Recently in Audiogon there have been a number of discussions about "improving" the B&W 802 D2, and a contributor found an excellent article from a self-proclaimed DIY speaker hobbyist Petr Kokourek on how they examined and re-thought the entire crossover. I thought this was a really great article for a number of reasons.
PKAudio provides an absolute wealth of information that is rare to find for commercial speakers, especially at this level/cost. The amount of tools, effort and courage it takes to dismantle a $12,000 pair of speakers is kind of breathtaking, unless you are Bruce Wayne and have a speaker lab in your basement.
I hope to help complete newbies to crossover design digest Petr's accomplishments more easily. Please read this article and then go take a look at Petr's thoughtful and thorough page on his work. I only wish I had the time, space and tools he does. What a great life he must have! Pictures from Petr are used with permission. Please contact him through his website if you are interested in his upgrades.
Crossover Upgrades
There are typically two types of crossover upgrades hobbyists do. The most common path is to replace original parts with "better" parts with varying degrees of success and return-on-investment. Essentially to do a 1:1 part swap without really understanding the crossover behavior, and perhaps removing protection circuits or L-pads when present.
The other approach, which Petr does with the B&W 802 D2, is to do a complete speaker analysis. More on that follows below. The famous Troels Gravesen also indulges in this approach with several vintage speakers including a predecessor to the D line the B&W 802 S3.
There is also another approach which is to go fully active. For these speakers the Hypex FA 123 or FA 253 are absolutely ideal. They each have three amplifier sections and outstanding DSP capabilities, which these speakers absolutely need.
What's Wrong with B&W?
If you are a B&W fan there is nothing wrong with them. Buy a pair and enjoy them for as long as you'd like. What they are not is neutral, so this makes them a very odd choice for mastering music or movies. If you wonder why Lord of the Rings has so many audibility problems, it's the final mix-down to a studio with B&W 802 D-series speakers. Of course, if you happen to be part of the 1% of listeners who have B&W 802s at home you don't understand why people are complaining. 😂
In addition to not being neutral, or smooth, these speakers also have significant problems with off-axis response as well, a problem shared by several Focal models as well. This off axis response affects imaging when not in the sweet spot as well as overall perceived tonal balance. Even if you put your head in a vice each listening session it's not an ideal situation.
The S3 version of the 802 line was first released in 1989. The 802 D series first came out in 2010. Lets take a little walk through memory lane to see how the house sound of B&W 802s has persisted for at least 20 years.
Speaker Analysis
Analyzing a speaker involves:
- Measuring each driver's unfiltered frequency response, impedance and delay curves in the cabinet.
- Tracing the crossover design
- Putting what you've learned about the crossover and drivers into a simulator like VirtuixCAD or XSim.
For the two examples below we take the work Troels or Petr have done from their complete speaker analysis to get into the sausage making. To learn more you can start with Joseph D'Appolitos excellent primer.
802 S3
Lets look at how Troels tackles the 802 S3s to see the pattern. After a complete analysis he is able to chart the final speaker response (in red) along with the contributions from each driver. Take a look at the original on the left vs. Troels' proposed solution on the right.
Clearly, Troels (and others) have a bone to pick with the aesthetic choices made with the original design. Troels flattens the midrange and woofer responses but keeping the same crossover points. The usable range goes from +-5 dB to about +- 3. A much more neutral and smoother response. This is not something you can achieve by simple part swapping. Like Kal Rbuenson's review of the original D, Troels also finds the voicing of the S3 quite forward.
Before moving on please notice that the tweeter response is the same. Troels leaves this untouched. Keep that in mind to compare with the changes Petr has to make for the D2.
802 D2
Now we switch back to Petr's data to see how he's chosen to tackle his situation. He has a lot of very good data but for this article I want to focus on just a few charts to help newbies out. Below we find the frequency response on axis, then every 10 degrees horizontally.
There is an excess and wide dip between 1 and 3k partnered with poor off axis integration between the tweeter and midrange which we see here as the deep valley a little above 2k at the bottom. There's also a pair of bumps and valleys above 3k.
I want to show more closely what Petr actually decides to do about this by putting beginning and ending charts next to each other like Troels did because it's key to understanding everything else. Here is the original, post analysis chart. We'll ignore the woofer in large part because it has the least number of issues.
Take a look at the midrange response, which is causing the bump around 600 Hz. Then note the two bumps to the right at about 4 kHz and 9 kHz.
Now lets compare to his proposed solution. For both response charts notice where the tweeter and midrange are crossing:
Petr has decided to move the tweeter crossover point down from around 4 kHz to 3 kHz. This doesn't sound like a lot, but look at the new dispersion chart:
Look at hos much smoother the off-axis response is at 2kHz and above in addition to the flattening of the overall response. Petr mentions later on that the new crossover takes a lot more parts... well, yeah!
The problem faced by B&W and Petr is the tweeter is just a real bear to manage with a passive crossover:
Notice the discontinuity at about 3 kHz. The answer to almost everything wrong with the speaker is found right there. If you aren't a crossover designer, let me explain a little bit.
Equalizing problems in a driver's response with passive crossovers takes parts, space and money. A shelving filter to flatten out the response around 3.5 kHz is a lot of work to get right, and a lot of parts, plus would eats away at driver efficiency.
B&W probably realized that by raising the "ideal" crossover point they could essentially avoid the discontinuity altogether. In other words, they avoid using the lowest frequencies of the tweeters capabilities so they don't have to equalize it. The poor off-axis response was a compromise to avoid fixing the tweeter and extending it's usable output to meet the midrange. I really have to wonder if this response is due to the "nautilus" loading of the tweeter being far less than ideal.
Petr comes up with a very usable solution by equalizing the tweeter (at least partially) to flat, increasing the high pass slope and lowering the crossover point. Now the midrange and tweeter dispersion matches much better.
In the tweeter's response there's still a bit of a hill around 9 kHz but the excess output at 4 kHz is completely dealt with. The faster roll-off will prevent the lower crossover point from overheating the tweeter.
Actively Better
Petr does a fine job of showing how very different crossovers can be, and exactly when you need to stop thinking about replacing parts and switch to a white-paper approach to crossover design. His finished crossover is also gorgeous and I encourage you to go look.
For me, the turning point in understanding his work was in looking at the raw tweeter response. Honestly it made me shudder to think of using a part like that in a passive speaker, and my very first instinct was that this is a speaker crying out for a modern, DSP based active amplifier.
The FA family of plate amps from Hypex has multiple choices for three-channel solutions which might even fit in the bottom plate of the D2, but are otherwise absolutely ideal for a project like this one.
My advice to anyone attempting this is to stick with Petr's overall goals in terms of balance and crossover points, but consider fixing up the tweeter and midrange slops so you can get a more ideal LR4 shape.
For anyone attempting their own crossover it may be best to measure at the midrange axis rather than the tweeter, as reviewers often comment that seems to be the correct listening location.
Philosophically Speaking
I really like what Petr did, but I have to step back and talk about the idea of "upgrading" an iconic speaker like this before I send you off to buy parts. Petr does a number of things:
- Equalizes the tweeter
- Equalizes the midrange
- Changes crossover points
- Builds entirely new crossovers from Jantzen
At every point he's disregarding the B&W sound. What he ends up with is essentially a box with B&W drivers and a completely new direction. B&W has been loyal to Mundorf caps for a while, and, like Magico, uses their high end caps in the high end speakers. That too is a significant part of the sound.
So, should you do any of these hacks? Are they better?
Well, for me, yes, Petr's aesthetic choices are what I consider correct and more pleasing. Should you? Well, if you don't mind hacking $12,000 speakers and making them unsellable and sounding like, oh, say Revels, sure! On the other hand, maybe a high-end kit speaker from Madisound or Solen.ca would be a better place to start?
