MFB Isobaric

[Duke]

While the sub can go into the midrange, I'll be cutting off at about 300Hz and let the actual mid driver take care of the rest.

I think that would be low enough.

I guess that brings up a good question: are isobaric speakers uni or bidirectional in their "use" of sound?

The radiation from the front of the outer woofer's cone would have the same pattern whether the enclosure is isobaric or otherwise.

Not too hard if I can help it. The main thing is it's about 10db shy on sensitivity compared to the mid/hi, but I'm only probably going to be running those at either 10 or 20 watts at most.

To a first approximation, you'll need about ten times as much power going into the woofer section in order for it to "keep up with" a midrange/tweeter section that is 10 dB more efficient. And note that a voltage sensitivity comparison does not translate directly into an efficiency comparison if the drivers have significantly different impedances.

I thought one of the main selling points of isobaric is it didn't need as much space?

That's usually how it gets "marketed". Ime there is a worthwhile improvement in bass quality resulting from using a larger airspace behind the inner woofer. This also increases the susceptibility of the outer woofer to over-excursion, which is an additional trade-off. That being said, imo unless minimizing enclosure size and/or maximizing excursion-limited SPL is the top priority, you might consider sizing the airspace behind the inner woofer more like you would if the outer woofer was not even there.

If this is to be a vented isobaric, I suggest including some provision for adjusting the port length, as I'm not confident that normal vented box modelling will be sufficiently accurate.

 

[AnalogSteph]

It's not so much the DACs (those are actually pretty good for audio) as the ADCs. Noise figures are way better on SAR than on delta sigma (typically).

Is your main objective recording bats? Typical decent higher-end audio ADCs will generally maintain their analog noise floor to 40-60 kHz before shaped noise takes over, sometimes even higher.

MEASUREMENTS: RME ADI-2 Pro FS R Black Edition as ADC.

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

EARLY LOOK: E1DA Cosmos ADC - affordable high performance measurements for the audio hobbyist!

A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.

archimago.blogspot.com

You realistically need about 110 dB(A) from the A/D-D/A chain, so if analog levels are well dialed-in you can get away with a good midrange part, even a CS5361. You would be appalled to see what real-life inexpensive active DSP speakers are using (think JBL 30*, ADAM T*V etc.)... the DAC-amps in there are generally good for 103 dB(A) or so, so no need for an ADC that's much better. Obviously the higher-priced ones ship with better converters as well.

 

[BKr0n]

What do you have connected in the analog domain that needs so much SNR anyway?

One thing I wanted to do with DSP was upsample and servo control. In order to get the upsampled data I would need a much more sensitive ADC. As far as using SAR DACs, there's actually a piece of gear that come to mind: https://www.schiit.com/products/yggdrasil Has been using SAR DACs since its inception.

I use the relatively ancient BoxSim.

Nothing wrong with old reliable software. There's a reason a lot of test equipment and even some aerospace stuff that uses older operating systems and software. It was made in a different Era.

The radiation from the front of the outer woofer's cone would have the same pattern whether the enclosure is isobaric or otherwise.

So if I'm reading that correctly, then it would be omnidirectional, and it's more about the enclosure at that point than the speaker?

To a first approximation, you'll need about ten times as much power going into the woofer section in order for it to "keep up with" a midrange/tweeter section that is 10 dB more efficient. And note that a voltage sensitivity comparison does not translate directly into an efficiency comparison if the drivers have significantly different impedances.

They're all 4 ohms so that shouldn't be an issue there. Is that 10 times to each woofer or just to the bass section itself?

Is your main objective recording bats?

I mean... it's technically still audio

Typical decent higher-end audio ADCs will generally maintain their analog noise floor to 40-60 kHz before shaped noise takes over, sometimes even higher.

Interesting. I figured it wouldn't even be that good. Not the first time in my life I've been wrong lol. I'm assuming they come in 32 bit models as well?

 

[voodooless]

One thing I wanted to do with DSP was upsample and servo control.

Why would you need upsampling for servo control? Bandwidth of the control loop is a few hundred Hz at best.

 

[BKr0n]

Why would you need upsampling for servo control? Bandwidth of the control loop is a few hundred Hz at best.

I meant upsampling of the audio signal. Sorry for the confusion.

 

[voodooless]

I meant upsampling of the audio signal.

Why not just sample the data at the native DSP rate? Any half decent DSP has a bunch of ASRCs. No special requirements are on the ADC side to make those work. They are usually more than transparent enough.

In order to get the upsampled data I would need a much more sensitive ADC

This doesn’t make sense to me.

In any case, if you want to do motion feedback, it’s probably easier to do it in the analog domain. You can get kits for fair money and they work pretty good.

 

[BKr0n]

Why not just sample the data at the native DSP rate? Any half decent DSP has a bunch of ASRCs. No special requirements are on the ADC side to make those work. They are usually more than transparent enough.

Ah OK. So I would first sample the analog input to the DSP. The DSP would upsample to the rate I would want to use (probably 768khz pcm). That would get sent to the speakers (preamp, power amp, etc.). A sample of that would be taken off of the speaker, stepped down, and looped back to the ADC at that higher sampling rate.

In any case, if you want to do motion feedback, it’s probably easier to do it in the analog domain. You can get kits for fair money and they work pretty good.

100% agree there. The MFB circuit would be separate from the amplification section. I've been looking at piratelogic, but i still want to take a little longer to study the circuit before I make a final decision on design.

 

[voodooless]

Ah OK. So I would first sample the analog input to the DSP. The DSP would upsample to the rate I would want to use (probably 768khz pcm).

Why would you do that?

 

[BKr0n]

That way I can do all the signal synthesis (crossover, filtration, etc) in digital, then send it out in analog.
Edit: I forgot a part lol. Upsampling for the sake of fidelity.

 

[BKr0n]

Although am thinking about segregating the bands in analog to simplify the design. It's easier to do a crossover than use multiple DACs... yeah good idea, me...

 

[Duke]

So if I'm reading that correctly, then it would be omnidirectional, and it's more about the enclosure at that point than the speaker?

Yes. Below 300 Hz that woofer will tend to be effectively omnidirectional, and a fairly small-faced enclosure wouldn't change that much.

They're all 4 ohms so that shouldn't be an issue there. Is that 10 times to each woofer or just to the bass section itself?

In the case of an isobaric it would be each woofer, because only the outer woofer is "seeing" the outside world. The inner woofer is mainly just modifying the response of the outer woofer.

 

[voodooless]

That way I can do all the signal synthesis (crossover, filtration, etc) in digital, then send it out in analog.
Edit: I forgot a part lol. Upsampling for the sake of fidelity.

That is nonsense

 

[BKr0n]

You would be appalled to see what real-life inexpensive active DSP speakers are using (think JBL 30*, ADAM T*V etc.)...

Yes I work with a guy who loves his JBL system. I keep trying to tell him there's better out there... *sigh*

Obviously the higher-priced ones ship with better converters as well.

Well if I do decide to go at least top shelf in audio ADCs, would one from TI or ESS with good specs be worth the engineering hassle?

Yes. Below 300 Hz that woofer will tend to be effectively omnidirectional, and a fairly small-faced enclosure wouldn't change that mmuch.

Ok I think that actually works out a bit better. By that logic, I could potentially put it wherever in the enclosure as long as the sound is routed out correctly and there's enough airflow. Yes?

In the case of an isobaric it would be each woofer, because only the outer woofer is "seeing" the outside world. The inner woofer is mainly just modifying the low-end response of the outer woofer.

So if for example, I sent 10 watts to the mid and 10 watts to the tweeter, each sub would still need potentially 100 watts a pop?

That is nonsense

Isn't that the point of a higher sample rate?

 

[D!sco]

There should really be a better formal guide to using VCAD as a pure sim tool, though Kimmosto has some very good guides for making driver measurements and using them in VCAD. I don't find this as helpful when determining which drivers to use. I'm gonna give you a quick walkthrough of my data sim workflow, because it isn't as easy as I'd like and took some time to come up with.

My usual workflow in VCAD is to start by setting up a bunch of folders. It's something like this:
<Project Name>
-->Driver Measurements
----><Driver1>
------>(Manufacturer Graph).png
------>(Frequency Response).frd
------>(Impedance).zma
----><Driver2>
...etc.
-->Box Simulations
----><DriverX>
------>(Vented Enclosure)
------>(Sealed Enclosure)
------>(Compact Enclosure)
...etc.
-->Merged Data
----><DriverX>
------>(Merged Measurements and Box Simulations)
...etc.
-->Diffraction Simulations
----><DriverX>
------>(An absolute pile of generated polar data)
...etc.

These folders are ordered top to bottom as the first to last simulation to accomplish.
1) Driver Measurements are done either with a mic or by using the SPL Trace in tools. I would save the FRD and ZMA data in separate folders for each driver.
2) I've already mentioned the Enclosure tool, but the value of the box sim cannot be overstated. You can save and restore different enclosure orders, driver arrangements, whatever. It's the best tool. But it does not replace actually constructing a box and measuring it.
3) In Tools, there is also the Merger. You will want to merge your "Far" driver measurements with your "Near" box sims. Don't forget to include the impedance from the box, as it will affect everything down the line. These should be exported in preparation for the next step. There's a lot of guesswork here, so don't take it too seriously. Any minute details under 500hz won't matter once the sound hits the room anyways.
4) The most important tool after box simulation is the front baffle. In many ways this a more important decision as it geometrically restricts the remainder of your enclosure choices. This also breaks a lot of driver arrangements. I recommend calculating each driver separately, even if they are an "array", like in an MTM or TMWW, as I find it better simulates impedance, cancellations, and diffraction. This tool does not help with waveguides. Those need their own measurements. Remember to place the microphone icon at the center of each driver placement, and select "Full Space" after uploading the merged data. It's also helpful to have a CAD drawing of the baffle for reference/adjustment and the necessary data points for driver size. It will probably be difficult to simulate an isobaric correctly because of this, as they act like one driver.

After that it's all crossover work. Don't forget to space each driver correctly to simulate dispersion. If you do it all right, it should look something like this:

 

[voodooless]

Isn't that the point of a higher sample rate?

No! The point of the higher sample rate is to capture more information, specifically at higher frequencies (> 20 kHz). It will not add more resolution or information density to the audible part of the audio. That's not how the sampling theorem works.

It seems like you are way over your head with all of this.

 

[Duke]

So if for example, I sent 10 watts to the mid and 10 watts to the tweeter, each sub would still need potentially 100 watts a pop?

That's how I see it, to a ballpark first approximation.

By that logic, I could potentially put it wherever in the enclosure as long as the sound is routed out correctly and there's enough airflow. Yes?

Yes.

Are you up for contemplating something really cool?

Calculate the floor-bounce dip frequency at your anticipated listening position using the calculator at this link: https://mehlau.net/audio/floorbounce/

Then put your woofer section in a SEPARATE enclosure from your mid/tweet section. Put the woofer section down on the floor, perhaps even against the wall. Then set your crossover frequency between the woofer and mid ABOVE the floor-bounce dip frequency. The mid/tweet section remains at normal height, again perhaps against the wall.

This approach (based on the work of Roy Allison) not only neatly sidesteps the floor-bounce dip, but it also gives your woofer some very effective boundary loading.

 

[BKr0n]

There should really be a better formal guide to using VCAD as a pure sim tool, though Kimmosto has some very good guides for making driver measurements and using them in VCAD. I don't find this as helpful when determining which drivers to use. I'm gonna give you a quick walkthrough of my data sim workflow, because it isn't as easy as I'd like and took some time to come up with.

My usual workflow in VCAD is to start by setting up a bunch of folders. It's something like this:
<Project Name>
-->Driver Measurements
----><Driver1>
------>(Manufacturer Graph).png
------>(Frequency Response).frd
------>(Impedance).zma
----><Driver2>
...etc.
-->Box Simulations
----><DriverX>
------>(Vented Enclosure)
------>(Sealed Enclosure)
------>(Compact Enclosure)
...etc.
-->Merged Data
----><DriverX>
------>(Merged Measurements and Box Simulations)
...etc.
-->Diffraction Simulations
----><DriverX>
------>(An absolute pile of generated polar data)
...etc.

These folders are ordered top to bottom as the first to last simulation to accomplish.
1) Driver Measurements are done either with a mic or by using the SPL Trace in tools. I would save the FRD and ZMA data in separate folders for each driver.
2) I've already mentioned the Enclosure tool, but the value of the box sim cannot be overstated. You can save and restore different enclosure orders, driver arrangements, whatever. It's the best tool. But it does not replace actually constructing a box and measuring it.
3) In Tools, there is also the Merger. You will want to merge your "Far" driver measurements with your "Near" box sims. Don't forget to include the impedance from the box, as it will affect everything down the line. These should be exported in preparation for the next step. There's a lot of guesswork here, so don't take it too seriously. Any minute details under 500hz won't matter once the sound hits the room anyways.
4) The most important tool after box simulation is the front baffle. In many ways this a more important decision as it geometrically restricts the remainder of your enclosure choices. This also breaks a lot of driver arrangements. I recommend calculating each driver separately, even if they are an "array", like in an MTM or TMWW, as I find it better simulates impedance, cancellations, and diffraction. This tool does not help with waveguides. Those need their own measurements. Remember to place the microphone icon at the center of each driver placement, and select "Full Space" after uploading the merged data. It's also helpful to have a CAD drawing of the baffle for reference/adjustment and the necessary data points for driver size. It will probably be difficult to simulate an isobaric correctly because of this, as they act like one driver.

After that it's all crossover work. Don't forget to space each driver correctly to simulate dispersion. If you do it all right, it should look something like this:
View attachment 320038

Thank you so much for this! Ok I got a much better place to start now.

No! The point of the higher sample rate is to capture more information, specifically at higher frequencies (> 20 kHz). It will not add more resolution or information density to the audible part of the audio. That's not how the sampling theorem works.

Well yeah and wouldn't that noise be just be distortion?

It seems like you are way over your head with all of this.

The only way to learn is to fail. Not knowing a fact is simply failing to know it already.

That's how I see it, to a ballpark first approximation.



Yes.

Are you up for contemplating something really cool?

Calculate the floor-bounce dip frequency at your anticipated listening position using the calculator at this link: https://mehlau.net/audio/floorbounce/

Then put your woofer section in a SEPARATE enclosure from your mid/tweet section. Put the woofer section down on the floor, perhaps even against the wall. Then set your crossover frequency between the woofer and mid ABOVE the floor-bounce dip frequency. The mid/tweet section remains at normal height, again perhaps against the wall.

This approach (based on the work of Roy Allison) not only neatly sidesteps the floor-bounce dip, but it also gives your woofer some very effective boundary loading.

Ha! I knew there was something I was missing! (Well there's a lot but my point is still valid). I figured there would be some sort of issue with bass and how it would travel.

I guess the cool thing is I wanted to make this about the size of a book case speaker. I would mechanically decouple the mid and tweeter from the sub in the same enclosure. Think the whole "room in another room" thing you would do constructing a soundproof room in a home.

If all I need to do is route the sound, then I can mess with not only the placement of the speakers within the enclosure, but it also means I have more options in terms of figuring out MFB

 

[voodooless]

Well yeah and wouldn't that noise be just be distortion?

What noise? There is no noise…

 

[BKr0n]

I mean that: wouldn't any superfluous data just become noise in the completed signal?

 

[voodooless]

I mean that: wouldn't any superfluous data just become noise in the completed signal?

No