NxtWave 8-VSB Equalizer Field Report: Lab and Philadelphia Field Test Findings

I went to NxtWave on Friday afternoon. I was not asked to sign any non-disclosure agreement. I visited their lab and the demo setup in CEO Matt Miller's office. I also got to sit in on some discussions with some other guests who HAD signed NDAs, which eventually made NxtWave uncomfortable (as it should have). I left NxtWave at my own pace; they didn't shoo me out.

My findings:

- The NxtWave equalizer is effective. In the lab, we adjusted to a condition of "completely closed eye" pattern (actually, for 8-VSB it's eight lines rather than an eye). For those of you who have seen demos of the Tektronix equalizer for SMPTE 259 built into the WFM 601 series of waveform monitors, the effect was similar: without equalization mush, with equalization eight distinct lines. NxtWave's goal is to be "best of class" (i.e., the best 8-VSB equalization). I do not have sufficient experience with all the alternatives to say whether they have met that goal, but I would say that this was the best 8-VSB demodulation I have yet seen.

- It is not magic, however. The present algorithm could not always handle a zero dB static echo (sometimes it could), for example. It also has some problems with very close-in echoes. The NxtWave people believe that they know what needs to be done to fix these problems and plan to do so in second-generation chips. There is, however, a practical limit to how much can be done at a certain level of power consumption and chip cost. The current package (100-contact plastic) draws about 1.2 watts. NxtWave's goal is for the chips to be able to live in PCMCIA cards. An increase in equalization complexity that raised the power level to over two watts would mean different packaging for heat removal. NxtWave, therefore, has to weigh what is technically possible with what makes business sense.

- There is no manual intervention required to handle different echo ensembles, nor was any manual intervention used in the "real-world" demo other than to change channels and orient the antenna.

- The ability to handle dynamic multipath was demonstrated in the lab. The appearance on the spectrum analyzer was that of a waving flag. The other guests and I both noticed, however, that the effect of the equalization on the dynamic multipath was not constant. At one point in each cycle, which appeared to be when the notch was in the vicinity of the pilot, the pattern closed a little. There was clear correlation between the pattern closing and the frequency of the dynamic multipath, but I cannot say for sure that it was at the pilot.

- There IS a sensitivity reduction caused by the equalizer, but it's not a lot. I saw demodulation at a C/N in the vicinity of 15 dB. With a ghost ensemble (I personally viewed ensemble "A" being tested), the reduction was on the order of 1.0 to 1.5 dB (C/N around 16.5 dB). NxtWave's specs submitted to the FCC claim the same reduction for ensembles "A" through "G."

- The field tests conducted by NxtWave (which I did NOT personally see) were conducted under conditions similar to those used in the non-Sinclair testing: van, 30-foot mast, log-periodic directional antenna on a rotator. The big difference is that there was no MPEG decoder. They tested based on "eye" (line) pattern. I believe that to be valid methodology (I also believe Sinclair used valid methodology in its tests -- in both cases one could argue about the equipment used and sites selected).

You may wish to take this entire section with a grain of salt because it's secondhand, but I think there's a very important point at the end that is quasi-first hand. I'll concentrate on just the Philadelphia field tests.

There were seven sites: Sites 1-4 were NOT in center-city locations; sites 5-7 were. Sites 1-4 were said to be successful on all four Philadelphia DTV stations. That means that, with the antenna rotated for maximum signal strength, a clear eight-line pattern was demodulated.

Sites 5-7, the inner-city sites, were NOT successful on all stations. It is said that CBS was okay at all sites, ABC less so, and NBC and Fox not receivable. That's in approximate descending order of transmitting-antenna height and effective radiated power.

Now comes the quasi-first-hand stuff: NxtWave personnel described what they called a "tunneling" effect at the bad sites, where a signal bounces back and forth between buildings, creating a very long ghost. We discussed this at length and eventually got around to what it looked like on a spectrum analyzer. We were at the demo setup at the time and had a spectrum analyzer in front of us so we could point and gesture and say things like, "It looked like that part, but maybe 10 dB peak-to-peak."

As a result, we came to the conclusion that what the NxtWave personnel had seen at the inner-city sites where some DTV stations could not be received very closely matched what I had seen once before. The place I had seen it was at the high-rise, Inner Harbor apartment site at the Sinclair tests in Baltimore.

Take this for what it is worth. What is odd about it is that the NxtWave personnel believe that problem to be caused by a very lengthy echo, which would supposedly cause a problem for COFDM if it exceeded the guard-interval length. Yet, there was clearly perfect and stable COFDM reception in the apartment; it was the 8-VSB receivers that failed. I was there. The Sinclair personnel couldn't understand why the 8-VSB had failed there, since there were no huge nulls visible on the analyzer.

- NxtWave plans to send a team to my apartment to record the DTV spectra and take them back to the lab for testing. They say they will tell me: a) whether they could demodulate the signals at all, and b) whether they think they could be demodulated cost (and power) effectively. I like the idea of the famous (infamous?) Schubin site.

They also reminded me (DUH!) that, even though I can't get DTV reception in my apartment with the receiver/decoder and antennas available to me, I COULD see what the spectrum looks like. As soon as I get a chance, I'll drag over an analyzer and have a look.

- Now to the meat: The "real-world" demo. NxtWave is using a thrown-together setup with different manufacturers for the different components (tuner, SAW filters, MPEG decoder, etc.). The MPEG decoder was not always working well (it appeared not to always refresh with I-frames), but I did not have any difficulty distinguishing between a decoder problem and a demodulation problem.

- I decided to compare apples and apples. When we walked over to the demo setup, an NTSC receiver was tuning in the CBS analog station in Philadelphia on rabbit ears, and the NxtWave setup was getting the CBS digital via a directional UHF antenna. The CBS analog looked awful. BUT the CBS analog is VHF channel 3. I asked for an analog UHF near the CBS digital (channel 26). We chose a Fox station (WTXF) on channel 29 and split the antenna three ways (to the analog TV, to the DTV receiving apparatus, and to the spectrum analyzer). The analog station came in superbly. I would rate the analog picture quality as "very-good to excellent." NxtWave personnel wandering by glanced at it and said, "Wow! How'd the analog get so good?"

- The strange antenna referred to in a previous posting about NxtWave is a Silver Sensor "Antiference" indoor set-top yagi from the UK. It was possible to use it to get not merely excellent analog reception but also perfectly stable DTV reception that was not affected by our walking around the room behind the antenna. The antenna DID have to be positioned, but that was pretty easy, and NxtWave's chip offers a tap-energy output that can be used as a positioning indicator (or for Charles Rhodes' crossed-dipole electronic-steering idea). The antenna was about six feet above the floor on a stand.

- We also tried a standard cheap UHF bow tie of the sort packed with every TV set in the U.S. (it was also one of the antennas used in the Sinclair testing). Its positioning was more critical, but it, too, could be positioned so as to provide absolutely stable DTV reception. There was not "just one" position in the room for it. Perfectly stable reception was possible from a height of only about four feet. The bow tie, too, was not affected by our merely walking around the room behind the antenna, but, with a metal chassis-grille plate, I COULD cause reception to fail at certain points (looking at the spectrum analyzer, I could also IMPROVE reception, with the plate acting as a back reflector).

- Please note that, while we were able to make the DTV reception fail (by misaligning the antenna or stalking it with the chassis plate), we never got the ANALOG to fail completely. Channel 29 looked great. When we tested ABC's digital at channel 64 and NBC's at channel 67, we switched to an analog Philadelphia station (WPSG - UPN) at 57. It wasn't as good as WTXF, but it, too, was quite watchable when the DTV failed. We passed other, perfectly good UHF signals between 29 and 57, too. It is clearly not the case that there will always be DTV reception where there is analog reception.

Please also note: From NxtWave's offices, the Philadelphia stations (analog and digital) are all essentially co-located. BUT, without moving the antenna, we also got good reception on channel 58, an analog New Jersey Network (PBS) station (WNJB) in New Brunswick, New Jersey, in the opposite direction (and probably farther away).

I'd say the Philadelphia stations are some 25-30 miles from NxtWave. Based on my previously driving and taking a train over a similar route, I'd say the terrain is relatively flat (some low rolling hills). There IS a road outside the demo-room window, but it is considerably lower (at no time was there any correlation between road traffic and reception). There are trees, however, across the road, which the signals probably have to pass through.

- At no time when there was successful DTV reception at the NxtWave demo did the spectrum analyzer display appear to be anything like what I saw in Baltimore. There were no multiple deep notches or slopes. Furthermore, while I could see myself affecting the spectrum display as I moved around the room, I had to look carefully to see the effect. In Baltimore, at the 10 West Lee Street site, we all felt we had to stop breathing lest we affect the test. The slightest hand motion was clearly visible on the analyzer, and, again, we were standing behind the antennas.

This is not meant in any way as a criticism of NxtWave's demo. They did not do anything to prevent my affecting the spectrum (although the antenna was positioned closer to the window instead of farther from it). It's just very clear that the sites were quite different. That's to be expected. I saw inner-city sites in Baltimore; NxtWave's offices are suburban.

- There WAS, however, a spectrum problem I noted at the NxtWave demo that I never noticed in Baltimore: impulse noise. I might not have noticed it in Baltimore because there I was looking only at DTV spectra. At NxtWave, when we were looking at ANALOG spectra, little spikes would pop up here and there amid the three NTSC peaks.

- Other items:

- NxtWave has never called their chip a "miracle" and objects to that characterization. I will cease to use it henceforth.

- NxtWave personnel seem to feel that diversity reception would be necessary for 8-VSB mobile applications.

- NxtWave has looked at a lot of DTV signals from a lot of stations. "Some of the signals we've looked at give us concern about the consistency of VSB encoders/modulators," said CEO Matt Miller.

- NxtWave's chip works fine with QAM as well as VSB.

That's about it. If anyone has questions for me, I'll try to answer them. You might also try Matt Miller ([email protected]).

Mark