Introduction to designing crossovers without measurement

diyAudio Moderator
Joined 2008
Paid Member
@ krafty. The least that is required would be a microphone, measurement software and a crossover simulator. It is necessary to know what you want and have a plan for how to achieve it.

For example, a good crossover might involve taking several measurements and combining them which means you'll want a scheme of consistent data formats and processing methods. There would seem to be some professional packages out there that can achieve this on their own, but as a DIYer I have found the need to write my own software for some of the processes.

However, it is possible to produce an intermediate level of crossover using single individual driver, in box measurements, or alternately a series of simulations (baffle etc). Tweaking might then produce a crossover that you can live with.

I am inclined to feel that what you measure and what you do with these measurements will affect the outcome of the simulation more than things like absolute accuracy and resolution, as I seem to reach a similar level using modern techniques to acquire single per driver type measurements, as I did when I used discrete sine tones and visual lissajous representations for phase components.

Still, I think it is an ideal place to start and I would suggest you familiarise yourself with the file extensions of *.FRD and *.ZMA. You should also choose a crossover simulator that you can get your head around, however some might be hard/easy to use and this doesn't necessarily mean they're bad/good.

You may also want some supplementary tools such as one called SPLTrace, which is used to convert an image (screen shot) of a response or impedance plot into FRD or ZMA format.

Along with your microphone you may/may not need supplemental electronics such as a pre-amp. You may want to learn how to control the noise floor on your setup, as well as how to buffer your computer from potential damage (no pun intended ;)).
 

Attachments

  • Lissajous_phase.png
    Lissajous_phase.png
    45.7 KB · Views: 577
Thank you for your guide! I have one question. When choosing an amplifier for the circuit, do I use the original impedance and wattage from the speakers themselves or do I account for the changes the resistors had on the impedance values and calculate a new total impedance of the system?
 
Help with tweeter attenuation and L-pad

Dear Mr. Allen,
I am a beginer in diy audio field and I have built my first pair of speakers with crossover components calculated from your wonderful post. I have been listening to the speakers for about a month now and comparing them to other speakers. I had to reverse the polarity on the tweeter as there appeared to be void in the audio spectrum ( also based on measurements with Holms with a webcam mike).
I have the impression that tweeter is a bit too bright for my liking. and want to reduce the level. I have following options, which one is correct and any of the options affect the crossover frequency? I have attached the schematic of my crossover circuit for your kind reference.
option 1. Placing an additional resistor (1 Ohm or 1.5 Ohm ) before the tweeter capacitor.
option 2. Placing the addtional resistor(1 Ohm or 1.5 Ohm ) in series with existing 3.75Ohm resistor after the capacitor.
option 3. reducing the R2 (the parallel leg of the L-pad) by 2-3 ohms.

Also in all L-pad calculations available online, the total ohms of the circuit is kept same as that of tweeter impedance. eg. for a tweeter having impedance of 6 Ohms, and if we want to reduce its level by 7db, the follwing are the values of L-pad
series resistance: 3.32 Ohms
Parallel resistance: 4.84 Ohms
Now net Ohms of the circuit = 4.84*6/(4.84+6) +3.32= 5.997Ohms.
However as per your tutorial, you handle series resistance differently, i.e. you only use it as a tool to attenuate tweeter level . Shouldnt the series resistance and parallel resistances be handled (changed) together so that the net load as seen by amplifier remains constant? If this is the correct method, shouldnt the tweeter crossover components be based on total ohms of the circuit? (as per your tutorial, you have based the tweeter crossover calculations on net impedenace based on tweeter impedance and parallel leg of the L-pad).

I hope I have been clear enough in converying the requirements.
regards
Prasi
 

Attachments

  • crossover final.jpg
    crossover final.jpg
    83 KB · Views: 496
Last edited:
Prasi,

You probably have situation with too much of spl at XO point.
Your part values for low pass need improvement in that respect.
Increase the L1 to 1,5 mH for start which should get you closer
to desired XO point of 2kHz. Since you are new to this hobby,
get rid of the Lpad for tweeter and use only series resistor
before high pass for attenuation. Start by inserting couple of
ohms and increase accordingly until you get it right.
 
Prasi,

You probably have situation with too much of spl at XO point.
Your part values for low pass need improvement in that respect.
Increase the L1 to 1,5 mH for start which should get you closer
to desired XO point of 2kHz. Since you are new to this hobby,
get rid of the Lpad for tweeter and use only series resistor
before high pass for attenuation. Start by inserting couple of
ohms and increase accordingly until you get it right.
Hi,
Appreciate your inputs. However, tell me why do you suggest 1.5mH for the low pass section? All the calculations available online for first order filter, for the woofer with impedance of 8ohms (nominal), the inductor value works out to be 0.64mH. As can be seen from my crossover sketch, I have also used a zobel circuit for impedance stabilization to 8 Ohms.

Noted your suggestion for L-pad removal, will definately try and see if it improves the speakers.
regds
Prasi
 
Hi,
Appreciate your inputs. However, tell me why do you
suggest 1.5mH for the low pass section?

Calculators are not the right tool for this kind of work.
If you opened up a high quality commercial loudspeaker
with an 8 ohm bassmid and inspected XO filter, you would
have found a large value inductor L1 somewhere in the
range around 2 mH. Usually this accounts for baffle step
issue.

These are simulation graphs for Vifa PL18 with realistic
FR. It uses a purpose only to have some kind of feeling
what happens with parts used. I am not saying this is
a solution for you, only an improvement over existing
filter. You can see that bass with 0,68 mH plays flat
way over 2 kHz. There is no alternative to measuring
the speakers if you want excellent results.
 

Attachments

  • Original.jpg
    Original.jpg
    138.9 KB · Views: 456
  • Modified.jpg
    Modified.jpg
    139.3 KB · Views: 439
Hi,
again thanks a lot. Regarding your suggestion to use large inductor, I will try that and see if it improves the things. Also, with the large inductor in series, tell me if I can also put a 5microF capacitor in parallel in the low pass section . I will then have 12dB/oct textbook circuit(close to linkwitz-riley alignment). Will this work?
regds
prasi
 
Dear Mr. Lojzek,
Thanks for the inputs. since i do not have measuring equipment, I have to settle for trial and error methods. measurement equipment (mic and sound card) import is a very costly affair in India.
Based on your suggestions (higher inductor and removel of L-pad), I have prepared some versions of the crossover, which I am attaching herewith.
first attachment:With low pass inductor increased to 1.27mH (air core DCR 1Ohm) with a addition of parallel cap of 5/10MicroF. It also retains the L-pad with attenuation of 8.4dB. The reason for retention of L-pad is to lower resonance peak of tweeter because of the parallel resistor of the L-pad. (allenB's tutorial)
second attachment: L-pad removed and added series resistance before high pass xover components. different resistors will be tried for best sound.
Third attachment: Added even bigger inductor 2.0mH( steel laminate core, DCR 0.23) to sufficiently take care of baffle step diffraction and cone break-up region above crossover frequency. The following link uses a very large inductor despite being 8 ohm woofer crossed over at 2200Hz. (Bifrost).
This version of xover again uses L-pad with an attenuation of 9.5dB (to account for the overall lower sensitivity).

Now I would like to have your views (others are also invited) on following,
1. Will low resistances in parallel leg of L-pad damage tweeter (in the two x-over versions that use L-pad ?
2. any other suggestions for improvement that i may have overlooked?
regards
prasi
 

Attachments

  • version2.jpg
    version2.jpg
    511.4 KB · Views: 443
  • version3.jpg
    version3.jpg
    480.1 KB · Views: 208
  • version4.jpg
    version4.jpg
    482.2 KB · Views: 204
Last edited:
For trial purpose 1,27 mH is fine, its dcr of 1,0 ohm could be
better though. Tweeter damage usually happens if one crosses
it very low with unsufficient filter steepness combined with
high listening volume setting.

Your 1,0 mH in shunt to tweeter is large and makes it receive much more
energy at lower frequencies than needed which makes it more prone to
damage.

Normaly one would use values between 0,15 and 0,45 mH in tweeter
filters, depending. It is unknown what the tweeter FR looks like, so we
are in dark. Lower Lpad values decrease the impedance, but would probably
be ok in your case. You can use larger values for Lpad and still have desired
attenuation with probably different looking FR, as I said we are in dark and
guessing.
 
Thanks a lot for the suggestions.

1. Low pass: Yes air inductor that I have, has high DC resistance , then I will go with steel laminate inductor of 1.5 mH / 2.0 mH (Trial and error again!). The Bifrost speakr (as your had indicated earlier )has a very inductor of 2.7 mH!!

2. Attached is the tweeter FR. If decreasing the value of shunt inductor of the tweeter will make the high pass slope more steep, then I will try with the original 0.6mH inductor.
thanks and regards
Prashant
 

Attachments

  • imp_bg_t26sg-lfs.jpg
    imp_bg_t26sg-lfs.jpg
    61.7 KB · Views: 521
Hi Prasi!
You seems to need to attenuate tweeter about 6db. Look at tweeters impedance plot, if you put paraller to tweeter same resistance as you tweeters max impedance is, what you get? You get 3db attenuation on resonance area and usually that resonance forms your highpass curve to more non-ideal form. Now you get for free slope near XO-point and you manage with one serias resistor just before tweeters "impedance forming paraller resistor". Got it?
 
Using the Fr from post254.
The treble rolls off fairly evenly @ 6dB/oct from ~ 2kHz.

If I were targeting an LR2 crossover at the same frequency, I would add a single pole @ 2kHz and get the 2pole roll off.
But there is a dip from 1.4kHz to 2.8kHz
How would I compensate for that?

Next question.
I want to target my LR2 to 3kHz
How do I compensate for the acoustic roll off that starts 1kHz lower than that?
 
Hi Prasi!
You seems to need to attenuate tweeter about 6db. Look at tweeters impedance plot, if you put paraller to tweeter same resistance as you tweeters max impedance is, what you get? You get 3db attenuation on resonance area and usually that resonance forms your highpass curve to more non-ideal form. Now you get for free slope near XO-point and you manage with one serias resistor just before tweeters "impedance forming paraller resistor". Got it?

Hi Touho-serkku,
If mean to say I should put a 11 Ohm resistor parallel to tweeter (divide the voltage) with which I will get a 3 dB attenuation and then put a series resistor to attenuate further 3 dB to have a total of 6dB attenuation. But that would necessiate recalculation of all xover components as it will change the load presented to the xover circuit, if I understand correctly. Attachment is from earlier post, which uses a L-pad to reduce tweeter level by 8.5bB (as woofer is 87dB and tweeter is 94dB+plus couple of db more)
 

Attachments

  • version2.jpg
    version2.jpg
    78.8 KB · Views: 527
Last edited:
Prasi, with 11 OHm parallel you get 3db attenuation only in resonance region. Then impedance in resonance is 11/2=5,50hm. on flat area it,s about 4,5-5Ohm. Now, put serial resistor about 4-5Ohm to get another 3db (or what you prefer)(resistance diverer by half). Then you get about 9-11Ohm impedance load for your computer calculated HP XO.
ANd I presume you have waceguide too, then decrease your serial capacitance value by half... (your waceguide increases your tweeter sound pressure many db in the lower section, compensate that with 1st order HP.
 
Hi Touho-serkku,
Thanks for the input. I will try out all the options suggested by you. Now the tweeter FR is 5-6 ohm( nominal). I made a simple calculation for L-pad which which will maintain net impedance of about 8 ohms ( not in resonance zone, there it will be 8.9 ohms).
Th following are the values worked out .
Rs: 5 ohms
Rp: 6 Ohms
Ztwtr: 6 Ohms (nominal)
It will give a net db reduction of 8.5dB, while maintaining a impedance of 8 ohms in the flat impedance region of tweeter.
It is based on attenuation network section of following article link Passive Crossover Network Design


Since I didnt want to have too low a resistance in the parallel leg of L-pad, I have done this. Will this work? I can even increase the parallel leg of L-pad while increasing the series resistance of the L-pad to have higher level of impedance. However going too high on the parallel leg will not flatten the resonance peak by large amount.
 
Hi Lojzek,
based on your suggestions, I simulated my existing xover in visaton boxsim and then modified the crossover circuit. As said by you, my earlier xover was a nightmare and had too much treble and spl at xover. the first image shows exactly that. The second image is the old xover. the third and fourth images are the fr of modified crossover and the modified xover itself. the sound is now very much balanced and pleasant to listen. Thanks a lot for taking interest in my first hi-fi attempt.
regards:):):)
Prasi
 

Attachments

  • old fr.png
    old fr.png
    506.7 KB · Views: 516
  • old xover.png
    old xover.png
    342.9 KB · Views: 488
  • modified fr.png
    modified fr.png
    409 KB · Views: 240
  • modified xover.png
    modified xover.png
    311.1 KB · Views: 245