Replacing Electrovoice DX38 to DX46: FIR vs IIR.

[kwingylee]

I recently upgraded by Electrovoice DX38 with a DX46. I have only been using the DX46 for several weeks and still trying to learn about it.

I decided on the upgrade after a friend of mine let me borrowed his for a week, and I was able to get one a decent price. The DX46 is about $1800 new, not cheap. I got one for less than 1/2 that from someone that had 2.

Here is what I like. Based on my transferring the DX38 settings onto it. So I am not using the FIR functions at all. I am still letting it soak.

1. Its a 3 way crossover. While the DX38 is a two way. The DX46 allows me to run a separate set of tweeters. I am currently running a set a Electrovoice T350 tweeters (acquired from this forum) with my TAD2001/LeCleach horn/Jub clones. So having this option is good. With a DX38, I would have to buy another one.

2. It has a lower noise floor. WIth the crossover at idle, I could barely hear the noise. While with the DX38, it was somewhat pronounced.

3. To my ear, it sounds more detail and more organic. (please don't throw stones). Maybe better DAC and opamps?

4. I like the USB/ETHERNET option to set it.

Here is what I do not like:

1. Too many lights: almost a night out in Hong Kong look.

2. I find IRIS-NET to be less user friendly than RACE, and documentation is dismal.

3. Can't easily make adjustments off the front panel control without using the edit function. With the DX38, you just turn a knob.

4. The most serious but I will figure this out....I don't know how to use the FIR function. From reading the pdf files, it seems that Electrovoice has predefined polynomials selected...other than that, I can't seem to locate any more information. And IRIS-NET is said to be the only way to set the FIR functions.

I am looking at suggestions for optimizing the FIR settings. Help anyone?

[Chris A]

PROs:

1. Its a 3 way crossover. While the DX38 is a two way. The DX46 allows me to run a separate set of tweeters.

I understand that advantage.

2. It has a lower noise floor. With the crossover at idle, I could barely hear the noise. While with the DX38, it was somewhat pronounced.

This is somewhat surprising, since I presently can't hear any hiss from my listening position, and have to put my head inside the K-402 horns to hear the quiescent hiss. Must be a difference in setups. - amps, preamps, etc.

3. To my ear, it sounds more detail and more organic.

This device is a newer design, but I'd expect the DC-One to also sound just as good since it was designed about the same time.

CONs:

2. I find IRIS-NET to be less user friendly than RACE, and documentation is dismal.

I presently jut use the front panel on the Dx38 and fat-finger in the settings, so I don't have much experience with RACE. However, I did get some experience with the HI-Q net System Architect application from HK when I set my Crown XTi's (subwoofers). I found that application to be quite easy to use.

3. Can't easily make adjustments off the front panel control without using the edit function. With the DX38, you just turn a knob.

I see that with the front panel controls. I've also heard the same comment from miketn about his DC-One - since it shares the same input and output channel gain control schema as the Dx46. Rotary pots are nice to have, but may be sources of problems for professional setups - the knobs probably get banged around and bent/broken.

4. The most serious but I will figure this out....I don't know how to use the FIR function.

The manuals really don't say much, do they? I would expect that you would need to instrument the crossover regions (low and high frequency) to see if you can correct any phase mismatches in the crossover regions between drivers/horns. I also expect that you'll need to get the delays reset at the same time. FIRs apparently are a bit more difficult to adjust without changing something else in the process (...I'm reading these comments online...).

[Greg Oshiro]

<snip>Here is what I do not like:<snip>

<snip>4. The most serious but I will figure this out....I don't know how to use the FIR function. From reading the pdf files, it seems that Electrovoice has predefined polynomials selected...other than that, I can't seem to locate any more information. And IRIS-NET is said to be the only way to set the FIR functions.

I am looking at suggestions for optimizing the FIR settings. Help anyone?<snip>

I just did some work for a speaker manufacturer client with the same concern. They wanted to import an FIR filter into an IRIS-NET controlled device. IRIS-NET needs a *.gkf file that appears to be a proprietary format. Hopefully EV will provide some kind of *.gkf creation utility. Unless, of course, the suits are trying to force the end-loser to buy EV speakers so the FIR is useful.....

Who me? Cynical?

[Chris A]

I just did some work for a speaker manufacturer client with the same concern.

Wow...big time bummer.

Chris

[kwingylee]

Greg:

if you are correct then EV truly suck and there is no reason to by a DX46 at all unless you are using EV speakers!
Seems I made a big oops... and assume the unit would do something that it was not meant to....

have you tried contacting their help line?

[Greg Oshiro]

Greg: if you are correct then EV truly suck and there is no reason to by a DX46 at all unless you are using EV speakers! Seems I made a big oops... and assume the unit would do something that it was not meant to.... have you tried contacting their help line?

I'll try calling EV tech support after the weekend.

IRIS-NET does allow building a very limited set of filters. In the FIR dialog, un-click the "BYPASS" button and click on the (edit: (I think) "Define filter" "GENERATE FIR") button. You can make some brick-wall approximation low-pass, band-pass and high-pass filters, but no detailed EQ.

I have only recently dabbled in FIR filters. I think building an FIR equalization file takes some serious math. What is it you hope to accomplish with FIR filters?

[kwingylee]

Thanks Greg...

Why I want FIR? I just want to hear what the hype is all about? The zero phase error, brickwall filtering etc...

[Coytee]

For us dummies... What is a FIR filter and what is the benefit of one?

[Chris A]

For us dummies...what is a FIR filter and what is the benefit of one?

A finite impulse response (FIR, pronounced "fur") filter allows you to correct for both amplitude AND phase of crossover response, and can be used to correct for non-ideal impulse response. FIR filters are also called moving average (MA) filters, and can have very steep filter slopes, but require a more analysis capability to implement well, and also have a longer digital delay to implement for lower frequencies than the more common digital filters (minimum phase filters) that you've used to date, such as those in the EV Dx38 and virtually all other digital crossovers, for instance.

Most filters that you've dealt with in loudspeakers today are infinite impulse response [iIR, pronounced "err"] types. For instance, Butterworth and Linkwitz-Riley are IIR filters. These are also referred to as autoregressive filters.

Chris

[kwingylee]

Greg:

I called EV help line yesterday and was told that the FIR settings have been developed by EV engineers only for certain EV speaker arrays. I cannot create my own settings for my speakers.

Its a sad sad they in Mudville.....

Should have looked into this before I bought the unit.

A DX46 could be on the market soon.

[Greg Oshiro]

Greg:

I called EV help line yesterday and was told that the FIR settings have been developed by EV engineers only for certain EV speaker arrays. I cannot create my own settings for my speakers.

Its a sad sad they in Mudville.....

Should have looked into this before I bought the unit.

A DX46 could be on the market soon.

In my very brief dabbling in FIR filters I have discovered that the major difficulty with FIR filters is creating the FIR filter coefficients from the desired frequency (magnitude & phase) response. The filter coefficients are a string of numbers that represent the sampled impulse response of the desired frequency response. The string is usually some power of 2 (256, 512, 1024, ...) long. If you have a way of determining the set of coefficients that you want in an ASCII file, there may be DSP boxes out there that you could program with the ASCII file.

I'm not aware of any software that can create the coefficent file other than the MATLAB scripts I've written, but I'm sure there's some software out there that's more user-friendly than MATLAB, but I haven't looked.

How were you planning to define the FIR filters for your system?

[kwingylee]

Maybe I was only naive'. Honestly Greg I was not expecting to have to whip out my Oppenheimer book to compute the coefficieintd myself.

I am not in audio business so I was unaware of the "extreme user unfriendliness" of these FIR capable devices: where real time user modifcation seem to be a dream. How would one do A/B comparison? How do you voice systems? This is crazy.

With the DX38, there was a lot of experimentation to get the sound that I want. That does not seem to be possible with the DX46 FIRs.

Are all current FIR capable active crossovers like this? The DX46 came out in 2010, and EV has only developed coefficient files for 3 of their speaker lines? This is almost a joke.

[Greg Oshiro]

Maybe I was only naive'. Honestly Greg I was not expecting to have to whip out my Oppenheimer book to compute the coefficieintd myself.

There are "standard" FIR filter types, but I think they have questionable value for speaker equalization and crossovers. The cool thing about FIR filters is that you can take a measurement of a speaker, do some math on the data and have a set of FIR coefficients that will flatten the *measurement*. Note that the set of coefficients is only good for the microphone position at which the data were taken. Another thing FIR filters can do is independently define the magnitude and phase characteristics of a filter (at the expense of signal delay). With analog and IIR digital filters, the magnitude and phase characteristics are closely interdependent (all-pass filters can change the phase and not the magnitude, but there are still limits). So anything useful for our purposes is significantly more complex than specifiying frequency, gain and Q for a bunch of PEQ sections, delays, LPF/HPF settings and gains as we do for analog and IIR digital speaker processing.

Long story short, FIR filters take a lot more work. Break out Oppenheim & Shafer and prepare to generate a string of at least 512 seemingly meaningless numbers that define the FIR filter.

I am not in audio business so I was unaware of the "extreme user unfriendliness" of these FIR capable devices: where real time user modifcation seem to be a dream. How would one do A/B comparison? How do you voice systems? This is crazy.

If a DSP box is sophisticated enough to let us split an input signal and run it through an "A" FIR and a "B" FIR, each with a mute switch, and sum the filter outputs, we can do an A/B comparison in real time. Voicing systems would be a slow process. It seems more suited to evaluation by measurement than by ear, where real-time twiddling is the order of the day.

Are all current FIR capable active crossovers like this? The DX46 came out in 2010, and EV has only developed coefficient files for 3 of their speaker lines? This is almost a joke.

Real-time twiddling of the frequency response of an FIR EQ filter probably exists in some software/hardware system, but I'm not aware of it. The IRIS-NET software will let you adjust some FIR filter parameters by dragging a point on a graph (on the FIR filter for the EV N8000, I haven't looked at the DX46). I'm kind of old school when it comes to EQ and crossovers, so my knowledge is limited here and I haven't had the motivation to research it. Maybe this thread will change that.

Have you looked at MiniDSP? Low cost, FIR capable, other forum members have experience with it.

[Modified-Tractrix]

I would be exhaustive for EV to build a FIR library for any possible permutation. Simply use the Remez Exchange Algorithm to find the coefficients, but be warned... chose your average weighting factor carefully for a relevant exponential average. the benefits of the FIR device should make any trouble getting the settings worth while.

This may be helpful.

http://www.mathworks.com/help/toolbox/signal/ref/fircls1.html

[Chris A]

Simply use the Remez Exchange Algorithm to find the coefficients, but be warned... chose your average weighting factor carefully for a relevant exponential average. the benefits of the FIR device should make any trouble getting the settings worth while. This may be helpful. http://www.mathworks.com/help/toolbox/signal/ref/fircls1.htm l

That's an interesting way to program your Dx46, assuming that you can easily transfer in your Matlab filter settings. I hadn't seen that particular algorithm before: the old way of programming FIR filters was extremely cumbersome.

The problem remains, of course, in instrumenting your speakers in-room to verify that what you are programming into the FIR filter design is what you intended to achieve.

If my Jubs had separate tweeters, I'd actually be interested in trying that one out for grins. However, the only crossover that my Jubs have is at ~400 Hz, The Dx38 filters used could be better phase-matched in the crossover region using the Dx46, but this isn't quite as critical as the crossover from a midrange driver to a tweeter.

My center Belle is another story: currently it is tri-amped using a Dx38 and just manages to integrate between the two Jubs.

Chris