Designing a 2-way bookshelf speaker

[Hari Iyer]

The phase is present in both SPL and impedance graphs. The solid lines in both cases are magnitude, the dotted lines are phase.

Ok saw the phase after zooming as it's very thin line which can be missed in casual review.

I have never designed a 4 the order crossover, but the filter phase at crossover frequency does not intersect @ 360 or 0 degree ( resize the phase range to view the phase at 360 Deg)

The impedance phase presents a potential shortckt at 2100 Hz where the phase angle is 70 Deg+ the impedance is around 3.5 ohms.

 

[tcpip]

... the filter phase at crossover frequency does not intersect @ 360 or 0 degree ( resize the phase range to view the phase at 360 Deg)

The impedance phase presents a potential shortckt at 2100 Hz where the phase angle is 70 Deg+ the impedance is around 3.5 ohms.

Yes, you're right. The two drivers are not 100% in phase sync in these measurements, at the crossover point. But they are fairly close, because I am getting almost a -20dB notch when I reverse one of the drivers.

 

[Hari Iyer]

Yes, you're right. The two drivers are not 100% in phase sync in these measurements, at the crossover point. But they are fairly close, because I am getting almost a -20dB notch when I reverse one of the drivers.

Typically there should not be a very deep notch when you reverse one of the terminals. If you get these, means driver response are been rolled off earlier than required. Try getting the phase accurate at the crossover point.

 

[tcpip]

Typically there should not be a very deep notch when you reverse one of the terminals. If you get these, means driver response are been rolled off earlier than required.

That's a very pedantic way of looking at it. I can understand that you can't always get a deep reverse-notch, but to say that it should not be there seems to me to be incorrect.

Try getting the phase accurate at the crossover point.

What "accuracy" are you referring to? The current design will not be audibly different from any other tuning of the crossover which will make the phase more "accurate".

 

[Hari Iyer]

Flat FR should not be aimed at the expense of filter phase linearity.

 

[tcpip]

Flat FR should not be aimed at the expense of filter phase linearity.

With respect, I disagree, specially in this case.( And in any case we are talking about phase coherence here, not linearity. The two drivers need to be in phase with each other, which is coherence. None of them are linear.)

If the two drivers are badly out of phase (worst case is 90 deg), then of course, it needs to be fixed. But a small phase mismatch between drivers is not a problem.

Finally -- and this is the clincher -- serious phase mismatch impacts the SPL in very visible ways. I've seen it in the first Asawari design. I had such bad phase mismatch that I had peaks and troughs 5dB tall in the aggregate SPL curve, and these peaks and troughs extended one octave on either side of the crossover point. I had to seek help and go for an asymmetric-slope crossover to fix it. It's all documented in the Asawari 1 pages. I challenge anyone to show me a flat, smooth aggregate SPL curve in a design with serious phase mismatch. It is practically and theoretically impossible. When you get a smooth aggregate SPL, and you get a fairly deep reverse notch, then the drivers are substantially in phase. Any further phase coherence will not be audible.

In fact, I'm surprised that you are emphasising phase coherence here, when just a little while ago, you said that a deep reverse notch is not very desirable. A deep reverse notch is the ultimate proof of near-ideal phase coherence.

 

[Hari Iyer]

Very deep notch indicate rolloff of the driver earlier than required. It's easy to get a deep notch if you rolloff LF earlier than the crossover point and HF later than the crossover point. But this will impact filter phase adversely and the speaker system will not be coherent. Another way of checking is the step response of the filter.

 

[kapvin]

That's a very pedantic way of looking at it. I can understand that you can't always get a deep reverse-notch, but to say that it should not be there seems to me to be incorrect.

What "accuracy" are you referring to? The current design will not be audibly different from any other tuning of the crossover which will make the phase more "accurate".

Hi Hari,

I think it depends on the crossover slope order. first order crossovers are 90degrees out of phase, so you should not get a deep null on reversing polarity of one driver. 2nd order and 4th order are 180degrees and 360degress (1 phase rotation) to each other so a reversing polarity (or in the case of 2nd order same polarity) should result in a null. a deep null points to very good phase tracking around crossover point. I think tcpip is fairly well set there as per the graphs.

minor phase tracking can be always adjusted by tilting the baffle to increase the tweeter offset.

 

[kapvin]

With respect, I disagree, specially in this case.( And in any case we are talking about phase coherence here, not linearity. The two drivers need to be in phase with each other, which is coherence. None of them are linear.)

If the two drivers are badly out of phase (worst case is 90 deg), then of course, it needs to be fixed. But a small phase mismatch between drivers is not a problem.

Finally -- and this is the clincher -- serious phase mismatch impacts the SPL in very visible ways. I've seen it in the first Asawari design. I had such bad phase mismatch that I had peaks and troughs 5dB tall in the aggregate SPL curve, and these peaks and troughs extended one octave on either side of the crossover point. I had to seek help and go for an asymmetric-slope crossover to fix it. It's all documented in the Asawari 1 pages. I challenge anyone to show me a flat, smooth aggregate SPL curve in a design with serious phase mismatch. It is practically and theoretically impossible. When you get a smooth aggregate SPL, and you get a fairly deep reverse notch, then the drivers are substantially in phase. Any further phase coherence will not be audible.

In fact, I'm surprised that you are emphasising phase coherence here, when just a little while ago, you said that a deep reverse notch is not very desirable. A deep reverse notch is the ultimate proof of near-ideal phase coherence.

ditto. sorry I did not see this and reacted to the earlier mail.

I think as I mentioned in my previous post; that Hari is bringing his experience of first order crossovers to bear here. In a first order crossover, if there is a deep null on reversing phase of one driver, there is something more than a bit wrong. (it should be more a broad valley)

hari has been as per my conversations, a first believer in first order crossovers, even using them on metal drivers.

best wishes

 

[tcpip]

hari has been as per my conversations, a first believer in first order crossovers, even using them on metal drivers.

Good to see you, old friend.

And yes, I know Hari's liking for 1st order. We've exchanged a few messages here and there. That's the one order I'm least comfortable with. I hadn't realised that he was referring to 1st order when commenting on my crossover graphs. Sorry, Hari.

Anyway....

BHAIRAV (MTM) values rationalised, tweaked (again!)
I know, I know, I keep saying "This is the final, and I'll just go out and order the parts", but I needed to get the values of the parts into some sort of sensible, nicely rounded arena, and this means I needed to re-check that tweaking the values don't throw the crossover off whack. So, I tweaked the values a bit to get a fairly decent reverse null. And here are the final, final, final, details.

And I double-checked the reverse null, and this is what I am getting:

So, the depth of the reverse notch tells me that the two drivers are very closely coherent in phase at the crossover frequency, at the specific point in space where the mic was kept when I took these measurements. (They will be in and out of phase as you move the mic away from that point.)

The images are pretty high resolution -- you'll have to click on each image to see the full details.

Now to tweak and freeze the Bhoopali (TM design) crossover.

BHOOPALI (TM design) tweaked and cleaned up

I needed to get the values of the components into sensible figures, so I did that. I also tweaked a couple of values to get a deep, sharp reverse notch (speaker designers love this notch! ). And finally, I tried to see if I could flatten the very gentle rise in the SPL curve and maybe tame the wide hump in the highest octave. So this is what I got. I have added one capacitor (C5).

I reduced the value of L2, because modelling showed that it's not having an impact on the SPL curve. I also added a kind-of 1st order high frequency low-pass filter to the tweeter (the C5) to cut the uppermost octave. I had originally added a classic RC filter, with an R in series and a C in shunt. I played with the value of R and realised that even a value of zero Ohms is giving me good results, so I removed the R altogether. I tuned the value of C5 by hand to get the right degree of flatness.

Finally, some playing with the other component values got me a good reverse null.

The SPL is flatter than before, and the last octave has been tamed too.

If you go back a few posts and see what the Bhoopali SPL curve looked like earlier, you'll see the gentle rising SPL through the entire range from 500Hz to the top end. That is now flattened -- the SPL is impressively flat, +/- 1.5dB or so, I think. And the top-end hump is no longer raising its head. You can still see the hump-like shape in the last octave, but it's not rising above the rest of the curve.

And the reverse notch is very smart indeed.

I think this is frozen now -- I'll order the parts.

Whew. It's been a good journey. My next posts will be once the components arrive. I'll ask DIYaudiocart to see if they'll make the inductors for me. For the capacitors, I will probably use CTR or DEC brand yellow coloured polypropylene caps commonly used for crossovers.

 

[Hari Iyer]

ditto. sorry I did not see this and reacted to the earlier mail.

I think as I mentioned in my previous post; that Hari is bringing his experience of first order crossovers to bear here. In a first order crossover, if there is a deep null on reversing phase of one driver, there is something more than a bit wrong. (it should be more a broad valley)

hari has been as per my conversations, a first believer in first order crossovers, even using them on metal drivers.

best wishes

In my designs, I never aim for more than -6dB notch at crossover, but aim for 100% phase coherence.

BHOOPALI (TM design) tweaked and cleaned up

I needed to get the values of the components into sensible figures, so I did that. I also tweaked a couple of values to get a deep, sharp reverse notch (speaker designers love this notch! ). And finally, I tried to see if I could flatten the very gentle rise in the SPL curve and maybe tame the wide hump in the highest octave. So this is what I got. I have added one capacitor (C5).

I reduced the value of L2, because modelling showed that it's not having an impact on the SPL curve. I also added a kind-of 1st order high frequency low-pass filter to the tweeter (the C5) to cut the uppermost octave. I had originally added a classic RC filter, with an R in series and a C in shunt. I played with the value of R and realised that even a value of zero Ohms is giving me good results, so I removed the R altogether. I tuned the value of C5 by hand to get the right degree of flatness.

Finally, some playing with the other component values got me a good reverse null.

The SPL is flatter than before, and the last octave has been tamed too.

If you go back a few posts and see what the Bhoopali SPL curve looked like earlier, you'll see the gentle rising SPL through the entire range from 500Hz to the top end. That is now flattened -- the SPL is impressively flat, +/- 1.5dB or so, I think. And the top-end hump is no longer raising its head. You can still see the hump-like shape in the last octave, but it's not rising above the rest of the curve.

And the reverse notch is very smart indeed.

I think this is frozen now -- I'll order the parts., I will probably use CTR or DEC brand yellow coloured polypropylene caps commonly used for crossovers.

If the DEC or CTR brand have MKT written on them, they are Polyester. Polypropylene have MKP written on them which is not easily available at LT road.

I will be redesigning my stock TL FS speakers with a concept of capacitor-less series crossover which will use only inductors. Planning to start that project next week onwards.

 

[tcpip]

If the DEC or CTR brand have MKT written on them, they are Polyester. Polypropylene have MKP written on them which is not easily available at LT road.

I don't really worry, for this project, whether these are MKP or MKT -- don't scare me. The tiny differences in quality are perfectly acceptable for a budget build like these two. What is more important is that those capacitors are made in India, are easily available at affordable rates, and are used by speaker builders in India regularly.

In fact, the common yellow caps I get in L.Road are labelled "DEC/MER". I just found the web page of the manufacturer: http://www.deccapacitors.net/htm/products.htm You'll see the "MER" category under "General purpose capacitors". DIYaudiocart has some of these listed, e.g.: https://diyaudiocart.com/Electronic...0V-Axial?filter=52,66&sort=p.price&order=DESC

If I was going to fuss over capacitor differences at this level, I wouldn't have started with Peerless India poly cone drivers...

(I have two other projects cooking, where I will be using imported capacitors. )

 

[Hari Iyer]

IME, these caps are very good as bypass, will not use them for filters though in a high-end solutions

 

[tcpip]

Is diyaudiocart.com currently active, do any of you know? I sent a mail a few days back about inductors -- no response. Sent a reminder after 2 days. Tried calling one of their listed phone numbers -- it kept on ringing.

Should I look elsewhere, you think? Any idea who else can supply a whole bunch of air-core inductors?

 

[Dr.Lakshay]

You could look at diyaudio parts.Com also. Have no idea on current status of diyaudiocart.com.

 

[tcpip]

I'll immediately check them out, thanks. TheAudioCrafts too have inductors, but those are ready-made. It's nice if one can get someone to make them to order -- you get more flexibility in values and wire gauges.

TheAudioCraft has said they'll make me the inductors I want. I'll now get busy procuring the capacitors.

I have just heard from TheAudioCrafts yesterday with the prices of the inductors. Some of the larger value inductors will have slightly different values -- I'm getting good quality stock inductors, but the flip side is that values will change a bit here and there. 1.1mH is becoming 1.0mH. 1.6mH is becoming 1.5mH. And 1.8mH is becoming 1.75mH (which I'm sure will be undetectable).

So, I'll now tweak the capacitor values if needed, to ensure that things fall in place well with slightly changed inductor values. So, there may be a bit of crossover value changes now. Once this is done, I'll have the final BoM of the capacitors and can order them.

In the meantime, am waiting for a final quote from TheAudioCrafts for total amount including shipping from Delhi to Bombay.

Things are crawling along...

 

[kapvin]

I have just heard from TheAudioCrafts yesterday with the prices of the inductors. Some of the larger value inductors will have slightly different values -- I'm getting good quality stock inductors, but the flip side is that values will change a bit here and there. 1.1mH is becoming 1.0mH. 1.6mH is becoming 1.5mH. And 1.8mH is becoming 1.75mH (which I'm sure will be undetectable).

So, I'll now tweak the capacitor values if needed, to ensure that things fall in place well with slightly changed inductor values. So, there may be a bit of crossover value changes now. Once this is done, I'll have the final BoM of the capacitors and can order them.

In the meantime, am waiting for a final quote from TheAudioCrafts for total amount including shipping from Delhi to Bombay.

Things are crawling along...

if its an air-core wound inductor, might make sense to oversize a bit and then "cut" to size. of course you'd need a LCR meter (don't some multi-meters offer that capability?). -- this also allow you to "tune" by ear if needed.

on its own, minor variations don't mean much but in a higher order filter design, the cascading impact could be meaningful especially if all the errors are in "one" direction. you could model and see what tolerances your can live with.

I am happy to get the benefit of your experience on locally purchased inductors. and then buyfor my long delayed build.

finally what modelling environment are you using? I downloaded VituixCad, and it looks really comprehensive, but It looks like it needs a lot of reading to understand and use it to its full functionality (what I am hinting at, is please use, learn, use and then .. teach )

 

[tcpip]

if its an air-core wound inductor, might make sense to oversize a bit and then "cut" to size. of course you'd need a LCR meter (don't some multi-meters offer that capability?). -- this also allow you to "tune" by ear if needed.

This is what I usually get from most sources... I find they're a bit over-sized. But this time, when I'm getting them from a brand which manufactures them using machinery to set values, I'm expecting to find very little differences in value. (TheAudioCrafts, who's supplying them to me, has decided to give me branded ready-made ones, and I guess the quality will be very good. 16 AWG or 16.5 AWG.)

And I don't need an LCR meter, because any impedance measurement jig can measure passive LCR. In fact, every time I test my impedance jig with a passive component. It's probably good till about 1%.

on its own, minor variations don't mean much but in a higher order filter design, the cascading impact could be meaningful especially if all the errors are in "one" direction. you could model and see what tolerances your can live with.

True. That's why I'm a bit worried this time, this being my first 6th order. But in the past, up to 4th order, there has been splendid match between my modeled behaviour and final completed measured behaviour without any struggle.

I am happy to get the benefit of your experience on locally purchased inductors. and then buyfor my long delayed build.

I would not suggest you wait. Go ahead and buy. I don't expect any problems.

finally what modelling environment are you using? I downloaded VituixCad, and it looks really comprehensive, but It looks like it needs a lot of reading to understand and use it to its full functionality (what I am hinting at, is please use, learn, use and then .. teach )

I'm using Unibox, my old favourite, for box modelling, and VituixCAD for my xo. I'll happily teach you everything I know about using VituixCAD. I'll guide you step by step from zero to a finished xo. I'll be happy to do screenshare with you, chat on the phone, and help you out. No issues at all. This offer applies to others on the forum too.

Even today, I get help from my seniors. We exchange CAD files, we send each other crossover designs. I'll just be extending that process to you.

Sorry, none of the crossovers I'm discussing in this thread are 6th order. I'm doing two other builds in parallel, and those are 6th order. Material for another thread. The ones here are simple 4th order acoustic.

 

[tcpip]

Inductors have arrived today.

Branded factory-made inductors shore look good. This is the first time that I'm not using hand-wound inductors.

Capacitors have been ordered. The project inches along. Just keeping you updated, in case you feel I've gone to sleep again.

 

[tcpip]

Capacitors have arrived. I decided to go with Audiophiler caps from Taiwan, supplied by TheAudioCrafts. These are MKP, not MER like the Indian ones made by CTR or DEC. And they're still not crazily expensive.

This weekend will be a busy one. Hope I get free time this weekend to finish the crossovers.