Help with my first speaker build (3-way)


I've been wanting to get into DIY for a while now, a few days ago i found a very good deal on some Scan-Speak elments that i couldn't resist. My DIY journey shall begin!

Drivers Scan-Speak
Tweeter (pair): D2905/950000
Mid (pair): W15/8530K00 Revelator
Woofer (pair): W21/8555-00

Got it all for under 340 EUR including a bunch of crossover parts as well, so not all too expensive for a first build i guess.

I've been fiddling with VituixCAD for a few hours now (I might also try Xsim later to see which one I prefer). Honestly, I dont really understand what I'm doing at the moment. I've assembled some components and tried adjusting their values until I achieved a smooth frequency response (within about +- 1-2 dB range?). However, it's clear that there's more to take into consideration. I would greatly appreciate a lot of feedback!

I plan to delve deeper into the enclosure features in VituixCAD and also fine-tune the x, y, z parameters for the drivers once I've settled on an enclosure design (I assume this affects phase?). According to WinISD, the woofer's f3 will be 29 Hz in a 50-liter enclosure. Maybe I should aim for a slight boost in the bass because of baffle step loss? Or will room gain compensate enough?

I suspect that the crossover point for the mid/tweeter might be a bit high. I suppose the crossover points are in phase due to the 6 dB summation?

My intention is to measure each driver within the actual enclosure and then base my design on the actual frequency response curves (or is that a bad idea?). For now, I'm primarily focused on learning the process of designing a crossover. I'm aware that my capacitor values seem rather high and could potentially be expensive, but the overall number of components appears to be relatively low, so perhaps it balances out.

I would greatly appreciate your feedback as I'm struggling to interpret the information presented by VituixCAD. Are there any red flags? Any positive/negative aspects that I should pay special attention to? Thanks for your time!


Skärmbild 2023-08-16 195532.png

Skärmbild 2023-08-15 214035.png


  • freqresponse.png
    23.3 KB · Views: 63
  • Skärmbild 2023-08-15 214012.png
    Skärmbild 2023-08-15 214012.png
    93.8 KB · Views: 54
  • Skärmbild 2023-08-15 213850.png
    Skärmbild 2023-08-15 213850.png
    79 KB · Views: 72
Sounds like a good start. Apart from the choice to pick a 3-way for a first time design. 3-ways are a lot more complicated than 2-ways. Mainly because getting useful input data is a pain.

Learn the measurement thing. It is crucial. And will teach you backgrounds for your crossover design work. Alas measuring bass and mid frequencies (<1kHz) becomes complicated quickly.

Learn VituixCAD on the way. Read user guidelines and backgrounds. For basic simulation Xsim and Boxsim are easier alternatives, VCAD has many features and in that is more complicated. If you are going to design a few systems only, refrain from VCAD.

Would you consider starting with a 2-way design? You could add two subs later, disguised as stands for the 2-way monitors you already successfully finished. Using an active (DSP) solution for the sub/sat integration will give you much room for improvement and learning, without fiddling with a complicated and expensive passive bass-mid crossover.

Just my 2 ct.
  • Like
Reactions: 1 user
That's a great first start. Don't be put off doing a 3 way first up. I did and it turned out very well. But it does take more effort.

Some observations.

By using Manufacturer FRD data (I assume) - you are using Infinite baffle on the drivers. As you noted - you are missing baffle step "rise" meaning the bass response is overstated. Your woofer will lose 3 - 4 dB sensitivity progressively below ~ 500Hz meaning your overall system sensitivity will be around 83-84dB otherwise you'll end up with a midbass lacking speaker (thin sounding). So either apply a simulated baffle step response to the manufacturer curve, or invest in a measurment microphone and do a quasi-anechoic measurement (nearfield, baffle response applied then farfield merge).

Second - your mid-tweeter could probably benefit with a higher order slope and lower crossover point. the tweeter is capable and you will offer better protection for it (and better off-axis / DI response).
Thank you so much for your input! I like the woodworking part, so worst case scenario is to rebulid if/when I discover some major flaw or that at 3-way is too big of a challenge. Any recommendation of software for simulating baffle step response?

I have thought about starting with a two way design, and maybe build an enclosure and redesign the crossover to implement the woofer later. I do, however, really like the idea of a floorstanders as i generally prefer them over standmoundts due to the "big" and impactful sound of a good floorstander.

Gonna try Xsim and see if I find it more intuitive for a beginner.

Again, thank you for your input!
8msec may not be low get adequate measurements for a 3 way if you are crossing 250Hz or lower. This will only give you a 125Hz lowest usable frequency response measurement, and crossovers are best served by at least 2 octaves of data either side of the XO point.

Unless you go outside and do a ground plane measurement or mount the speaker very high off the ground (both require an equal distance in all directions to be reflection free - e.g. a warehouse / gym / hall with high ceilings), a quasi-anechoic measurement is your best option. This splices nearfield and port, applies baffle step resposne then a merge with a gated farfield measurement (e.g. 5msec which we can achieve in most listening rooms)
Any recommendation of software for simulating baffle step response?
VirtuixCad will do so in the diffraction tool.

Each driver can be simulated in fullspace
on the baffle size used.
You can establish the listening position.
By leaving microphone in same
position for all drivers. Then phase will be correct.
Usually assume tweeter to be listening position.

Rather revealing process since baffle step
becomes much more easily understandable.
Since it can be seen visually rather quickly
with modern software.

Tweeter has Fs of 550 Hz and woofer 20 Hz
Having only Briefly viewed the datasheets.
They do seem like rather nice impressive drivers.
Such low Fs for tweet and especially 20 Hz for woofer
is pretty nice for 8"

In Virtuix cad use the library examples for crossover
coils/caps. You can adjust impedance according to
driver impedance curve and get much more ideal values.
Since crossover points usually are not at " nominal" values.
And the impedance curve is of course visible right on the
  • Like
Reactions: 1 user
8msec may not be low get adequate measurements for a 3 way if you are crossing 250Hz or lower. This will only give you a 125Hz lowest usable frequency response measurement, and crossovers are best served by at least 2 octaves of data either side of the XO point.

Unless you go outside and do a ground plane measurement or mount the speaker very high off the ground (both require an equal distance in all directions to be reflection free - e.g. a warehouse / gym / hall with high ceilings), a quasi-anechoic measurement is your best option. This splices nearfield and port, applies baffle step resposne then a merge with a gated farfield measurement (e.g. 5msec which we can achieve in most listening rooms)
View attachment 1203696
My suggestion to go for a 2-way comes back to mind...

The merging of responses certainly is viable, but prone to calculation errors. That is why I suggest going for a longer time frame. 8ms is enough for crossing at 250Hz if you roughly combine far field measurements with near field (to check for anomalies). At 250Hz the far field only serves to get the level right, which -being under the baffle step and in room contribution territory- is a bit arbitrary anyway. Furthermore, I would consider crossing W to M at frequencies near the baffle step. This offers addressing both the BSC and the X-over to mid in one step.

Crossing passively at 250Hz sure will bring some more issues: big expensive coils, impedance flattening of the woofer to counter the 100Hz bump, big expensive foil capacitors for the mid and impedance flattening of the resonance of the mid to assure proper working of the highpass to name a few.
Thank you all for your valuable input, I have a lot to learn I see!

I have no issues measuring the drivers in the cabinet outdoors far from the ground.

Im about to start buliding the cabinet, is there anything I should know before i start making the baffle (driver placement)? My idea is to place the drivers as close as possible, just a normal tower speaker. I've read that some prefer to place the woofer close to the floor while others advocate otherwise. Visually, I'd rather place all the drivers close to eachother meaning the woofer will be just under the mid. I will round over the corners with a router (only the sides, not the top).
Whether you build a 2 or 3 way speaker, you invariably take a quasi-anechoic measurement for the mid-woofer / woofer. Some will argue you don't need to worry about bass measurements for a 2 way speaker for doing the crossover, but I'd argue if you haven't correctly allowed for baffle step compensation, how do you know you are designing the mid-woofer low pass correctly? and therefore accurately assessing your sensitivity? (target SPL once baffle step losses are taken into account).

Therefore the same measurements are required 2 way or 3 way (in the case of the 3 way - you just have another driver to do quasi-anechoic measurements for)

It certainly is likely that for a 3 way however, passive crossovers will be more costly with lower crossover points, and also the bass hump problem (~ 100Hz you can see in a transfer function) - caused by the necessarily large value series inductor inline with the woofer to create the lowpass. I've decided to address this by going fully active, however many people go hybrid (active woofer-mid, passive mid-tweeter). This arguably brings the best of both worlds.

In terms of woofer placement vertically (from floor), the closer to the floor the more boundary re-inforcement (unsure how significant this is and this itself can be difficult to model) but higher the first dip/peak problem (floor bounce at listening position).

In reality - I think you are fine placing wherever it is aesthetically pleasing, knowing room effects will swamp low frequency response and the dip/peak problem for floor bounce is likely to be above the passband for the driver anyway (i.e. filtered out). The main thing is to try and keep the woofer to less than a 1/4 wavelength distance from the midrange at the XO point between the two. Usually quite easy to do in a 3 way.

Some on-line calculators for floor dip/peak at listening position:
Hi there, im also doing 3-way speaker and with same bass driver that you have.
Some things i have learned:

-BSC Your bass driver has sensitivity of 87db, your mid has 86db. You will need 3-6db more sensitivity for bass driver to compensate for baffle step.You already have 1db, so if you put your speaker very close to backwall, you might get away with this doing nothing, but i don't know without simulating.. Maybe someone could?

Baflle step happens when lenght of the soundwave becomes tiny as you go up from bass to midrange, that it stars to become directional and bounces from the front baffle. Effect is that upper range is amplified, and you bass is not so without bsc you will have bass levels playing 3-6db less than mid_high.
You have few options:

-Go for active bass section(i do this with hypex fa251), example of this is Troels MUN17:

-Double the bass drivers in series, you will gain +6db sensitivity for bass driver section, so your new sensitivity for bass is: 93db

-Pad the mid with resistors,(this is cheap & easy, just make sure to use right big wattage resistors as troels so they will stay cool (check this out with your prototype playing music for few hours), and check you don't cross to bass lower than 450hz. Example of this is troels Jazzman:
Due to low efficiency of the bass driver, some attenuation is required for the midrange, hence paralleled 10 watt MOX
resistors in midrange section. These will stay cool even at high playback levels."

If you put your crossover to bottom of the speaker to "free air" (and make wooden sides to hide it) this could work very well. You probably could cross bass then by 350hz. Look troels revelator 851:
Last edited:
  • Like
Reactions: 1 user
Your Tweeter (pair): D2905/950000 give you options for crossover point.
Jeff Bagby used higher model 990000 with 1.8khz crossover point, usually problem with lover crossover points are increased distortion, intermodulation and harmonic distortion.

I would recommend trying out 3.0khz crossover point since your mid can take it, lr2 or lr4.
But model things out and see what you got with different crossover points, buy basic crossover parts and try out if you like it..

Here is troels project where he uses your mid, look it trough for hints how to make crossover:
Hi again!

First of all, thank you for all your input, I am slowly learing more and more!

I've done some baffle step simulation i VituixCAD and have made two new crossovers with some bass boost under 500 Hz. The next step i to assemble the cabinets and see how far off I am.

Out of the the two designs below, which one do you think would be a better startning point (and why)?
Theres more "overlap" between drivers in "B" and it has a lower crossoverpoint for the tweeter (2.2k).
Alternative "A" has its tweeter crossed over at 3,4 KHz and less "overlap" between drivers. Freq. response within +-1 dB from 500 Hz. I dont know what "prefrence rating" in Vituix is, but alternative A gets a higher score (8.4 vs 8.15). Alternative A "looks better" to me. Is this a good starting point?

The tweeters Fs is 550 Hz.

Alt A

Alt B


  • 450 n 3k 2.png
    450 n 3k 2.png
    13.9 KB · Views: 22
  • 500 n 2.2k 2.png
    500 n 2.2k 2.png
    14.2 KB · Views: 21
Sanity check time.

You are running 2 21w in parallel so your sensitivity will be 87 + 6 dB = 93dB

Minus 3 to 4 dB baffle step losses (allowing for front wall reinforcement).

Your target SPL should be around 89dB

Why is your target so low?

Getting the right sensitivity is the first step in crossover design. Or all values will change again