Hi Steve,
You may find this book of interest too. The book only covers analog active crossovers, but chapters 1 through 7 are applicable to IIR DSP versions of active crossovers such as yours. You'll find some interesting comments in a section entitled "The advantages and disadvantages of active crossovers" in chapter 1. Denibeni may find item No6 of interest too.
The Design of Active Crossovers - Douglas Self - Google Books
Peter
You may find this book of interest too. The book only covers analog active crossovers, but chapters 1 through 7 are applicable to IIR DSP versions of active crossovers such as yours. You'll find some interesting comments in a section entitled "The advantages and disadvantages of active crossovers" in chapter 1. Denibeni may find item No6 of interest too.
The Design of Active Crossovers - Douglas Self - Google Books
Peter
Steve, you might consider reading some more basic scientific articles (even online sources) or books about acoustics, the interference of waves from multiple sources, etc. The concept of phase and superposition of signals is in principle the same for sound, light, etc. any electromagnetic radiation really. These fundamentals are essential to understand how sound from the loudspeaker drivers will combine at the listening position, intereact with the room, its size and its boundaries, and so on.
Last edited:
Charlie, I have no problem with the physics of acoustics. I think that I have expressed before: My problem is with DSP and especially Hypex Filter Designer
What each filter does? How apply? How it affects the signal? And how one may fix something but spoil something else? Etc
For example, I gather that AllPass filters have something to do with Phase?
This is a filter option in Hypex FD. When I asked Hypex how to use, they replied that they don't have that knowledge!!!!!!!!!!!!! It's in their programme!!
I did have a problem with measurement with ARTA. Your article helped me understand how to use ARTA though I have many more questions to understand every option, the answers were not necessary to get me started.
Steve
This really shouldn't be this difficult, and I feel for you. Honestly, I think that Hypex FD is very confusing and definitely not for the first-timers. It is confusing and non-intuitive for more experienced users too. As good as their hardware is, this HFD sucks IMO (and they always promise a very soon update and improvement, for years now). I was considering myself their DSP unit but gave up after working with the demo HFD. There, I said it: the King is naked!
May I suggest that you first make a new set of measurements with ARTA: woofer near+far (1 m) and splice them, OR if that is confusing, only the far; then the mid and the tweeter. One measurement each at 1m. On axis only. Each measurement at 1m distance, at each driver's height (move the mic accordingly). Don't worry about minor details, feel free to use longish gate and smooth them reasonably.
With this set of measurements, use the fabulous VituixCAD simulation software. Create a 3-way active x/over. You can choose several platforms to work on (Behringer DCX, MiniDSP etc.... and yes, even Hypex!).
Then you can simulate with your measurements virtually everything: LP, HP, BP, AP, EQ, delay, phase reversal... This way you could see and learn what each does. Take your time. I'll repeat that: take your time.
After that, well, you could get back to HFD....
Last edited:
Draki
Then I could write a manual for everyone else.
You're right, it shouldn't be that difficult to use.
I'm glad that I'm not the only one that finds Hypex FD non intuitive etc.
Their hardware is good. I have now got my DSP system sounding better than the passive with the same loudspeaker drivers. The measurements, although a lot better, still don't look brilliant. Therefore I know that there is room for improvement.
Yes, I have to take my time. Problem is that there are more pressing commitments in life, so I may have to park this project for a while. A good manual with step by step procedure would have speeded this up
Then I could write a manual for everyone else.
You're right, it shouldn't be that difficult to use.
I'm glad that I'm not the only one that finds Hypex FD non intuitive etc.
Their hardware is good. I have now got my DSP system sounding better than the passive with the same loudspeaker drivers. The measurements, although a lot better, still don't look brilliant. Therefore I know that there is room for improvement.
Yes, I have to take my time. Problem is that there are more pressing commitments in life, so I may have to park this project for a while. A good manual with step by step procedure would have speeded this up
Tweeter Focal Audiom TD5
Mid ATC SM75-150S
Woofer Focal Audiom 15 WX
And some nice drivers those are! Sigh....
Hi Steve,
Yes, flatten the curve first (ideally to about 1.5 octaves past the crossover frequency), then apply the LR4 Lowpass. After this, use ARTA's target curve to compare your results with. The two responses should closely overlap to about 1.5 octaves past the crossover frequency.
Please keep in mind my comment in post No.103 (last paragraph) that the accuracy of a nearfield response is limited at high frequencies. This limit is dependant on the drivers diameter! For your 15 inch driver, an accurate response can only be guarenteed to about 332Hz!
Peter
Hi.
Now have my living room back for measurements etc.
I have bought and read Duhazi's suggestion of:
Accurate Sound Reproduction Using DSP, by Mitch Barnett.
Good but doesn't really help with the workings of HFD.
What I have gleaned from the book though which is relevant to my task:
Target curve should be flat (horizontal) upto 1KHz, then slope down to -6dB at 20KHz.
I'm going to ignore this and aim for all horizontal because if the recording has taken this into consideration, then would be compensating twice. I don't know therefore going to go a neutral flat horizontal.
More from the book:
"Design a digital crossover that does not introduce any frequency or time domain distotions. (seems logical, but how?) In addition to the benefits of bi-amping or tri-amping, digital crossovers open the door to driver linearization and time alignment" ( I thought this was possible passively too??)
"Linearize approx. +/- 2dB, Time alignment, usuallydelay Mid 1000 samples and tweeter 200 samples" (again I hope someone can explain this to me?)
Apart from reading the book, I have had the time to leisurely read all the posts again. I have at last come to the conclusion that my method may be wrong and that YSDR, Peter, Charlie and perhaps others are correct.
My method was to apply the crossovers straight away and then try to linearize the drivers. Why? To cut out of the equation the response beyond the crossover point so that there is less linearization, less filters/biquads etc.
Apparently this could cause drivers to be out of phase and also perhaps have "suck out" (I think it's called that) at crossover frequency. (there's not much suck out apparent in my total measurements though???)
The correct way (someone is going to say, says who?) is to linearize each driver to +/- 1dB first and to at least 1 or 1.5 octaves beyond crossover frequency. Once response of each driver is linearized, the crossovers can be applied and the resultant response should follow that of a Target Curve. If so, there should be no suck out at crossover and adjacent drivers should be in phase. IS THAT SO?
In the HFD manual: "For best crossover filter results, Linkwitz Riley are a good choice.............LR4 results in 360º phase shift - low output with 1 period delay"
I suppose LR8, 720º phase shift??
"The main advantage of LR filters is, when using them as low pass and high pass crossover filter (with same cut-off freq) your total filter response will be flat"
Now have my living room back for measurements etc.
I have bought and read Duhazi's suggestion of:
Accurate Sound Reproduction Using DSP, by Mitch Barnett.
Good but doesn't really help with the workings of HFD.
What I have gleaned from the book though which is relevant to my task:
Target curve should be flat (horizontal) upto 1KHz, then slope down to -6dB at 20KHz.
I'm going to ignore this and aim for all horizontal because if the recording has taken this into consideration, then would be compensating twice. I don't know therefore going to go a neutral flat horizontal.
More from the book:
"Design a digital crossover that does not introduce any frequency or time domain distotions. (seems logical, but how?) In addition to the benefits of bi-amping or tri-amping, digital crossovers open the door to driver linearization and time alignment" ( I thought this was possible passively too??)
"Linearize approx. +/- 2dB, Time alignment, usuallydelay Mid 1000 samples and tweeter 200 samples" (again I hope someone can explain this to me?)
Apart from reading the book, I have had the time to leisurely read all the posts again. I have at last come to the conclusion that my method may be wrong and that YSDR, Peter, Charlie and perhaps others are correct.
My method was to apply the crossovers straight away and then try to linearize the drivers. Why? To cut out of the equation the response beyond the crossover point so that there is less linearization, less filters/biquads etc.
Apparently this could cause drivers to be out of phase and also perhaps have "suck out" (I think it's called that) at crossover frequency. (there's not much suck out apparent in my total measurements though???)
The correct way (someone is going to say, says who?) is to linearize each driver to +/- 1dB first and to at least 1 or 1.5 octaves beyond crossover frequency. Once response of each driver is linearized, the crossovers can be applied and the resultant response should follow that of a Target Curve. If so, there should be no suck out at crossover and adjacent drivers should be in phase. IS THAT SO?
In the HFD manual: "For best crossover filter results, Linkwitz Riley are a good choice.............LR4 results in 360º phase shift - low output with 1 period delay"
I suppose LR8, 720º phase shift??
"The main advantage of LR filters is, when using them as low pass and high pass crossover filter (with same cut-off freq) your total filter response will be flat"
Hi Steve, welcome back!
Here are few possible answers to your questions:
- A flat target curve as you said is a good starting point. You can fine tune later if you feel so.
- Avoiding time domain response distortion requires FIR filtering (instead of IIR), which the Hypex products doesn't have (yet?). But FIR filtering is not 'perfect' either because it introduces a larger total time delay to the whole response which may be a problem with e.g. movies where the sound and the picture needs to be in sync.
- Linearizing the total response after applying the crossovers not necessarily causes dips in the response (as you see with your result), but you may not use the maximum gain from the interaction of two or more drivers, and this may cause additional unnecessary distortion and stress to the drivers.
- If you linearize your drivers above and below the crossovers point and then apply an LR type filter you can get a maximum possible gain and flat response from the interaction of two or more driver if the phases matches at the desired axis. To get phase matching at your desired (listening)axis you can fine tune with the channel delay in HFD.
- You are right, LR4 filters are 360 degree out of phase (1 period delay for the lower frequency driver), LR8 filters are 720 degree out of phase (2 period delay for the lower frequency driver). 1 period delay may not necessarily audible, but naturally, the 2 period delay have more chance to be audible.
Here are few possible answers to your questions:
- A flat target curve as you said is a good starting point. You can fine tune later if you feel so.
- Avoiding time domain response distortion requires FIR filtering (instead of IIR), which the Hypex products doesn't have (yet?). But FIR filtering is not 'perfect' either because it introduces a larger total time delay to the whole response which may be a problem with e.g. movies where the sound and the picture needs to be in sync.
- Linearizing the total response after applying the crossovers not necessarily causes dips in the response (as you see with your result), but you may not use the maximum gain from the interaction of two or more drivers, and this may cause additional unnecessary distortion and stress to the drivers.
- If you linearize your drivers above and below the crossovers point and then apply an LR type filter you can get a maximum possible gain and flat response from the interaction of two or more driver if the phases matches at the desired axis. To get phase matching at your desired (listening)axis you can fine tune with the channel delay in HFD.
- You are right, LR4 filters are 360 degree out of phase (1 period delay for the lower frequency driver), LR8 filters are 720 degree out of phase (2 period delay for the lower frequency driver). 1 period delay may not necessarily audible, but naturally, the 2 period delay have more chance to be audible.
Last edited:
Hi Dene, Thanks.
So I have decent woofer and mid measurements but you weren't too impressed with my tweeter measurement with high pass at 1KHz.
The XO freq for tweeter was to be 2KHz though you mentioned this was low. I have found Focal speakers with my tweeter with cut off freq at 2.2KHz and 2.5KHz. I could raise to one of these?
But for the tweeter measurement I am not comfortable to measure without high pass. I don't think that I can go much lower than 1KHz. See attached manufacturer's curve.
As you know, I have taken measurements with ARTA. Any advantage in taking measurements in HFD?
So I have decent woofer and mid measurements but you weren't too impressed with my tweeter measurement with high pass at 1KHz.
The XO freq for tweeter was to be 2KHz though you mentioned this was low. I have found Focal speakers with my tweeter with cut off freq at 2.2KHz and 2.5KHz. I could raise to one of these?
But for the tweeter measurement I am not comfortable to measure without high pass. I don't think that I can go much lower than 1KHz. See attached manufacturer's curve.
As you know, I have taken measurements with ARTA. Any advantage in taking measurements in HFD?
Yes, you can rise the crossover frequency of tweeter/mid for better power handling and lower distortion. Your mid dome used to 3.8kHz (if I remember correctly) in commercial designs anyway but i recommend lower than that if your tweeter can handle. Guess between 2.2-3kHz.
Measuring the tweeter's frequency response for crossover development without any high-pass filter is pretty normal method at any decent speaker designer/manufacturer , including Focal i bet. Just don't do too loud, it is really not necessary to do it loud, it can be done very quietly and the tweeter have no problem to handle that.
One advantage of HFD measurement is that you need to measure once per driver for frequency response correcting. But for best result you need more measurements to check off-axis responses and ARTA have the advantage here for measuring, overlaying and averaging more measurements like the off-axis responses.
Measuring the tweeter's frequency response for crossover development without any high-pass filter is pretty normal method at any decent speaker designer/manufacturer , including Focal i bet. Just don't do too loud, it is really not necessary to do it loud, it can be done very quietly and the tweeter have no problem to handle that.
One advantage of HFD measurement is that you need to measure once per driver for frequency response correcting. But for best result you need more measurements to check off-axis responses and ARTA have the advantage here for measuring, overlaying and averaging more measurements like the off-axis responses.
With REW and propably with ARTA you can set lower limit for the measurement sweep/noise. A protective serial capasitor for the tweeter might be a good idea, if the driver is rare or expensive. Remember to measure off-axis response of each driver in the baffle too! To catch distortion play at least at 90dB at 1m, preferably 95dB.
Each driver's response should be flat at least 1 octave past intended xo point, to get good phase mach and sound. Doing this first enables you to easily try several different xo points and slopes.
After setting crossovers, check timing/delay by playing two drivers at the time, also with reversed polarity on the other. Check spl response and step response.
Each driver's response should be flat at least 1 octave past intended xo point, to get good phase mach and sound. Doing this first enables you to easily try several different xo points and slopes.
After setting crossovers, check timing/delay by playing two drivers at the time, also with reversed polarity on the other. Check spl response and step response.
Last edited:
Looks pretty flat down to 2kHz, so if you use a 2,5kHz LR4 crossover for example, you get good phase and frequency response down to 2kHz, which is not so bad.
But if you want better phase and frequency response, apply a Shelf2 -> LowShelf -> Center freq 1000Hz -> Q=0,7 -> Gain=10dB filter + your choice of LR crossover obviously.
Of course, this is only true, if you have the exactly same tweeter frequency response like in the datasheet.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.