AVS Forum banner

1 - 14 of 14 Posts

·
Registered
Joined
·
39 Posts
Discussion Starter · #1 ·
Most speakers have a crossover that splits the incoming signal into two parts: a high-frequency part that goes to the tweeter, and a low-frequency part that goes into the woofer.

The crossover components that do this work need to be able to handle high power and large currents because they perform the splitting at the speaker after the signal has been amplified by the amplifier.

Furthermore, these components are affected by the non-constant resistance of the speaker drivers, unless the drivers themselves are compensated by Zobel networks.

A better way to perform the crossover function is to split the signal at line-level first, before the signal is amplified by an amplifier. This requires twice as many amplifiers, because each speaker driver is now powered by it's own amplifier.

Thus, I decided to work on bi-amping my Dynaudio Audeince 40 speakers.


The first task is to measure the actual (electrical) crossover characteristics of the loudspeaker.

The Dynaudio Audience 40 speaker, according to its specifications, has a 1st-order crossover for both tweeter and woofer at 1.8kHz.

As we will see, that is hardly an accurate description of what actually happens!

The method I used to measure the actual crossover characteristics is known as a transfer function.

A test signal is generated, amplified, and sent to the speakers.

Test leads are connected to the tweeter and to the woofer, and the signals at the drivers are measured.

The transfer function measurement takes the response at the tweeter or woofer, and divides it by the measured stimulus at the input to the speakers.

This way, any non-uniformity in stimulus (both magnitude and phase) is canceled out, leaving only the response of the speaker's internal crossover network.




The stimulus used was a uniform chirp signal with 16,384 discrete tones at a sample rate of 128kHz.

The tones covered the span from 3.9Hz to 57.6kHz, with a separation between tones of approximately 3.9Hz.

The stimulus was amplified using a Technics amplifier, and fed to the Audience 40 speaker (one speaker at a time).

The signal at the input of the speaker was measured (the reference channel of the transfer function measurement), and the signal across the tweeter or the woofer was measured simultaneously (the signal channel of the transfer function measurement).

The resulting transfer function (both magnitude and phase) is the transfer function of the speaker's crossover.


The first speaker is Column B, and the second speaker is Column C.

The crossovers in both speakers show very similar characteristics, and can be closely approximated by an ideal 1st-order high-pass filter (Column D).

However, the crossover frequency shown is at 4kHz! This is a lot higher than the specified 1.8kHz.

Also of note is the resonance in the woofer (around 60Hz) that makes it into the tweeter's response curve!

Also, the maximum amplitude of the fitted high-pass filter is down 3.5dB (0.667x); this is likely due to the tweeter being more sensitive than the woofer, so it needed to be shelved.


 

·
Registered
Joined
·
39 Posts
Discussion Starter · #2 ·
The next graphs shows the measured crossover response of the woofer.

Again Column B is the first speaker, Column C is the second speaker, and Column D is a special fit.

Note, however, the compressed vertical scale – it barely spans 20dB!

The low-pass filter is significantly less than 1st-order (6dB/oct, or 20dB/decade), and manages only maybe 8dB/decade of attenuation!

Since both speakers perform almost identically, I must assume this was intentional.

I do not think a 1st-order low-pass crossover, say at 250Hz, for the woofer would crossover well to the tweeter, which appears crossed at 4kHz!

Thus the challenge is to create a low-pass filter that mimics the very mild crossover that the woofers seem to have.





The fit in Column D was achieved by summing two low-pass filters in parallel.

The first low-pass filter is a 1st-order filter at 500Hz with an overall gain of -3.22dB (0.69x).

The second low-pass filter is also a 1st-order filter, but at 9kHz and with an overall gain of -10.2dB (0.31x).

This fit gives more bass response (below 100Hz), but less mid-bass response (around 200Hz), but we're talking a
 

·
Banned
Joined
·
15,491 Posts

Quote:
Originally Posted by ianchan1970  /t/1524557/bi-amping-the-dynaudio-audience-40-speakers#post_24537373


My next step is to build the line-level crossovers with the fitted characteristics
Or just get one of these or the equivalent:
http://www.rane.com/ac22s.html
 

·
Banned
Joined
·
3,677 Posts
Wouldn't the woofer need to be crossed significantly lower than 4khz as it would likely have started beaming much before that? Also, I am confused as to how you are implementing the two parallel LPF's? What does the sum of the two LPF's come out to?
 

·
Registered
Joined
·
39 Posts
Discussion Starter · #5 ·
The sum of the two (parallel) LP filters results in the data shown in Column D.

The gain and crossover frequencies of the two filters were selected to give the closest possible match to the measured woofer crossover response.

At low frequencies (eg 20Hz), the gain of the two filters are 0.69 & 0.31, and they add up to 1.0 (0dB). The phase of both LP filters there is also 0 degrees, so the combined phase remains at 0 degrees as well.

As frequency goes up, there magnitude and phase will change differently in the two LP filters, but if you do a sum, you will get the magnitude and phase shown in Column D.


I have built an opamp circuit that does just this, and the response is just as predicted.

I have been doing some preliminary listening with one biamped speaker this way.

The differences between normal and biamped speaker are so far surprisingly subtle.

There is slightly greater clarity and definition in the biamped speaker, but it is by no means a night and day difference.


I will post more data and pictures later, when I hopefully have 2 biamped speakers to listen to!
 

·
Banned
Joined
·
3,677 Posts
I am on my iPhone and don't see a column D.

What do you mean by the measured woofer crossover response? Do you mean the measurement of the frequency response of the woofer with its stock crossover?


Also, what do you mean by the gain of the two LPF's? What is gain?


So have you actually done any changes in the crossover yet? If so, how does the new, modified crossover sound? Wouldn't it be easier, and perhaps better to go active?
 

·
Registered
Joined
·
39 Posts
Discussion Starter · #7 ·
Yeah, sorry I wasn't clear.

The measured response of the tweeter and woofer are with the built-in crossover, and that behavior is what I am trying to mimic using line-level, active components.

So yes, I am making an active crossover, and then amplifying the high-frequency sounds for the tweeter and the low-frequency sounds for the woofer separately.


What was really surprising was that the actual built-in crossover response was very far from the nominal 1.8kHz crossover specified in the manual.


The "Column D" I am taking about is the legend in the graph I posted.

There are 3 traces, labelled Column B, C & D, which correspond to the response from one speaker, the second speaker, and the theoretical active crossover response.
 

·
Registered
Joined
·
39 Posts
Discussion Starter · #8 ·
The line-level crossover circuits were implemented with OP2227 opamps with +/-12V wall-wart power supplies.

The circuit diagram is as follows:



The circuit shown is for one channel (left or right), and it actually has an input buffer, which is not shown.

The parallel low-pass filters are summed in the 3rd opamp U3, with a gain of -1.025x and -0.464x for the 500Hz and 9kHz LP filters respectively.

The high-pass filter is at 4kHz, and is unity gain.

The high-pass filter should really be a gain of 0.667x, but I decided to instead gain up the LP filters by 1/0.667 instead, so the 0.69x and 0.31x gains become 1.034x and 0.465x.

A photo of the actual circuit is shown here:



So how do the Dynaudio Audience 40's sound biamped?

Well, I still only have one amp, so I can only A/B one speaker at a time, but the biamped speaker is definitely less veiled.

Whereas the normal speaker can sound slightly congested in the midrange, and “speaker-bound,” the biamped speaker sounds more free of the confines of the speaker. Ambiance is better conveyed, and you can hear “deeper” into the recording. Textures feel more real, especially the brass instruments from Stereophile's Test CD 2 (St. James' Infirmary).

Now, it is not a night-and-day difference. If you were listening to them casually (e.g. while cooking or chatting with friends) you likely won't notice the difference. But while relaxing with a glass of wine to Diana Krall or Norah Jones, the biamped speaker definitely brought a smile to my face.


My one speaker is currently biamped with the Zamp v3, but I have two Topping TP20 (Tripath-based) amps on the way. I hope they will sound just as good as the Zamp, otherwise I will have to spring for the pricier Zamp.

As fun as this project is, it is only an exploratory project into bi-(tri-)amping.

My ultimate goal would be to tri-amp my floor-standing Audience 82's (with Zamps driving mid & treble).

I'll live with the biamped Audience 40's a little bit before deciding if the effort of tri-amping the Audience 82's will be worth it...

My better half is already kind of annoyed with the mess she's had to endure (and still endures) with the ongoing biamping project.
 

·
Banned
Joined
·
3,677 Posts
Wouldn't an easier way to go about this be ripping the stock crossover out, and using something such as a MiniDsp or DCX into a decent 2-channel amp? Another possibility would be to use a Behringer iNuke1000dsp as it has built in active crossover DSP/EQ capability. Now granted, the iNuke won't work in this fashion with a 3-way speaker, but, a MiniDsp would be an excellent choice.
 

·
Registered
Joined
·
8,854 Posts

Quote:
Originally Posted by ianchan1970  /t/1524557/bi-amping-the-dynaudio-audience-40-speakers/0_50#post_24541728


The differences between normal and biamped speaker are so far surprisingly subtle.

There is slightly greater clarity and definition in the biamped speaker, but it is by no means a night and day difference.
If you emulate the existing speaker level transfer function at line level actively, are you really surprised there's little difference? That's what I'd expect.
 

·
Registered
Joined
·
4,832 Posts
I agree, if executed correctly any perceived differences should by subtle... especially in a quality loudspeaker such as this.
 

·
Registered
Joined
·
39 Posts
Discussion Starter · #12 ·
I did not want to change the tonality of the speakers, but improve the quality of the sound.

I want to do this by removing possibly distorting passive components in the path (especially the iron-cored inductor in series with the woofer), and give the amp far more control over the drivers by driving them directly.


A valid question is what would Dynaudio engineers have done with the crossover if they were making an active speaker from the start?

I wish i knew!

It probably wouldn't look like the less-than-first-order woofer crossover.... But I don't have any data to tell me otherwise.
 

·
Banned
Joined
·
3,677 Posts
What do you mean by the amplifiers having more control over the drivers? Are you saying that you believe the passive crossover somehow affects how the amplifier drives the actual drivers??
 

·
Registered
Joined
·
39 Posts
Discussion Starter · #14 ·
I think the passives do reduce the amplifier's damping factor (control) of the drivers.


Take the woofer, which normally has a big inductor in series with it as a low pass filter.

Past the crossover frequency, when the amplitude at the driver is reduced by 10x (-20dB), the inductor impedance is, say, 40 ohms, compared to the woofer's at 4 ohms. The amplitude to the woofer is reduced, but so is the amplifier's control over the woofer because it has to go through 40 ohms of the inductance!

Doing the low-pass at line level and then amplifying, only the amplitude at the woofer is reduced, but not the amplifiers damping factor.


Even in the pass band, I measured the dc resistance of the inductor as 0.4 ohms, in series with the woofer.

That's equivalent to 100ft of 16awg wire.

But doing low pass at line level, and connecting the woofer directly to the amp, I bypass the inductor altogether, resulting in better damping factor.


So yes, I think the active crossover allows (much) better control of the drivers.
 
1 - 14 of 14 Posts
Top