|
|
Rank: HIFI Veteran
Groups: Member
Joined: 11/25/2008
Posts: 529
Location: UK
|
mat wrote:kengale wrote:mat wrote:kengale wrote: single bar cores have loads of leakage inductance which will have to be allowed for somewhere else in the crossover design.
Hi Ken, Could you elaborate on this comment please. mat. We're only taking about transformers here, not single-winding inductors: when you wind a transformer it has a defined parallel inductance (normally referred to as the "primary inductance") and a series inductance which is caused by the primary and secondary coils not being perfectly coupled with each other (the "leakage inductance"). With gapped E-I cores the leakage inductance can be very low, < 1/100 of the primary inductance or even less. With bar or rod cores it will be substantially more than this, and with air cores so bad that you can't really make transformers this way. If you are using a defined-inductance transformer in a 3-pole crossover such as the HP one in the original LS3/5a the leakage inductance could spoil the crossover performance as the only other reactances required in the crossover are capacitive. Hence the use of an air-gapped transformer core in this design. They have to be air gapped because without the gap the tolerance on the inductance is far too wide (20% or worse): with air gaps typically around 0.1mm - 0.3mm this tolerance can be brought down to 2% or better, as well as making the performance a lot more linear. In the sonar transducer designs I am currently working with we often require defined-inductance transformers AND defined-value series inductors in the same circuit - our inductor suppliers, who are a bit of a whizz at this sort of thing, can often supply us with one component which does both at the same time and has the right turns ratio, right primary inductance AND right-value series inductance, thus saving us a lot of space and sometimes more importantly weight. Hi Ken, being a mechy you'll have to excuse my rather clumsy understanding of electronics! So we have have a bar inductor with a tapped winding to step down the tweeter, say o.5mH acting as a shunt inductor for the tweeter, tapped at 0.3mH to step down the tweeter level. Are you saying that with this type of solid ferrite core bar inductor, the attenuation of the tweeter will vary with power or with frequency, or neither? mat. There will be spurious inductance in series with the transformer which will affect the frequency response in ways you wouldn't expect if you just treated the transformer as being perfect. You would have to model the crossover properly to work out if this will give unexpected roll-offs or resonant peaks or both. Much better to use a gapped core-pair, where the spurious inductance will be low enough to be ignored. And remember that the inductance of a transformer or inductor winding is proportional to the number of turns SQUARED - thus a 0.5mH primary tapped at 0.3mH only gives you an attenuation of the square root of 0.3/0.5 = 0.77, not 0.6. Similarly the impedance ratio (the impedance seen by the input compared with the load impedance) is proportional to the turns ratio squared, not directly to the turns ratio. All this has to be taken into account when calculating the values of any capacitors in the crossover etc.
|
|
Rank: HIFI Guru
Groups: Member
Joined: 11/12/2008
Posts: 312
|
thanks ken,
in practise they are a bit of a moving target, change the tap and it changes the impedance so the cap values need to change. Using non gapped inductors you can step the level up and down up to about 3dB. At attenuation of more than around 1.5dB the frequency response can tend to pivot around a point, say 8k for example, and its quite tricky to get that level to change. This may be due to the leakage inductance you are talking about (?)
mat.
|
|
Rank: HIFI Guru
Groups: Member
Joined: 11/12/2008
Posts: 312
|
On the subject of inductors across bass drivers: Not sure about this one, the usual is to use a large series cap to protect the midrange, you can also use a shunt inductor in front of this to make the filter second order. Typical values of cap would be 40mic+ typical inductance would be 3mH+, the larger the value the lower the frequency they kick in at. You'll also probably need a resistor either across the cap or in series with the inductor to adjust the Q factor of the filter.
mat.
|
|
Rank: HIFI Veteran
Groups: Member
Joined: 11/25/2008
Posts: 529
Location: UK
|
mat wrote:
On the subject of inductors across bass drivers: Not sure about this one, the usual is to use a large series cap to protect the midrange, you can also use a shunt inductor in front of this to make the filter second order. Typical values of cap would be 40mic+ typical inductance would be 3mH+, the larger the value the lower the frequency they kick in at. You'll also probably need a resistor either across the cap or in series with the inductor to adjust the Q factor of the filter.
mat.
A little bit of work with a calculator, rather than taking a "typical value" plucked out of the air, is a much better way of approaching this sort of thing. First look up the dc resistance of each of the speakers. For 1st order, assuming a crossover between bass unit and mid, series inductor for bass unit is R/(2*pi*f) where f is the required crossover frequency. Series capacitor for treble is 1/(2*pi*f*R). Connect the units in phase. For 2nd order, series bass inductor = twice the above value, parallel bass capacitance = half the above value, series treble capacitor = half the above value, parallel treble inductor = twice the above value. Connect the mid in antiphase to the bass. This give you filter Q's of 0.5 and a flat response across the crossover. No additional R's needed anywhere. simples.
|
|
Rank: HIFI Guru
Groups: Member
Joined: 11/12/2008
Posts: 312
|
not so simples.....
the crossover works on the impedance of the drive unit not the resistance. Off the shelf crossover values rarely give the expected results in practise. If you can accurately model it with software using the impedance curves, measured half space spl, simulated room acoustics, and box dimensions, then that can be a good starting point. The best way is to use typical values based on the impedance then use your clio system to find out what's really going on and optimise.
Typical values are plucked from experience not from the air, and are only a starting point for measurement and listening based iteration. Unfortunately neither a calculator or the most sophisticated computer model tells you much about how a speaker might sound in real life.
|
|
Rank: HIFI Veteran
Groups: Member
Joined: 11/25/2008
Posts: 529
Location: UK
|
mat wrote:not so simples.....
the crossover works on the impedance of the drive unit not the resistance. Off the shelf crossover values rarely give the expected results in practise. If you can accurately model it with software using the impedance curves, measured half space spl, simulated room acoustics, and box dimensions, then that can be a good starting point. The best way is to use typical values based on the impedance then use your clio system to find out what's really going on and optimise.
Typical values are plucked from experience not from the air, and are only a starting point for measurement and listening based iteration. Unfortunately neither a calculator or the most sophisticated computer model tells you much about how a speaker might sound in real life. Bass-to mid crossovers are usually OK for using the resistance for a starting point - I usually use full spice models of the drive units. As long as the bass-mid crossover is around the usual 500Hz-ish mark the actual acoustic performance is usually pretty well in accordance with the theoretical. More of a problem using the simple model for mid to treble, because the treble usually has fundamental resonance not that far down from the intended bandwidth, and all sorts of interactions take place, as well as cabinet edge effects etc. Using the simple model for the mid area, 500Hz and 4ohm speaker gives 80uF and 1.3mH for a first order crossover, 40uF and 2.6mH for maximally flat 2nd order. I expect these will turn out to be about right. I agree that deciding what's best in the bass is a real problem because a flat-response half-space is probably not what you need - it would in most circumstances be very bass-heavy in the sort of listening rooms I can afford! though I have in the past designed large active horn-loaded systems for auditorium and open-air use, where half-space is about right.
|
|
Rank: Administrator
Groups: Administration
, Member, Moderator
Joined: 9/19/2008
Posts: 1,158
Location: UK
|
Good points made above which mirrors my experiences from my early Teenage experimentations to recent projects. It seems that some manufacturers rely too much on CAD;they laud it, and don't spend enough time actually listening to their designs properly. I could hear design errors in them and still wonder just how sensitive I am to phase errors as well. I will experiment with this when time permits. One design I worked on started with quite a complicated network; I simplified it, and it still wasn't totally right but it sounded more fun. Guess the drivers used matched the latter's designed network better? From experience I have noted that matching drivers to networks is more than just looking at the physical parameters of the driver, and indeed some that look like they would work well with a first order don't sound right in such an implementation and vice versa. There is far more involved in designing crossovers and getting them to sound right than many, including some experienced designers and makers appreciate, that is what makes it fun 
"Quicquid Nitet Notandum"
|
|
Rank: HIFI God
Groups: Member
Joined: 9/18/2008
Posts: 1,013
Location: Greece
|
darkmatter wrote:It seems that some manufacturers rely too much on CAD My hunch is that too many people, especially so on DIY fora, rely on software without having mastered the underlying concepts. It's like someone pretending to know arithmetic when all they know is how to operate a calculator.
Sumer is icumen in!
|
|
Rank: Administrator
Groups: Administration
, Member, Moderator
Joined: 9/19/2008
Posts: 1,158
Location: UK
|
zonepress wrote:darkmatter wrote:It seems that some manufacturers rely too much on CAD My hunch is that too many people, especially so on DIY fora, rely on software without having mastered the underlying concepts. It's like someone pretending to know arithmetic when all they know is how to operate a calculator. I totally agree having rescued some DIY buddies from disaster, "But the computer says....." 
"Quicquid Nitet Notandum"
|
|
Rank: Administrator
Groups: Administration
, Member, Moderator
Joined: 9/19/2008
Posts: 1,158
Location: UK
|
Very appropriate from page 7 in particular his is a great read and I would loved to have been there. http://www.hificritic.com/scene/news.aspxhttp://www.hificritic.com/downloads/BrightonReport.pdfSimon 
"Quicquid Nitet Notandum"
|
|
Rank: HIFI Guru
Groups: Member
Joined: 9/18/2008
Posts: 479
Location: Melbourne
|
OK, another update: The Supravox may be called a full range, but I could not use it this way (unassisted) as it is significantly lacking in the highs. I put in a newly made test baffle and measured and listened to it. The highs, from about 5K were down about 10 - 15 db, and my baffle gave a 9 db peak centred arount 500 Hz. A broad notch filter and some thick F11 1/2" felt smoothed the reponse, but still left the missing highs. I tried adding the Seas tweeter on the rear out of phase from 3KHz (as recommended on DIY Audio), but this made almost no difference. The tweeter was swapped to be forward firing, in phase, from 6K, and the Supravox was padded down with a 2 Ohm series resistor, but no low pass filter. This now sounds quite well balanced, although, it will require fine tuning and there is the little matter of integration to the bass. The final solution may well be to have an aditional rear firing tweeter as well, as the back wave sounds very dull. I will also try a hard crossover to Seas at 3K A couple of photos (sorry my camera is awfull!)   Cheers David
|
|
Rank: HIFI God
Groups: Member
Joined: 9/18/2008
Posts: 1,013
Location: Greece
|
6 kHz has a wavelength of 5.7 cm. With this kind of distance between the two drivers, and this deep a shelf between them, you are having horrid beaming and cancellation lobes from at least one octave below cross and all the way upwards. How on earth did you ever come by this arrangement? The only way to realise your design goals would be a fairly deep horn for the tweeter and a far lower crossover frequency (if the tweeter can take it).
Sumer is icumen in!
|
|
Rank: HIFI Guru
Groups: Member
Joined: 9/18/2008
Posts: 479
Location: Melbourne
|
zonepress wrote:6 kHz has a wavelength of 5.7 cm. With this kind of distance between the two drivers, and this deep a shelf between them, you are having horrid beaming and cancellation lobes from at least one octave below cross and all the way upwards. How on earth did you ever come by this arrangement? The only way to realise your design goals would be a fairly deep horn for the tweeter and a far lower crossover frequency (if the tweeter can take it). The tweeter is not crossed over at the moment, it is just supporting the "full range" from 6K up. I will be experimenting with a complete crossover at 3K This is just a prototype baffle. I will get the tweeter closer for the final version, and the offset may not be as great. However, I don't hear anything horrid. Put it on the Clio today and with the mic on axis between the mid and tweeter (at 1M), it measures (1/3 octave) +/- 1.5db from 250Hz to 15KHz and -6 by 20K Hz. With the mic on axis with mid, there is a suck out around 6K caused by the baffle offset to the tweeter. I just receive a pair of vintage Saba 4" Alnico paper cone tweeters in the post - they look like new. Am having a listen to one now, quite impressive for its age - very dynamic and sweet, though not as nice as the T35.
|
|
Rank: HIFI Guru
Groups: Member
Joined: 11/12/2008
Posts: 312
|
Hi David,
The project looks like a lot of fun!
Adding some of your felt to the step in between the woofer and tweeter may help bluff some of the Q of your 6k notch.....or it may make it worse, wider and more noticeable. Have a play.
When you say you're not crossing over the tweeter, presumably you are protecting it with a cap, 2.7mic should be in the ballpark for a first order.
The 500Hz peak is a bit of a puzzle, I would be surprised if its the baffle, on small baffles like this edge effect tend to be higher at 2k+. Could be the tweeter resonance if its high Q and there is no ferrofluid damping?
rear tweeter effects can be quite subtle and may only show up once the speaker is properly tuned. Many 'omni's' are directional above 2k, there's that one from cornwall who's name i forget. Top mounted woofer, lightweight NXT inspired TL cabinet, and top mounted forward firing ribbon tweeter.
best of luck!
mat.
|
|
Rank: HIFI Novice
Groups: Member
Joined: 7/12/2009
Posts: 82
Location: the netherlands
|
by making the baffle octagonal, the distance between cone and edge becomes more uniform over the cone radius (ultimately a circular baffle would have a uniform distance), and thus effects will be concentrated on a narrow region of frequencies
it might be better to put the driver in a rectangular baffle, that might look simple, but is one way of distributing the cone-baffle edge distance and thus distributing its effects
|
|
Rank: HIFI Guru
Groups: Member
Joined: 9/18/2008
Posts: 479
Location: Melbourne
|
frank23 wrote:by making the baffle octagonal, the distance between cone and edge becomes more uniform over the cone radius (ultimately a circular baffle would have a uniform distance), and thus effects will be concentrated on a narrow region of frequencies
it might be better to put the driver in a rectangular baffle, that might look simple, but is one way of distributing the cone-baffle edge distance and thus distributing its effects Hi Frank, Thanks for your thoughts. I know that the octagonal shape looks like it is closer to circle than a rectangle, but until you make the rectangle quite large and have the driver very asymetrically placed, it does not make a big difference to the baffle peak, which has to equalized (either actively, or) with a broad notch filter. I have spent hours modelling all this in "Edge", and "Dipole Design". If the rectangle gave a real advantage I would use it. Since the differences are small, I am trying to mke the speaker look smaller and and more asthetically pleasing. Here is my model in edge, including an allowance for the thickness of the baffle.  Here is what edge says the response will be:  Regards David
|
|
Rank: HIFI Guru
Groups: Member
Joined: 9/18/2008
Posts: 479
Location: Melbourne
|
mat wrote:Hi David,
The project looks like a lot of fun!
Adding some of your felt to the step in between the woofer and tweeter may help bluff some of the Q of your 6k notch.....or it may make it worse, wider and more noticeable. Have a play.
When you say you're not crossing over the tweeter, presumably you are protecting it with a cap, 2.7mic should be in the ballpark for a first order.
The 500Hz peak is a bit of a puzzle, I would be surprised if its the baffle, on small baffles like this edge effect tend to be higher at 2k+. Could be the tweeter resonance if its high Q and there is no ferrofluid damping?
rear tweeter effects can be quite subtle and may only show up once the speaker is properly tuned. Many 'omni's' are directional above 2k, there's that one from cornwall who's name i forget. Top mounted woofer, lightweight NXT inspired TL cabinet, and top mounted forward firing ribbon tweeter.
best of luck!
mat. Thanks Mat. There is felt on the baffle step, and just a little around the tweeter - getting the right amount makes a big difference.  The cap on the tweeter is a 4.0uf + 0.68uf, which gives it an F3 of just under 6KHz. The resistor on the mid is now 1.68 Ohms I am listening to the single 1.5 way open baffle as I type this, and I have to say that it exceeds my expectations given the little Denon system driving it. Obviously there is no bass and the balance is a little bright, but the coherence and dynamics are wonderful. Tomorrow, I am taking the test box (currently 170L) to be cut down in size to 130L and braced.  I will then be able to play about with integrating the bass with the OB head unit. Regards David
|
|
Rank: HIFI God
Groups: Member
Joined: 9/18/2008
Posts: 1,013
Location: Greece
|
dcathro wrote:trying to mke the speaker look smaller and and more asthetically pleasing If this is one of your design priorities, it seems to have gone the way of the rest of them, i.e. a bit awry, if you don't mind my saying so. Of all speaker manufacturers I can recall, only Vandersteen seem to follow this path, and they have been careful to hide everything away in a big sock, for obvious reasons.
Sumer is icumen in!
|
|
Rank: HIFI Guru
Groups: Member
Joined: 9/18/2008
Posts: 479
Location: Melbourne
|
zonepress wrote:dcathro wrote:trying to mke the speaker look smaller and and more asthetically pleasing If this is one of your design priorities, it seems to have gone the way of the rest of them, i.e. a bit awry, if you don't mind my saying so. Of all speaker manufacturers I can recall, only Vandersteen seem to follow this path, and they have been careful to hide everything away in a big sock, for obvious reasons. As usual ZP you are just so full of positivity! It will look a lot different in the final version.
|
|
Rank: Administrator
Groups: Administration
, Member, Moderator
Joined: 9/19/2008
Posts: 1,158
Location: UK
|
dcathro wrote:zonepress wrote:dcathro wrote:trying to mke the speaker look smaller and and more asthetically pleasing If this is one of your design priorities, it seems to have gone the way of the rest of them, i.e. a bit awry, if you don't mind my saying so. Of all speaker manufacturers I can recall, only Vandersteen seem to follow this path, and they have been careful to hide everything away in a big sock, for obvious reasons. As usual ZP you are just so full of positivity! It will look a lot different in the final version. Obviously some can build a perfect looking/performing speaker right from the off , like you David I too enjoy the prototyping journey Simon
"Quicquid Nitet Notandum"
|
|
Rank: HIFI God
Groups: Member
Joined: 9/18/2008
Posts: 1,013
Location: Greece
|
What's the CSD like on the Clio?
Sumer is icumen in!
|
|
Rank: HIFI Guru
Groups: Member
Joined: 9/18/2008
Posts: 479
Location: Melbourne
|
darkmatter wrote:Obviously some can build a perfect looking/performing speaker right from the off , like you David I too enjoy the prototyping journey Simon Thanks Simon, Can you offer any thoughs or suggestions?
|
|
Rank: HIFI Guru
Groups: Member
Joined: 2/2/2009
Posts: 267
Location: uk
|
Really enjoying this thread. For layout dont think you have to put the tweeter above the mid. Put them side by side and "-hand" the speakers. For example.... http://dahlquistspeakers.com/ for DQ10s. A couple of pages about cabinet diffraction http://stereophile.com/reference/704cutting/
|
|
Rank: HIFI Guru
Groups: Member
Joined: 11/12/2008
Posts: 312
|
dcathro wrote:frank23 wrote:by making the baffle octagonal, the distance between cone and edge becomes more uniform over the cone radius (ultimately a circular baffle would have a uniform distance), and thus effects will be concentrated on a narrow region of frequencies
it might be better to put the driver in a rectangular baffle, that might look simple, but is one way of distributing the cone-baffle edge distance and thus distributing its effects Hi Frank, Thanks for your thoughts. I know that the octagonal shape looks like it is closer to circle than a rectangle, but until you make the rectangle quite large and have the driver very asymetrically placed, it does not make a big difference to the baffle peak, which has to equalized (either actively, or) with a broad notch filter. I have spent hours modelling all this in "Edge", and "Dipole Design". If the rectangle gave a real advantage I would use it. Since the differences are small, I am trying to mke the speaker look smaller and and more asthetically pleasing. Here is my model in edge, including an allowance for the thickness of the baffle. Here is what edge says the response will be: Regards David Hi David, So the 500Hz peak is in the modelled output and the measured in the real thing? Unless I am being particularly slow this morning I'm not sure why this is! On small box, 7" wide baffles, edge effects come in at around 2k+, so to be at 500Hz, the baffle would have to be 28" wide..... So it must be due to the open baffle as below 500Hz it starts to cancel. (If you uncheck the open baffle box in your model does it still appear?) Does anyone know why an open baffle exhibits this peaking behaviour at a frequency lower than the one predicted by the wavelength to the edge? mat.
|
|
Rank: HIFI Veteran
Groups: Member
Joined: 11/25/2008
Posts: 529
Location: UK
|
mat wrote:dcathro wrote:frank23 wrote:by making the baffle octagonal, the distance between cone and edge becomes more uniform over the cone radius (ultimately a circular baffle would have a uniform distance), and thus effects will be concentrated on a narrow region of frequencies
it might be better to put the driver in a rectangular baffle, that might look simple, but is one way of distributing the cone-baffle edge distance and thus distributing its effects Hi Frank, Thanks for your thoughts. I know that the octagonal shape looks like it is closer to circle than a rectangle, but until you make the rectangle quite large and have the driver very asymetrically placed, it does not make a big difference to the baffle peak, which has to equalized (either actively, or) with a broad notch filter. I have spent hours modelling all this in "Edge", and "Dipole Design". If the rectangle gave a real advantage I would use it. Since the differences are small, I am trying to mke the speaker look smaller and and more asthetically pleasing. Here is my model in edge, including an allowance for the thickness of the baffle. Here is what edge says the response will be: Regards David Hi David, So the 500Hz peak is in the modelled output and the measured in the real thing? Unless I am being particularly slow this morning I'm not sure why this is! On small box, 7" wide baffles, edge effects come in at around 2k+, so to be at 500Hz, the baffle would have to be 28" wide..... So it must be due to the open baffle as below 500Hz it starts to cancel. (If you uncheck the open baffle box in your model does it still appear?) Does anyone know why an open baffle exhibits this peaking behaviour at a frequency lower than the one predicted by the wavelength to the edge? mat. It's not a diffraction effect your'e looking for - it's when the path length from the cone on the front to the edge, and then from the edge to the cone on the back equals 1/2 a wavelength. Because the rear radiation is reverse phase this then gives a summation peak. If you assume speed of sound = 340m/sec then for 500Hz wavelength = 340/500m = 680mm, 1/2 wavelength = 340mm, which is the total path length front and back, edge-to-cone = 340/2mm = 170mm: pretty near to 7" ! Basically this baffle is simply too small. And the drive unit is a bit big. In a large rectangular baffle you would be able to spread the peak a bit and drop it to a sensible frequency.
|
|
|
Guest |
