When specifying leakage inductance you need to be clear which winding you are measuring it from. The winders I dealt with specified leakage inductance with all leakage reflected to the primary. So, 80mH on secondary of 75:1 step-up would have been quoted as 14uH. I have gotten best results when I provide winding instructions (ie wire gauge, turns per layer, number of layers, insulation between layers…etc) rather than specifications, and then do the measuring myself.
Last edited:
I know that suitable step-up arrangements for hybrids can be put together using small toroids. However, if you are going to be getting something custom wound anyways, you might consider a design that would require only one transformer per speaker. In my experience, suitable step-up transformers for segmented designs with capacitance < 200pF is most easily attained using a dual bobbin UI core transformer. See attached pic of Acoustat Spectra 11 transformer, and coil layout arrangement to get an idea what I am talking about. If the winder you are talking with can wind a UI core (5 in x 3 in) and 1 in^2 core area with M6 grain oriented steel, it is pretty easy to get 120:1 ratio with core handling capability of 40Vrms (200W/8ohm) @ 150Hz without any fancy winding geometry. Unless costs are significantly different in your area of the country, you should be able to get it wound for ~$100 which is likely not much different from what you would pay for the number of toroids you are considering. I can help provide winding directions if you decide this is something you want to pursue.
Attachments
Funny, I was looking at this kind of design after reading thru the threads on the guys winding their own transformers.... I didn't pursue it because I am way out of my element in knowing the details of building transformers and what I should be asking for.
I will ask the guy and see if he could do this. Basically, the spec sheet for a quote is in the attachment to my prior post... I would be grateful for giving me the info I should ask for (winding instructions etc...).
Actually winding arrangement depends greatly on the frequency range to be achieved, breakdown voltages dielectric choice etc. and some requirements are rather self-contradictory. To lesser the capacitance you need to have shortest turn length for the hottest part of winding i.e. the ones that are connected to the stators. To do so you have to have them near the core. Thus requires really good insulation as well as spacing to the core/primary. Since it highly hard to do in real life, HV HF x_former has tapered secondary i.e. winding width becomes smaller and smaller layer to layer as well as creepage distance (the area without wires) becomes bigger and bigger. Increased layer to layer spacing or insulation thickness is also a plus. Every layers has to start from the same side of winding period.
It's clear that even experienced "winder" could not get it right at the first attempt so $100 seems rather small. Quality of materials and craftsmanship as well as cleanness at the shop and impregnation equipment plays very big role in a process. I would of use special insulation paper and really simple wax or paraffin impregnation.
I do not advise on PET or PP film use for insulation due to static electricity and subsequent dirt/dust contamination. Nevertheless simple paper (somewhat yellowish) not the bleached high density one and simple potting in clear wax using water bath and appropriately sized metal do the miracles.
Almost forgot: I successfully used 2mil teflon + paper combination. 1 layer + some overlap (1/2") is more then enough not to eat too much of winding window. Paper from coffee filers will do. Paper prevents damage of teflon, holds things together well and takes up liquid wax easily. Do not submerge coil fully from the very beginning - it will impede air removal. Submerge gradually and let the capillary force do the trick.
It's clear that even experienced "winder" could not get it right at the first attempt so $100 seems rather small. Quality of materials and craftsmanship as well as cleanness at the shop and impregnation equipment plays very big role in a process. I would of use special insulation paper and really simple wax or paraffin impregnation.
I do not advise on PET or PP film use for insulation due to static electricity and subsequent dirt/dust contamination. Nevertheless simple paper (somewhat yellowish) not the bleached high density one and simple potting in clear wax using water bath and appropriately sized metal do the miracles.
Almost forgot: I successfully used 2mil teflon + paper combination. 1 layer + some overlap (1/2") is more then enough not to eat too much of winding window. Paper from coffee filers will do. Paper prevents damage of teflon, holds things together well and takes up liquid wax easily. Do not submerge coil fully from the very beginning - it will impede air removal. Submerge gradually and let the capillary force do the trick.
Last edited:
I don’t disagree with your comments and certainly some amount of trial and error is required to design transformers working at the extreme of what is possible in terms of bandwidth and power handling. However, the simple transformer configuration of using a separate primary and secondary for each phase with the cold end of the secondary near the primary(connected together as CT) works well for ESL step-up transformers when the capacitive load is not too great. Audiostatic uses this technique as does Quad on their ESL-63. The only difference is that we are putting both primaries on the same core so when they are put in parallel the combined primary inductance is not cut in half.
In this configuration, the primary-to-secondary capacitance and core-to-secondary capacitance is very small…typically < 15pF for step-up ratios > 100:1. The only remaining parasitic parameters are the secondary winding capacitance and the leakage inductance both of which are easy to calculate for this simple geometry. If you take a 1 in^2 (6.5cm^2) core and put 50 -60 primary turns on the core with ~2”(5cm) winding width and then 11 or more layers of secondary each separated by 6mil(0.15mm) of insulation with turns per layer chosen appropriately based on secondary wire gauge(I usually use 32 - 36AWG), you will see that there is plenty of margin for a 120:1 design for frequencies > 150Hz even if the winder isn’t perfect and no special insulation material is used.
Perhaps I should have stated up front that this isn’t just theory. I have had 5 different transformers wound with this configuration that matched calculations for parasitics nicely. Perhaps I got lucky and found a winder that knows what he is doing and can follow build instructions.
In this configuration, the primary-to-secondary capacitance and core-to-secondary capacitance is very small…typically < 15pF for step-up ratios > 100:1. The only remaining parasitic parameters are the secondary winding capacitance and the leakage inductance both of which are easy to calculate for this simple geometry. If you take a 1 in^2 (6.5cm^2) core and put 50 -60 primary turns on the core with ~2”(5cm) winding width and then 11 or more layers of secondary each separated by 6mil(0.15mm) of insulation with turns per layer chosen appropriately based on secondary wire gauge(I usually use 32 - 36AWG), you will see that there is plenty of margin for a 120:1 design for frequencies > 150Hz even if the winder isn’t perfect and no special insulation material is used.
Perhaps I should have stated up front that this isn’t just theory. I have had 5 different transformers wound with this configuration that matched calculations for parasitics nicely. Perhaps I got lucky and found a winder that knows what he is doing and can follow build instructions.
Last edited:
Hmmm...perhaps. You would need to let them know it is UI core with 3" x 5" external dimensions and 1" stack height using M6 laminations. Also be sure to mention there are two sets of these coils, one on each long leg of the core; perhaps show them the pic of the Acoustat transformer. If price is agreeable, then we can get more details on the wire and insulation types they have readily available and dial in the exact turn numbers and insulation thicknesses.BTW, is this enough build instructions that I can get one quoted from a winder.
I finally got around to updated the spreadsheet calculator:
ESL_line_sectioned_DIY_v3
I used you design values as a test case for the newly added esl_seg_ui parameter outputs in the spreadsheet for Configuration 1 and 2. The results are shown in attached pics. As you can see, differences in dispersion are only in the top octave. With the larger middle section of Configuration 1, you can see nulls in the top octave. This is what golfnut’s 17mm maximum width rule is meant to avoid.
Attachments
Thanks so much for supporting us tinkering nerds!
I was actually thinking about this last night, the segmented ESL progressively attenuates the high freq signal from the center to the edges. So it kinda of makes sense that the dispersion is unaffected at lower frequencies.
What I am trying to understand is why attenuating the hi freq signal toward the edges improve dispersion and the "head in a vise effect"? Is this because of the hi freq nulling you are showing/referring to?
BTW... still waiting on supplies to show up but should be starting my build soon. Also waiting to hear back from a couple of winders to get quotes on the transformer (which I am starting to realize is the single most import component in this rig).
So I tinkered around with your revised spreadsheet (which, again, is awesome) and this is what I come up with (see attached).
Question I have for folks is, is there a rule of thumb or something figure out if I stretch this size mylar by x% its resonance frequency is around X?
I thought about making a stretching table which I could get an approximate resonant frequency by exciting it with a signal generator.
I noted that for hybrids, the lower the Fs, the quicker the low end response dives down (trying to keep it +- 3dB). So I kind of came up with 125hz as my ideal.
Any concerns/thoughts?
Question I have for folks is, is there a rule of thumb or something figure out if I stretch this size mylar by x% its resonance frequency is around X?
I thought about making a stretching table which I could get an approximate resonant frequency by exciting it with a signal generator.
I noted that for hybrids, the lower the Fs, the quicker the low end response dives down (trying to keep it +- 3dB). So I kind of came up with 125hz as my ideal.
Any concerns/thoughts?
Attachments
well the resonance is mostly dictated by the width of the panel(if your panel is X times bigger in length then width that is, if its square it does not mater ofcourse). a res of 125 hz is pretty easy to dive under when your panel is wide. and maybe even impossible to achieve (to reach or go over)when it is wide enough. in my opinion resonance in a big panel used for hybrid is never a problem with the roll off. with fullrange panels the resonance dictates the lowest it can go until it rolls off. but when using a big panel as a hybrid with usually lower spacer distances (to increase spl) stability might be an issue, for instance a quad panel measures approx 60*20cm and has a res of 65-75 hz under HV. this is with the maximum tension a 3 micron film can handle, you might get a little higher if you wanted but not much. so max resonance drops pretty fast when you decrease mylar thickness and width of panel. it is possible to achieve higher tension with thicker mylar.
So if your panels are anything near that mark or bigger it is almost the max you can get, so lower then you are aiming for witch is good.(70 hz is based on 3 micron mylar, it wont allow(much) more stretch before it tears on these panels)
(this is based on the stators from a quad panel it might be a bit different with wires, but not dramatically)
i think what you mean with the quicker the low end will go down is extending the frequency response with the resonance frequency, and critically damping it until it extends the low end instead of having a hump. this can work. but just my to cents it is more easy to cut it before or over damp it. not extra low end but getting the resonance and the damping right so it extends is not that easy. for a first project i would not recommend it but.... if you want and personally i never liked the sounds of the resonance frequency to be honest (this could be between my ears) (even in the quads it sounds sloppy sometimes, witch could eb my room to so it is hard to say)
So if your panels are anything near that mark or bigger it is almost the max you can get, so lower then you are aiming for witch is good.(70 hz is based on 3 micron mylar, it wont allow(much) more stretch before it tears on these panels)
(this is based on the stators from a quad panel it might be a bit different with wires, but not dramatically)
i think what you mean with the quicker the low end will go down is extending the frequency response with the resonance frequency, and critically damping it until it extends the low end instead of having a hump. this can work. but just my to cents it is more easy to cut it before or over damp it. not extra low end but getting the resonance and the damping right so it extends is not that easy. for a first project i would not recommend it but.... if you want
and personally i never liked the sounds of the resonance frequency to be honest (this could be between my ears) (even in the quads it sounds sloppy sometimes, witch could eb my room to so it is hard to say)
Last edited:
What I was saying is the lower the Fs, the quicker it rolls off from 200hz or so (in my config).
This was touched on briefly in another thread…another-segmented-esl-Post#53
Here is a simple way to think about cause for the nulls you see at some off axis angles. When the listener is off-axis of a wide segment, you can imagine that different parts of the diaphragm are different distances from the listener. So, the sound from those different parts would arrive at the listener’s ear at different times…ie shifted in phase. For a given segment width and off-axis angle, there will be frequencies were the summation of output from all the different parts of the segment will be zero due to this phase shifting. The wider the segment, the larger the phase shifts and the lower in frequency the first notch will be for a given off-axis angle.
For those who like mathematics, you can start from the sinc function(which defines directivity of rectangular segments) in the following post and derive that the first notch will form at f = c / (W * Sin(angle)), where c = speed of sound, W = segment width.
http://www.diyaudio.com/forums/plan...-dispersion-planars-how-much.html#post2264055
The directivity of a segment is caused by the phase shifts mentioned above, so is dependent on how its width compares to the wavelength of sound it is radiating. Once frequency increases to where the wavelength of sound is smaller than the segment width, the directivity increases steadily the higher in frequency you go. Basically the radiated sound energy is bundled more and more tightly on axis rather than dispersing it broadly. The result is “beaming” of high frequencies straight ahead with little radiated off to the sides. To avoid this directivity increase completely, you would need to halve the width of the panel radiating sound for every octave increase in frequency. If we did this with line source ESLs, we would wind up with a frequency response that falls -3dB/octave at the top end. This could be corrected with EQ in the crossover. What golfnut’s AES paper provides is a set of equations to define simple segmentation schemes that result in flat response for line source ESLs with directivity that increases slowly with frequency.
If you are trying to cross your hybrid over unusually low or want maximum output with lower power amplifiers, then you are certainly right. Otherwise, the 75:1 step-up provided by the toroid pairs CharlieM mentioned work just fine, especially if you plan to use a DSP crossover where you can easily tweak things to your liking.
It is tempting to try and take advantage of resonance to keep the low end response of a finite line source from rolling off so quickly. However, if you don’t plan to use acoustic damping mesh to lower the Q of the resonance, you need to put the resonance frequency 1.5 – 2 octaves below where you plan to crossover. Otherwise, you will run into issues with overdriving the panel at the resonance peak and the diaphragm hitting the stators. You may recall CharlieM wrestled with this a bit on his last build and solved the problem by using the combination of a sharp crossover and notch filter.
thinking-about-segmented-wire-stator-esl-Post#177
thinking-about-segmented-wire-stator-esl-Post#186
True enough, it is tempting. I'll try to shoot for a resonance of 75hz but still not sure how I am going "tune" the diaphragm to that
. I guess I will shoot for a 1% stretch and see what happens. I do have room equalization (adyssey) in my pre-processor so I will be able to eq out some of the problems.
Also, as I have learned, theory doesn't reflect reality when you put a speaker in a room. In addition to all the other problems you get, you tend to get a in-room boost for lower frequencies.....so I will probably be OK.
Im just stating that a res of 75 might be hard to achieve depending on the width, there is only so much you can do with stretching. Big panels have low res stretching helps but there is a limit, especially if you want it to be stable over time. You could also tune it like magnepan does, if you would have a panel the size and weight of a magnepan you need insane amounts of stretch to reach even 50 hz. Ofc they do use the resonance at there advantage, devide an conquer to fill in the mid low dip and extend the low end. So they got 2 resonances depending on how many tuning knobs used.
Thickness of Mylar is key here. You can't stretch 3 micron hard enough to reach 75 on a big panel 6 micron is a different story. I maybe ever when my stretch jig is finished do an extensive amount of tests so we can see relation between size and res. and max res for a certain thickness . Would be very helpful. Time is not on my side
Btw very nice looking stators !!!
depending on how many tuning knobs used. Thickness of Mylar is key here. You can't stretch 3 micron hard enough to reach 75 on a big panel 6 micron is a different story. I maybe ever when my stretch jig is finished do an extensive amount of tests so we can see relation between size and res. and max res for a certain thickness . Would be very helpful. Time is not on my side
Btw very nice looking stators !!!
Last edited:
Hi bengel,
If you are using 1/16" rods, nylon edging grommets are an alternative to bolts. They work very well for keeping them aligned and straight. They are made by Panduit, are about 12 3/4" long, and can be purchased through Mouser and other vendors. They look like mini dental teeth. I am building ESL's using multiple small panels using the grommets, but have also used them for larger single panels.
Below is a picture using the grommets and one of a small single panel. You've got quite a project there, but it's a lot of fun.
If you are using 1/16" rods, nylon edging grommets are an alternative to bolts. They work very well for keeping them aligned and straight. They are made by Panduit, are about 12 3/4" long, and can be purchased through Mouser and other vendors. They look like mini dental teeth. I am building ESL's using multiple small panels using the grommets, but have also used them for larger single panels.
Below is a picture using the grommets and one of a small single panel. You've got quite a project there, but it's a lot of fun.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.