Definitely not for low frequency response, as core excitation is very low. At high frequency, what behavior change should I expect with higher, realistic voltages, frequency response wise? Dielectric absorption will kick-in? More distortion?
This transformer is intended for SE amplifiers.
But P-P, especially class AB is a different story and definitely not simpler. Not only one tubes sees a quarter of the impedance, but leakage inductance between halves should be minimized as well.
Yes..antenna design before the use of modern day FEA software was a blistering migraine.. Hand calculations could take 20 to 30 pages of de'Broglie wave equations...curl and divergence matrices of near and far fields...ect..ect.. All starting from Maxwell's equations the first time you do this... Then lots and lots of field testing with antennae directivity patterns...this data is then used to modify your design equations to make them more accurate ..... Then you have the guys that design wave guides for satellite communications with Traveling Wave tubes....that's altogether another can of worms ....
Last edited:
This is what happens, when you neglect details. You have to learn it the hard way.
No worries, we'll take it from the humorous point of view! Pardon me for the reaction, it was just a reflex kick due to people who love to distort basics and claim them to be their magic realm for the sake of building a larger egocentrism. There's nothing wrong in hearing differences between cables and wooden blocks, I do that as well. Just don't try to explain it with "direction of electron flow".
Oh.....my bad !!!! I ran my mouth before reading in more detail... I saw the words WILLIAMSON and then assumed...LOL So this is a Single Ended ???? Very interesting..... Now I am starting to see the light.....LOL One thing I preach that many refuse to believe, is that capacitors are the most NON-Linear creatures in an amplifier equal with tubes..... Transformer capacitance is subject to a huge changes depending on the DC voltage applied as well as the magnitude of AC signal superimposed on the DC.. I have many years of plots to show this... As the AC voltage is increased, the di-electric constant changes, thus increases capacitance with larger applied signals....there is a also a frequency dependency as well... Some materials such as TEFLON have a negative coefficient for di-electric constant....especially with temperature..
Last edited:
Now this is some golden knowledge... But it's another question if you're willing to share it, haha
Now this is some golden knowledge... But it's another question if you're willing to share it, haha
Build the amp...
Sweep the frequency response of the amplifier at low power, middle power and full power output...
Look at your BODE plot for your -3dB corners and work the math backwards and wa-la you have the delta in your capacitance....since leakage inductance is fixed based on geometry..
Inductance will vary but not that much due to the significant gap for SE...
I wasn't referring to the Williamson amplifier but to the technique used in the output transformers which is valid for both PP and SE. In reality the Williamson solution is not a good one if 8R and 4R are the desired secondary loads.
Alex would probably be better off following the solutions proposed by Radford and Quad in the 50ies however efficiency and frequency response will never be exactly the same for all taps. Nonetheless this is a good compromise for commercial purposes because ins't too complicated.
Briefly, Quad and Radford had 4 secondary windings of n turns each (this could be 1 layer or, better, multiple layers in parallel). The combination for 4R, 8R (really 9R) and 16R where:
1) 2 groups in parallel each compsed by 2 windings in series in series for 4R (this was called 2n)
2) 2 windings in series forming a 2n and then the other 2 in parallel between each other and in series with the 2n for 8R (this was referred as 3n).
3) all in series for 16R (called 4n).
This is actually the one used by Sowter if I am not wrong.
Obviously the nominal 8R connection is not ideal as the other two (that will give the same results each with its own load) having about 10% more losses in terms of Rdc and a bit more losses at high frequency because of the non-uniform distribution of the m.m.f among the different windings. But nonetheless is good commercial solution. Surely better than simple taps where increase in losses is typically as bad as 25-30%....
Is there an ideal (or close to ideal) solution at all? Yes but only if one sticks to two impedances, say 8R and 4R that are the most common. It's a bit more complicated but not too much. The whole point is in getting the same current circulating in all the turns. This means that for a target primary impedance not every combination of primaries + secondaries will work. Actually if there is one is already a win! Non-negotiable conditions are: 1) maximizing the use of the winding volume to minimize copper losses and leakage and 2) maintaing equal number of turns for all all the secondary windings (otherwise bye bye extended bandwidth).
The general rule is to have the possibility of forming a number of 2 different sub-groups of EQUAL secondary turns that will be all in parallel for the final connection using all the turns. These 2 sub-groups are in a ratio between them that falls between 4/3 and 3/2. The former will give an impedance ratio of about 1.78 ( 7.1R and 4R, for example) and the latter 2.25 (9R and 4R). The 3/2 combination is that of the Radford/Quad scheme. This means that secondary windings have to be first combined by layers rather than by entire windings.
This is easier to make if I have at least 6 primaries and 5 secondaries or more! But is also possible (but not every time) where I have 5 primaries and 4 secondaries and even 4 primaries and 3 secondaries.
Why 4/3 to 3/2 interval? Because the secondary turn ratio for the 4R and 8R secondaries is SQRT(2) and those two ratios are two practical close limits. In fact 4/3<SQRT(2)<3/2.
So all one needs to do is make the secondaries, including the number of layers, where I can I achieve a ratio the interval 4/3-3/2 that use all the turns. The higher the number of windings the closer I can get to the the ideal case of SQRT(2).
On the other side the number of windings WILL be limited by the copper loss as I do not want to waste too much space for insulators....as always there are compromises (or best balance if you will).
Last edited:
There is way too many assumptions made here..... First establish, at the least, a plate load and desired bandwidth.... Then derive your target design numbers....ie Min Inductance, max Leakage, and Max capacitance..
An optimum winding geometry can then be calculate based on these values...
Execution of winding geometry should take into account PROXIMITY losses ....Multiple paralleled windings tend to have cross currents due to Proximity affects... In some cases I have employed winding transpositions to equalize these AC losses.... Ideal cases of SQRT (2) are not needed... since the plate load in reality is a moving target due to speaker impedance curve, Don't expect your 8 Ohm and 4 Ohm speakers to have the exact curves or same damping...so plate loads can move a little depending on tap used...Tube to tube plate resistance isn't that tight either....so no need getting OCD about the plate load....
An optimum winding geometry can then be calculate based on these values...
Execution of winding geometry should take into account PROXIMITY losses ....Multiple paralleled windings tend to have cross currents due to Proximity affects... In some cases I have employed winding transpositions to equalize these AC losses.... Ideal cases of SQRT (2) are not needed... since the plate load in reality is a moving target due to speaker impedance curve, Don't expect your 8 Ohm and 4 Ohm speakers to have the exact curves or same damping...so plate loads can move a little depending on tap used...Tube to tube plate resistance isn't that tight either....so no need getting OCD about the plate load....
Too many? Where?
There are only two assumptions (I call them non-negotiable conditions) because they are essential for max bandwidth: 1) Use all the winding volume and 2) secondary groups always must have equal turns. The target is to have equal current in all turns.
What you list is the usual starting point which is valid if one secondary (or the cheap taps). For a 8R/4R secondary in the 4/3 to 3/2 range (that is surely not ideal) few times works and most times needs to be modified....2 or more iterations....
I do not expect 4R speakers and 8R to be the same I expect the transformer to work as specified as he will be selling a transformer with defined specs. Your idea of rough approximation is valid for inexpensive transformers. Indeed it is exactly the way it is in reality. If you buy a Monolith Magnetic there is no or very little approximation between the two impedances....
There are only two assumptions (I call them non-negotiable conditions) because they are essential for max bandwidth: 1) Use all the winding volume and 2) secondary groups always must have equal turns. The target is to have equal current in all turns.
What you list is the usual starting point which is valid if one secondary (or the cheap taps). For a 8R/4R secondary in the 4/3 to 3/2 range (that is surely not ideal) few times works and most times needs to be modified....2 or more iterations....
I do not expect 4R speakers and 8R to be the same I expect the transformer to work as specified as he will be selling a transformer with defined specs. Your idea of rough approximation is valid for inexpensive transformers. Indeed it is exactly the way it is in reality. If you buy a Monolith Magnetic there is no or very little approximation between the two impedances....
Being OCD about tubes is a joke... Consistently find me a 1% tube to match the data sheets and then you have a valid point about being OCD about the plate load... There is no plate load...it's exist in fantasy and on paper... The speaker is all over the place not to mention the speaker is an elliptical load ..brand A to brand B 8 ohm speakers will not even be the same... You can't define transformer specs as a single constant unless you provide a graph of the Inductance over signal level and provide capacitance graph over signal levels...since these are variables over operating conditions..You could however specify the leakage as well distortions at specific frequencies and operating levels..
I never mentioned "rough" approximations of plate load....just a slight difference, say 4K load at 8 Ohms and a 3.92K at 4 Ohms would be an example... I design, wind and sell very wide band transformers to customers all over the world, but refuse deal with customers who just don't get it...
Even if you have equal turns in your sections...they are not equal in the AC sense due to differences in PROXIMITY effect at different layers...so you will still encounter cross currents in the section with equal turns... Need different solutions to solve that issue..
There ar times when I do not have a full window fill...it optimizes that way do to using wire gauges that meet the skin depth....larger wires to fill the last percentage of the window in some cases would hurt the design...
I never mentioned "rough" approximations of plate load....just a slight difference, say 4K load at 8 Ohms and a 3.92K at 4 Ohms would be an example... I design, wind and sell very wide band transformers to customers all over the world, but refuse deal with customers who just don't get it...
Even if you have equal turns in your sections...they are not equal in the AC sense due to differences in PROXIMITY effect at different layers...so you will still encounter cross currents in the section with equal turns... Need different solutions to solve that issue..
There ar times when I do not have a full window fill...it optimizes that way do to using wire gauges that meet the skin depth....larger wires to fill the last percentage of the window in some cases would hurt the design...
Last edited:
Clearly you have never done it. That is certain. If things are done right you get EXACTLY the same low frequency performance and the same high frequency response. Buy and test a Monolith Magnetic on 4R and 8R, please!!!!
Selling a transformer as top quality with the approximation used in a cheaper transformer is a joke! Would you buy a Ferrari with the internals of a common saloon car?? I wouldn't.
And that's what you get if the ratio is in the 4/3 to 3/2 range. Do some maths please. You don't get it sorry. It's not rocket science and it doesn't mean lot of complications....
I know how to calculate skin effect and proximity effect and what they do. Don't forget that if the wire gets too large Rac increases quickly.....Don't need lessons on how to make a transformer.
There is really nothing to discuss here because there is nothing new. I have only given a general idea because I know Alex can do the calculations and design on his own. It's just a SLIGHTLY different approach because of two secondary impedances instead of one. And it's mostly done on paper without winding anything if one knows what he is doing....
50AE, your SE application would effectively have one primary winding end and one secondary winding end at signal ground, as well as the transformer core. Is that how you set up your measurement scheme?
That would be interesting to see (ie. a photo) - the only info I've come across on the Savage OT for Williamson amp is the advert for 3C67A in the WW reprint collection of Williamson articles. Were the wedges causing the middle leg of the E lams to bulge out so that the outer curve of the middle core leg better matched a curved former (rather than have a pure square profile former) ?
That would be interesting to see (ie. a photo) - the only info I've come across on the Savage OT for Williamson amp is the advert for 3C67A in the WW reprint collection of Williamson articles. Were the wedges causing the middle leg of the E lams to bulge out so that the outer curve of the middle core leg better matched a curved former (rather than have a pure square profile former) ?
Yes... I am pretty sure that is the transformer I was referring to... Have not seen it in many years... I have seen it in person...my professor was cleaning out the university lab and wanted to give it to me as a gift, cause he knew how much I liked building Williamson amps at that time... He got sick and did not return back top the University so I never really knew what happened to the OT...
Last edited:
.. otherwise the m.m.f gets unevenly distributed across the sections, due to uneven voltage drop across the windings, correct?
I was thinking, if one wishes to go really fancy about it, why not wind a layer of thinner wire on top of the secondary and parallel connect it, to equalize Rdc with the secondary of the bottom. A thinner wire should fit nice between the gaps of the thicker round wire secondary. Of course one might use thicker wire, but I'm not sure if the commercial availability will give so much freedom.
About proximity effect, is this the main limitation of HF power output in transformers? How evil it is? Is the use of litz wire justified then, although its lesser filling factor?
Your posts made complete sense and thanks! Right now I'm winding a transformer with 3/2 subgroups for 4R and 8R connection. Patrick Turner has diagrams on his website, but shamefully, I missed these.
Last edited:
Alex you only need Litz wire if plan to use diameter above 0.5 mm and want no less than 100 KHz bandwidth, roughly speaking.
If you look at the tables of Litz wire manufacturers you will see the "equivalent" you need for a certain target bandwidth and low loss. If you exaggerate with the number of strands you are going to get more Rdc. As always is a matter of balance between opposite requirements.
The 3/2 scheme will give you 4R/9R or 3.56R/8R. Given the typical loudspeakers impedance for nominal 8R and 4R I would design on the 8R secondary and accept a bit lower impedance on the nominal 4R as the higher impedance has typically more variation....
If you look at the tables of Litz wire manufacturers you will see the "equivalent" you need for a certain target bandwidth and low loss. If you exaggerate with the number of strands you are going to get more Rdc. As always is a matter of balance between opposite requirements.
The 3/2 scheme will give you 4R/9R or 3.56R/8R. Given the typical loudspeakers impedance for nominal 8R and 4R I would design on the 8R secondary and accept a bit lower impedance on the nominal 4R as the higher impedance has typically more variation....
Last edited:
I meant that if you want 5K, for example, then do the secondary as 5.6K-9R or 25:1 ratio anyway. The reason is that most 4R speakers have minimum Z above 2.7-2.8R while there are plenty of 8R rated speakers that go down to 4R.
I have seen someone using Litz also for primaries in high power amps but I have never do amps powerful enough to require it. For secondaries requiring large wires, yes. The latter happens more frequently.
Multifilar you can use as well if you already have it and the bonding doesn't eat too much space.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.