3-way to active - Hypex FA253 - learning project

Would appreciate any feedback on it (from yourself as well as others), and similarly, I would like to understand reasons for your approach (why you picked these xo frequencies, slopes and EQs).
My approach depends on personal taste, and I like simplicity. I prefer shallow slopes because all the speakers I prefer have shallow slopes. But it's not all about filters and topology, the selection of good drivers is more important.
I aim to have a flat FR on axis with a good PIR and SP responses, so I choose drivers, xo frequencies and filter slopes accordingly, starting from the point that the midrange must have a very good extension (250-2500Hz).
Basically, I use the optimiser to set up my intended filter, like this

e.g., Midrange: red is measured FR, blue is with filter applied, gray is with filter and PEQ/shelving applied.

1748859849858.png


Do the same for all the drivers and then optimise all the blocks to get a good compromise between on axis, polar FR, PIR and SP.
 
Last edited:
Will you be measuring again at the church?

How repeatable are your SPL levels during measurements? For example, my AVR volume is usually set to the same number for sweeps. I think I use ten in REW, and seventy on the AVR. The exact numbers don't matter, but it allows me to go back and check something later.
 
I may do another try in the church or maybe outdoors. I do like how much cleaner the measurements come out.

Yes, I took a note of gains n the chain: REW(-17dB), computer (max) and Hypex (0dB) during the church session. The mic distance was also measured at 1m +/- 3cm.
 
So, you can mix and match measurements from each session. That's an advantage. You have a lot of good data, but part of it is not so good. (The last mid measurements above 1k for example.. But the measurement below 1k is excellent.)

Do you see the baffle step in the measurements of both the mid, and woofers? There's a rise in both starting at about 500hz and increasing to 1k.
 
Last edited:
No, I'm wrong about the mid measurement.

Take a very close look at the low end of the mid and woofer measurements taken at the church. They are virtually the same! I suspect that the woofers were playing during the mid measurement. I would not use that mid measurement at all.
 
My approach depends on personal taste, and I like simplicity. I prefer shallow slopes because all the speakers I prefer have shallow slopes. But it's not all about filters and topology, the selection of good drivers is more important.
I aim to have a flat FR on axis with a good PIR and SP responses, so I choose drivers, xo frequencies and filter slopes accordingly, starting from the point that the midrange must have a very good extension (250-2500Hz).
Basically, I use the optimiser to set up my intended filter, like this

e.g., Midrange: red is measured FR, blue is with filter applied, gray is with filter and PEQ/shelving applied.

View attachment 1467922

Do the same for all the drivers and then optimise all the blocks to get a good compromise between on axis, polar FR, PIR and SP.
With that software, can you select a target response, and just show the target and raw response together on the same graph? Then can you manually make filter changes?
 
The magenta target response is the one you can create with the Optimizer tool by setting the various filter values (Type, slope, Fc, etc.) in the relative window.
Beside you can import all the FR you want as overlays (dash-dotted lines) and use those as target responses. Then you manually choose the component and its relative value in the crossover tab.

The big advantage is that you do not have to open a new window to modify the parameters (you select the component by clicking on it and use the mouse wheel to modify its value). All the values modifications are shown on the graph in real time.
You can also use the Optimizer tool to automatically get a desired target response but I am not good at it.
 
  • Like
Reactions: AlmaAtaKZ
Nearfield measurements can be used in VCAD sim, if baffle layout mic location and driver locations in all xyz dimensions are correctly set. Other way is to measure farther away with fixed mic position and 2-way souncard, then impulses carry the delay that is caused by pathlengths and also baffle diffracion effects.

When doing measurements to confirm simulation, real mic should be positioned as the virtual mic in simulation - same distance and height relative to drivers.

When I set up dsp without simulation, I set the mic height between mid and tweeter, at 60-100cm depending on how first reflections appear. First I measure and adjust only mid and tweeter, when short gating can be used. In a tall 3-way bass driver(s) measuring and matching need more distance and longer gating, preferably outdoos and groundplane.

The graph below shows time domain as step response of this project 2nd xo version, measured at 1,5m indoors. Step is another way to look at impulse, it shows polarity of drivers and delays between "ways". Now we can see that all drivers have negative polarity and that T and M integrate well. Intergration is analyzed from how valleys between peaks integrate, in case ot 4th order. Woofer peak comes too late, so mid and tweeter shlould have approx. 100us (0,1ms) delay. Floor reflection etc. appear a bit after 3ms. https://www.audiosciencereview.com/forum/index.php?threads/step-response-does-it-really-matter.1999/

View attachment 1450414
@Juhazi , could you please help to take this advice in:
  • how to I generate a view like this? this is REW, right? or is this VCAD?
  • how do I interpret the M and T delays from this? I do not see where the 100us comes from and why.

Would appreciate.

In a few days I will do the 4th attempt at consistent measurments and will do VCAD simulation and filters/EQ set up based on these.