Building a 3-way TL speaker

[kapvin]

1- To keep measurement artifact to bare minimum I did all these measurement after 11:30PM to 2:30pm to keep the mic and speakers at the same position without moving them even an um. So any artifacts would be because of other reasons beyond my control. I took many measurements ( more than 4) for each change so that it can be averaged out.

2-the wires are inside the boxes where the temperature changes will be minimum due to the isolation from the ambient.

3-ime most drivers integrate quite well in the horizontal axis SPL wise, that's why you don't get too much variation in level when you move in the horizontal axis. But most drivers where there are timing issues the SPL changes when you move in the vertical plane. If the level when you sit lower to the listening axis or in in the listening axis or when you stand up changes then it's an indication of timing issues or integration between drivers. The trick is to get them same at all listening positions in both the horizontal and vertical planes.
Time shifts causes phase errors which will increase sharpness in the highs and make the vocals very shrill asking you to reduce the volume to reduce the overlaps between drivers.

I have kept the crossover frequency in multiple of wavelength of each drivers so that any overlaps between them will add in-phase atleast at the crossover frequency.

One major point noted was earlier I needed to toe-in the speaker around 5 Deg to 10 deg to get perfect center imaging in my room. After applying the appropriate delay the toe-in is not required and the speakers are now parallel to the wall. Also the sweet spot has widened and I get the same center image in all positions in my room

Hi Hari,

That's not exactly what i meant.

1. as per your calculation; the velocity of signal propogation through your cable material is 1400m/s (delta distance /delta time). now that's some fraction (0.0004%) of C or Velocity Factor of 0.000004. typically VF for a metallic conductor is 0.4+ in which case cable length should not make a perceptible difference. therefor the measurement night be due to the artifacts, or even due to small shift in speaker / mike. Or is there something I have missed

3. I think i should clarified more about horizontal axis. its easy enough to sum drivers of a speaker up over a longer distance. but the concern is when distance is less (like typical mumbai bedroom? where you get 6-7ft of space. that's when it becomes a challenge. vertical axis is always a challenge

best wishes

Kapil

Hari can you explain how one determines what specific "speed" one wants or should want for power supplies and speaker cables?

Ultimately, it all boils down to conductance which is the reciprocal of resistance. So isn't a simple answer for using a conductor (for the areas aforementioned), the highest conductance aka the lowest resistance? Or am I missing something here?

hi, Keith, my physics is really really rusty, but delta conductance will only impact signal amplitude. Unless the conductor also has some inductance and/or capacitance, in which case there could also be a load and frequency change is amplitude and also a consequent phase shift. What hari is trying to do is reduce VF (speed of conductance) by changing the material. My very rudimentary maths tells me that even if you change VF by 99% (which is a lot) it still wont materially cause a delay of the magnitude written.
But of course, my physics is very old and very basic. So i am trying to learn here too.

 

[Kannan]

Good evening Hari. How are you actually measuring the time difference as every measurement will add latency unless you do a loopback using a Y adapter to compensate for the delay it takes for the signal to reach the mic from one speaker so that the difference is compensated for both speakers for every following measurement.
If it is done purely in the analogue domain, then the delay will vary each time as you have no control over the starting pulse due to time shift.

 

[keith_correa]

hi, Keith, my physics is really really rusty, but delta conductance will only impact signal amplitude. Unless the conductor also has some inductance and/or capacitance, in which case there could also be a load and frequency change is amplitude and also a consequent phase shift. What hari is trying to do is reduce VF (speed of conductance) by changing the material. My very rudimentary maths tells me that even if you change VF by 99% (which is a lot) it still wont materially cause a delay of the magnitude written.
But of course, my physics is very old and very basic. So i am trying to learn here too.

Oh, I understand what he is/was trying to achieve. I think!
My question was not directed towards understanding what he was trying to do but more towards a specific portion of what he said which I had quoted when I asked my question.

 

[sandeepsasi]

Hi @Hari Iyer,

Appreciate your findings! I have a different opinion about the physics behind your observations, especially in the way you relate the drift velocity of electrons in metals to the group delay of frequencies you observe in the step response.

The drift velocity is a constant of proportionality between mobility and electric field, that happen to have the same SI unit of velocity. In fact the drift velocity is quite small (~ 1/10000 m/s for conductors). If you take the reciprocal, you can see that this translates to a delay of 10,000s to travel a distance of 1m. This means that, in mains wiring at home, our lamp will glow only an hour after we switch it on, if the switch is 8m away from the lamp.

To approximately estimate propagation delays of traces in PCBs, we usually use the velocity of light as a reference. The delay is typically 3.33ns/m. This delay increases drastically with parasitic caps and edge rate of the driver or source. This number however sets a theoretical limit on the length of a wire at a certain switching frequency. For example, to send a clock through a 1m wire, the min delay of rising + falling edges = 3.33ns + 3.33ns = 6.66ns. This puts an upper bound of 150MHz for the clock rate. Practically, parasitic R/L/C limit this frequency to a much smaller value, due to delay of a few tens of ns. The pure velocity of propagation of electrons in different materials as well as the drift velocity will only bring about a few ns of change in the actual signal delay. At speeds of 340m/s of sound waves, this will only translate to a few micro-meters of difference in the times of arrival of different frequencies of sound waves, and for sound waves of wavelengths in the range 17m ~ 17mm, this should hardly matter. However, the parasitic R/L/C associated with different materials and wires made out of them can vary the delay by larger magnitudes across various bands of frequencies. The wires act as a filters in their own right, can affect the group delay of audio signals by larger magnitudes and this can have audible and measurable effects on sound.

With regards,
Sandeep Sasi

 

[Hari Iyer]

The drift velocity depends upon the cross-section area, molecular weight, density, temp and may be other factors which i am not aware. What i am aware is of what i have found by my measurement. Again there can be thousands of reasons to tell that my measurements might by faulty and i dont know how to measure etc. etc. and i am okay with that. At the end of the day its all about what you get the outcome as the end result. I am happy that by using aluminium for the mid-range, silver for tweeter and copper for woofer, i am getting a much balanced sound both measurement wise and subjectively and I call it a day.

@sandeepsasi the delay as estimated by you in micrometer is not correct, the delay is in millimeters in my case and below is the working.

Assuming my measurement is accurate - IR delay is 18microsecond - This will translate to a wavelength of 6.192mm considering speed of sound as 344m/sec @ 25 deg temperature.

While adding the delay circuit with R,L IR delay measured was 12microsecond. This will translate to a wavelength of 4.128mm considering speed of sound as 344m/sec @ 25 deg temperature.

If by using different type conductors if my IR delay is reduced to say 1microsec. This will translate to a wavelength of 0.344 mm considering speed of sound as 344m/sec @ 25 deg temperature.

I dont know if there is a delay if you measure at 11:30 PM at night and at 12:15 PM at night. That will be another discovery if everything else remain the same.

Good evening Hari. How are you actually measuring the time difference as every measurement will add latency unless you do a loopback using a Y adapter to compensate for the delay it takes for the signal to reach the mic from one speaker so that the difference is compensated for both speakers for every following measurement.
If it is done purely in the analogue domain, then the delay will vary each time as you have no control over the starting pulse due to time shift.

The REW tool has a loopback function which you can apply to your measurements.

Hi Hari,

That's not exactly what i meant.

1. as per your calculation; the velocity of signal propogation through your cable material is 1400m/s (delta distance /delta time). now that's some fraction (0.0004%) of C or Velocity Factor of 0.000004. typically VF for a metallic conductor is 0.4+ in which case cable length should not make a perceptible difference. therefor the measurement night be due to the artifacts, or even due to small shift in speaker / mike. Or is there something I have missed

3. I think i should clarified more about horizontal axis. its easy enough to sum drivers of a speaker up over a longer distance. but the concern is when distance is less (like typical mumbai bedroom? where you get 6-7ft of space. that's when it becomes a challenge. vertical axis is always a challenge

best wishes

Kapil

1. I did not understand this calculation. I am a noob.
3. I did not understand the challenge in the horozintal axis as in 6-7ft of space there is hardly any moving space. Place the speakers in golden ratio and it should be done imo provided you dont have lobing between drivers due to incorrect driver spacing or crossover frequency consideration.

 

[prem]

With this speaker, with the earlier crossover with cheaper caps, you used the same internal wire for all drivers?

I am not a tech guy but intuitively what you are doing doesn’t sound right to me. To me it appears to be a case of band aid.

Mundorf caps have a bit of smiley equalisation. One doesn’t try to equalise that. If you wanted neutral, you should look at some other caps.

 

[sandeepsasi]

The drift velocity depends upon the cross-section area, molecular weight, density, temp and may be other factors which i am not aware. What i am aware is of what i have found by my measurement. Again there can be thousands of reasons to tell that my measurements might by faulty and i dont know how to measure etc. etc. and i am okay with that. At the end of the day its all about what you get the outcome as the end result. I am happy that by using aluminium for the mid-range, silver for tweeter and copper for woofer, i am getting a much balanced sound both measurement wise and subjectively and I call it a day.

@sandeepsasi the delay as estimated by you in micrometer is not correct, the delay is in millimeters in my case and below is the working.

Assuming my measurement is accurate - IR delay is 18microsecond - This will translate to a wavelength of 6.192mm considering speed of sound as 344m/sec @ 25 deg temperature.

While adding the delay circuit with R,L IR delay measured was 12microsecond. This will translate to a wavelength of 4.128mm considering speed of sound as 344m/sec @ 25 deg temperature.

If by using different type conductors if my IR delay is reduced to say 1microsec. This will translate to a wavelength of 0.344 mm considering speed of sound as 344m/sec @ 25 deg temperature.

I dont know if there is a delay if you measure at 11:30 PM at night and at 12:15 PM at night. That will be another discovery if everything else remain the same.

Hi @Hari Iyer,

Drift velocity does not depend on the cross-sectional area; it depends on the length though. Link,

Drift velocity - Wikipedia

en.wikipedia.org

The drift velocity is in the order of 1/10,000 m/s for conductors, as shown in the example in the link. This speed cannot be related to the speed of a signal through a conductor. The delay I estimated (3.33ns for a conductor of length 1m, at 300MHz signal frequency), using the speed of light as a reference, is often used as a rule of thumb and in back of the envelope calculations, in high speed PCB designs to estimate the delay of a trace.

The signal delay in the order of ns will not translate to a group delay of us, as you have measured. The changes in drift velocity from material to material can only account for signal delays in the order of ns. However, the parasitic R/L/C associated with wires can introduce group delays in the order of us, for various band of frequencies.

The point I was trying to make was that the effect you observed is not due to the drift velocity of the material, but due to the parasitics associated with a different wire that you have used, which in turn will have some contribution from the material itself. The wire that you have found to give best measurement is essentially an inverse filter that has compensated for the imperfections introduced by rest of the system - by delaying the band of frequencies that have arrived faster while using a different wire, if rest of the chain is responsible for this delay or by acting as a linear phase filter so that the entire audio spectrum is equally delayed if the rest of the system introduces no group delay.

With regards,
Sandeep Sasi

 

[Hari Iyer]

Hi @Hari Iyer,

Drift velocity does not depend on the cross-sectional area; it depends on the length though. Link,

Drift velocity - Wikipedia

en.wikipedia.org

The drift velocity is in the order of 1/10,000 m/s for conductors, as shown in the example in the link. This speed cannot be related to the speed of a signal through a conductor. The delay I estimated (3.33ns for a conductor of length 1m, at 300MHz signal frequency), using the speed of light as a reference, is often used as a rule of thumb and in back of the envelope calculations, in high speed PCB designs to estimate the delay of a trace.

The signal delay in the order of ns will not translate to a group delay of us, as you have measured. The changes in drift velocity from material to material can only account for signal delays in the order of ns. However, the parasitic R/L/C associated with wires can introduce group delays in the order of us, for various band of frequencies.

The point I was trying to make was that the effect you observed is not due to the drift velocity of the material, but due to the parasitics associated with a different wire that you have used, which in turn will have some contribution from the material itself. The wire that you have found to give best measurement is essentially an inverse filter that has compensated for the imperfections introduced by rest of the system - by delaying the band of frequencies that have arrived faster while using a different wire, if rest of the chain is responsible for this delay or by acting as a linear phase filter so that the entire audio spectrum is equally delayed if the rest of the system introduces no group delay.

With regards,
Sandeep Sasi

This is what i noted from the same source,

Drift speed
The mobile charged particles within a conductor move constantly in random directions, like the particles of a gas. (More accurately, a Fermi gas.) To create a net flow of charge, the particles must also move together with an average drift rate. Electrons are the charge carriers in most metals and they follow an erratic path, bouncing from atom to atom, but generally drifting in the opposite direction of the electric field. The speed they drift at can be calculated from the equation:

I =nAvQ

where

I is the electric current, n is number of charged particles per unit volume (or charge carrier density, A is the cross-sectional area of the conductor, v is the drift velocity, and Q is the charge on each particle.

I think we are discussing on two different subject, you are talking about the velocity of the electron in a conductor. I am taking about the delay in IR measurement in "Air". For me its immaterial to know what is the absolute velocity in various conductor types and if the difference is 1/10000000 of a meter. Its enough for me to know that when i use different material type (silver, copper, aluminium) there is a difference in the arrival times as measured in the IR. Its not important for me to know if its because of drift velocity, speed of electron or conductivity. I am not doing any research or PhD on this topic nor i am trying to build an ultimate speaker on planet "Earth". Nor i am saying that this speaker will beat any speaker costing 100 times the value. Probably speakers costing 1/10th this value may out-perform them in all the parameters and i am ok with that too.

With this speaker, with the earlier crossover with cheaper caps, you used the same internal wire for all drivers?

No, the wires used earlier were different in guage and type.

I am not a tech guy but intuitively what you are doing doesn’t sound right to me. To me it appears to be a case of band aid.

How does it matter if the application is a band-aid or rocket science if i am able to get the objective measurement and subjective listening to my satisfaction. The end result should be noted and not the means to achieve that. Its a myth that Aluminium is not a good conductor for audio. Sometimes that may be the only solution available. Btw, the 40K+ Beryllium tweeter uses a copper clad Aluminium wire in its voice coil to reduce weight, silver lead wires and gold plated terminals.

Mundorf caps have a bit of smiley equalisation. One doesn’t try to equalise that. If you wanted neutral, you should look at some other caps.

I could find only subtle difference between the EPCOS and the Mundorf capacitors that i used. Both are MKP type. IMO, EPCOS are meant to be used in challenging environmental situations like submerged pumps, Air-conditoner etc and these equipment's are known to generate surge requirements of 50+ Amperes. So for the EPCOS to play audio signal will be child's play and they will be "yawning" when i use them in audio crossovers. i seriously doubt if the Mundorf can be used in an Air-conditioner.

 

[prem]

In layman terms what you are doing is recessing the mids with an aluminium wire, cutting the bloom and balancing mids and bass better and using a silver wire to improve high frequency. If that approach works for you, good.

 

[Hari Iyer]

Imo we are discussing too much on drift velocity, speed of electrons, different conductor type etc etc etc etc.

In the book Gospel of Sri Ramakrishna there is an episode where Swami Vivekananda (SV) was discussing, debating and arguing about the existence of GOD along with Ramakrishna Parahamsa (RP). The debate went on for hours together and finally RP told SV that - "All your argument about existence of GOD will come to an end when to meet him directly". SV asked RP - Have you seen/met GOD. RP replied - Yes, i have seen/met him and talk to him as i talk to you and he listens". RP further told SV - A pitcher when filled with water will make noise when its empty or getting filled and all noise will stop when the pitcher is full. Later when SV himself discovered GOD he went into absolute silence / samadhi and all his arguments stopped,

Similarly, all discussions about this speaker will automatically stop once you listen to these speakers and we can continue discussing/ debating on the topic till the end of millenia and we may not progress even an inch. I am planning to ship the speakers around mid of March and FMs in Mumbai can listen till that time after which they will be in Hyderabad and you will need to ask the owner.

 

[prem]

Hari, the Mundorf caps take a lot of time to break in. I suggest give about 200 hrs before you tweak anymore. After 200 hrs, you may want to experiment with some other wires.

 

[mpw]

I agree with prem

 

[Hari Iyer]

Hari, the Mundorf caps take a lot of time to break in. I suggest give about 200 hrs before you tweak anymore. After 200 hrs, you may want to experiment with some other wires.

Ok, it's impossible for me to give that many hours on them before mid march. They must have crossed probably 10 to 15 hrs by now max.

I agree with prem

Have you listened them before and after break-in by any chance?

 

[prem]

Hari, I have had mundorf caps in one of my earlier speakers. And it went through a lot of changes. Hence suggested letting it run in.

But I can understand your limitations and the inability to run it in for 200 hrs

 

[Hari Iyer]

Hari, I have had mundorf caps in one of my earlier speakers. And it went through a lot of changes. Hence suggested letting it run in.

But I can understand your limitations and the inability to run it in for 200 hrs

Is this change only in subjective domain or also in objective measurements?

 

[sandeepsasi]

This is what i noted from the same source,

Drift speed
The mobile charged particles within a conductor move constantly in random directions, like the particles of a gas. (More accurately, a Fermi gas.) To create a net flow of charge, the particles must also move together with an average drift rate. Electrons are the charge carriers in most metals and they follow an erratic path, bouncing from atom to atom, but generally drifting in the opposite direction of the electric field. The speed they drift at can be calculated from the equation:

I =nAvQ

where

I is the electric current, n is number of charged particles per unit volume (or charge carrier density, A is the cross-sectional area of the conductor, v is the drift velocity, and Q is the charge on each particle.

Hi @Hari Iyer,

The current is also I = V/R and R = pl/A. When current, I in the above equation is substituted by VA/pl, the terms A will knock-off from either sides of the equation. And rest of the terms in the equation equates to mobility, u which is a fundamental property of the conductor and V/l = E, the electric field intensity. Thus,

Drift velocity, v = Mobility, u x Electric Field Intensity, E
= u * V / l

Thus, if the same voltage is applied across a thick conductor and a thin conductor of the same length and material, the drift velocity will be the same in both, though the currents are different.

With regards,
Sandeep Sasi

 

[prem]

Hari, mine are all subjective. But changes are huge. So am sure must be affecting the objective measurements too.

 

[Hari Iyer]

Hari, mine are all subjective. But changes are huge. So am sure must be affecting the objective measurements too.

Ok, i am a poor judge of break-in results as my memory is very weak what i heard say one month ago or two month ago. I can make out changes only within a few hours or probably a day or two. For me the break-in is more with my ears than the equipment as after a while my ears gets adjusted to whatever i hear. Though i have not made a deep analysis of changes objectively say between day one and day 60 while i have made no changes in anything other than playing music, i am not sure if the objective measurements are any different. More over its very difficult for me to get the same environmental settings to measure at home between day 1 and day 60. These things are possible only in a lab where nothing other than measurements happen and no one can intrude.

Hi @Hari Iyer,

The current is also I = V/R and R = pl/A. When current, I in the above equation is substituted by VA/pl, the terms A will knock-off from either sides of the equation. And rest of the terms in the equation equates to mobility, u which is a fundamental property of the conductor and V/l = E, the electric field intensity. Thus,

Drift velocity, v = Mobility, u x Electric Field Intensity, E
= u * V / l

Thus, if the same voltage is applied across a thick conductor and a thin conductor of the same length and material, the drift velocity will be the same in both, though the currents are different.

With regards,
Sandeep Sasi

Refer my post # 116. I am not doing a PhD on this topic and do not intend to as it will take me no where. Of-course if someone is willing to fund the project then i can consider. Please send me a PM.

 

[mpw]

Have you listened them before and after break-in by any chance?

There is enough on the net on this.

My Merlin speakers have duelund caps and the manufacturer suggested 500 hours

I see here that good advise is being taken with reluctance and if measuring is all that interests you then calling FM's for listening sessions doesn't make sense to me at least.

 

[Hari Iyer]

Do capacitors really need to break in?

I always thought this was complete BS. I figured if there was any break in it would take seconds, not hours. Speakers, sure. Built enough guitar cabs to know they need the surrounds and cones broken in to sound their best. I finished refreshing some AR3a speakers. New caps, l-pads, etc...I...

audiokarma.org

Capacitor Burn-In/Break-In

The subject of capacitor burn-in generally elicits a fight between so-called subjectivists, who claim that capacitors sound better when broken in, and so-called (typically self-styled) scientific types, who claim the subjectivists are loonies and have no evidence for their claim. For my part, I...

www.diyaudio.com