NHT Xd system and other digital active speaker systems

[Daverz]

After the original post, I decided to try some passive speakers again, the Buchardt S400. After 8 months with the Buchardts, fine as they are, I was feeling a bit restless, so put back the Xds. These have an immediacey, vibrancy, and color to the midrange that I miss with all my speakers and they float images in space better than any speakers I've heard (albeit a limited spectrum of them). I guess I'll use the Xds as long as the electronics still work. I don't think it's cost effective to try to replace all the electronics.

I wonder if I can EQ in the same "vibrancy" for other speakers. Looking at the Soundstage measurements, there's a slight shelf between 2kHz and 4kHz.

 

[Deleted member 4708]

I have a 2.2 NHT XD system for the past 10 years. Spent 3k$ (NHT discounted them when they gave up on this system) and couldn't not ask for more. Since I am running 2 subs I installed the 135Hz crossover (compared to 110 Hz) to the subs. The only problem with this system is the number of cables you have to run. Everything else is perfect imo.
In many ways this was the end of my hifi OCD - I soon understood how many sub-par recordings are out there.

 

[Daverz]

I got these new on closeout for $3k circa 2008 (original price was $6k). I hauled them out of the side room yesterday and hooked everything up.

I finally assembled everything needed to replace the NHT electronics (XDA unit) and the broken XDW subwoofer:

* The HiFiBerry DAC8x 8-channel DAC hat for the Raspberry Pi 5 for the 6 channels needed if the digital crossover (subs, mids and tweeters).

* 2 Fosi V3 aps (just the single-ended ones, not the V3 mono) for 4 channels of amplication. Super cheap, but sound fine.

* 2 SVS 1000 Pro subs. Overkill pricewise for this little project, but I already had them to use with my Buchardt S400s.

I used mdsimon's very easy to follow instructions to set up the DAC8x with CamillaDSP on the Pi.

The first thing I wanted to do was create crossover filters for these. If I could directly measure the transfer function of XDA processor, that would be very easy, but I don't have the equipment for it.

Instead what I did was a set nearfield measurement with a calibrated micripohone of each driver from a 6 cm distance on each driver axis. First with the XDA driving the speakers, and then directly connected (so no filters) using one of the Fosi amps. Then I used these nearfield measurements to "reverse engineer" the XDA filters for each driver assuming the basic starting point of Linkwitz 8th order filters.

There's a question of what the actual crossover point is for the tweeter. John Atkinson at Stereophile mentions 2 numbers: 2100 Hz and 2500 Hz. My own nearfield measurements (and his plots) suggest 2300 Hz is a better match.

My final results look very much like the transfer functions measured by Audioholics (I didn't measure the XDA's sub output).

I then used Acourate's "Delay Measurement with Loopback" to measure the unfiltered delay between the mid and tweeter (the mid is 4 samples behind the tweeter at 48000 samples/second) and adjusted the crossover delays to compensate. (The same for the subs, but that's dependent on the subs and their location.)

FInally, here are the in-room responses for XDA (without subwoofer) and DAC8x + Fosi (left channels; right channels tell a similar story):

Differences in the low end are because I used 150 Hz for the sub/mid crossover frequency and applied some PEQ filters in the SVS sub. Note the nearly flat in-room response between 2 - 10 kHz, due, I assume, to the directivity of these speakers. The speakers are well out in the room, and I sit 2.35 meters away from them.

Subjectively the audible result is excellent. The original system had what I'd call a "dry" sound in the upper mid/lower treble that is gone with the Dac8x and the little Fosi amps. Without having to go through the ADC stage in the XDA, the noise level is much lower and the sound seems more transparent. The XDA produces audible hiss at the listening position (consequence of a noisy ADC and very high gain), while I have to put my ear a couple inches from the tweeter to hear light hiss in the DAC8x + Fosi system.

 

[jhaider]

Interesting. Does anyone know of a site where you can enter a FR target curve and it spits out EQ parameters?

Maybe one of the headphone EQ sites can be adapted to this task?

I'm thinking about using my XDS as nearfield surrounds. Truthfully I'll probably just use the Rick Craig designed passive xovers that came with them and a 2-channel amp, but I do have a decent 4-channel DSP amp on hand (Powersoft Mezzo 604) that could be used as well.

 

[Daverz]

Interesting. Does anyone know of a site where you can enter a FR target curve and it spits out EQ parameters?

Maybe one of the headphone EQ sites can be adapted to this task?

I'm thinking about using my XDS as nearfield surrounds. Truthfully I'll probably just use the Rick Craig designed passive xovers that came with them and a 2-channel amp, but I do have a decent 4-channel DSP amp on hand (Powersoft Mezzo 604) that could be used as well.

REW can do that if you have a measurement mic.

 

[Daverz]

I finally assembled everything needed to replace the NHT electronics (XDA unit) and the broken XDW subwoofer:

A few updates to my post:

* About the competing tweeter crossover numbers, I missed this bit in Atkinson's followup in Stereophile where he got some new filter files from NHT with crossover changes:

"[...] crossover point between tweeter and midwoofer moved up a hair (to 2.4kHz from 2.3kHz) to increase power handling [...]"

NHT Xd active loudspeaker system John Atkinson January 2006

John Atkinson wrote about the NHT Xd system in January 2006 (Vol.29 No.1): When Kalman Rubinson reviewed NHT's $6000 satellites+subwoofer loudspeaker system in the November 2005 issue (pp.105–115), he was very impressed, concluding that "The NHT Xd is the best thing to come down the pike in a...

www.stereophile.com

So I was right about 2.3 kHz being the stock crossover point, but this also suggests experimenting with a 2.4 kHz crossover point. As far as I know, these new filter files were never made generally available. Nor I can't find the additional 135 Hz crossover files that were available at the time (damn linkrot!)

* I missed that the mid/tweeter crossover slope is steeper than the subwoofer/mid crossover: Linkwitz 12th order fits better here than 8th order. So the midrange crossover is asymmetrical (8th order on the left side and 12 order on the right side).

* When I remeasured the midrange/tweeter timing difference it was 3 samples at 48kHz rather than 4 samples. This is with the Dirac pulse method in Acourate, which may be sensitive to noise. I should probably confirm with @mitchco's time alignment method. The crossover filters attached delay the tweeter by 3 samples @ 48kHz.

* I've since replaced the Dac8x with a Motu Ultralite Mk5.

I've attached the 150/2300 Hz crossover files that I created. The filter files are all 65536 samples in double (64-bit float) format and for a 48kHz sampling rate. They are

XO1L48.dbl, XO1R48.dbl: generic Linkwitz 8th order crossover files at 150 Hz (no attempt to correct the Xdw subwoofer output, since I don't use an Xdw).
XO2L48.dbl, XO2R48.dbl: midrange crossover at 150 and 2300 Hz. Linkwitz 8th order on low end and 12th order on the high end.
XO3L48.dbl, XO3R48.dbl: tweeter crossover at 2300 Hz. Linkwitz 12th order.

It's interesting to note Atkinson's description of the Xds that "there was a somewhat reticent quality to the highest frequencies". I noticed this, too, with the original Xda electronics. The issue is not there with the CamillaDSP/Ultralite Mk5/Fosi system. The treble sounds open and extended. I suspect the A/D was the bottleneck in the Xda.

However, even with different electronics and subwoofers (a pair of SVS SB-1000 Pros) I have had trouble getting a subjectively balanced integration between the Xds midranges and the subwoofers as Atkinson describes with the original system. Double basses tend to have a somewhat hollow sound, for example. I've carefully aligned with subwoofers with the mains using the sinewave convolution method:

https://www.audiovero.de/freedownload/Time_Alignment_of_Drivers_by_Sinewave_Convolution.pdf

but the lack of integration is still a bother. (I'm not sure how to verify this with measurements). I have the subwoofers against the front wall about a meter behind the mains, which may be too much distance at this high a crossover point? Or maybe it's just inherently difficult to integrate a subwoofer with the small midrange?

Apart from this issue, this is the best stereo reproduction I've heard in my room.