The Thing to Know About DSP Processors - FYI

Many, if not most CD players on the market today have DSP manipulation of some sort (bass, treble, EQ, time alignment, crossovers). It is considerably less expensive and often times more effective for a manufacturer to manipulate sound signal within a product in the digital realm. So, most manufacturers do. Also many enthusiasts that have a large amount of time and money invested in their systems have purchased outboard DSP devices.
The thing to know about DSP processors is that they are inherently phase problematic devices. The other thing to know is how to over come this characteristic. You can, you just have to know how.

If you have a DSP device in your system (most every one does) there are some things that you should know before deciding on your settings.
1.     Why certain crossover settings sound better than others but with out consistency from car to car or even within a system from channel to channel.
2.     Why measured time alignment settings don’t always work out as they should.

First let’s address the crossover problem. In theory digital active crossovers will not have the same phase shift characteristics that their passive counterparts have. But they do have phase shift it is only of a less predictable nature than their passive counterparts. The reason for this is that as a signal is processed in the digital realm it is delayed by a few fractions of a millisecond (digital latency). To make matters worse this effect is accumulative. The more you manipulate the signal more the delay you introduce into that channel. Very, very few products have any sort of compensation circuit built in to address this issue. The reason for this is two fold. First, it’s expensive and complex to implement in design. Second, from the engineers (often somewhat narrow) perspective if the end user can identify this problem they are most likely capable of compensating for the problem.

So what’s happening here is that unwanted time delay is occurring, and the amount of time delay is very unpredictable because it is dependant on both the design of the processor as well as the settings by the end user.    

You might find that various crossover slopes have quite a profound effect on the sound. Aside from the slope its self, the speakers in this channel are most likely changing phase in relation to all of the other speakers in the system due to the delay syndrome. You need to ask yourself, is this why this setting sounds better. Is it the setting, or the phase relationship?

Here is how you compensate.
In the event that your processor has an incremental time delay feature you simply play with your delay settings per channel after you have set them using physical measurement. The amount of unwanted delay introduced by your processors design will most likely only amount up to a single wave length or less at 250Hz (mid bass high pass). This means that the delay is plenty to mess things up royally, but easily compensated for by the very processor that introduced it to begin with.

Start with the speaker closest to you (usually left front). Listen to a track that you are intimately familiar with and work the delay setting for this channel back and forth around your predicted level until you get the sound that you are looking for. Then move on to the next closest speaker. It’s worth mentioning that as frequency rises past the mid range (1-2K) phase becomes less relevant in the grand scheme of things. Always strive for phase coherency, but if you are off by a few cycles yet in phase relative to opposing drivers you will not likely be able to hear the difference, even if you try.

At 6Khz, wave length is 2.26”. So every 1and 1/8th” difference in distance or less than 1 tenth of one millisecond of delay you will be 180 degrees out of phase!

At this point it should be evident why estimated (tape measurement) time alignment values might not work out the way that you would expect them to. Don’t give up! Time alignment values based on distance measurements are easy to obtain and are a very valuable piece of knowledge. You just need to append to that the knowledge of possible delay caused by the processor to achieve total digital nirvana.

Every .1 millisecond of delay = 1.4”
Use this to determine your initial settings before your listening tests.

Listen, listen, listen, and have fun!