After many years of building passive crossovers I knew there had to be a better solution.  15 years ago, enter the EV Dx34 DSP crossover.  Dialing in a crossover with this unit was a breeze compared to constantly swapping passive components.  No more having to have to deal with the chain reaction of tweaking passive components, swap a coil, then a cap needs to be swapped, attenuate, mix and repeat, etc, etc, over and over until the crossover is just right. Active analog crossovers could be just as difficult as their passive counterparts, but at least I didn’t have the passive, soul sucking (power), components inline with my amplifier.

    With an active crossover in my audio system now I could really get all aspects of the crossover just right. After progressing from analog active, to a dedicated hardware DSP crossover, I ran across Allocator by Thuneau. This is a VST crossover that has standard filters that can stacked to achieve up to 42dB/oct slopes, along with EQ and delay functions. Using Allocator was simple and fun. I measured each driver in my speakers, then created targets using LEAP CrossoverShop. After the targets were created I tweaked Allocator to fit the targets as best as I could. When the targets were dialed in I adjusted the delay of each driver for the best time alignment. Presto, a great crossover was realized. Now what the heck is that giant block diagram below? -The next evolution in my stereo.

    After using Allocator for while and messing around with Reaper DAW combined with various VST EQs to make crossovers, I decided to try making DSP crossovers using convolution VSTs. Convolution is simply combining two signals resulting in a third signal.  An example is using a music signal combined with a speaker correction filter and the result of the two is sent to the speakers.  

[Input Signal]->[Crossover Impulse Response]->[Filtered Response][1]

You like da juice???

VST Convolution

    Generating convolution filters is fairly straightforward.  A frequency response (FR) of each driver needs to be measured.  Using a user determined target response, the intended FR for each driver (crossover points) and system FR, the target FR of each driver is divided by the measured FR to generate a transfer function to match the target FR.  The drivers target FRs consists of high pass, low pass or band pass acoustic slopes.  The system FR is the overall response of all of the drivers combined, the speaker’s frequency response.  I use Audiolense to generate filters. I also have Acourate, but I prefer Audiolense’s interface and I’ve had much better success with its filters.  In the future I’ll create a page on generating filters with Audiolense.

Signal Flow

   One of the issues with running a setup like this is how to route the audio.  In a typical setup all of the Windows applications are routed through the Windows sound driver.

[Application]->[Windows Sound]->[Sound Card]->[Stereo]

With a DSP convolution setup the system needs a way to route the source audio to the convolver so the audio can be processed.  In the past I used to use two sound cards, one for the source applications and the other for the convolver.  It is possible to use the same sound card for output and input (loopback the S/PDIF IN and OUT) but it’s not very reliable. I routed the S/PDIF output of the source sound card to the convolver sound card.