So, I’m learning on the job and built an active 3 way speaker which I’m really not happy with! The thing I’m not happy with is the balance. I’m still learning tonnes when it comes to building speakers and I love it but I get frustrated as I’m not quite sure how to get it right.
I end up tweaking and tweaking and tweaking some more until it sounds just as bad as when I started lol. The tonal balance is just bloody terrible.
It’ll be great to get some advice on how you guys go about balancing your speakers ie what tools you do use, which might help me in my quest for a great balanced speaker in the end.
Thanks for any advice given.
Sal
I end up tweaking and tweaking and tweaking some more until it sounds just as bad as when I started lol. The tonal balance is just bloody terrible.
It’ll be great to get some advice on how you guys go about balancing your speakers ie what tools you do use, which might help me in my quest for a great balanced speaker in the end.
Thanks for any advice given.
Sal
When you try to "tweak" a speaker you often up going in circles. Siegfried Linkwitz called it the "circle of confusion" (this was talking about speakers using in mixing recordings, but the problem is the same). You are using recordings and your ears to judge the sound. Recordings vary in their tonal balance, and hearing is adaptable to different tonal character over a short period of time. This often means that you do not have a reliable reference point to hone in on. After listening to one recording it might sound OK, then you switch to another and it sounds terrible.
If and when I "tweak" my crossovers, it would only be as the final step in a long process. I would only be making very small adjustments here and there to what already is known to work. I use listening, but I make sure to take breaks during which I don't listen to anything, and the process is repeated over several days until I am satisfied that I have struck the right balance with a variety of music types.
What you (likely) need is a way to get very close to "done", e.g. 95% of the way to perfect, without using your ears. The best way (in my opinion) to do that is with measurements and modeling. To do this you will need a microphone (and possibly a mic preamp) and a computer. A crossover design software program should be used. This will combine the measurements and other info into a good model of the loudspeaker as seen from the listening position. Then you can add the crossover to the model and tweak all the values until you get to that 95% point. At this point you can build or implement the crossover and do any final tweaking if you feel it is necessary. There is some knowledge and skill required during this process, and you must take the "right" kind of measurements to be able to have a good quality and accurate model.
But maybe I should ask you to back up a minute and explain a bit more about what you build and how:
Is it a kit, is it based on an existing design, or did you create it from scratch?
Are you using a passive or active/DSP crossover? Please share details about it?
Please provide a physical description of the speaker or post pictures so we know how large it is, where the drivers a positioned on the baffle, etc.
Have you build speakers before, or is this your first go at it?
You need some measurement capability. One of the biggest issues is that the woofer is typically the least sensitive driver so you need to dial down the other drivers or it won't sound right. Once you get the driver levels balanced things will sound better and the you can go for your "house curve".
Rob
Rob
And this is why I bloody love this site. Thank you so much for all the info!
I’m at a gig right now but will reply with any questions once I’ve digested all the info.
I use a dbx driverack which has an RTA so it’s useful but as you’ve pointed out about tweaking and tweaking, it’s clearly not the right approach.
I’m at a gig right now but will reply with any questions once I’ve digested all the info.
I use a dbx driverack which has an RTA so it’s useful but as you’ve pointed out about tweaking and tweaking, it’s clearly not the right approach.
...and how to interpret the measurements in terms of what your ears don't like. For example, it is possible that one tweaks a loudspeaker to almost perfectly flat anechoic on-axis SPL response, and it still sounds TERRIBLE. The trick is to understand what the measurements (don't) tell you, and how they relate to what you hear. I am trying to learn this art since a few decades now.
Two important lessons I learned so far is (1) to also look at off-axis response, which tells you how the speaker interacts with the room (2) to arbitrarily change things in the x-over and check how this changes the measurements and the perceived sound.
True, very true. Don't neglect phase.
I was tweeking small bookshelf, based on aurum cantus ac130 and horn loaded fountek ribbon, meant to be used with subs, and finally achieved ruller flat on axis, almost perfect off axis behaviour, and flat phase...now that's one neutral sounding well behaving no fatigue sound. All it needs is better subs to sit on.
I was tweeking small bookshelf, based on aurum cantus ac130 and horn loaded fountek ribbon, meant to be used with subs, and finally achieved ruller flat on axis, almost perfect off axis behaviour, and flat phase...now that's one neutral sounding well behaving no fatigue sound. All it needs is better subs to sit on.
It would be ruler helpful as to what did you " tweak " , resistor values , caps to alter the crossover point . Did you put in a Lpad ? What type of crossover is in it , e.g. 2nd order , 1st order ( not with a fountek surely) or what ?
so just to shed some light on my situation:
I've built a pair of TH18 tapped horns which have come out amazing and sound excellent so i wanted to then build a pair of tops to go with them but also be able to use these tops by themselves when needed.
Each box is fairly large at 650mm tall, 450mm deep and 400mm wide. the box houses 1x 18sound 12nd930 woofer, 1x BMS 4552 tweeter with 18sound waveguide and 1x faital pro M5N8-80 mid-driver.
I then thought i'd try something a little different and go for a point source type design and put the tweeter and midrange drivers in front of the woofer which is on the baffle but in free air (this was an after thought, hence why the free air) this is most probably the start of my issues!
I've gotta say that i used Xsim to try and build a crossover for these drivers and to be fair, i got a great result but went the active route in the end.
Anyway, i've decided to build another pair of boxes in the next month or so which will house 1x 18sound 12nd930 woofer and 1 Beyma 5CX200FE coaxial driver, both of which will be on the baffle in the traditional position of woofer sitting beneath the coaxial.
I should add i'll be using Beyma's crossover for the coaxial driver and will use my DBX driverack to then crossover the Beyma and the 18sound driver. I reckon using Beyma's crossover will give me a much better base to work with as well.
So is there any way of simulating my box without building it just yet?
I've built a pair of TH18 tapped horns which have come out amazing and sound excellent so i wanted to then build a pair of tops to go with them but also be able to use these tops by themselves when needed.
Each box is fairly large at 650mm tall, 450mm deep and 400mm wide. the box houses 1x 18sound 12nd930 woofer, 1x BMS 4552 tweeter with 18sound waveguide and 1x faital pro M5N8-80 mid-driver.
I then thought i'd try something a little different and go for a point source type design and put the tweeter and midrange drivers in front of the woofer which is on the baffle but in free air (this was an after thought, hence why the free air) this is most probably the start of my issues!
I've gotta say that i used Xsim to try and build a crossover for these drivers and to be fair, i got a great result but went the active route in the end.
Anyway, i've decided to build another pair of boxes in the next month or so which will house 1x 18sound 12nd930 woofer and 1 Beyma 5CX200FE coaxial driver, both of which will be on the baffle in the traditional position of woofer sitting beneath the coaxial.
I should add i'll be using Beyma's crossover for the coaxial driver and will use my DBX driverack to then crossover the Beyma and the 18sound driver. I reckon using Beyma's crossover will give me a much better base to work with as well.
So is there any way of simulating my box without building it just yet?
This is only an insignificant choice in the bigger process of making a crossover.
Yes.. can't build a crossover without knowing what it is supposed to manipulate.
I added 3 other steps in my xo building.
first there is the research of the drivers,
then the box size
then models are made, measured, room in taken into consideration, volume levels etc
I build the actual final products with damping etc
listen and measure for issues, if issue, go back to prototype, fix it, rebuild or fix final design
do many types of XO, listen for what sounds best, where the xo can cross without issues, beam and thd and imd
then do final xo accordingly, measure again, should be spot on.
0 need to do ear, only to hear how low/high the drivers sound nice, this is done by using multiple type of xo, listening random music and hear any problems which are not obvious in the thd curve.
there is also a great deal of research to get the desired best response curve, just inputing a straight line or getting the response flat is never a good idea, better stay within +/- 2 db or 3 db and take what sounds best there,
it sounds trivial but this process takes many years
first there is the research of the drivers,
then the box size
then models are made, measured, room in taken into consideration, volume levels etc
I build the actual final products with damping etc
listen and measure for issues, if issue, go back to prototype, fix it, rebuild or fix final design
do many types of XO, listen for what sounds best, where the xo can cross without issues, beam and thd and imd
then do final xo accordingly, measure again, should be spot on.
0 need to do ear, only to hear how low/high the drivers sound nice, this is done by using multiple type of xo, listening random music and hear any problems which are not obvious in the thd curve.
there is also a great deal of research to get the desired best response curve, just inputing a straight line or getting the response flat is never a good idea, better stay within +/- 2 db or 3 db and take what sounds best there,
it sounds trivial but this process takes many years
Last edited:
one mistake people make is to have the xo right , but wrong.
like the xo point should be for the best dispersion AND where both drivers are getting tired. if at the xo point one driver is making more distortion than the other it is a failed xo.
also it is plain dumb to start filtering the best driver when the other one has higher distortion at the same frequency. Or if the distorted driver is not filtered enough and the peaks of resonances and distortion are rising faster than the filter curve, this is typical for midranges and tweeters.
As the midrange is filtered and tweeter takes over, the mid range is 2n or 1st order and start screaming distortion inside the tweeter, and vice versa. this is trivial but it needs to be properly designed regarding at least 2n and 3rd order THD, it is a time consuming process.
like the xo point should be for the best dispersion AND where both drivers are getting tired. if at the xo point one driver is making more distortion than the other it is a failed xo.
also it is plain dumb to start filtering the best driver when the other one has higher distortion at the same frequency. Or if the distorted driver is not filtered enough and the peaks of resonances and distortion are rising faster than the filter curve, this is typical for midranges and tweeters.
As the midrange is filtered and tweeter takes over, the mid range is 2n or 1st order and start screaming distortion inside the tweeter, and vice versa. this is trivial but it needs to be properly designed regarding at least 2n and 3rd order THD, it is a time consuming process.
As I’m buying Beyma’s dedicated crossover, surely that’ll make life easier, right?
Then it’s a question of crossing over the woofer with the coaxial.
These speakers are for a live band so room acoustics can’t really be taken into consideration as it’ll constantly change.
Gotta be honest, a few of the last comments have confused the shite out of me.
Once I’ve built the box, is using pink noise a good way of measuring? Or a mix of pink noise and music?
What do I look for in the pink noise response?
Then it’s a question of crossing over the woofer with the coaxial.
These speakers are for a live band so room acoustics can’t really be taken into consideration as it’ll constantly change.
Gotta be honest, a few of the last comments have confused the shite out of me.
Once I’ve built the box, is using pink noise a good way of measuring? Or a mix of pink noise and music?
What do I look for in the pink noise response?
Sort of. Can you measure the drivers in the box/baffle, both on axis and off axis, without any xover filters? You could feed these data into a speaker CAD software, draw the xover schematic and simulate the system result. Vituix CAD comes to mind.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.