Optimizing NHT Xd systems with REW

[Keith Shaw]

Hello folks-

I’ve been experimenting with REW for a few months now, and I thought I would track my progress in optimizing my two NHT Xd setups in a thread here on the Shack. My setup so far is pretty modest- a Dell XPS M1530 laptop, a Radio Shack 33-2050 SPL meter, and some assorted cabling. I’ve had fun experimenting with my setup, though, and I hope to keep learning and making the system sound better. By the way, if I’m treading on some rules or customs in the way I describe things or show measurements, please let me know. I haven’t been reading the forum all that long, so I’m sure there are things I’ll need to adjust.

For some background on the Xd concept itself, here’s a link to the Stereophile review of the NHT Xd. There used to be a page on the NHT site about the system, but it seems to have gone missing (no doubt a casualty of the system’s discontinuation). One wrinkle of optimizing this system, outside of the active nature of its crossovers, is the ability to load different filters via the XdA’s USB port. I’ll show some measurements of the standard filters NHT offers below.

Here is one picture of my main system using 2 XdS satellites and 2 XdW subwoofers. The actual location of each can vary according to the week and my latest measurements.



My optimization strategy has essentially three pieces:

1. Adjusting the relative positions of the XdS/XdW modules for proper phase response in the short crossover range. The crossover point I’m usually using is high enough that the units must be essentially collocated, and the crossover slopes are very steep.
2. Positioning the XdW units to best interact with room modes in the bass region.
3. Where EQ is needed, custom filters from the folks at Audio Consulting Services (a group of former NHT engineers who are pretty handy at creating custom filters for the XdA module).

NHT keeps a page of Xd filters that they offer. According to their literature, they stack up this way:

240305 – The factory filter. Crossover supposedly around 110 Hz, flat to around 27 Hz on the low end.
040805-1 – Same low end, but the crossover point has been raised to 135 Hz to improve the dynamic capabilities of the XdS.
040805-2 – Also the 135 Hz crossover, but the XdW module has been EQ’d to be flat to 20 Hz.

First is a comparison between the two low end options. True to their claim, you do see the “regular” filter begin to roll off around 27 Hz with the 040805-2 filter continuing unabated towards the 20 Hz point.
Yellow = 040805-1
Purple = 040805-2



You can see, especially on the yellow, that the room has a pretty serious hump at 30 Hz, and I’ll show more on that in a bit.

Next is a system response across the crossover region for the 040805 vs. 240305. My placement has been setup for the former, so it’s possible with some shuffling around of the units the response through the crossover region on the latter might be improved.
Purple = 040805
Green = 240305



Looking a little further at the crossover range, here is a plot showing the separate responses of the XdS satellites, the XdW subs, and the combined system (040805 filter).
Purple = whole system
Blue = XdW’s
Yellow = XdS’s



I marked where 135 Hz is located, and you can see that perhaps I ought to be putting “135 Hz crossover” in quotes from this point forward. It sure looks more like a 150 Hz crossover point to me. I’m not entirely sure where that nasty 150 Hz dip in the XdS response (yellow) is coming from. It doesn’t really line up with any room modes I’m aware of.

Speaking of room modes, you may have seen the system dip that occurs around 122 Hz. I believe this is a 2nd order height null, as my listening height (and the microphone position) for my reclined chair is almost exactly at this height. Here are the modes I get some plugging in the room’s measurements:



Absent from that list, though, is the large 30 Hz hump you can see in almost all my measurements. Here’s what I looks like on a waterfall plot. It’s not pretty- you can see I’ve used a 700 ms window, and it’s still hanging around:



I suppose that will probably be one of the top candidates for some EQ down the line.

I certainly have a lot more measurements, but I’ll try to restrict the posting to images that show some progression or pattern. Much of the time, I appear to be randomly moving things around to see the result, both from listening and from REW. I have tried some XdW positions that correspond to limiting the excitation of certain room modes from the list above, with only modest success.

I hope the information is interesting, and I plan to update the thread as I get other plots to show.

Regards,
Keith

[brucek]

Have you experimented with placing the sub modules into the corners to achieve some room gain and extension? Even with the high crossover, I don't think there will be any localization issue, given their stereo nature.

the large 30 Hz hump you can see in almost all my measurements. Here’s what I looks like on a waterfall plot. It’s not pretty- you can see I’ve used a 700 ms window

Generally, with a waterfall you want to maintain the scaling of your measurements, so, use a vertical scale of 45dB-105dB, and a horizontal scale of 15Hz-200Hz. Use the LOG scale instead of the LIN.
The 30Hz problem is a modal resonance (whether the perfect world calculations say so or not) and may need to be EQ'd out.

Any sharp dips you may be concerned with are not a problem. You simply cannot hear them...

brucek

[Wayne A. Pflughaupt]

Very interesting, Keith, thanks for sharing. Wow, those are stunning speakers! The Stereophile write up is pretty impressive, too. A shame NHT discontinued them - I expect that all speaker systems will be like this eventually.

Frankly, I don't see any compelling reason to scale your waterfall graphs the same way as a frequency response graph, because you aren't gleaning the same information from them. If you want to see how the signal decays all the way down to the room's noise floor, you'll want to use a lower graph floor (and it looks like you have a really quiet room!). If you only care to see what it looks like at the lowest levels you'll typically listen to, then raise the graph floor. However, there's no good reason to raise the upper limit above 105 dB. LIN mode is easier to analyze below ~40 Hz, but squeezes everything above that point like an accordion. LOG mode is the opposite - easier to analyze above ~40 Hz, while somewhat compressing below that point. I find both modes to be useful.

It looks like ~30 Hz is the only serious room mode you have, and perhaps a couple of minor ones at ~40 and 60 Hz.

Be interested to see what you have to show us next!

Regards.
Wayne

[Keith Shaw]

Thanks for the comments, Bruce and Wayne.

Bruce, I had never actually considered even trying the XdW units in the corner. Since I'm almost always running the higher crossover point, I'd simply confined my placement options to areas right around each satellite. But of course, now I'm curious....

So here's what I get. The first thing that surprised me was where the corner placement effect was strongest. Essentially no impact at all below 70 Hz? That seems... odd.

Both graphs using 040805-2 filter (135 Hz crossover, flat to 20 Hz response)

Red - Standard XdW placement
Green - Each XdW moved to its respective corner



The best way I understand the impact is that the room mode excitation that's most different between the two placements is the width mode (for the other two modes, the XdW units weren't very far from the front wall, and they were already sitting on the floor). For the width modes, the 1st and 3rd order waves have a null at the center; this is, my centered listening position (and mic position) are a null point for two major modes now being excited. The 2nd order wave, 73 Hz (see the room modes posted earlier), has its peak in the listening position. Sure enough, that is where the additional output begins, though it remains quite high through the 120 Hz area. Possibly a good configuration for someone craving a lot of mid-bass.

I checked the two crossover options at this configuration as well. The 110 Hz crossover is probably more suited to the spacing, at least according to NHT's documentation.

Green - Corner XdW's, 135 Hz crossover
Purple - Corner XdW's, 110 Hz crossover



For several frequency ranges, the two filters almost seem to swap places back and forth, heh.

Lastly, here's a waterfall comparison from the original response curve (same 135 Hz crossover).

As before:
Red - Standard XdW placement
Green - Corner XdW placement



As you'd expect from a minimum phase system, the added output in the 70 - 120 Hz range also elicits a fair bit of extra ringing. It might be intersting to run the comparison again with the mic notably off-center to see how much the pattern changes. Maybe it will show some extra output in the 40 - 70 Hz range. Any bets?

Keith

[Jack Hidley]

Keith,

No change in low frequency response below 70Hz, when you moved the XdW to the corner makes sense.

Assuming that you were already within 4' or so of the corner: At 70Hz, the wavelength is 16'. You were already close enough to the corner to excite any axial standing wave modes very efficiently. 4'/16' is 1/4 of the way from the end of the 1st standing wave mode. This is very close to the pressure peak.

A similar thing occurs with respect to wall reflections also. The comb filtering you get from the XdW being 4' from the corner is a function of the phase shift that the reflection causes. At 70Hz, this is a 90 degree shift. About a 3dB drop, assuming that there was only one perfect specular reflection from a single surface that was magically 4' away no matter where you looked. This doesn't happen in the real world, so you never get a 3dB drop with a calculated 90 degree phase shift. Remember, for our purposes sounds almost always radiate in spheres, so you always have multiple reflections with different path lengths. As the frequency drops below 70Hz, the phase shift keeps decreasing, since the path length of the reflection is a constant 4' and the wavelength keeps increasing.

It's almost always a myth that putting a subwoofer in the corner increases the low bass output. It does increase, and really screw up, the upper bass output however.

With regards to the CSD plots for the two XdW locations. Remember, in a CSD plot, if the ridge is parallel to the time axis, it is the system's energy decaying from some resonant behavior. If the ridge is not parallel to the time axis, then the ridge is due to a reflection (which will still decay monotonically or nonmonotonically, since the energy source powering it is decaying). In your green CSD it looks like there is some definite diagonal behavior between 70 and 90Hz going on.