Designing a 2-way bookshelf speaker
- Thread starter tcpip
- Start date
just use the simple battery trick
I've done the battery trick for my Peerless HDS tweeters. It worked. [emoji2]
keith_correa
Well-Known Member
The one lug which is wider is +ve
Oh I am using it reverse, let me fix it
That's the reason you should measure step response if possible.
Yes, this is correct. There are even very faint "+" and "-" marks on the plastic bases of the terminals.
I started out with Step 2 of speaker design. As a recap, the steps in my speaker design process, simplified, are:
So, I'm now starting Step 2.
I took some impedance measurements of the Peerless India TL26 tweeters and faced my first disappointment. Here is what the graphs look like for the two tweeters I bought:
Look at the graphs, see the peaks. The resonant peak is in the region of 1 KHz. The published documentation, both on diyaudiocart and theaudiocrafts, says that the Fs is 596 Hz. This huge discrepancy is not just because of piece-to-piece inconsistencies -- the published data is plain wrong.
This discovery makes this tweeter as boring as the old Peerless India SR10DT. The primary reason I had found this tweeter interesting was because I thought I'll get the ability to cross it over quite low, maybe at 1.6KHz to 2KHz. Now I don't think I should cross over less than 2.5KHz, 3rd order or 4th order. I could have done this with the venerable SR10DT too. Luckily, I've bought this tweeter to go with the 6.5" Peerless India midbass driver whose claimed SPL curve is quite smooth and extended. If my measurements of the actual SPL (see Step 5 above) even roughly matches the published data, then I'll be able to use the TL26 even with a crossover frequency of 2.5KHz.
I don't suspect any inaccuracy in my ARTA LIMP measurement setup. I tested it with other drivers and with passive components (capacitors, resistors), and the impedance is spot on.
Now we know why I don't like using Peerless India drivers. This sort of inaccuracy is not worth the time and trouble. And I won't buy any more TL26 in future -- better get some affordable Vifa or SB Acoustics tweeters instead.
Finished doing the T/S parameter measurement of the 5.25" drivers:
I'll start enclosure modeling for the MTM now, which will use these drivers. And I need to finish T/S parameter measurement of the 6.5" Peerless India driver which I'll use for the TM design.
This is the first project I'm using ARTA for. I like it.
I paid up some two days ago, and I intend to stick with it now.
- Step 1: Study published (and other) data about drivers, select the drivers for your design
- Step 2: Measure driver impedance, and from it, derive
- Thiele/Small parameters for woofers
- resonant frequency for tweeters
- Step 3: Design the enclosure. For this, I use Unibox.
- Step 4: Construct the enclosure, including all finishing, polishing, etc. Fix the drivers on the enclosure.
- Step 5: Take SPL and in-box impedance measurements. The impedance measurements taken in Step 2 are not useful after Step 2. The new measurements are needed. For this, I use ARTA now, with a mic.
- Step 6: Design the crossover using the SPL and in-box impedance. For this, I will now use VituixCAD, Xsim, or maybe Speaker Workshop.
- Step 7: Build the crossover.
- Step 8: Start a period of listening tests and tuning the crossover. This involves a series of iterations of listen-measure-modify the crossover.
- Step 9: Freeze the design and write up all details.
So, I'm now starting Step 2.
I took some impedance measurements of the Peerless India TL26 tweeters and faced my first disappointment. Here is what the graphs look like for the two tweeters I bought:
Look at the graphs, see the peaks. The resonant peak is in the region of 1 KHz. The published documentation, both on diyaudiocart and theaudiocrafts, says that the Fs is 596 Hz. This huge discrepancy is not just because of piece-to-piece inconsistencies -- the published data is plain wrong.
This discovery makes this tweeter as boring as the old Peerless India SR10DT. The primary reason I had found this tweeter interesting was because I thought I'll get the ability to cross it over quite low, maybe at 1.6KHz to 2KHz. Now I don't think I should cross over less than 2.5KHz, 3rd order or 4th order. I could have done this with the venerable SR10DT too. Luckily, I've bought this tweeter to go with the 6.5" Peerless India midbass driver whose claimed SPL curve is quite smooth and extended. If my measurements of the actual SPL (see Step 5 above) even roughly matches the published data, then I'll be able to use the TL26 even with a crossover frequency of 2.5KHz.
I don't suspect any inaccuracy in my ARTA LIMP measurement setup. I tested it with other drivers and with passive components (capacitors, resistors), and the impedance is spot on.
Now we know why I don't like using Peerless India drivers. This sort of inaccuracy is not worth the time and trouble. And I won't buy any more TL26 in future -- better get some affordable Vifa or SB Acoustics tweeters instead.
Finished doing the T/S parameter measurement of the 5.25" drivers:
I'll start enclosure modeling for the MTM now, which will use these drivers. And I need to finish T/S parameter measurement of the 6.5" Peerless India driver which I'll use for the TM design.
This is the first project I'm using ARTA for. I like it.
I paid up some two days ago, and I intend to stick with it now.
Last edited:
I started out with Step 2 of speaker design. As a recap, the steps in my speaker design process, simplified, are:
- Step 1: Study published (and other) data about drivers, select the drivers for your design
- Step 2: Measure driver impedance, and from it, derive
For this, I use ARTA LIMP now, with an impedance measurement jig.
- Thiele/Small parameters for woofers
- resonant frequency for tweeters
- Step 3: Design the enclosure. For this, I use Unibox.
- Step 4: Construct the enclosure, including all finishing, polishing, etc. Fix the drivers on the enclosure.
- Step 5: Take SPL and in-box impedance measurements. The impedance measurements taken in Step 2 are not useful after Step 2. The new measurements are needed. For this, I use ARTA now, with a mic.
- Step 6: Design the crossover using the SPL and in-box impedance. For this, I will now use VituixCAD, Xsim, or maybe Speaker Workshop.
- Step 7: Build the crossover.
- Step 8: Start a period of listening tests and tuning the crossover. This involves a series of iterations of listen-measure-modify the crossover.
- Step 9: Freeze the design and write up all details.
So, I'm now starting Step 2.
I took some impedance measurements of the Peerless India TL26 tweeters and faced my first disappointment. Here is what the graphs look like for the two tweeters I bought:
Look at the graphs, see the peaks. The resonant peak is in the region of 1 KHz. The published documentation, both on diyaudiocart and theaudiocrafts, says that the Fs is 596 Hz. This huge discrepancy is not just because of piece-to-piece inconsistencies -- the published data is plain wrong.
This discovery makes this tweeter as boring as the old Peerless India SR10DT. The primary reason I had found this tweeter interesting was because I thought I'll get the ability to cross it over quite low, maybe at 1.6KHz to 2KHz. Now I don't think I should cross over less than 2.5KHz, 3rd order or 4th order. I could have done this with the venerable SR10DT too. Luckily, I've bought this tweeter to go with the 6.5" Peerless India midbass driver whose claimed SPL curve is quite smooth and extended. If my measurements of the actual SPL (see Step 5 above) even roughly matches the published data, then I'll be able to use the TL26 even with a crossover frequency of 2.5KHz.
I don't suspect any inaccuracy in my ARTA LIMP measurement setup. I tested it with other drivers and with passive components (capacitors, resistors), and the impedance is spot on.
Now we know why I don't like using Peerless India drivers. This sort of inaccuracy is not worth the time and trouble. And I won't buy any more TL26 in future -- better get some affordable Vifa or SB Acoustics tweeters instead.
Finished doing the T/S parameter measurement of the 5.25" drivers:
I'll start enclosure modeling for the MTM now, which will use these drivers. And I need to finish T/S parameter measurement of the 6.5" Peerless India driver which I'll use for the TM design.
This is the first project I'm using ARTA for. I like it.
This is precisely the reason i dont go for Peerless drivers anymore. You can reduce the resonant peak by using a series notch at the resonant frequency. It works quite good with this tweeter. You can design one and measure again with the resonant trap. You will be surprised with the result.
Even though this drivers uses some internal damping material, ferro fluid and a cavity to reduce the resonance, it actually does not help in reducing them. Hence it will be wiser to used the resonant trap no matter what the data sheet mentions.
Why should I put any impedance compensation network to cut the resonant peak? It won't get in the way of my using the tweeter.
The resonant peak doesn't bother me. The fact that the peak is at 1KHz where the published spec says it is at 600Hz bothers me.
Agreed, the peak is way below the design frequency of your 2.5KHz. The compensation will usually be required if you are using a first order filter. A flattened impedance is good for maintaining the crossover frequency more accurately than a peaking one. If you are designing the crossover frequency at 2.5KHz then you should then consider the measured impedance @ 2.5KHz if you are not using the resonant trap and not the published result. Also in your case, both the tweeters have a serious mismatch in its measured impedance which you will need to factor in.
Most drivers do not change their behaviour at all by run-in --- this is yet another audiophile myth. Some very specialised subset of drivers (those which are pushing the boundaries of the materials they use for their cones and surrounds) do change their sound, but I doubt that even those change their T/S parameters. Remember, at this point, I'm trying to measure some very basic parameters of these drivers, nothing very subtle or complex. Burn-ins don't impact these parameters at all. I've never seen any changes with burn-in for all the cases where I've tried this.
Tried plugging in the T/S parameters for the M13NH drivers into Unibox for some box modelling. The sensitivity is unusually high -- I'm hoping to get some loud, clear sound from these little boxes. I set the configuration to "2 drivers in parallel" and played around with box size etc for a bit.
With sealed boxes, the roll-off is way too early, as one can expect from this sort of Qts and Fs. When doing sealed boxes, I like a box Q less than 0.7, more like 0.5-0.6. If I try this, I get rolloffs with F3 in the 200Hz region and F10 in the 80-100Hz region. Not worth it. So I tried vented.
This is what I am getting with a 15L box:
The F3 point (3dB down from the flat portion) is hardly an indicator of what you hear in-room. Most designers believe that anywhere between F6 (6dB down) to F10 (10dB down) is what you'll get as the realistic in-room response, because the room will strengthen the low frequencies with its own resonance etc. So, as this graph shows, the F10 here is in the 40-50Hz region. Not bad at all for a small enclosure with such small drivers. A 15-litre box is really quite a small stand-mount.
WIth a 12L box, I get a slightly smoother shaped curve:
I think I'll play around a bit more with the numbers and then finalise on something in this broad region. I'll use one port of 1.5" diameter (you can see 3.75cm port diameter here).
With sealed boxes, the roll-off is way too early, as one can expect from this sort of Qts and Fs. When doing sealed boxes, I like a box Q less than 0.7, more like 0.5-0.6. If I try this, I get rolloffs with F3 in the 200Hz region and F10 in the 80-100Hz region. Not worth it. So I tried vented.
This is what I am getting with a 15L box:
The F3 point (3dB down from the flat portion) is hardly an indicator of what you hear in-room. Most designers believe that anywhere between F6 (6dB down) to F10 (10dB down) is what you'll get as the realistic in-room response, because the room will strengthen the low frequencies with its own resonance etc. So, as this graph shows, the F10 here is in the 40-50Hz region. Not bad at all for a small enclosure with such small drivers. A 15-litre box is really quite a small stand-mount.
WIth a 12L box, I get a slightly smoother shaped curve:
I think I'll play around a bit more with the numbers and then finalise on something in this broad region. I'll use one port of 1.5" diameter (you can see 3.75cm port diameter here).
The Unibox model for the low frequency behaviour is in this spreadsheet. If you're familiar with Unibox, you can go straight to the spreadsheet and play with details. If you're not familiar, remember to use a version of MS Excel of MS Office 2000 or later and remember to switch off all restrictions on macro execution within MS Excel to allow this sheet's macros to work.
You'll see many aspects of this design which are less than perfect, but then no designs are. The aim here is to get the best output from (i) this set of midbass drivers, and (ii) listening to music, not "Home Theater", at modest volumes in a small Bombay apartment's rooms.
Once the Unibox model tells us that we can operate between 12L and 15L to get reasonable results, I start working on a second spreadsheet to help me calculate internal volume of my enclosure, starting with my dimensions and details. This lets me arrive at a desired internal volume and connect it with external dimensions, number of braces, etc.
This second spreadsheet is something I created in 2006 when I designed the first Asawari. It has data and assumptions which fit my approach to enclosure design. For instance, I use 25mm MDF sheets for the outer shell of the enclosure and 20mm (well, 18mm) plywood for braces. I use two sheets of MDF for the front baffle, to make it rock steady (these are the tricks which give my speakers their phenomenal soundstage accuracy). All those assumptions are there in this spreadsheet's calculations.
From this second spreadsheet, I arrive at a design whose outer dimensions are 48cm tall, 22.5cm wide and 27.5cm deep. These dimensions assume, for instance, that I don't have veneer on this enclosure -- I just have laminate. With these dimensions, I arrive at an internal volume of 13.7 litres. This will give me something like 13 - 13.5L of internal volume when I add the crossover components into the box. So, I'm well within the ballpark 12-15L I wanted to be in. I'll now use XFig to make the box design drawings the way a draughtsman does, and hand them to my carpenter.
You'll see many aspects of this design which are less than perfect, but then no designs are. The aim here is to get the best output from (i) this set of midbass drivers, and (ii) listening to music, not "Home Theater", at modest volumes in a small Bombay apartment's rooms.
Once the Unibox model tells us that we can operate between 12L and 15L to get reasonable results, I start working on a second spreadsheet to help me calculate internal volume of my enclosure, starting with my dimensions and details. This lets me arrive at a desired internal volume and connect it with external dimensions, number of braces, etc.
This second spreadsheet is something I created in 2006 when I designed the first Asawari. It has data and assumptions which fit my approach to enclosure design. For instance, I use 25mm MDF sheets for the outer shell of the enclosure and 20mm (well, 18mm) plywood for braces. I use two sheets of MDF for the front baffle, to make it rock steady (these are the tricks which give my speakers their phenomenal soundstage accuracy). All those assumptions are there in this spreadsheet's calculations.
From this second spreadsheet, I arrive at a design whose outer dimensions are 48cm tall, 22.5cm wide and 27.5cm deep. These dimensions assume, for instance, that I don't have veneer on this enclosure -- I just have laminate. With these dimensions, I arrive at an internal volume of 13.7 litres. This will give me something like 13 - 13.5L of internal volume when I add the crossover components into the box. So, I'm well within the ballpark 12-15L I wanted to be in. I'll now use XFig to make the box design drawings the way a draughtsman does, and hand them to my carpenter.
I will never use a 1st order crossover in any situation other than situations where a crossover is not necessary at all. I'm referring to 1st order acoustic, not 1st order electrical.
Let me explain.
A first order low-pass will apply to a midrange or midbass driver only when the upper end of the driver is so extended and so smooth that I don't need to suppress the upper regions at all. What sort of driver will this be? The only answer is a very good full-range driver, which I can play without a crossover. Remember that with a 1st order low-pass, I'm allowing my midrange or midbass driver to be clearly audible even four octaves beyond Fc, because at 4 octaves out, it's down by only 24dB. In fact, it'll be audible right till the end of its natural frequency curve.
A first-order high-pass will apply to a tweeter when it has the construction to handle low frequencies thrown at it, and still not blow. A first-order high-pass hardly cuts the low frequencies, and an average tweeter will be damaged if it's played at anywhere remotely near its rated power with such a filter. (That's why power rating measurements on tweeters are done only after applying a 3rd order or 4th order filter first, if you read the fine print. It protects them from blow-outs.) Therefore, what sort of tweeter will be able to handle a 1st order high-pass and still survive? My guess is that some of those (very expensive) larger ribbons, which can handle frequencies down to 500Hz at full power, will survive with a 1st order at 3KHz or so. In essence, here too, we are allowing the tweeter to be audible almost till the bottom of its frequency range, just tapering the bottom end down a bit.
Therefore, in my eyes, a 1st order filter is not a "crossover", because there's no frequency beyond the Fc where the sound actually "crosses over". Both drivers practically remain audible through their entire audible frequency ranges, playing in parallel, and their responses are just being shaped by the crossover, very gently. These need very special drivers to sound good. I accept that a lot of paper cone or poly cone midbass drivers can be driven with just a 1st order low-pass (LP), but then those drivers can also be driven without a crossover. (Hence my initial remark -- I'll use a 1st order where I actually don't need a crossover at all.)
If I do a 1st order crossover, I'll be restricted to only these very special drivers, or else I'll have to live with the stresses, strains, resonances, and distortions which come when a driver is pushed beyond its "comfort zone". I choose to increase my choice of drivers, by using "normal" drivers, and I choose to use my crossover to shape their responses such that they are restricted to their "comfort zones" and can perform at their best. That's what crossovers are for, IMHO. I can imagine using a 2nd order LP for a very, very well-behaved mid-bass driver. And in the case of tweeters, I never use anything less than a 3rd order slope, because I want to protect tweeters against damage.
I want to do good engineering first, and esoteric audio second. I find no merit, for instance, in a design which sounds wonderful, but where I can only play the music upto 2 Watts of power, where an accidental turning up of the volume will blow the tweeter. I also don't see any point in using a 7" midbass driver in such a way that it is audible even at 8KHz, so that is beaming sound (all drivers beam at frequencies too high for their diameter) merges with the widely dispersing sound from the tweeter to create weird dispersion profiles.
I know that 1st order crossovers have a huge fan following. To each his own, I guess.
In the graphs, the mismatch is very visible. In actual use, I am not sure this mismatch will make an audible difference. We need to see how much the impedances differ in the range in which the tweeter will be used, and that will be probably above 2.5KHz. I think in that range, the difference will be small enough to be ok.
However, the impedance of one tweeter is showing some irregularities, the curve is not as smooth as it should be. I'll wait to take the SPL measurement of that tweeter, and if I see corresponding differences in the SPL curves, then the tweeter is defective and I'll need to buy another one. One more good reason to not buy Peerless India drivers.
Last edited:
shafic
Well-Known Member
Hi,
Interesting explanation. I have a question hope you can help. I have a set of 7" mid bass drivers and tweeters. One of the tweeters is giving the reading g of 4 ohms and other one? is 8 ohms. Mid bass drivers are both 8ohms. I am planning to make book shelves using these.
Will these different reading of the tweeters cause any problem to the amp? And how will it affect the sound quality?
Regards.
Interesting explanation. I have a question hope you can help. I have a set of 7" mid bass drivers and tweeters. One of the tweeters is giving the reading g of 4 ohms and other one? is 8 ohms. Mid bass drivers are both 8ohms. I am planning to make book shelves using these.
Will these different reading of the tweeters cause any problem to the amp? And how will it affect the sound quality?
Regards.
How did you measure this? In real life no impedance measurement of a driver gives exactly 4 or 8 Ohms.
Are you going to build speakers using these drivers? If yes, you will have to design and build different crossovers for the two channels. They won't affect the amp, though.
It is probably with the multimeter
Order your Rega Turntables & Amplifiers from HiFiMART.com - India's reputed online dealer.
Similar threads
