Active speaker DSP test bed -- enough interest?

[neildavis]

I've got a couple of active speakers in mind for the PE design team, so I tried to come up with an easy way for others to make their own active speaker designs. I proposed a simple DSP test bed that anyone with basic electronics skills could build, and was going to write a series of articles that explained the technology and how to use it. The articles would supplement the PE write-up for the active speakers, so anyone would have the information they needed to "roll their own". The test bed idea is in a PowerPoint briefing at this link. The articles would probably follow a sequence like this:

1.Active Speakers – why?
2.Digital processing basics
3.DSP devices for active speakers
4.ADAU1701 and SigmaStudio
5.The microprocessor:how to control the DSP
6.Design tools for active speakers (ASD)
7.Case study #1:a small subwoofer
8.Case study #2:a 3-way design
9.Wireless Active Speakers
10.Case study #3:A line array with DSP
11.Future directions:going beyond the ADAU1701

But I'm not going to bother with these articles if there isn't enough interest. So that is my question--are there enough people interested to make this worth my time, and if so, what are the problem areas that need the most explanation. I don't intend to provide a thorough explanation of the theory, math or programming, as there are many good resources for getting this technical background either online or at a nearby school. I feel what is missing is how to apply this theory to making active speakers, so that would be my focus. But that's still a broad topic, and I would like to hear ideas about how to make this effort more useful. There have been some PM's going back and forth on this topic with a couple of members who previously expressed interest, but I need to hear from others...

BTW, does anyone know how to start a blog with this version of the forum software? I tried following the Help, but that didn't work, and it looks like all of the blogs were started with the old software.

[DanP]

Man Neil, I'd be very interested in articles 2 - 5, especially 4 and 5. I know how much work this would be, but I hope you move forward with this endeavor as yours is a unique perspective and knowledge base in this hobby.

Dan

[isaeagle4031]

Man Neil, I'd be very interested in articles 2 - 5, especially 4 and 5. I know how much work this would be, but I hope you move forward with this endeavor as yours is a unique perspective and knowledge base in this hobby.

Dan

Agreed on all points. Im working on a few PA installations now, so more in depth data is a real interest to me.

[killa]

I am sure people would be interested in the basics (mainly how to design and implement active xo) but I think a lot of people (myself included) just like an easy way to create a crossover. I might do a write up on converting a passive speaker to an active speaker soon. It will show how to measure the passives FR and how to manipulate the filters to match. If you would want to use it you are free to do so. I am no expert but have a pretty good idea of an easy way to accomplish my goal.

[Millstonemike]

Great introduction / write up. I briefly scanned the entire content and one item stuck out: Noticeable lag ... less than 1/2 second (DSP) delay. Would such a delay preclude HT use?

I ask because I have a commercial shelf system with significant DSP delay. It can not be used with video because of the mis-match between the video and audio timing (e.g., lips move, sound comes out later).

[augerpro]

I'd really like this too Neil. Especially the wireless write up for a soundbar I'm making.

[neildavis]

Great introduction / write up. I briefly scanned the entire content and one item stuck out: Noticeable lag ... less than 1/2 second (DSP) delay.

No--this is the amount of time it takes the little 8-bit microprocessor to calculate the floating point numbers for the biquad coefficients. I just pointed it out because usually you don't do complex math in an Arduino, and I was impressed that it worked at all. But once the coefficients are calculated and loaded into the DSP, there is essentially no delay. The ADAU1701 does 1024 instructions every audio sample time, so the only delay is the filter settling time and maybe one audio sample (at 48K sampling that's 21 usec).

BTW, the Arduino can't do double precision floating point, but it doesn't matter--32-bit single precision is perfectly adequate for the ADAU1701. That's because the ADAU1701 uses 28-bit coefficients, and 32-bit precision is fine for that. I checked about 10 values against the coefficients that Charlie Laub calculates in his ACD spreadsheet with double precision, and the hex values were never off by more than 1, and usually they were the same..

[neildavis]

... but I think a lot of people (myself included) just like an easy way to create a crossover.

That's the goal: have the Design Team project be something that can be easily downloaded or purchased at low cost, and then have the DSP test bed for the people who want to explore the "innards" so they can design their own. I don't know yet whether that is really possible, but that is the goal.

[killa]

I ask because I have a commercial shelf system with significant DSP delay. It can not be used with video because of the mis-match between the video and audio timing (e.g., lips move, sound comes out later).

Interesting. I know the minidsp has about 1 to 2ms of delay but most TVs (led/lcd) have about 20ms and up so at least in this particular instance the dsp should be faster than the TV.

[bcodemz]

Great introduction / write up. I briefly scanned the entire content and one item stuck out: Noticeable lag ... less than 1/2 second (DSP) delay. Would such a delay preclude HT use?

I ask because I have a commercial shelf system with significant DSP delay. It can not be used with video because of the mis-match between the video and audio timing (e.g., lips move, sound comes out later).

The DSP in most commercial systems should not cause any significant delay, unless it uses FIR filters and uses it at low frequencies, but this is something you won't see in any sub $10000 system.

What is more likely the case is that you have a system that is wireless in some way, like a Sonos. They intentionally delay, say 0.5 seconds, to give themselves enough buffer to have enough stability to account for wireless delay/issues so you don't hear a dropout every single time the wireless signal is degraded from some interference.

[ernperkins]

Great work and I'm definitely interested.

[dld]

very interested.
thanks!

[Kevin K.]

Really hoping you follow through with this Neil. I'll be taking the "Active" plunge in the next couple of weeks on the CBT arrays. Rick will be helping, otherwise, you know... Not everyone has an opportunity for a private tutor so this write up could be very useful and informative. I will be doing more active designs in the future and a simple two way will probably be first on the list.

A question for you and/or any of the other designers. Do you think a person needs to be fluent in "passive" design before they try "active" designs? "Cart before the horse" so to say.

[killa]

A question for you and/or any of the other designers. Do you think a person needs to be fluent in "passive" design before they try "active" designs? "Cart before the horse" so to say.

No. Active filters are very easy to do. You basically just tweak the response where you want it. I personally just make changes and measure the result. Some people might use programs to spit out filter files but i like to do it on the fly.

[neildavis]

A question for you and/or any of the other designers. Do you think a person needs to be fluent in "passive" design before they try "active" designs? "Cart before the horse" so to say.

I sure hope not, because I don't consider myself fluent in passive crossover design . I understand passive designs, but I tend to avoid them, so I haven't had much practice with them. Passive crossover design is a specialized art with some unique challenges, and it is not necessary to master that art in order to do active designs.

As far as I'm concerned, active is a lot easier. The main reason is that you don't need to be as concerned with the driver response, as active designs often use steeper crossover slopes, and as a result the system response is dominated by the filter response. That makes driver selection a lot easier. Also, you don't need to worry about the driver impedances interacting with the crossover, so there is typically less interactive "tweaking" to achieve the target response. And multi-way active speakers are usually much easier than similar passive designs, as you can often use textbook filters to get very good results. But all of this will get discussed in detail in that first article: 1. Active Speakers – why?