|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
50.202.202.186
In Reply to: RE: New "redesign" of LXmini. posted by Lee M on October 14, 2016 at 10:17:35
The story of how we got here. (It's not what you think, really).
Beanstalk Audio started out as a custom speaker shop mainly dealing in subwoofers. We built some really cool stuff (Tardis sub, Art Deco transmission line, etc...) but have always felt the draw to move into full range. We found the LXmini and after building a pair and being blown away by the sound stage and imaging, we started developing our first version based on SLs innovations. We focused on making it compact (for studio and nearfield use) sturdy (build quality and durability is very important to us) and address a couple small things we saw had room for improvement (cabinet resonance, basket vibration, etc...). We took those developments out to Linkwitz and had a nice listening session. We discussed pricing, licensing, marketing, and a lot of technical details. He said we should consider the design ours and basically go forth and see what the market says. I should be clear here that if SL had said not to continue on, we would have gone down a completely different path. We were lucky enough to get back to Maryland a week before CAF and get the last table on the floor. Could not get a room with any amount of cajoling. The response was super positive on the design front. Many people recognized the LXmini heritage and we had some great conversations about radiation patterns and speaker orientation and whatnot. But, no one (aside from a couple short sessions after hours) got to hear them. It is important to note here, that no one is buying them either. We were priced at $2295, which was right in line with the German builder of the LXmini (who had done no engineering work whatsoever) and honestly not high enough to build a dealer network on, but we really wanted to be a value leader and were willing work with a lower margin in the hopes of spreading the open baffle word. The reason, we found out over the next few weeks, was primarily the external active setup. No one wanted to spend money on the extra amps and cabling and they didn't trust this "DSP" thing in the first place. They also didn't like to have to run things through the DSP box at all, especially if they had a turntable. When I talked about "pain" above, this is what it means. If it is hard for a customer, they will go look for something easier, even if it doesn't sound as good.
Ok, we thought, maybe we just need to cast a wider net. By a stroke of luck, we managed to get a demo in front of two of the most influential audiophile reviewers and a musician/producer who is very well known in the audiophile world. Went up to New York firmly convinced that these guys would see the genius of SL's work and be immediately blown away. Already had the celebration dinner planned and everything. Here's the problem, they weren't blown away. The guys who knew and hugely respected SL's work and really wanted to like it...didn't. They were completely and maybe a bit brutally honest and gave us feedback that we really did not want to hear. First, they said the system was too complicated for them to write about. Still a strong believer in active crossovers, I tried very hard to make "our" case. But, then, the bomb was dropped. The sound character, they said, wasn't at the quality they expected. It sounded "hi fi" and not "high end". It didn't have the clarity they were seeing in comparatively priced systems. Yes, the sound stage was great, but it just didn't have the detail it should. The nail in the coffin was when the guy who wrote, performed in, recorded, and mastered a particular recording said it didn't sound right. That things were missing. That it just wasn't great.
Well, shit. Our first reaction was solid denial. SL was way ahead of his time. They just didn't understand what they were hearing. yada yada yada. That lasted about a day. Then, we started to wonder; what if they had a point. These guys had no incentive whatsoever to lie to or mislead us. They genuinely wanted to help us. Maybe, just maybe, they were right. Maybe there were gaps we were not seeing. Maybe the guys who had invested tens of thousands of dollars in their DACs and amps and sources should not be expected to just start from scratch.
Double shit. In order to address their concerns and (hopefully) have a chance at actually bringing something like this design to market we had to do two damn near impossible things; markedly improve on the work of a genius who has spent decades living and breathing this concept and build a passive version that sounds like the active one. Oh, and on top of that, no one from overseas understands our whimsical Beanstalk branding so throw in a complete redesign and rebrand as well. And do it all between early August and RMAF. Two months.
What the hell. We will probably fail, but we decided to give it the old college try. Here is what we found. Siegfried, if you read this, I really hope you don't take this as an affront in any way. We had no choice but to look for holes, and sadly, we found them.
1. Linear phase PEQ filters are bad. They introduce pre-ringing and general fuzziness that, once you hear, you can't unhear. MiniDSP? All FIR. All linear phase. Plus a less than stellar DAC chip and output stage. When the three wise men heard bad things in the DAC, they were 100% right. Proof: for the LXmini owners, go into the high channel PEQ. Disable everything but the 2.5k notch. Listen for clarity. Re-enable. Listen again.
2. The lack of a tweeter was really hurting the clarity. Bringing it in in our omni configuration helped a great deal with being able to play complex pieces with a lot of layering. Sting's Soul Cages sound bad? Sarah's Black and White vocals muddy? That is why.
3. Even with the tweeter, it wasn't enough. Cheaper systems were killing us on detail and resolution. We tried everything to get the FU10RB up to snuff and it just was not happening. So, driver evaluation time. We tried every full range possible and it wasn't until the Fostex Sigma that we finally found what had been missing. Detail. Clarity. Resolution. But, not out of the box. It takes 40-50 hours just to start sounding smooth and doesn't really come into its own until over 100. So, add in 100 hours of break in to our manufacturing cost. Awesome.
4. Passive crossovers suck. Audiophile grade ones even more. Moving to passive meant not only redesigning the cabinet (again) to fit these huge inductors in, but starting from scratch on the entire system configuration (crossover points, slopes, phase, everything). All those items that matter a great deal and SL had years to work out just right, we had to do in weeks, passively, and they had to be better.
5. Wave nulling around the edge of the full range was causing dips and thinness in the vocal range that could not be DSPd out. Mechanical baffle only real solution. After about 100 iterations, we have the solution pretty well figured out.
6. As an added bonus, the design was dated. Modern was the way to go. Won't even get into the R&D work it took to get the look we now have.
We had a number of other improvements along the way, but they really are too many to go into. What we ended up with is a system that cost significantly more to produce (more than 50%, so, great, now we have to raise our price as well, which is something you never do) but...and this is a big but...sounded really, really good. Better than all of our reference systems. Better than we ever thought possible. Even better, the passive and active versions were almost indistinguishable. We had done the impossible. Yay.
The end result? Well, RMAF went amazingly well. People who had heard both (or owned LXminis/521s/Orions) were stunned at the difference. Press is excited, especially internationally. People actually want to buy our stuff now.
And, sadly, SL thinks we are on the express train to crazy town. We made changes that he would not have. We introduced (optional) baffles that he thinks are a step backwards. We use a system architecture that is very non-traditional. We moved away from the SEAS driver that he helped develop. How could you not see our massive set of changes and not think we had lost the way. All we can do from our end is continue to extend our invitation for SL, and anyone else, to hear what the end result is and judge for themselves.
Davey (patents),
B&W and a number of other speaker companies are seeing a flood of clones with very low quality drivers and crossovers show up on Ebay/Amazon, etc. IP protection (primarily design patent, but possibly utility on some of our inventions) is about the only reliable way to fight that. Having clones that die after a year and might catch fire along the way is not good for anyone. I genuinely hope people use what we have as a starting point and keep pushing the concept forward. That is what innovation is all about.
Follow Ups:
Good golly.
You're in error starting with Item #1. The miniDSP/LXmini scheme is definitely NOT linear-phase. All FIR? Absolutely not. All PEQ filters are minimum-phase IIR types.....same as analog. Maybe you are confused regarding the recon filter implementation? If you're starting with a basic misunderstanding like this, I'm not sure what to make of any further conclusions.
Should I continue?
My goodness.
Dave.
You are saying all PEQ are minimum phase? Everywhere or just the minidsp? If the PEQ in the minidsp are indeed minimum phase, and I personally would not believe that without seeing the source, then that is actually worse. That means the filter implementation is so bad that they sound like linear phase. Either way, please continue...
PEQ filters can be either, but not so with the basic/cheaper miniDSP units. Only their platforms with more computing power can implement linear-phase filters.
If you don't believe it, a very simple bench test can reveal this behavior.Regardless, (using your over-simplified conclusion) it's actually better not worse...IMO. Minimum-phase PEQ correction filters (applied in this context) actually yield a net linear-phase result, whereas a linear-phase correction filter would yield a non-linear-phase result. Think about that a bit. :)
However, it's arguable that the whole concept of worrying about phase distortion with amplitude corrections is even meaningful.Phase distortion (and the audibility of it) is an interesting side discussion not (directly) related to this. Suffice it to say, there are multiple opinions on it.
Anyways, obviously you fellas can build your design however you like. I don't have any control over that, but what I can/will do is correct mischaracterizations and misinformation regarding the existing LX designs from Linkwitz.
Dave.
Edits: 10/15/16
I probably gave too much credit to MiniDSP in being able to implement FIR with their processing power. Let's just assume I am wrong and they are IIR, which are much more efficient to implement.
As you mention, though, the phase effects from every type of filter are complex. Additionally, different algorithmic implementations have different side effects (usually distortion, but phase and resolution can be affected as well). In theory, correcting the SPL response also corrects phase at the same time. Win win. In reality, the "correctness" of leveling phase correction is probably not as perfect as the theory states and the distortion and other byproducts of having the filter there at all are a real problem.
So, what we observed was a negative effect of having filters in place. We observed them in crossovers, shelf, and peaks. We did see that moving to high precision 32 bit floating minimum phase filters sounded much better (clarity) than whatever the implementation was on the three different MiniDSP products we had. But, by using a driver that did not require any filters (other than a small 2.2k notch) we removed the possibility of introducing distortion or unwanted phase effects and have found that the resulting clarity is far more important than having a ruler flat SPL graph. Additionally, even without a single filter, our ABX tests put the DAC implementation in the MiniDSP dead last against our available test units. The 2x4HD is significantly better then the 2x4 and 4x10, but still not at current audiophile standards. The standard bearer, in my opinion, is the currently unknown dspmusik.
This is actually a perfect illustration of the difference between SLs approach and ours. SL has complete understanding of the theory behind the relationship between SPL and phase and has a mathematically perfect solution to correct both at the same time. He is 100% right that that is the ideal solution and any deviation from that is a step in the wrong direction. The problem is, the implementation of that solution (on the DSP hardware side) is not perfect and the side effects of attempting the solution are audible and problematic.
So, what do you do? You could wait for the DSP implementation to catch up with the theory and eventually have a correct solution that sounds perfect. Pros: You are proven right. Cons: May take a while to get there. In the meantime, you are getting your ass kicked in clarity by KEF and other manufacturers that take a more pragmatic approach. Or, you could go back to the fundamental nature of the problem, which is the use of a driver that requires so much PEQ in the first place, and redesign a system that does not rely so heavily on filters. Pros: You might be able to make something subjectively sound better. Cons: Everyone thinks you don't understand the theory and therefor your solution must be wrong.
The thing is, we do not have the same design goals as SL. He is making a transducer that converts electrical impulses to sound waves. We are making a speaker that sounds as good as possible given the limitations of driver construction, analog and digital filters, and all the other factors that go into speaker design. A lot of times this means having a solid grasp of the theory and then being willing to throw that out the window and try a bunch of stupid shit until you find something that works better. As my dad likes to say, If it's stupid, and it works, it isn't stupid.
So, throw the theory out the window and try a bunch of stupid shit until it sounds/works better? I'm glad your design acumen is not squishy or anything. :)
Take care.
Dave.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: