Rerurn to Romy the Cat's Site

Hi,

Romy the Cat wrote:

This subject of digital crossovers became quite popular during last few years and many audio-people give up and go for a simplicity and painlessness of digital crossovering.

Most readily available digital X-overs come from a Pro-Audio background. Even if we ingnore the many inherent items that will compromise the performance of such devices, there is a crucial problem linking them into a home audio system, namely the level issue.

Let us assume we have a "state of the art" digital Pro-Audio crossover with perfect DSP and good quality converters allowing per output a dynamic range of 114db (which is not bad if not the best achievable) or a real 19-Bit resolution. The full scale level of that X-Over is for arguments sake at +22dbu or 9.8V RMS. Let us say we use a Driver of 105db/1W/1m and we connect between X-Over and speaker a Lamm ML-2 with a nominal input sensitivity of 0.775V (or 0dbu) for 18W Output.

Now, the resulting would have AT FULL POWER from the ML-2 a SPL of 118db @ 1m. The noisefloor and resolution limit of our digital X-Over would be at -114db re +22dbu but only -92db re 0dbu, the point where the amplifier is likely to clip. Now if we assume that normal listening levels for a big orchestra at FFFF is something like 105db peak @ 1m with this system or 13db lower than "full tilt" of the Amplifier we now have the system "resolution" (not neccsearily the noise floor though) at -79db with reference to the highest peaks during use.

In other words the music begins to turn to crap at levels around 80db below max....

In other words still, we have just succeeded in making our system a digital system with around 13 Bit actual resolution.

Now, the solution. Insert ONE AD-1 after EVERY output of the digital X-Over. Adjust the level on the AD-1 such that at the maximum levels normally listened to you retain around 6db headroom on the peak reading meters build into any such Pro Audio Digital Crossover (PADXO from here on). You may find that the PADXO output will clip before your amplifiers but if you never listen that loud, what the heck?

So, our system is now fitted with AD-1 between PADXO and ML-2. With an SPL of 111db/1m the PADXO reports "over". So, the full dynamic range of the source is retained and any "artifacts" from the X-Overs internal processing etc are suddenly for any given listening level around 33db lower than before.

Of course, most PADXO's (especially the Behringer DCX-2496 - see footnote 1) have analogue stages and powersupplies that are open to severe questioning and in fact suitable upgrades can increase dynamic range and improve the subjective soundquality to a very large degree.

EVEN THEN the only 32-Bit DSP system in most PADXO's may present additional problems that may become audible, as one would really need 64-Bit or larger floating point number crunching ability to ensure complex filter algorythms can be implemented without running out of resolution. So, the more you EQ and the steeper the Crossover curves the more problems you generat for the DSP.

So, affordable PADXO's are a pretty good tool and used well can work very well, but they will (not yet) match a really well implemented line level passive X-Over/EQ system. That said, if you implement the PADXO well into the system and you have done the neccesary mods the difference becomes somewhat notional and for most semi-demoronised audio morons perfectly livable.

Ciao T

Footnote 1 - The analogue stages of the Behringer DCX2496 and DEQ2496 are a textbook example how NOT to do it.

Their old DEQ8024 Digital Equaliser has excellent designs of analogue stages only compromised by parts quality, there a simple Op-Amp buffer with a "electronic transfomrer" servo type circuit was used in the input stage, placing just 1 Op-Amp in unity gain mode is series with each polarity for a balanced signal and for an unbalanced signal there would in effect be another Op-Amp as inveter on the non-driven input. The output stage had again 1 Op-Amp per polarity with another electronic servo circuit, there was a single Coupling Cap per polarity in the Input and equally in the output. Muting was via relais. About as minimal a stage count as you can get. The AD & DA chips where a little iffy, sadly, especially the DA.

Now their new stuff input circuitry first has a pair of coupling electrolytics to one Op-Amp to turn the signal from a balanced one to SE and to attenuate it, then another coupling cap to another Op-Ampto re-amplify this signal and yet another Op-Amp to regenerate a balanced signal to drive the ADC.

But the "Piece de resistance" is the Output. Here they take the nice balanced output signal from our DAC and turn it SE with an Op-Amp. They then couple the signal via an electrolytic cap to another Op-Amp and implement muting via muting transistors (YUCK) between the two. Following these two Op-Amp's are TWO MORE, to make a balanced output from the single ended signal. Just for good measure all resistor values in the circuit are pretty low, so the distortion from the Class AB output stage of teh Op-Amp's used is maximised.

If one wanted to, the DCX could be modified to accept SE inputs and outputs with a minimal number of highest possible quality Op-Amp's (OPA627 or similar on input, LM6171/72 r similar on output) or even without any Op-Amp's if we employ suitable transformers to replace all the solid state circuitry.... Either way result would be much improved and not require silly levels of attenuation to match HiFi gear as well.

Rerurn to Romy the Cat's Site