House Curve relationship to crossover targets

[Dmanstarr]

I’ve noticed when a house curve is used, when adjusting the target crossover slope in EQ, it changes the slope from the classical shape it would be if the target curve were set to flat.

Does this preserve the summing and phase relationship between the HP and LP’ed, adjacent drivers using the same house curve, assuming they perfectly match the adjusted slopes?

 

[deercreekaudio]

I’ve noticed when a house curve is used, when adjusting the target crossover slope in EQ, it changes the slope from the classical shape it would be if the target curve were set to flat.

Does this preserve the summing and phase relationship between the HP and LP’ed, adjacent drivers using the same house curve, assuming they perfectly match the adjusted slopes?

Given the assumption that you're either building a two-way active or a 2.1(2) system,then:

1. Generally crossovers are established first where you attempt to match the levels between the high frequency and low frequency drivers and also get a seamless amplitude transition.
2. Then using REW to correct or apply a house curb to the composite response, the overall response is modified and no changes made to the crossover characteristics
3. Here is a Tech Blog and Video that address this topic:
   1. Crossover Basics using miniDSP Device Console
   2. Using Key Measurements to Verify Basic System Setup

 

[Dmanstarr]

That’s not what I’d doing. I’m using eq to more closely match the measured acoustic slopes to the classical electrical crossover slopes.

All I need to know if if the represented target slope when modeling crossovers in in this way in the EQ section with a house curve applied considers the phase relationship relative to adjacent slopes, because it does modify the slope relative to the curve selected.

It would be strange if it didn’t do that, but I want to be sure the represented targets maintain phase coherence.

 

[John Mulcahy]

The target is a magnitude response, it doesn't have phase. Whether your outputs sum acoustically as intended at your desired response distance depends on the overall effect of the crossover and EQ filtering you apply and the response of the driver. A measurement will show that. Why add a house curve if you are trying to match a crossover response?

 

[Dmanstarr]

But changing relative, adjacent crossover slopes does affect phase when the adjacent drivers play together.

If the measured curves match the classical curve chosen (as represented with the crossover targets in the EQ section with no house curve applied), they will sum predictably.

When a house curve is used and the drivers are playing at different relative levels, this changes things.

I am trying to set a target for the slope of the crossover given that I am not building a flat response at the speaker system level in an anechoic space, but rather a response that correlates to the house curve at the listening position. This is what must be done in a car for correct perceived sound reproduction.

I am not the first person to model crossovers with REW in this way. What I want to know is whether the shift observed in the target curves corresponds to what would be seen modeling with a flat target and applying the eq for a house curve after the fact. There are reasons in a car that this would be easier to do “backwards”. For one, preserving available eq bands (given that some will be spent matching the curve). For another, reducing trial and error.

 

[Dmanstarr]

Also, if there’s no reason to have a house curve adjust the targeted crossover, why even have the feature in EQ at all?

If the answer is something other than maintaining the complementary nature of adjacent crossover slopes, I’d like to know what it is.

 

[John Mulcahy]

What I want to know is whether the shift observed in the target curves corresponds to what would be seen modeling with a flat target and applying the eq for a house curve after the fact.

The target is the (dB) sum of the crossover shape and the house curve. If the corresponding EQ is applied in series that is exactly the result it would produce, there isn't a way for it to be anything different. From a system perspective it is no different to having an input whose spectrum matched the house curve. That might not hold if you were to split the house curve effect into the parallel chains of the LP and HP paths, depending on how the house curve component was implemented in each case.

The reason I asked why you want a house curve is because you said you wanted to match the acoustic slopes to electrical crossover slopes, which have no such adjustment.

 

[Dmanstarr]

I should have been more clear. The intention is to match the measured acoustic slopes at the listening position. with eq after electronic crossover is applied to get things close in the measured response.

If the shift in the target slope in the EQ caused by the house curve input coincides with the slopes needed to maintain the phase relationship at the crossovers, I can just eq in the first place to match the adjusted slopes on each side of a driver pair.

Eq would be applied to each driver individually to match slopes shown in REW with said HC selected in REW. My hope is that they would sum correctly as they would with no HC and the driver pair eq’d flat to a flat target curve. My lack of knowledge regarding how eq near crossover points can screw up things is what makes me unsure if this is a good way go do it.

I understand the first part of what you said about the sum represented. When you say split the house curve into *parallel* chains, what do you mean?

In my mind, if the eq applied is individual to each driver in addition to the electric crossover effects, and thereby the measured curves match the predictions of the target crossover curves with HC selected, that should keep the crossovers complimentary with the sum aligning with the house curve at the xvr point.

The only way this wouldn’t be was if there is something I’m missing about how the house curves adjust the slopes. Is there?

 

[John Mulcahy]

As above, the target is the (dB) sum of the crossover shape and the house curve. The effect that has on the slope depends on the shape of the house curve through the sloped regions.

By parallel I mean applying house curve EQ in each driver's separate path. If that is done the overall effect of the house curve EQ becomes the acoustic sum of the house curve EQ in each path. That means the phase response of that EQ in each path becomes a factor in how the sum will behave. One way to ensure the house curve elements sum coherently would be to duplicate them in each path, but that doubles the house curve-related filters. The alternative is to apply any house curve EQ globally, i.e. to the signal before it is split by the crossover. That also has the advantage that the house curve element of the response can be changed independently of the crossovers. Crossover-related EQ would be best applied before making any house curve adjustments.

 

[Dmanstarr]

Very well explained. Thank you. The second way is what I wanted to do, but I don’t have the input mapping options to use global eq, only individual drivers (10 PEQ per plus crossover) annoyingly. So I guess I am forced to duplicate per side near crossovers.

IF I understand correctly, it seems if once the acoustic outputs per driver match the target curves with xvr slopes in the “EQ” section using the same HC, I should then be able adjust by adding identical filters to each driver’s eq at the crossover to fully meet the curve at the xvr pt with both playing while maintaining coherency, say by adding 3db at the xvr point with Q 1.6-ish when using LR24 slopes (assuming arrival times are the same) to compensate for the 3db notch in power response at an LR xvr pt. This regardless of eq filters applied to match the slopes on either driver and only applied to that driver.

If this isn’t right, I guess I still don’t get it. :/

 

[John Mulcahy]

Pretty much. I would apply any adjustment to achieve the desired crossover response without any house curve applied, could use nearfield measurements of each driver for that.