Notifications
Clear all

What do you prefer? Phase matched vs. Superflat Crossover  

  RSS

DIYAudiphileElliottBridge
(@diyaudiphileelliottbridge)
Estimable Member Registered
Joined: 12 months ago
Posts: 145
15/12/2019 3:08 pm  

So I have come to the point in my atmos design where I have simulated a full space response with both the tweeter and the woofer.

This includes diffraction simulation of both drivers on the rounded-edged baffle I have planned, and for the woofer, it was combined with the simulation of the Acoustic Suspension box I designed as an enclosure. Don't worry about the mixed order crossover the phase matching is just fine (I have checked it), it also tilts the response of the tweeter towards the listener when further off-axis (which gives the same effect of controlled directivity without the nasty lobing or missing the upper end when off-axis - as far as I am aware from the research I have done, I am the first person to use this technique so don't steal it. 😀

Anyway, to get to why I wanted to post this: I have two capacitors that when swapped either give a more flat response in the crossover area, or a perfectly matching phase response between the drivers.

Perfect phase
Better linearity for crossover

Now, I would like to know peoples opinions on:

1. Which they prefer in general and why.

2. Which they think would be better for this circumstance (home theatre use (e.g. needs good directivity control for Dolby Atmos), and of course why.

Hope to hear some replies from you all soon. 😀


Quote
123Toid
(@123toid)
Illustrious Member Admin
Joined: 3 years ago
Posts: 933
15/12/2019 7:39 pm  

Just a quick note. I use odd order crossovers all the time. In fact most of my two way designs use a third order on the tweeter and a second order on the woofer. There can be a lot of benefits for this.

As for the two, I prefer the imperfect phase relationship or if the two graphs. Your phase is not bad at all on that graph and the off-axis response should be minimally affected (if at all).  However there is a bigger issue here than your phase.

From 600hz-1200hz you have a pretty, what will most likely, be an audible rise. I think I mentioned this before, but I'd look to increase your inductor or do a rlc notch filter instead of the zobel. This area is a high risk area that should really be dealt with. I don't think you'll be happy with your results if you don't.


ImCoKeMaN liked
ReplyQuote
DIYAudiphileElliottBridge
(@diyaudiphileelliottbridge)
Estimable Member Registered
Joined: 12 months ago
Posts: 145
16/12/2019 6:55 am  

@123toid

That sounds great Toid, thank you for your comment. 🙂

I also remember you mentioning it before, however a lot has changed in the crossover and the responses on the graphs are now much more accurate (because they have simulated full space response)

As for the rise, I think you are referring to the double peak area, this part of the response is due to the behaviour of the driver and more specifically diffraction due to the deep, thick rubber surround. This bumped up area is only 1.5db greater than the surrounding area. If we are taking this approach wouldn't the peak between about 1.4 and 2.1khz need the same/similar treatment.

Also as you may have noticed the graph that you preferred (which I do agree with because that annoying dip is less prominent) is the one with that extra peak, the graph with better phase matching but with that annoying dip, doesn't actually have that second peak!

See my difficulty here 😀

Also that zobel is an incredibly important area for this crossover to work, otherwise they do not combine very well at all. Because I would need to keep the zobel this either means I increase the inductor's value (which I'll test when I get home), or add further complexity to the crossover, which I am not particularly keen on. This is because I am trying to make my ultimate budget design, £15 tweeter, £32 woofer, ~£30 crossover. The crossover is really expensive as I want accurate components with great linearity, e.g. Poly caps + air coil inductors. 3 extra high quality components using a series/parallel notch filter will increase the price a lot. I was wondering what you all thought about these points I have explained. I really appreciate your help Toid. 🙂


ReplyQuote
DIYAudiphileElliottBridge
(@diyaudiphileelliottbridge)
Estimable Member Registered
Joined: 12 months ago
Posts: 145
16/12/2019 7:27 am  

I am actually quite worried now, because even though this will predominantly be a home theatre speaker I will also be using this for a lot of music, and something I have just read had said:

"MID-RANGE 500 – 2000 Hz
Boosts in this range can make an instrument prominent in the mix. Be careful while mixing here as too much of 500-1 kHz can make your instrument sound muddy, and too much of 1-2 kHz can create a tinny sound."

This means I have to choose between the BBC dip which will make my speaker not market as well (+/- 2.5db) rather than the +/- 1.5db found on the response we discussed.

The reason this makes me concerned is the quote above makes the latter response sound quite bad.

Not only do we have the muddiness of the midrange but with the response we discussed (the one that doesn't have a BBC dip) we have a peak that describes as making a speaker sound tinny, I definitely don't want that!

With how difficult this decision is; if we do not come up with any other conclusions, then I will choose to get everything for one crossover and then swap the two capacitors between each other and test which I prefer.

I know it is a lot more stuff I have added, I just want to add everything I have learned related to this topic.


ReplyQuote
DIYAudiphileElliottBridge
(@diyaudiphileelliottbridge)
Estimable Member Registered
Joined: 12 months ago
Posts: 145
16/12/2019 2:02 pm  

I have just tried changing the inductance value, but unfortunately, it has no effects on that peak. 🙁

I think in order to sort out that annoyance I would have to add more components and create either a parallel or series notch filter, which I am not really up for unless it creates a big enough difference to the audio, will a 1.5db peak be noticeable to even the trained listener?


ReplyQuote
123Toid
(@123toid)
Illustrious Member Admin
Joined: 3 years ago
Posts: 933
16/12/2019 10:27 pm  

@diyaudiphileelliottbridge

Increasing your inductor should change the crossover point. However, if that isn't working, just change your zobel to a notch filter. You may have to change your values, but there is no reason that shouldn't take care of it, by adding just 1 more component. 

To answer your question, yes, I believe it'll be audible. Most of your graph is around 85db, while part of the graph is close to 88db or 3db higher.  This may not be an issue in some areas, but this particular area is of pretty high concern.


ReplyQuote
Share:
%d bloggers like this: