Step-up transformer design

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi,

the easiest way to test an stepup is like this:

1. Take a labtop or PC and connect a Soundcard via USB (M-audio, US122L....).
Some soundcards have integrated mic-amplifier.
2. Download Arta software
3. create a simple made impedance measurement setup using an additional amp
4. Connect your amp to the primary winding of your stepup.
5. first you measure your impedance without a secondary load like an ESL

The graph will tell you a lot:

- Bandwidth is within phase angle change zero (low end) to zero (high end) degree
- e.g. if your upper bandwidth show limitations lower than 15 Khz secondary windings have too much parasitic capacitiy.
- Q of the bandwith filters is visible by steep or flat phase change angle
- To check compatibility with your esl load, just connect a simple Capacitior having same value like the panel to the secondary windings and repeat measurement.
- You will easily see, how the additional load will reduce the upper bandwith, will change the Q of high-band filter
- On the low end, arta will calculate the primary inductance, which need to be at least 5mH used from 200 Hz up to 50mH used for fullrange-ESL

Capaciti
 
Last edited:
here is the complete study i did in its entirety and corrected. so that nobody missed any thing.lots of info here and some basic theory on using (toroid)power transformers to drive esl's.I hope you enjoy. jer
How it started: Hi Jer,
I'm curious about my speakers' impedance too, since I have no way to measure it and I don't know how to calculate it. My Carver amps drive them with no problem but I'd still like to know their impedance. Anyway, I'm using the same tranny setup as Mavric: Tandem 230v/2x6v toroids wired backwards with inputs parallel and outputs in series to give a theoretical 76:1 step-up; although I'm told the step up is somewhat less in praxis (68:1). My panels are 12x48 inches with 1/16" d/s spacing. Also, I have a 1-ohm resistor in front of the trannys, just to be safe. Calculating final impedance vs. transformer ratio: okay. Here is the formula, first you calculate the area 12"(l)*48"(h)*.51(percent open area)=293.76 sq. inches and (1/16"d/s*2) .0625*2=. 125 then you enter these values into (search) "capacitance calculator" with dielectric constant of 1 for air you get 528.2978 Pico farads, then you (search) "reactance calculator) and enter 528.297pf (or rounded if you wish) and enter highest freq. of 20khz and you get 15.063kilo ohms, your turns ratio is 230/6=38.3*2(because the secondary is in series)=76.6:1, final impedance is 15063ohms (capacitive reactance)/5867.56(turns ratio 76.6 squared)=2.567ohms+1ohm(resistor)=3.567 ohms final impedance mileage may vary and be lower, as I didn’t account for the stator coating ,but that’s basically it. 3.56 ohms at 20 kHz, which also means 35.6 ohms at 2 kHz, and 356 ohms at 200hz. Don’t be alarm by the higher impedances as at doesn't bother today’s amplifiers. But as you know the lower ones are the critical ones that cause exessesive current draw, instabilities and amplifier failure. I hope that helps. Jer How impedance affects the amp: are you familiar with op amps (short for the term, operational amplifier)? A power amp is but nothing more than a great big op amp capable of producing a larger amount of voltage and current swings than an 8-pin i.c. Can, but the principle is the same. Negative feed back forces the circuit to drive more or less current to the load regardless of the output resistance to achieve the voltage required which is determined by input voltage times the gain of the amplifier. For instance if your amp has a gain of 30, then with a 1v input it's supposed to produce a 30v output, ac or dc. Now if you have let say a 1000 ohm load across the output, ohms law says V/R = I so then your amp only has to provide .030 amps (or 30 milliamps) of current to produce 30v across the load. Now if you were to put a 1 ohm load across it's output, due to ohms law the amp will be required to produce 30 amps of current to get 30v across the load and according to ohms law power = volts squared divided by resistance so 30*30/1=900 watts.your average amplifier cannot produce that much current and this is were clipping occurs and this is why amps fail when such a low impedance load is hooked to them, and why high power amps have so many output devices per channel. That is, to deliver massive amounts current when need. Which is the reason i'm very skeptical about car amps as I’m sure that there are some that actually produce that kind of power but I find it very hard to believe that the cheaper ones do and do it cleanly. But that’s an argument I wish not to start, at least not here. I hope that helps you understand. Jer

Discovery: hey guys check this out, as I was looking over my $20 200watt toroid transformers that I got on clearance from parts express I noticed a 13.8 volt winding so I did the calculations and 4 of these would give me 8 ohms on my 7.75"x22" panels. So I fired up the signal generator and scope and here's what I got, 65hz-18khz flat -1db points, 46hz-25khz -3db, 30hz-31khz -6db, 18hz-50khz -12 db. Pretty cool huh? Since I have 10 of them I’ll give them a try. I just might undo the secondary windings and rewind a new 13.8 with some taps to change ratios. By getting rid of all the extra windings I don't need it should reduce the interwinding capacitance (to zero to start with) and extend the flat response past 30khz anyway here's some pics of one with a 2hz to 20khz .6sec sweep test. While testing I got a clean 400v to 500v sine wave throughout the audio pass band and a good looking square wave from 300hz to 10 kHz in the critical range, I saw no signs of saturation or distortion even when loaded with a 130m 6kv capacitor and 3ft of coax cable. Incredible! The scope graticule reads as follows from the center 50v per division 5 points per division, 9 = -1db, 8 = -3db, 5 = -6db, 2.5 = -12db,the two markers are set at -3db for reference. Okay I’m going to go and build my new frames now. Stay tuned and I’ll keep you posted! Jer
Attached Thumbnails see “material for esl post # 147”

Description of test subject: I had intended too use them for a lm3886 type chip amp I got 5x lm4780ta chips as free samples from national semiconductor when they were first released. All I had to do was pay the shipping which was $14.however the main secondary winding is 35v and would yield 49v in order to use two of them I would need a regulator to drop 7v as the maximum is -+42v.but I could use one of them to power 4 of them in a bridge parallel configuration too achieve 200+watts.there are several other discrete amp circuits I could use, but you can't beat the simplicity and quality of a $6 to$7 68watt power op amp the lm4780 has two on one die for $9 (they went up through the years) check out their website for more sample info. My goal at the time was to use 8 of them in a bpa configuration as 3 or 4 of would yield more than 300 to 400 watts into 4ohms and I was looking to be able to go down to 2ohms.just another project that got put on the back burner till now, all intended for my esl's. I did allot of studying on transformer technology only to find out that it is really quite simple once you understand the basics. Most of the work is already done for you using a power transformer. The only difference between an audio transformer and power transformer is basically two things audio ones have a higher quality core material and are design to go down to 20hz. To do this you have to use a bigger core than power transformer of the same power rating. Today’s toroidal cores are becoming more and more the norm by today’s standards and use a higher quality steel than the old E-I style, which makes them so darn expensive. Now here's little something that allot of people don't realize is that for every doubling of frequency the power rating is doubled and vice versa. So my 200watt transformer at 60hz now becomes a 400watt transformer at 120hz and so on. Which is why your 600watt computer power supply has such small transformers, because they’re running at 40khz+. Get it? Now, in order to get a flat response from 20hz to 20khz becomes tricky you start the design using 20hz as a factor than 60hz which roughly requires 2 to 3 times more core area and/or 2 to3 times more inductance on the primary. That is the basics of transformer design. Most all of the factors are pretty much reciprocals of each other. On the other end of the scale to achieve a high frequency response, interwinding capacitance (the capacitance between the windings) becomes the limiting factor (also core material in some cases). I hope that helps you to understand a little better. There are allot of websites out there that go more in great detail if you wish to know more, but don't be afraid to ask me anything if you wish. Jer

Theory worked: hey, my theory worked! 10 turns of wire is all it took. Now all I need is a bias supply and these baby's will be driven harder the ever before 750V plus out of one 120V winding and there's two of them. I just saved myself six awesome power transformers. You’d better build another pair of ESL’s for Quadra phonic or surround sound! Because you truly only need one doughnut each. I was using my awia cd changer at full tilt before clipping. As it has a very sensitive shutdown protection circuit. If I had some bias on the speaker my ears would have started to bleed at that level! What do you think? Here’s the proof. Jer
Attached Thumbnails see “material for esl post #152”




The study and new test results: Okay, everybody here is the new test results. My equipment is as follows, Hitachi v-425 oscilloscope, tenma model 72-455 signal generator .for the previous sweep test the following was used, Phillips 16 bit pcs706 sound card, set at a 96khz playback rate, mackie 32-8 mixer with a confirmed 0hz to 20hz response, and, is well within and exceeds factory specs with a aiwa cx-na707 digital audio system for an amplifier, all eq’s and signal processing are set to off. I hereby state that the data presented is to the best of my abilities and are as measured. I did not plan to take any pics at this time but if requested, I will be happy to repeat the test and post the results again with the pictures. The original pics posted of the sweep test were confirmed that it is the same of the amplifier, as suspected, by using a 10-ohm resistor for the load. I have chose to use the two 120V windings as any other winding posed an improper load for the signal generator which resulted in a distorted sine wave. also the two windings were matched. the rest were left open .test #1 generator on one 120V winding and the scope on the other 120V winding. test #2: the scope was connected to the new custom 10 turn winding all else is the same. Test #1 results are as follows: -2db 10hz, -1db 20hz, -.1db 60hz, ruler flat 0db 100hz-50khz, -1db 115khz, -3db 200khz, -6db 300khz, -8db 400khz, -6db 500khz, -12db 600khz, -7db 850khz, -12db 1mhz. Test #2 is as follows: -2db 10hz, -1db 20hz, -.1 60hz, flat 0db 100hz-50khz, +1db 115khz, bump of +7.5db 305khz, dip of .7db 450khz, bump of +9db 650khz, +2.75 db 1mhz.test #1 shows possibly the effects interwinding capacitance above the audio range. It also reflects the loading of the signal generator due to the self-resonance’s of the transformer it self, as evident in test #2. Test #2 shows no degradation of frequency response due to interwinding capacitance only the high frequency bumps due to the self-resonances. This is confirmed by the leading edge ringing of a square wave. Although I had discovered that my signal generator did not produce a very symmetrical square wave below 100hz but was very acceptable anywhere above 100hz.after finding this only flaw with my generator I rechecked the input signal to the transformer and compared it to the output with the scope and I assure you that what went in did infact come out .the transformer produced both square and sine waves without any evidence of added distortion or saturation of the core, even at a higher power as previously tested(except for the square wave leading edge ringing common in all transformers) throughout and exceeding the audio pass band. These test results are the best of any transformer I have test to date power or audio. However it is the very first toroidal I have tested so far and being a 200-watt core I did expect good results but not the impressive ones I got. I expect similar result from a smaller core due to the higher quality materials used than the E-I cores of yesteryear. When I acquire some different samples of sizes and/or brands I will test them and post their results also. I have proven that it is possible to use reverse engineering by easily adding a new winding to determine the existing turns to create a custom ratio transformer to drive an esl panel instead of using two or more devices cutting down the risk signal degradation from using more than one device per panel due to poor manufacturing tolerances and cutting cost for the average diy’er as we all know that transformers cost their weight in gold. It’s no joke! My next tests will be to study the effect of frequency response with a change of turn’s ratio by adding and deleting turns from the new custom primary. Jer
More random tests: I did find a few more transformers laying around today to mess with. 1# a toroidal somewhere between 100 to 150 watts I’m guessing because it is in a steel shielding case and is about the same outside dimensions as the previous one I tested. The results were roughly the same, almost identical. Except the resonate peak was at 840khz and had less ringing on a square wave which resulted in a much cleaner waveform up to 120 kHz. 2# an E-I tube type power transformer that failed miserably, it looked like a band pass filter tuned to peak of 225hz and not even at the 60hz it was designed for. It would not even produce a square wave. 3# E-I tube type output transformer for push pull el-34’s (6ca7) showed good square wave from 10hz to 600hz with much much more ringing at 175khz .the sine wave test was flat 10hz to 2.2hz with a constant slope from 2.2khz to +8db at 20khz continuously rising onto 100khz and up. Which tells me why this amp sounded harsh and trebly, yet with a warm low end, when I used it for my guitar (which I liked, but not for music) this is the same transformer I used to drive my mini esl driver when I described what I heard (I don't remember if it was this thread or another) coming from it. All of the tests made no difference in response by swapping the primary for the secondary or vice versa except for the loading of the signal generator level which remained constant regardless of frequency, changing only the voltage ratio. Once again a toroidial power transformer wins hands down and also it was unmodified. any suggestion and thoughts on this is welcomed .jer p.s. i would like to thank charlie and mavric for asking the question they did, which inspired me to do this study.thanks again guy's and all of you diy'ers on this site.thanks, jer.
 
Last edited:
Transformer winding jig

Hi, enclosed is a picture of my transformer winding jig. Not fitted with digital counter or magnetwire roll yet.

It will be driven by a battery drill and with the digital counter it will be easy to count turns.

The coilformer in the picture will be relieved of its winding spacer before use.
I will use Silivio Klaic's transformer calculatior to give me a pointer on the number of turns, and are working on sorting out interleaving at the moment.

Regards

Bent
 

Attachments

  • IMG_0405.jpg
    IMG_0405.jpg
    856.7 KB · Views: 1,530
Hi,

the high frequency transformer/ESL limit is in a first approximation due to leakage inductance of the tranny and not its interwinding capacitance.
In a first approximation it calculates to:
f-3dB(hf)= 1/[2*pi*SQR(Ll*Cspkr-refl) with:
pi=3.141.....
Ll= leakage inductance in Henry
Cspkr-refl= Capacitance of the panel in Farad, reflected to primary side of the tranny. The reflected capacitance is the capacitance of the panel multiplied by the square of the turns ratio.

The interwinding capacitance forms a RC-Filter together with the sum of the amplifiers output resistance and the trannie´s primary winding resistance. Those resistances are usually below 1Ohm, so that the RC-filter´s bandwidth reaches far beyond audio bandwidth.
Since interwinding capacitance is defined as inter-winding, it is not transformed (up) by the trannie´s turns-ratio but enters the calculations with a factor of 1. This means it is and it stays of rather small value, compared to the reflected panel´s capacitance.
So while it indeed affects the bandwidth, the role of the interwinding capacitance is small. The main factors are leakage inductance and reflected panel capacitance.

jauu
Calvin

ps: geraldfry.fr could You do us a favour and bring Your threads into a readable, understandable und proper form?
Just had a quick look over parts of Your treaty. Your capacitance calculations of the panel looks reasonable on first glance, but in practise the capacitance is very close to a massive sheet, especially when the panel is coated with a material featuring an epsilon value >1. So You can leave out the openness factor completely. Your panel will measure around 1nF. As such the resistive impedance value @20kHz will be close to 1 Ohm.
 
Last edited:
Calvin, thanks for the extra info as i did not go into extreme detail. " Final impedance mileage may vary and be lower, as I didn’t account for the stator coating ". I’m sorry that you some times miss certain details in my posts. I do the best I can. If it is because the font gets smaller I do not know as of this moment why it happens, I don't know. I some times have a problem reading some of your posts as well. Such as when you use the character U in place of what should be a V or a (und) and so on. I realize the language barrier but I don't ridicule you. Do I? Now, I’m no grammar expert but I did spend over 4 hours correcting and editing several post together, had I left it in it's original format it would have looked extremely silly. No? Now, using the open area percentage is an important factor and was based upon theory, and, theory doesn't always apply exact in the real world. As I plan to investigate the issue further. As the hole size may be a larger factor than just open area alone. There is allot to be studied here on several issues. I enjoy reading about ideas and dreams on this website, and, have doing so since it was first created. On my travels I have also found allot of speculating and what if's with some very good suggestions along the way. But I very seldom find any solid proof of things by testing. The whole scope of my article was to prove that a toroid power transformer can be used to drive esl’s (which has already been done), and, that if one could not find one with the so desired turns ratio that by creating a new primary one could infact obtain a ratio desired, with great performance, with readily available parts. It was not about ground up design. That will be another study at another time in the future. Therefore I do appreciate your info. I did not mention reflected capacitance and only used reflected reactance as impedance to explain in a simple manner that is understandable for someone. I am not a newbie and going into L,R,C losses would only confuse some one who barely knows the basics of transformer. I have rewound many E-I cores myself and had planned to do a ground up design in 2003 at a time when large toroid cores weren't readily available (especially here in the states). Now the opposite is true. Not to mention the poor quality steel still used in E-I cores of the recent past. Sure it is a great learning experience, but I would hate to see anyone go through all the trouble of rewinding an E-I transformer only to have a mediocre product or one that doesn’t work at all. If you have ever wound a E-I core you know that it takes special care and special measures in order to handle the 2kv to 5kv or more required to drive a esl panel without it breaking down. And as you know that leads to more losses not to mention sealing the laminates so that they don’t vibrate (and they do), but not so sealed that you can’t take it apart in case you have to start over. I have yet to see any where in these threads a recipe for I given core size (as core sizes are all standardized) that works! What I have presented here is a process that is simple and easy and needs further investigating. As I have stated before any suggestion and oppions are welcome, and, if there is anything that I have stated is unclear, I will be happy to explain them to the best of my ability and discuss them. Have a nice day .jer
 
Hi Gerald,

If it is because the font gets smaller I do not know as of this moment why it happens, I don't know.


No I think it would be very beneficial if you could just split your texts up into smaller paragraphs. Many of your posts are just one huge block of text and my eyes are struggling to keep track of where I am while reading. At some point, I just have to give up because it it too tiring.

So at least from my POV I believe your posts would get more of the attention they deserve if you could just insert some empty lines here and there.

Thanks
Kenneth
 
yes,i understand. i did try to split them up as i was having porblems with ms word and i'm not very proffciant with it yet.i'll try to do better,sorry.having a pc since the 80's this the first year i have forced my self to write documents and use email.sad huh?so i am still learning and i'll try to do better.i see your point and i don't like the way they look either.i just get very frustrated sometimes when i just spent (a few times) 4 hours making a document ,in order to hit a wrong button or a bsod occurs losing everything .just so i can start over!aaaaahhh,practice practice practice . jer
 
thanks for the tips jonas.is there an easier way to insert a new line with out having to hit the spacebar 1000's of times.also i do normaly use the forum window, but there have been times when i have lost hours of typing and had to start over.trying to remember everything can be quite difficult and very time consuming for such a simple task.you'd think that it could be recovered some how.but i haven't figured it out yet. cool winder, wachara!
 
Last edited:
Now here's little something that allot of people don't realize is that for every doubling of frequency the power rating is doubled and vice versa. So my 200watt transformer at 60hz now becomes a 400watt transformer at 120hz and so on.

My experience is that for every doubling of frequency the power a transformer can handle before core saturation increases by a factor of 4. Looking at the formula for Peak AC flux density this makes sense because the AC flux density is proportional to applied voltage, not power; and power is proportional to voltage squared.

Bmax = E x 10^8 / (4.44 x f x N x A)

Bmax = Peak AC flux density in core (Gauss)
f = frequency
N = number of primary turns
A = core area cm^2

Note that Bmax is inversely proportional to the number of primary turns.
Typical Bmax for modern transformer iron is about 12,000 gauss at the onset of saturation

Using a core area of 2.25 in^2 (14.5 cm^2) we can calculate core saturation levels for a range of frequencies.
I chose this core size because it is a fairly typical size used for full range transformers, and is probably pretty close to the core size of your toroidal power transformer.

Looking at the plot summaries for 60 turn and 10 turn primaries, you can quickly get an idea why you can't just reduce the primary turns to get a high step up ratio. As you reduce primary turns, the onset of core saturation comes at higher frequencies and lower voltage levels.
 

Attachments

  • Core_60.gif
    Core_60.gif
    87.3 KB · Views: 1,131
  • Core_10.gif
    Core_10.gif
    91 KB · Views: 893
A quick question; should not a transformer be a 8ohm load on the primary side for the amplifier?

All spec's I have seen has stated 8ohms on the primary side.

regards
Bent

Hello Bent,

The quick answer is no. What ESL transformers have you seen specs for that state 8ohm?

For an ESL step up transformer the load impedance that the amplifier sees is a rather complicated function of the winding resistance and inductance, leakage inductance, winding capacitance, inter-winding capacitance, as well as the capacitive load of the ESL panel.

At low frequencies(below 500 Hz), the dominant parameters are primary winding resistance and inductance. At 0 Hz, the impedance would just be the winding resistance. As frequency increases, the primary inductance increases and thus the load impedance the amplifier sees. The higher the number of primary turns, the higher the inductance. The inductance is also a function of the permiability of the core which itself is a function of flux density(input voltage) and frequency.

At high frequencies(above 2kHz), the dominant parameters are leakage inductance, winding capacitance, inter-winding capacitance, and the capacitive load of the ESL. For typical ESL designs, impedance falls as frequency increases above 2kHz until you reach the resonant frequency between the leakage inductance and the sum of the capacitive loads. At resonance the impedance is usually quite low...less than 1 ohm. Beyond this resonance, the impedance rises again. Most ESL designs put a low value resistor(1 to 4 ohms) in series with the primary to damp this resonance.

Between these two frequency ranges, the impedance is rather high, usually peaking at 100 ohm or more.

The brief comments above are by no means a complete description of the parameters involved in ESL step up transformer design.
They are only meant give you an idea of how complicated the load impedance of an ESL and it's step up transformer are.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.