(For Mechano24 design with Scan Speak drivers click here LINK)
(For Mechano323 3-way design with AMT driver click here LINK)
(For Mechano23 design with Scan Speak and SB Acoustics drivers click here LINK)
Not so long ago, I made a small 2-way speaker design with the following goals: f6 around 40Hz with 8 liters BR enclosure, generally balanced on-axis response and close to linear frequency power response.
Experience gathered with that design led me to a conclusion that maybe it could be done better and cheaper. And here is the new version, codemame "Mechano22".
It is based on relatively inexpensive set from Dayton: ND25FW-4 tweeter with waveguide and DS135-8 midwoofer.
Box dimensions: HxWxD: 290x174x260mm, made of 18mm birch plywood. The Box is filled with polyester fiber.
Both drivers are placed in the middle of the front panel, with centers 75mm (tweeter ) and 195mm (woofer ) from the top edge.
The woofer has a flat basket edge, which, according to the manufacturer's note, allows application without flush so it wasn't implemented. On the other hand, woofer mounting hole was chamfered on the inside.
Pic1. System on-axis and power response characteristics in comparison with target lines. For power response calculation +-180deg measurements were taken in H and V planes (10deg step):
Pic2. frequency response for DS135-8 (orange) and ND25FW-4 (gray):
Pic3. Crossover schematic:
Pic4. VituixCad, schematic and 6-pack for double check:
Pic5. Filters assembled on a scrap of plywood (for 1 unit):
Pic6. Units assembled (need sanding and waxing ):
Pic7. Measured on-axis dBspl (both units, 5dB/div). f6=40Hz if I did not messed up the measurements.
Pic8. Measured system impedance (both units).Nominal impedance is 6R (considering 5R minimum at 200Hz):
List of components used (for one unit):
Dayton Audio DMPC4.3 / 4.3 uF / 5% / 250 V / Polypropylene capacitor MKP
Capacitor Jantzen Audio CrossCap 0.47uF / 400VDC / 5% / MKP /
Dayton Audio DMPC8.2 / 8.2 uF / 5% / 250 V / MKP 2 pcs.
Capacitor Jantzen Audio Cross-Cap 9.1uF / 400VDC / 5% / MKP
Dayton Audio DMPC-2.7 / 2.7 uF / 5% / 250 V / MKP polypropylene capacitor /
Ty-Ohm ceramic resistor 22ohm / 22R0 / 20W 5% SQP
Ty-Ohm ceramic resistor 4.7ohm / 4R7 / 5W 5% SQP
Ty-Ohm ceramic resistor 100ohm / 100R0 / 5W 5% SQP
Ty-Ohm 12ohm / 12R0 / 5W 5% SQP
Jantzen Audio core coil 0.68mH / cylindrical / 0.162ohm / dr.1mm Fe 0.021kg / dia.20, length 40mm /
Jantzen Audio core coil 0.82mH / reel / 0.22ohm / dr 0.8mm Fe 0.056kg / dia.24 length 33mm /
Jantzen Audio Air coil 3.7mH / 3.25ohm / dr 0.5mm / dia.33, length 15mm
Jantzen Audio Air coil 2mH / 3.21ohm / dr 0.4mm / dia.31, length 8mm
Jantzen Audio air coil 0.12mH / 0.26ohm / dr 0.7mm / dia.27 length 8mm
Attached files: VituixCAD *.vxp and XMachina *.xmp with spinorama.
(For Mechano323 3-way design with AMT driver click here LINK)
(For Mechano23 design with Scan Speak and SB Acoustics drivers click here LINK)
Not so long ago, I made a small 2-way speaker design with the following goals: f6 around 40Hz with 8 liters BR enclosure, generally balanced on-axis response and close to linear frequency power response.
Experience gathered with that design led me to a conclusion that maybe it could be done better and cheaper. And here is the new version, codemame "Mechano22".
It is based on relatively inexpensive set from Dayton: ND25FW-4 tweeter with waveguide and DS135-8 midwoofer.
Box dimensions: HxWxD: 290x174x260mm, made of 18mm birch plywood. The Box is filled with polyester fiber.
Both drivers are placed in the middle of the front panel, with centers 75mm (tweeter ) and 195mm (woofer ) from the top edge.
The woofer has a flat basket edge, which, according to the manufacturer's note, allows application without flush so it wasn't implemented. On the other hand, woofer mounting hole was chamfered on the inside.
Pic1. System on-axis and power response characteristics in comparison with target lines. For power response calculation +-180deg measurements were taken in H and V planes (10deg step):
Pic2. frequency response for DS135-8 (orange) and ND25FW-4 (gray):
Pic3. Crossover schematic:
Pic4. VituixCad, schematic and 6-pack for double check:
Pic5. Filters assembled on a scrap of plywood (for 1 unit):
Pic6. Units assembled (need sanding and waxing ):
Pic7. Measured on-axis dBspl (both units, 5dB/div). f6=40Hz if I did not messed up the measurements.
Pic8. Measured system impedance (both units).Nominal impedance is 6R (considering 5R minimum at 200Hz):
List of components used (for one unit):
Dayton Audio DMPC4.3 / 4.3 uF / 5% / 250 V / Polypropylene capacitor MKP
Capacitor Jantzen Audio CrossCap 0.47uF / 400VDC / 5% / MKP /
Dayton Audio DMPC8.2 / 8.2 uF / 5% / 250 V / MKP 2 pcs.
Capacitor Jantzen Audio Cross-Cap 9.1uF / 400VDC / 5% / MKP
Dayton Audio DMPC-2.7 / 2.7 uF / 5% / 250 V / MKP polypropylene capacitor /
Ty-Ohm ceramic resistor 22ohm / 22R0 / 20W 5% SQP
Ty-Ohm ceramic resistor 4.7ohm / 4R7 / 5W 5% SQP
Ty-Ohm ceramic resistor 100ohm / 100R0 / 5W 5% SQP
Ty-Ohm 12ohm / 12R0 / 5W 5% SQP
Jantzen Audio core coil 0.68mH / cylindrical / 0.162ohm / dr.1mm Fe 0.021kg / dia.20, length 40mm /
Jantzen Audio core coil 0.82mH / reel / 0.22ohm / dr 0.8mm Fe 0.056kg / dia.24 length 33mm /
Jantzen Audio Air coil 3.7mH / 3.25ohm / dr 0.5mm / dia.33, length 15mm
Jantzen Audio Air coil 2mH / 3.21ohm / dr 0.4mm / dia.31, length 8mm
Jantzen Audio air coil 0.12mH / 0.26ohm / dr 0.7mm / dia.27 length 8mm
Attached files: VituixCAD *.vxp and XMachina *.xmp with spinorama.
Last edited:
You should remove the 100 ohm resistor. Values that high in this placement are not much different than infinite, and it's directly across the amplifier with with shunt coil as an LR circuit. This means it has potential to run hot, even at 100 ohms.
The 22 ohm shunting in the woofer is also not a good idea. I see that you at least used a high power model in that spot, but a larger value would be better to reduce current through it, say 40-50 ohms.
The 22 ohm shunting in the woofer is also not a good idea. I see that you at least used a high power model in that spot, but a larger value would be better to reduce current through it, say 40-50 ohms.
Thanks for your comment. In practice, playing music at levels that sound loud does not even make r1 resistor warm. But let's assume feeding up the speaker with constant power 50W (next to damage) then the voltage on 22R resistor (r1) would be 20V RMS so there would be a dozen of Watts to dispatch (some power headroom available). Doing the same estimation for r3 (100R) there would be a few Watts to dispatch (again some power headroom). And this is a theoretical situation that's not very likely to happen.
Last edited:
They sound transparent and balanced, very clean midrange even at high volume levels but then some bass compression occurs (will try wider BR port). Mid bass seems to be a half step forward, something that not very evident on system frequency response. Maybe I'll experiment a bit with room placement .
Even so, bypassing a high-pass with a resistor is just partially defeating the high-pass to where it could allow the tweeter to play music from the bass range albeit a bit attenuated.
And it's about the resistance load on the amplifier, not the drive of voltage or loudness level. You have LR circuits across the amplifier in both networks. The low pass coil and 22 ohm make a loop on the amp itself, as does the 100 ohm and the tweeter shunt coil.
This is not good practice as far as circuits go. The amp will drive those loads, and the resistors could get hot.
Tweeter filter attenuation in the bass region is almost 60dB which seems to be quite ok.
Resistors shouldn't get overheated if the power dispatched is within working range of the components.
I used the following estimation of Power dispatched by r1 r3.
Power handling for DS135-8 Pmax=50W RMS
Nominal Impedance Znom=8R
Umax=(Pmax*Znom)^0.5=20V RMS
For the purpose of worst case assume coils are shunt. Then r1 and r3 are directly connected to the voltage source (i.e. amp), so
Pmax[r1]=Umax^2/r1=18W (r1 working range is up to 20W)
Pmax[r3]=Umax^2/r3=4W (r3 working range is up to 5W)
And let me point out again, these calculations are for artificial case that is unlikely to happen in normal use.
I'm speaking from experience, not from calculations. Place an Lpad before a xover, and same diff. It gets hot! The coil is seen as a virtual short. The same issue will occur with full bass bandwidth entering the LR circuit and heat them under direct drive from the amp.
Whether the attenuation looks to be good or not, bypassing a cap with a resistor allows the possibility for a DC pulse to enter a tweeter.
I'm not buying your example. They are weak points in your topology and should be changed.
Whether the attenuation looks to be good or not, bypassing a cap with a resistor allows the possibility for a DC pulse to enter a tweeter.
I'm not buying your example. They are weak points in your topology and should be changed.
Of course, there is always a way to adjust for improper design methodologies and improve or retain the sound achieved. His graphs look well thought out and measured.
Allen, he looks accomplished enough to get this far, and I feel he is a capable person to do his results one better. I know I would not leave them like that were it me and I was informed and now know better.
Allen, he looks accomplished enough to get this far, and I feel he is a capable person to do his results one better. I know I would not leave them like that were it me and I was informed and now know better.
I'm not devoted to any particular approach in diyaudio, I'm still experimenting trying different paths. This time the path was "I'll try to do something flat like a ruler". Removing r3 without changing anything else would have no audible effects I think. It creates c.a. 1 dB difference on the characteristics. But these changes are at the region which already had some variations, so the response would be +-2.5dB from the target line, not +-1.25dB like the original with r3 present and the "ruler flatness" goal would be more questionable. Then instead of removing r3 I thought of adding 1.2u in series with r3 and this is resolving all problems related to r3 and making no visible changes on the system response characteristics. But that would be an extra component in a crossover that is already a bit complicated. Considering that power estimations and stop band attenuation of the filter seemed to be ok, I didn't see any reason for searching further modifications and eventually kept the original version for implementation.
Currently, seeing that issue of resistor application can excite discussions , I will try to find some time and follow AllenB suggestion of taking out r3 and looking for a solution changing other component values.
First I have to correct myself. Everywhere I referred to r3 in previous posts, I meant r4 actually, i.e. the 100R resistor in the HP filter. Unfortunately post editing is disabled.
Pic1. Changes to the HP filter.
The effects are shown in the Pic2.
Pic2. system characteristics: initial (sky blue), removal r4 (yellow), c5 and r3 modification (red)
The effect is very good, the linearity of the characteristics is kept and one component less is used.
In such situations, when a certain good solution is known but the algorithm not finding it, I always think about the causes.
It seems to me that I know the answer for this case, if someone is interested, I will present it briefly.
First, the changes to the system characteristics between the initial and the modified version are negligible and do not affect the selection.
The answer is probably in the component application strategy. The resistance value r3 after C5 modification should be 11 ohms. The machine was configured in such a way that it should only use the values from the E12 series and this value is not there. On the other hand this value can be easily achieved by connecting two E12 components for example 10+1 in series or 2x22 in paralel. But that would be two components instead of one, so from the point of view of the machine, there is no profit anymore. And there is one more detail that makes this new solution seem even slightly worse for the machine than the initial one.
During the synthesis of circuits, the cost of the components is added up. If the component costs are not loaded then the cost is estimated in terms of their value but only for coils and capacitors. The higher the inductance / capacitance value, the higher the estimated cost. In the case of resistors, the cost is per piece, regardless of the value.
Returning to the analyzed case, the cost of the resistors has not changed because two resistors 100R and 12R have been replaced with two other resistors, e.g. 10R and 1R.
But the capacity of the C5 has increased, so the total cost for the modified version is higher, so for the machine it less attractive than the initial version. But it all depends on the machine configuration.
As for modifying the tweeter filter, I removed r4, changed the c5 value to 3.3u and readjusted r3 resistance to best match the target line. The best value for r3 turned out to be 11R.
Pic1. Changes to the HP filter.
The effects are shown in the Pic2.
Pic2. system characteristics: initial (sky blue), removal r4 (yellow), c5 and r3 modification (red)
The effect is very good, the linearity of the characteristics is kept and one component less is used.
In such situations, when a certain good solution is known but the algorithm not finding it, I always think about the causes.
It seems to me that I know the answer for this case, if someone is interested, I will present it briefly.
First, the changes to the system characteristics between the initial and the modified version are negligible and do not affect the selection.
The answer is probably in the component application strategy. The resistance value r3 after C5 modification should be 11 ohms. The machine was configured in such a way that it should only use the values from the E12 series and this value is not there. On the other hand this value can be easily achieved by connecting two E12 components for example 10+1 in series or 2x22 in paralel. But that would be two components instead of one, so from the point of view of the machine, there is no profit anymore. And there is one more detail that makes this new solution seem even slightly worse for the machine than the initial one.
During the synthesis of circuits, the cost of the components is added up. If the component costs are not loaded then the cost is estimated in terms of their value but only for coils and capacitors. The higher the inductance / capacitance value, the higher the estimated cost. In the case of resistors, the cost is per piece, regardless of the value.
Returning to the analyzed case, the cost of the resistors has not changed because two resistors 100R and 12R have been replaced with two other resistors, e.g. 10R and 1R.
But the capacity of the C5 has increased, so the total cost for the modified version is higher, so for the machine it less attractive than the initial version. But it all depends on the machine configuration.
I finally replaced the BR ports, and that made a huge change to bass. The port I have used so far was described as 3.5cm in diameter, but it turned out that 3.5cm is only on the outlet side. On the inner side it was much more narrow, I had to cut it's length to 10cm to get 45Hz. Now the port is Monacor MBR-35 adjusted to 14cm length. Same tuning as before but works much better with higher volume levels.
This parallel LC circuit performs a midrange equalization. Removing it causes 700Hz bump on the characteristics, and degrades the characteristics in the crossover region.
There is a Pats Express C-note design LINK wit the same drivers, but in a different housing and a simpler crossover. Some problem with this design is a 700Hz bump on the characteristics, which is suspected to color female vocals (despite this, C-note has good opinions and is recommended). M22 is free of this drawback.
Note that a coil with appropriate high resistance has been selected for this parallel LC circuit. High resistance provides suitable Q factor (and makes the component cheap by the way).
Comparison of electrical LF filter characteristics (blue) and shifted spl measurement of the midwoofer in my housing (purple).
Comparison of the LF filter characteristics before (blue) and after (green) LC removal.
Effects of LC removal on sys spl and power response characteristics (~700Hz bump and crossover region degradation).
Before: red and blue. After: green and yellow.
There is a Pats Express C-note design LINK wit the same drivers, but in a different housing and a simpler crossover. Some problem with this design is a 700Hz bump on the characteristics, which is suspected to color female vocals (despite this, C-note has good opinions and is recommended). M22 is free of this drawback.
Note that a coil with appropriate high resistance has been selected for this parallel LC circuit. High resistance provides suitable Q factor (and makes the component cheap by the way).
Comparison of electrical LF filter characteristics (blue) and shifted spl measurement of the midwoofer in my housing (purple).
Comparison of the LF filter characteristics before (blue) and after (green) LC removal.
Effects of LC removal on sys spl and power response characteristics (~700Hz bump and crossover region degradation).
Before: red and blue. After: green and yellow.
Last edited:
@XMechanik could you please attach the xmachina files used? I'd like to see how the ways and design task was done. I struggle with the user interface, so it would be nice to have an example.
Attached XMachina design file for Mechano22. In the file I left only one task and few solutions (the one implemented among them), but of course it was not like that with the first setting it came up with something interesting. I removed several versions of tasks and several dozen solutions to clear the field for action, if anyone want or want to use the file as a starting pattern for another design (or try to synthesize his own Mechano). Feel free to experiment. Getting useful results when starting new design may not be that easy, especially if working on spinorama. Sometimes it's needed to try several different settings before something interesting comes out, but it's also sort of a fun, especially if you like to experiment.