited 1/6-1/20. My own experience is that I can't hear the difference!

 

[h.g.]

Regarding octave smoothing, you are quite correct that smoothing may mask certain room modes, but you will also be aware that it is questioned whether the human ear & brain actually 'perceive' unsmoothed frequency response curves, since there is a significant amount of perceptive 'smoothing' introduced by our brains. Many appear to use between 1/6 to 1/20 octave smoothing to approximate what we hear; I don't take a consensus viewpoint on this, but find that 1/6 smoothing works well for me.

If you were to play a tone with a signal generator and manually move the frequency around you would not hear what is shown in your plot. What you would hear would follow the actual transfer function rather than one with a lot of useful information averaged away. The perceived loudness of a note from a musical instrument would also tend to follow the actual transfer function rather than an averaged one. It is when you play noise with equal energy at all frequencies that a case might begin to appear for averaging away some of the information in the actual transfer function to better represent what is heard.

If you are using parametric IIR filters to reduce the low frequency room modes doesn't that first require the low frequency modes to be well resolved in order to fit the parametric curve to each one and determine the inverse set of coefficients for each one? Or are you doing things differently like, for eample, the equivalent of a 31 band graphic equalizer?

Again I am not criticizing the use of DSP which is pretty much unavoidable for anyone with an interest in high fidelity. I am nitpicking at the details.

 

[Drewan77]

Great to see others embracing this.

After years playing around via software, Mini DSP & finally DEQX, I have the in-room, listening position response below: +1dB -2dB 10hz-23khz

(Twin subs & Open Baffle speakers with digital crossovers, phase & time alignment, Earthworks M23 mic)

 

[VanDerGraaf]

Thanks for highlighting this, very interesting. It makes the idea of trying out the MiniDSP 2x4HD quite alluring.

 

[Jagged24]

Thanks h.g. - your points are well taken & explained as follows.

It is when you play noise with equal energy at all frequencies that a case might begin to appear for averaging away some of the information in the actual transfer function to better represent what is heard.

I did indeed use a pink noise sweep normalized for equal energy; hence the case for using an averaged transfer function to better represent what is heard.

If you are using parametric IIR filters to reduce the low frequency room modes doesn't that first require the low frequency modes to be well resolved in order to fit the parametric curve to each one and determine the inverse set of coefficients for each one?

I defined the LF room modes at different resolutions using 1/20 to 1/6 smoothing to fit the parametric IIR filter coefficients, but could not hear any difference. I settled on 1/6 smoothing as the best compromise.

Again I am not criticizing the use of DSP which is pretty much unavoidable for anyone with an interest in high fidelity.

I'm glad you emphasize this point - it is the main reason that I started this thread. Most discussions on this forum focus on changing equipment as the means to achieve better sound quality, but I think DSP deserves more attention. My experience is that even a modest investment in DSP can have a surprisingly large impact on sound quality. I hope that the discussions on this thread and others like it will motivate more Wammers to have a go. (I'm glad to see Andrew has just come in to say the same thing).

 

[Drewan77]

Nothing made the impact I was looking for until I started playing around with DSP a few years back. I am nearly 60 & finally have a system & room that is damn near perfect as shown in the image I posted (to me, it really does sound as good as the plot suggests).

The hifi graveyard in my loft (& tons of gear I have sold on) is testament to the money, time & frustration I had wasted over the years before this....

Andrew

 

[SergeAuckland]

Thanks h.g. - your points are well taken & explained as follows.

I did indeed use a pink noise sweep normalized for equal energy; hence the case for using an averaged transfer function to better represent what is heard.

I defined the LF room modes at different resolutions using 1/20 to 1/6 smoothing to fit the parametric IIR filter coefficients, but could not hear any difference. I settled on 1/6 smoothing as the best compromise.

I'm glad you emphasize this point - it is the main reason that I started this thread. Most discussions on this forum focus on changing equipment as the means to achieve better sound quality, but I think DSP deserves more attention. My experience is that even a modest investment in DSP can have a surprisingly large impact on sound quality. I hope that the discussions on this thread and others like it will motivate more Wammers to have a go. (I'm glad to see Andrew has just come in to say the same thing).

I agree that DSP can make a significant difference, but it's worth doing after treating the room as much as one is able to, to get the RT60 down to something sensible (which DSP can do nothing about) and positioning one's loudspeakers and listening seat for the flattest response. 

I accept that if loudspeakers have to be positioned less than ideally, and/or  the room isn't ideal, then DSP can turn a sonic disaster into something almost acceptable, but that is trying to turn a sow's ear into a silk purse. 

Much much better by far to get the room/loudspeaker interface good first, then apply DSP to correct what's left. 

DSP  works, and works very well, but my concern is that it gets used instead of getting the room right rather than once the room's right. 

I'm lucky, I suppose, that my in-room response without applying DSP correction is pretty good, so haven't felt the need to apply it, but it did take a lot of Rockwool and loudspeaker placement to get it that way, depriving my wife of a sitting room along the way.

S

 

[Jagged24]

I agree that DSP can make a significant difference, but it's worth doing after treating the room as much as one is able to, to get the RT60 down to something sensible (which DSP can do nothing about) and positioning one's loudspeakers and listening seat for the flattest response.

DSP  works, and works very well, but my concern is that it gets used instead of getting the room right rather than once the room's right.

I certainly agree with you, and indeed, the first thing I say in starting this thread is: "As Serge and others point out in the other thread, DSP is not a panacea that will overcome truly awful room acoustics or poor placement, nor will it replace proper acoustic treatment. But if – like most of us on this forum – you don’t have a dedicated listening room, which limits options for speaker placement or room treatment, then DSP is well worth a try IMO."

In my case, room treatment was not a viable option. I suspect that many others will be in a similar situation.

 

[Tony_J]

Any views on the difference (from a SQ point of view) between the HD and the regular versions of the mini dsp? Would I hear a difference?

 

[Jagged24]

I haven't tried the MiniDSP 2x4, only the HD version, but suspect that audible differences would be minimal if any.

But the HD version has considerably better DSP capability with its more powerful floating point processor, and also offers better EQ options. In addition, there is the possibility to upgrade to Dirac Live in the future.

These improvements make it worthwhile to go for the HD version IMO - the difference in cost is ~£75 I think. Keith at Purite stocks these units and would know.

 

[Tony_J]

I haven't tried the MiniDSP 2x4, only the HD version, but suspect that audible differences would be minimal if any.

But the HD version has considerably better DSP capability with its more powerful floating point processor, and also offers better EQ options. In addition, there is the possibility to upgrade to Dirac Live in the future.

These improvements make it worthwhile to go for the HD version IMO - the difference in cost is ~£75 I think. Keith at Purite stocks these units and would know.




 

Thanks.