B. you are right, one cannot just put an autoformer in the place of an Lpad. I remodelled the whole mid and high path and use only R’s parallel to the drivers, so not in the “signal path” (whatever that means 
). Now resonances caused by driver inductance and X-over C’s and L’s are damped in that way or by the amp.
I bought two really good new autoformers on eBay from Germany for about 220 US$ actually meant for Altec X-overs. I save some money on the smaller C’s and L’s in the paths. I’m not someone that believes that audio wires or different kind of polycaps make a big difference. I never really heard much improvent, mostly not at all by changing already good audio-caps. But the autoformer really made a difference. To me it was an improvement because I like clean and dynamic sound. The stereoimage seemed less wide but I hear more resolution. Like it is an active monitor now. I can also play louder. Voices never hurt my ears anymore at high levels. Would that be the control of the amp?
). Now resonances caused by driver inductance and X-over C’s and L’s are damped in that way or by the amp. I bought two really good new autoformers on eBay from Germany for about 220 US$ actually meant for Altec X-overs. I save some money on the smaller C’s and L’s in the paths. I’m not someone that believes that audio wires or different kind of polycaps make a big difference. I never really heard much improvent, mostly not at all by changing already good audio-caps. But the autoformer really made a difference. To me it was an improvement because I like clean and dynamic sound. The stereoimage seemed less wide but I hear more resolution. Like it is an active monitor now. I can also play louder. Voices never hurt my ears anymore at high levels. Would that be the control of the amp?
It seems like an autoformer is only going to have one or two possible taps of interest when designing a crossover. Am I wrong in thinking that one could just measure the driver impedance & response with the autoformer plus swamping resistor connected to it and use that as a new “lumped” driver to sim with? Since there’s only one or two taps of interest, one would only have to compare two “lumped” drivers.
When simulating varying the outer capacitor, considering it is shorted across the Voltage source. For varying the inner capacitor you might either remeasure for each change, or find an equivalent circuit using transformers. This may call for a general purpose circuit simulator.
What you have written is mostly about FUD without any proof or verification:
Fear, Uncertainty and Doubt – Wikipedia
It seems that you have not understood autoformers or and maybe have never used a good one. Resistor or not perfect linear devices.
For the given example of a 2nd order filter the best way is to place the coil after the autoformer and the cap before. With the swamping resistor before the AF you can adjust the roll-off. The biggest difference is that an AF has a very small Zout to the CD compared to a Lpad.
Was brought here because I was curious about autotrans in crossover. I'm no expert in this area but ehre are some thoughts after reading this thread.
1. Saturation issue - I think the autotrans should be insert after the 1st cap so that the cap can cutoff the lower freq into the network.
2. Software modelling - if we reduce the autotrans to passive component with inductance and resistance, it is effectively a L-pad with LRseries and LRparallel instead of Rs & Rp.
So a autotrans of 1mH and 1R would be something like LRs = 0.1mH+0.1R and LRp = 0.9mH+0.9R at the 90% tap.
Probably not as straight forward as the above but I think it would be close.
3. Just like the L-pad isn't a potentiometer, the autotrans should have taps that are customised and the taps 'hard-coded' after proper design.
Anyway, these are just some thoughts to add on to what was discussed in this old thread. Would really love to see some simultions curves with actual L+R values of a autotrans
1. Saturation issue - I think the autotrans should be insert after the 1st cap so that the cap can cutoff the lower freq into the network.
2. Software modelling - if we reduce the autotrans to passive component with inductance and resistance, it is effectively a L-pad with LRseries and LRparallel instead of Rs & Rp.
So a autotrans of 1mH and 1R would be something like LRs = 0.1mH+0.1R and LRp = 0.9mH+0.9R at the 90% tap.
Probably not as straight forward as the above but I think it would be close.
3. Just like the L-pad isn't a potentiometer, the autotrans should have taps that are customised and the taps 'hard-coded' after proper design.
Anyway, these are just some thoughts to add on to what was discussed in this old thread. Would really love to see some simultions curves with actual L+R values of a autotrans
Last edited:
Agree
I have modeled autoformers in LTSpice with good success. An autoformer is one inductor with one or more taps. To simulate, you need to have two series inductors that are mutually coupled. Since the output voltage is based on turns ratio, your 1mh example should be .01mH and .81mH.
DCR is tricky since the first 10% of the windings are probably closer the the core, and each of the remaining 90% of the turns get longer.
The Crites 3636 autoformer can attenuate the voltage up to 12db, in 1db increments.
I made a pair not so long time ago. I remember that I used some old M85 035mm type laminations from Siemens choke. Gap was 1mm but with oposite stacking showed around 1/3. The laminations was from my findings from higher relative permeability. That is important because someone can use less windings and have more space for wider wire diameter, to decrease Rdc of the wire.
That is from importrance because of the small one digit load. The wire was 1.3mm dia, Rdc was 0,2 ohm or a bit less, and total inductance was about 25Hy
parallel resistor say 47ohm, from input end to gnd is a must (like in the traditional designs for this type of attenuation).
All in all the sound in general was totaly diferent and significantly better. And the impedance curve was better.
I can recommend this attenuator/filter.
some photos at local forum.
I can send MatCad sheet for calculating no of turns and other things (I dont have it in this computer )
That is from importrance because of the small one digit load. The wire was 1.3mm dia, Rdc was 0,2 ohm or a bit less, and total inductance was about 25Hy
parallel resistor say 47ohm, from input end to gnd is a must (like in the traditional designs for this type of attenuation).
All in all the sound in general was totaly diferent and significantly better. And the impedance curve was better.
I can recommend this attenuator/filter.
some photos at local forum.
I can send MatCad sheet for calculating no of turns and other things (I dont have it in this computer
)
Very nice work! And glad that you also experienced better sound quality.
I really believe it’s not the component that we hear. There is nothing wrong or non-linear with a resistor, but there really is something wrong with adding serious resistance between driver and amp.
When I remodeled my speakers, putting all the components in Boxsim and taking away the L-pad, the response was not only higher sensitivity but also a rippled response with high peaks and also with dangerous lows in the impedance curve. The x-over stored energy, that was damped and flattened by the L-pad resistors. I first had to totally redesign the x-over with different values to get to the target curve (the reference, the as-it-was response) without using resistors and without getting peaks in the amplitude and the corresponding dips in the impedance.
I have a strong feeling that such “low-Q” filters without stored energy already sound better. Next thing is to add the parrallel L value for the autoformer. That could cause a resonance, but depends on the value. When it is very large, this shouldn’t have to be a big problem. Always measure the impedance of the built x-over to check. One can add a parallel resistor (15 Ohms like in the old days), that doesn’t stand in the way of the amp controlling the driver.
So, in general, I think designing X-overs without series resistors is already making the speaker sound better.
I really believe it’s not the component that we hear. There is nothing wrong or non-linear with a resistor, but there really is something wrong with adding serious resistance between driver and amp.
When I remodeled my speakers, putting all the components in Boxsim and taking away the L-pad, the response was not only higher sensitivity but also a rippled response with high peaks and also with dangerous lows in the impedance curve. The x-over stored energy, that was damped and flattened by the L-pad resistors. I first had to totally redesign the x-over with different values to get to the target curve (the reference, the as-it-was response) without using resistors and without getting peaks in the amplitude and the corresponding dips in the impedance.
I have a strong feeling that such “low-Q” filters without stored energy already sound better. Next thing is to add the parrallel L value for the autoformer. That could cause a resonance, but depends on the value. When it is very large, this shouldn’t have to be a big problem. Always measure the impedance of the built x-over to check. One can add a parallel resistor (15 Ohms like in the old days), that doesn’t stand in the way of the amp controlling the driver.
So, in general, I think designing X-overs without series resistors is already making the speaker sound better.
Last edited:
Thanks. I put some additionl photos showing the way for extracting taps. I used copper foil (isolated and tined). And solder at the wire striped isolation. Not thet problem I was expected. But the windings stacked much better...
The key point are core laminations. Simly could not be common for the "refrigerators", nas to be from higher permeability - and bigger size.
Yes indeed You right. Energy loss in L-pad is main factor about the sound. I could not believe that the difference is that huge... Anyway i didnt used a lot L-pad but Briget T resistor attenuation network, which sounds a way better than L-pad. But inductive att. is better. Too much thermal energy, from usefull input is wasted, making a sound "dumped". I can also recomment Bridged T net as better thaen Lpad. Only have 2 resistors more but impedance is not that dependant from frequency. @ resistors have same value as Load (6.8 - 8,2 ohms ets) other 2 maling the att.
please try wouldn be dissapointed. google for on-line calc
(I made mistake about values, 25mHy is total inductance, and I think that value of parallel R is 47ohms or higher...)
The key point are core laminations. Simly could not be common for the "refrigerators", nas to be from higher permeability - and bigger size.
Yes indeed You right. Energy loss in L-pad is main factor about the sound. I could not believe that the difference is that huge... Anyway i didnt used a lot L-pad but Briget T resistor attenuation network, which sounds a way better than L-pad. But inductive att. is better. Too much thermal energy, from usefull input is wasted, making a sound "dumped". I can also recomment Bridged T net as better thaen Lpad. Only have 2 resistors more but impedance is not that dependant from frequency. @ resistors have same value as Load (6.8 - 8,2 ohms ets) other 2 maling the att.
please try wouldn be dissapointed. google for on-line calc
(I made mistake about values, 25mHy is total inductance, and I think that value of parallel R is 47ohms or higher...)
Hi Zoran. How to calculate the number of turns of wire for the first and second autoformer coils, e.g. if we are talking about a simple attenuator with one tap option?
The output voltage is directly proportional to the number of output windings divided by the number of input windings If there are 1000 input windings and 500 output windings, the output voltage will be half the input voltage, or -6db.
An important thing to remember is that this is an inductor in parallel with the driver, and the inductance of the output windings will create a 2nd order high pass filter when a capacitor is connected to the input taps. Using the 25mh example, the inductance of the -6db taps will be 6.25mh.
An important thing to remember is that this is an inductor in parallel with the driver, and the inductance of the output windings will create a 2nd order high pass filter when a capacitor is connected to the input taps. Using the 25mh example, the inductance of the -6db taps will be 6.25mh.
These little chokes with so few turns of such a heavy wire it is impossible to reach 25Hy, either you are measuring these in a wrong way or I really have to start reading physics all over again!I made a pair not so long time ago. I remember that I used some old M85 035mm type laminations from Siemens choke. Gap was 1mm but with oposite stacking showed around 1/3. The laminations was from my findings from higher relative permeability. That is important because someone can use less windings and have more space for wider wire diameter, to decrease Rdc of the wire.
That is from importrance because of the small one digit load. The wire was 1.3mm dia, Rdc was 0,2 ohm or a bit less, and total inductance was about 25Hy
parallel resistor say 47ohm, from input end to gnd is a must (like in the traditional designs for this type of attenuation).
All in all the sound in general was totaly diferent and significantly better. And the impedance curve was better.
I can recommend this attenuator/filter.
some photos at local forum.
I can send MatCad sheet for calculating no of turns and other things (I dont have it in this computer
Thanks for your response!
I understand the general logic of the attenuator, about the proportions between windings. I have more of a practical question. Is it possible to empirically choose some number of turns to test the operation of such an attenuator, or how to calculate it correctly? For example, at least 500 turns or at least 1000 turns?
I wish there was a quote button on the last post of a thread!
Not sure if I follow your question. Maybe my answer will help clarify...
The number of total turns doesn't matter, just the ratio. You can use 100 turns input and 50 turns output and get -6db, but the inductance would be so low that the capacitor would be connected to almost a dead short at low frequencies.
Let's start with this calculator and see where it goes.
http://www.sengpielaudio.com/calculator-amplification.htm
Since windings equal attenuation, enter 1000 in the Input field and -4 in the Level change field. Click on Calculate value 2. You'll get 630.957 in the Output field. So, if you have 1000 turns on the input, and 631 turns on the output, you will get -4db.
Sorry if I'm insulting your knowledge, but does that help?
Not sure if I follow your question. Maybe my answer will help clarify...
The number of total turns doesn't matter, just the ratio. You can use 100 turns input and 50 turns output and get -6db, but the inductance would be so low that the capacitor would be connected to almost a dead short at low frequencies.
Let's start with this calculator and see where it goes.
http://www.sengpielaudio.com/calculator-amplification.htm
Since windings equal attenuation, enter 1000 in the Input field and -4 in the Level change field. Click on Calculate value 2. You'll get 630.957 in the Output field. So, if you have 1000 turns on the input, and 631 turns on the output, you will get -4db.
Sorry if I'm insulting your knowledge, but does that help?
In such cases, phrase can be quoted.
Thank you, yes, it was useful, I'm a little slow in such things, but I began to understand better. Now I need to think more
Do I understand correctly... I can buy a transformer inductor of the value I need for the filter. Then disassemble it, count the number of turns, and wind two serial coils with the same wire?
I understand that it is easier to buy a ready attenuator. But first, they are quite expensive. Secondly, it is always interesting to DIY