How low is too low? (minimum tweeter crossover point theories)

[Vance Peate]

Hello everyone,

I want to discuss how to determine the minimum crossover frequency for a HF transducer.

This discussion is focused on finding the minimum crossover point, not the reasons for finding it (I am aware that the minimum crossover point for a HF transducer is not always the optimal point to cross).

Here's an example I'd like to present:

I am currently using an SB26STWGC-4 in a 2-way with a Volt BM165.1. For the purpose of this discussion, let's disregard midbass characteristics; I am solely interested in determining the safe minimum crossover point for the tweeter. I want the system to have the least amount of distortion at relatively high SPL.

To determine the minimum crossover point of the SB26STWGC-4, my approach would be:

1. Consider the type of transducer (soft dome, hard dome, AMT, planar magnetic, etc.).
2. Study and consider the frequency response graph and recommended operating range of the SB26STWGC-4.
3. Ensure that the crossover point is at least an octave higher than the Fs (1560Hz (780Hz Fs *2)).
4. Determine where harmonic distortion / nonlinear distortions start to occur**?
5. Take into account the Xmax**?

Points 4 and 5 are a little vague for me. Any help with this would be greatly appreciated.

Thanks,
Vance.

 

[Moondog55]

Point 3 depends very much on the steepness of the crossover, one octave above and I'd be looking at 4th order and unless I'm using digital XOs I much prefer second order.
Looking at the Volt I can see why you'd want to cross low
Limit of my experience/knowledge

 

[Audio>X]

The points of these first two posts basically cover everything > and are correct.
* for whatever reason, I have found second order XO to be very 'musical' 🙂

 

[mattstat]

I want the system to have the least amount of distortion at relatively high SPL.

4. Determine where harmonic distortion / nonlinear distortions start to occur

For home hi-fi, this is typically the most important aspect if you want to get to a real answer. Common tests for it are harmonic distortion at levels of 86 and 96 dB at 1 meter. If you want more output than that, you should test at a higher level with your intended crossover or use a test with a limited frequency range to minimize the chances of driver damage.

In practice the level is often set with the microphone at 1 meter, then the microphone is moved much closer to the diaphragm of the driver being tested. This gets the level up high enough that background noise doesn't significantly contaminate the measurement (since you're often looking for distortion products that are more than 40 dB below the reference level and homes aren't silent). Typical measurement microphones are good up to approximately 120 dB, but staying below 115 dB is sensible.

While rules of thumb like your other points can also be used, they have limitations.

 

[Vance Peate]

Thanks for the insights.
Can someone explain the below graphs and how to determine 'where distortion starts':

Distortion 2.83 volts, 315mm:

 

[Shaun]

Looks like 1kHz is the knee. But could you increase the size of your pics, please. These are illegible.

 

[mattstat]

From what I think I see, distortion is rising for some harmonics below 2 kHz (the green line and lighter yellow - 4th and 5th, I think). The levels look relatively low, but they are increasing, so if you want the least distortion you may want to stay above that range.

5. Take into account the Xmax

Forgot to address this earlier: volume displacement is typically what you want to look at (Sd x Xmax). You can calculate maximum output with that, so is a good thing to check to if you think you're going to be near the driver's limits (high output systems or those that are using drivers at lower frequencies than typical).

 

[stv]

'where distortion starts':

for simplicity and clarity you could hide all distorsion graphs except H3 (and eventually H5).
that's the most revealing one and also one of the more annoying audible ones.

 

[Vance Peate]

Looks like 1kHz is the knee. But could you increase the size of your pics, please. These are illegible.

Apologies!
Here are the larger files:

 

[PeteMcK]

from the graphs, it can be seen that the H3 rise confirms the old 'octave above Fs' rule of thumb, remembering that any reactive components in the Xover need to be well above the Fs resonance as well.

 

[Scottmoose]

Unless you flatten the Fs impedance rise to ensure the filter behaves consistently in that regard.

A couple of general points, FWIW:

• All other things being equal (they never are, but...) a tweeter with greater LF extension, or at least more output low down & measured with the same method will automatically have higher HD in that BW; you should be able to use it as low though once corrected
• Dealing with the tweeter alone is interesting, but a bit academic; in practice the minimum usable frequency heavily depends on what you're partnering it with -you can't really separate them out. Since some woofers obviously handle more power than others, or have higher / lower distortion than others, then purely in terms of lowest practical filter frequency you ideally need to be picking your filter frequency based on that, as well as what the the tweeter can handle, using the maximum dynamic peak at the listening position as the guide. Essentially, you want the woofer / midbass / whatever to run out of steam first.

 

[mayhem13]

With the Volt woofer, your have a lot of options as to where to crossover to a tweeter. Although it’s marketed as a 180mm driver, it’s actual radiating surface is closer to a 150mm so I would expect the off axis response to be pretty constant up to around 3khz.

Here‘s a tip for you…..before considering where to cross as a design guideline based on measurements, place the woofer in an enclosure and just listen to it all by itself without any components for a bit….find its voice and then consider what you don’t like or are lacking

 

[Moondog55]

It drops off a cliff at 3 k tho, with a substantial hump just before that drop-off. WAG a second order at 2500 might work on it,

 

[mayhem13]

Look at the scale of the graph….2-3db bump isn’t much and the rolloff is very smooth. I’d throw a cap on it for a 1st order electric overlapping the acoustic and call it done.

 

[Moondog55]

How much smoothing is on that woofer graph tho?
The total result will depend on what XO is used on the tweeter and how much overlap there is.
You may be right tho, and a simple XO is often preferable and a cap to ground on the woofer may be all that's needed but the OP wanted the tweeter addressed and I'd still opt for second order at 2500
I don't like the look of the tweeters dip at 15K but I don't hear much above 17k anymore so it probably wouldn't bother me too much

 

[davidsrsb]

I would look at using a 28 mm tweeter. The lower Fs makes matching the midwoofer easier. Extension to 20 kHz is not much use to most of us here.

 

[Vance Peate]

Thanks for all the insights here everyone, much appreciated.

A little bit about this project:
3-way sealed design.
The Volt BM165.1 will be used with the SB (SB26STWGC-4) or Wavecor (TW030WA11) tweeter. The Wavecor is quite incredible, with an Fs of 410Hz, greater directivity and FR.

The low end will come from dual opposed (not isobaric) shallow woofers in an acoustic suspension setup. I have toiled with BR for too long... I could never capture the same transients of a sealed system accurately. I will be using a Hypex FA253 (employing FIR filters). Once the tweeter is decided upon, I will build out the filters and EQ. The system will be phase-coherent and should deliver some interesting results.

so I would expect the off axis response to be pretty constant up to around 3khz.

Yes, the off-axis response behaves fairly well, up to approximately 2,600Hz.
I am currently exploring the lowest minimum crossover point purely out of interest, however I do want to keep the point low for directivity reasons.

I like the concept of aligning everything in the physical domain and investing in the right drivers to avoid relying heavily on DSP.

Look at the scale of the graph….2-3db bump isn’t much and the rolloff is very smooth. I’d throw a cap on it for a 1st order electric overlapping the acoustic and call it done.

True. I think I'll try playing with the natural roll-off. However, there are some minor baffle diffractions around 2600hz, so hence the need to cross somewhat lower.

 

[weltersys]

I am solely interested in determining the safe minimum crossover point for the tweeter. I want the system to have the least amount of distortion at relatively high SPL.

Vance,

It would be helpful to define "relatively high SPL".
A spoon dropping from 1/2 meter on a ceramic plate can be over 115dB peak SPL at one meter, hands clapping over 120dB peak.

4. Determine where harmonic distortion / nonlinear distortions start to occur**?
5. Take into account the Xmax**?

Xmax is the maximum excursion for linear voice coil travel, the point where distortion will start to increase rapidly due to the voice coil leaving the magnetic gap.
Xmax is measured one way, the SB26STWGC-4 has 1.2mm peak to peak linear travel, 0.6mm Xmax.
Total harmonic distortion generally reaches ~10% at Xmax.
10% distortion is reached when harmonics are -20dB below the fundamental frequency.

Depending on your sensitivity to distortion, 10% be considered objectionable with some types of music, especially above ~800Hz.
That said, some people don't mind distortion at three times that level, and others notice 1/3 that level.

Thanks for the insights.
Can someone explain the below graphs and how to determine 'where distortion starts':

Distortion 2.83 volts, 315mm:

Reduce the SPL by 10dB for one meter equivalent.
At one meter, the driver has reached ~10% distortion at ~390Hz at ~75dB.
Reduce level by -6dB for each doubling of distance, at 2 meters 69dB.

A doubling of excursion is required to increase SPL by 6dB.
Four times the displacement (excursion) is required for each halving of frequency to maintain the same SPL.

At the drivers Fs of 780Hz, at 1meter we could expect it to achieve around 10% distortion at 87dB.
An octave above (1560Hz) at 1meter we could expect it to achieve around 10% distortion at 99dB.

If we use this excursion calculator
http://www.baudline.com/erik/bass/xmaxer.html
we might expect 3-to 4dB more output than the above when the driver goes a bit past Xmax:

At any rate, if you are picky about distortion, and like accurate transient response, don't cripple the system by crossing the tweeters too low to reproduce them with linearity.

Art

 

[Vance Peate]

At any rate, if you are picky about distortion, and like accurate transient response, don't cripple the system by crossing the tweeters too low to reproduce them with linearity.

Like with anything loudspeakers, it's about finding the right balance of compromises. Thanks, Art. This information is very helpful.

 

[mayhem13]

How much smoothing is on that woofer graph tho?
The total result will depend on what XO is used on the tweeter and how much overlap there is.
You may be right tho, and a simple XO is often preferable and a cap to ground on the woofer may be all that's needed but the OP wanted the tweeter addressed and I'd still opt for second order at 2500
I don't like the look of the tweeters dip at 15K but I don't hear much above 17k anymore so it probably wouldn't bother me too much

We DIYers have been burned by mfgr smoothing……I hear ya! One would ‘hope’ a premium driver mfgr like Volt wouldnt do that…….