e-Lips

[richidoo]

While I'm waiting for analog filters, I figured I would try to figure out why AudioLense failed so badly at the G2G, and maybe try out Acourate, another PC based filter generator for convolver.

I spent some time setting up AudioLense again tonight, ran the measurements, generated filters, targets convolutions, etc. I saved it and played it, all 4 outputs are playing full range sound, no crossovers at all. I could have sworn that it was filtering (albeit no very well) last month when I was first using it. Apparently not. Turns out that I was imagining all the "changes" I was making.

When saving the final filters a warning message pops up saying:

"The dummy filters can be used to verify that the playback path, including filtering is working.
The filters attenuates the signal by approx 10 dB, so you know it's working when the sound level is lower than usual.
You should be aware that the dummy filters has full bandwidth, and tweeters may be damaged by high sound presure level.
You can get the real filters by purchasing an Audiolense licence at www.juicehifi.com"

I read that before, and interpreted it that the demo filters are full function to show the power of audiolense but are attenuated too much for practical use. But I learned tonight that the dummy filters really are dummies, just like me. They are just passthrough, with no filtering whatsoever. No crossovers, no EQ, no room correction. I told the developer that his warning message needs to be more clear to say that the filters do NOTHING except passthrough to see if you are capable of setting up all the computer audio crap but give no indication whether the processing is worth a damn. (in so many words...)

So I will try Acourate to see if its demo actually processes the audio. Its documentation is horrible, so almost every step that needs clarification requires a question posted on the user forum. Ugh. The only instruction is tutorial videos which are too simple. But supposedly it sounds good, just like AudioLense. 

Just another day in DIYland

[tmazz]

You would think that the developer would make the demo a fully functioning version that is time limited so that you would set it up, get to like it and then want to buy it when the demo period expired. If all of the filters are nonfunctional, just what is it that they are trying to demo? 

You can't make this stuff up. 

[richidoo]

Ease of use, I guess, that's the biggest benefit compared to his competitor Acourate.
AudioLense produces filter files that can't be timed out because they are used in a different program, so he has to handicap the filters somehow, but making them completely lame seems overkill. Maybe insert loud background hiss or 120Hz hum that can be listened to for eval, but intolerable in the long run.  I understand his predicament, but a clearer message, even on the download page or during setup would be helpful. I should have been more skeptical, then my pessimistic attitude would interpret his warning correctly. No biggie, on to the next thing.  I would just buy it except he wants 390EU for the crossover version, and I would like to hear it before investing. Only in audio can you charge like that, but it's a very limited audience so that's what it has to cost to exist. I will spend some time with the free version "DRC," whose filters run on Linux only, I think. I was hoping to avoid Linux, but it's something I need to learn eventually.  Rather just listen to music if you know what I mean.

[tmazz]

a) it would seem to me that if they put a time limited license on the main program you could use the filters as intended and when the eval period was over the whole program would simply shut itself down. But having never seen the program myself there could be something that I just don't understand. 

b) I know what you mean about Linux. I have had a copy of Red Hat and a tutorial program on my shelf  for ages andevery weekend I say to myself, "next weekend will be the one when I sit down and learn Linux." 

[richidoo]

I know what you mean about Linux. I'm just intimidated to start learning a new thing. I think that is the definitition of OLD!  boooo     OK now I have to do it.    The newest linux installs like Ubuntu are perfectly easy and run really well for standard jobs. Now getting them to work with audio drivers and signal processing with homemade apps is another story. But people seem to be having success with it so it's not impossible. It's just down to how much time you're willing to devote.

The AudioLense program is like a Paint program. You use it to create an impulse.wav file. Then the impulse file is used in a separate open source DSP program (convolver) as a pattern with which the audio stream is manipulated. So making AudioLense 30 day expiration would still allow someone to get busy and finish his creation of impulse files before it expires, then he doesn't really need AudioLense anymore. Yet another reason it shouldn't cost $500, but that's free market and he's been doing it a while, answering lots of support questions from compududs like me.

He chose to handicap the filters by making them flat impulse with -10dB attenuation. At first i thought the attenuation was the handicap, because it increased noise floor if you turn up the preamp or increases digital noise if you crank the 10dB back in digitally. But the -10dB is just so you can tell the difference between the original stream and the convoluted output, to make sure it is actually working.

I like calling it convoluted.

[richidoo]

I told my wife about the dummy filters. That they don't let you hear whether the result is any good but only serve to let you know whether you are able to do the setup of the computer audio, convolution engine, multichannel sound card, etc. She said "then they are aptly named, they filter out the dummies. "  haha     

[richidoo]

After going through the same routine with Acourate, the competitor to AudioLense, I was told it too is not able to produce real filters to evaluate the sonic performance of the software. The filters it produces proprietary and only for use within Acourate for demonstrating how it works, not for actually using in a convolution engine.  At least that only took 3 days.

There is a open source filter maker called DRC. It is quite arcane. Not sure I'm up for that, but I might try it to see how it goes.

http://drc-fir.sourceforge.net/
http://www.duffroomcorrection.com/wiki/Main_Page

I also found another commercial vendor making correction software, in Germany, called Dirac. They look very promising, and expensive. I think they are stand alone black boxes like DEQX, but they say they make software too.
http://www.dirac.se/


Meanwhile the parts for the analog filters have shipped. Maybe hear it this time next week with any luck.
Rich

[tmazz]

Meanwhile the parts for the analog filters have shipped. Maybe hear it this time next week with any luck.
Rich

We can only hope.....

[richidoo]

Yes, definitely looking forward to that moment. (get the fire extinguisher ready)

Parts express (Fedex) and digikey (postal) won the shipping race - both arrived today (2 days!) Mouser with UPS is by far the most expensive and slowest, but the best online ordering and customer service.

I started looking at DRC, it doesn't look too bad, but I don't see how to create crossover filters. I think I can create them in RoomEQWiz, not sure.

[richidoo]

I found another FIR crossover plugin, this one for foobar, from Japan. It has provision for baffle step correction built right in. Not tried it yet though.
http://www.aedio.co.jp/download/index_e.html
Cool website, these guys are passionate about their audio. Translate it from Japanese with google.
http://www.aedio.co.jp/

I contacted Dirac, they do have some products coming out soon which will be more suited to audiophiles. The products available now using their processing algorithms are for the film industry, or high end auto mfgs. They use mixed phase filters, the best of both worlds, IIR and FIR. I think that means linear phase when it's important in the bass, and minimal phase when it's important in the uppers, and fading between them gracefully. There are some white papers on their site, I've not read them yet.

Meantime, while I wait for UPS to bring my parts, I have the subwoofer set up, tweeked and sounding very good. I played a couple symphonies this morning. It sounds good. Plays loud with good tone, except for reflections. Great imaging and tone! I love these Feastrex drivers when they are happy. Listening to Bruch Violin Concerto 3 right now. This is the first time I have ever heard them play symphonic music well enough to listen without major distraction. Should be better with OB bass drivers.

I am running the feastrex drivers full range in open baffle at high volumes, they are hanging in there. The builder told me it's hard to hurt them as long as you back off when you hear bottoming. I only heard that once today playing Bromberg "Wood" a little too loud. But the symphonies were fine even at too high a volume. With a little rolloff and bass driver support, the Feastrex should be able to run with no crossover, or maybe a 1st order cap.

Just this moment my wife hands me a box from Mouser. Time to get crankin again. Crossovers, acoustic panels, and painting the stuff that's already installed.
Rich

[richidoo]

I got the new foobar plugin installed and running. I had to paste the entire userguide into google translate to be able to read it from Japanese in google warped english. That was good for some good laughs.

I was surprised at how good it sounds! The sound quality of foobar DSP seems better than AudioMulch, at least in the bass. This filter has the analog outputs hard coded into the plugin, and foobar has no multichannel routing, so I am forced to us the analog outputs of my soundcard, even on the Feastrex. The feastrex is not quite as see through as with the Buffalo, but very good. A slight bit of dynamic compression is audible, but that might be the current chip amp because I have heard it before. This is the best that the speakers have sounded yet.

The filters in this crossover are FIR, with linear phase throughout the entire bandwidth, so there is no phase error and no group delay at all.

Now i just have to adjust the settings. There are 3 settings. The crossover frequency, is the -6dB crossing freq. The Taps adjusts the size of the filter. Taps are the number of samples in the impulse, so it's the length of the math pattern used to effect the filters. This also adjusts the slope somehow that I don't understand by manipulating the math of the filter. The length of the tap controls the slope of the filter. The filters are not symmetric, but larger taps make them more symmetric.  The final setting is the baffle step correction, which is actually 3 settings, freq to begin attenuation,  freq to end attenuation and attenuation amplitude. It superimposes a baffle step shelving filter on top of the crossover filters to cut the upper freqs, effectively boosting the lower freqs to compensate for the cancellation. Every speaker has this, but open baffles are especially audible because it goes up higher in freq due to the pressure cancellations at the edge of the open baffle. 

Probably need RTA or FR sweep to set the baffle step compensation accurately, especially with 2 different size drivers crossing down low.

In other FIR filters I have been using a long filter length is necessary to insure accuracy of the bass signal. This filter uses relatively very short filters, even as short as 500taps! The impulses used in the other convolver filters are 130000 taps, about 1.5 seconds long at 44.1kHz. So I am worried that the slight woofiness of the bass is due to the short filters. But I don't know enough about the math to know for sure if that's possible in this app where the filters are very simple, pure math, no recorded impulse, just made by the setup buttons. The programmers seem like they are smart enough to not make that error. But even with a very short impulse of only 4001 taps, there were some dropouts occasionally, while there was less when the filter was even shorter. I just don't understand what it's doing. That's OK.

I started with 125 Hz, which sounded best on the Behringer with Linkwitz-Riley 4th order. But on the steeper slopes allowed by FIR this was too chesty, fuzzing up the midrange. The bass drivers sounded too slow.  Lowering to 50Hz cleaned things up, but lost some beef. I have been listening to the Feastrex drivers full range all week with the subwoofer at hig volumes so I'm not worried about 50Hz cross being too low. There is pretty steep slope below that so they are safe. I moved it up to 75Hz which seemed about right. The woofer drops out completely by 200Hz, while the midrange slopes off much more gradually, looks like 4th order slope.  That was 9pm, time for election TV! I will continue to experiment with crossing freq and slopes to fine tune it.

But even with the low crossing, the bass is not clean enough for me yet. If it rises above 150 it clouds the Feastrex. Even playing by itself, it is not as etched as it should be. These drivers have extremely low inductance, extremely powerful motors and easy suspension, with Qts=.7. So it should have the ability to sound very clear. I am using a good amp which gave one of the clearest bass sounds ever on the Ushers, but I think the long runs of RCA zipcord to the woofers, and also for the wiring hookups to the 4 separate coils, then connected to the drivers with spring clips might raise impedance so much that the damping is shot to hell. I've got some better speaker hookup wire coming, and will solder everything together and use posts and SC3 SCs to the drivers. If it doesn't sound good then, I can try notching the resonant peak of the drivers, before looking for other drivers. But I think there will be a big improvement with the wires and connections. These drivers are supposed to rock. We'll see, I have high standards after the Eton carbon fiber hexacone drivers in the Ushers. They cost about the same, but Etons are not designed for dipole.

All the parts for the analog filters are here, but I was too busy to work on it last two days. I got a lot of different size resistors for experimenting with crossover points, but it requires 40 solder joints to change.

[Face]

While building my last pair of speakers I was using cheap zip cord(tinned M-Cable?) to test the woofer and it didn't sound right either.  I then switched to something better, and even with with spring clips holding everything bass tightened up immediately.  I also found using cheap brass binding posts and brass nanners(monoprice) as adapters degraded sound quality more than spring clips. 

[richidoo]

Thanks Face, that's good to know that you found the spring clips to be OK. That jibes with my experience too. This morning I ran a single pair of drivers in stereo, with their dual coils zipcorded in parallel, but using 5ga JPS speaker wire for the main run, the bass really came to life. Detailed and punchy like it should be. Looks like the 15 feet length of 14ga zip was the killer.
This stuff: http://www.amazon.com/Rca-Speaker-Wire-100-Gauge/dp/B002HPNDDW/ref=sr_1_2?ie=UTF8&qid=1288887152&sr=8-2

These are the posts used on the drivers:


The voice coil braid solders to the end of a threaded rod or tube that screws into the bottom of the spring post.

[richidoo]

Analog filters are halfway built. It will be LW 4th order at 112Hz.

[Response Audio]

I really like the look and layout of this speaker. When do we get to see the full, finalized layout with finish on the design.

Maybe I missed it with 8 pages to read through but what will be your final crossover design. Active or passive?