I'm seriously contemplating winding some SE transformer for a couple of big projects. Though both of the projects I'm considering are triode-based so that the tap is not an issue, I'm wondering how to place the ultralinear tap with an interleaved construction fpr optimal coupling. Placement at 50% is a no-brainer, but what about 40% or other arrangements?
wrenchone said:
A conventional practice of good workmanship in making transformers is to place any taps at the end of a layer---
if you pull a tap from say the middle of a layer--- your going to have to wind over the wire exiting and re-entering that layer--- to the spot where the tap was pulled. This puts some mechanical stress on the wire--- and to prevent shorts these leads exiting and re-entering the layer are typically covered with some insulation to protect them electrically as well--- the achilles heel of this approach is that it also leads to increased leakage inductance.
So--- it is generally considered best (unless there would be some vastly overriding requirement) to pull any taps at the beginning or end of a layer.
Hope this helps,
MSL
Crossover insulation is not so much of a problem in terms of increased leakage when the wire in question is smaller than 30 AWG. One could also use tape with 1 mil (0.025 mm) thickness for crossover duty without unduly compromising the insulation.
The point may be moot, as the primary sections will have to be multiple layers to accomodate the required number of turns to satisfy requirements for sufficient primary inductance and bias current capability. I'll have a wide range of tap options by pulling a tap off of one of the intermediate layers.
I was looking at a EI 1 1/8" lamination, with a 1.5" thick stack (standard bobbin). For a transformer with 150 mA peak capability and 50 H inductance, this comes out to about 5000 primary turns. I will have to review what other transformer makers offer in that size, as I was going by the rule of thumb of making the primary impedance to equal the load impedance at 20Hz.
The point may be moot, as the primary sections will have to be multiple layers to accomodate the required number of turns to satisfy requirements for sufficient primary inductance and bias current capability. I'll have a wide range of tap options by pulling a tap off of one of the intermediate layers.
I was looking at a EI 1 1/8" lamination, with a 1.5" thick stack (standard bobbin). For a transformer with 150 mA peak capability and 50 H inductance, this comes out to about 5000 primary turns. I will have to review what other transformer makers offer in that size, as I was going by the rule of thumb of making the primary impedance to equal the load impedance at 20Hz.
wrenchone said:
You might want to give that rule of thumb some thought. What you're effectively saying is that at 20Hz the reactance of the transformer's primary inductance draws the same current as the load current, but in quadrature. Or, to put it another way, the output stage has to supply 1.414 times as much current. That will increase the distortion of the output valves. Further, the increased current in the transformer takes you closer to saturation, and at saturation, there is no primary inductance. All in all, you need to be very careful as you approach saturation.
Now, you're in the position of knowing what your load (loudspeaker) is. If it's a bookshelf loudspeaker, it's probably a closed box with a resonant frequency of 60-80Hz. If so, its impedance at 20Hz is going to be quite close to the DC resistance of the voice coil and all of the previous considerations apply. On the other hand, if your loudspeaker is a big reflex, it might have its port tuned to 30Hz, making the impedance at 20Hz quite a bit higher than the DC resistance of the voice coil, and leaving the load on the output valves at 20Hz simply the output transformer primary inductance.
So, if you've got a big reflex box, then your rule of thumb is probably fine, but you might want to rethink if you have a small closed box.
hey-Hey!!!,
You can put in more than one tap too. Go for ~20% and around 40%. The end-of-layer placement is a good idea, and exactly where your taps go will depend on how many layers you have to build. With 5k turns, I suspect you'll have more than 10 layers so the increment will be at most 10%.
cheers,
Douglas
You can put in more than one tap too. Go for ~20% and around 40%. The end-of-layer placement is a good idea, and exactly where your taps go will depend on how many layers you have to build. With 5k turns, I suspect you'll have more than 10 layers so the increment will be at most 10%.
cheers,
Douglas
Yvesm said:Can't believe your are suggesting that we must target a larger primary inductance because small loudspeakers are unable to correctly load the amp (nor providing any sound at lowest frequencies)
What about a low cut somewhere, preferably before the OPT
Happy April 1st.
Hello Yves, I take your point that it seems a contradiction to need more primary inductance for a loudspeaker that's unable to reproduce the frequencies where that primary inductance matters, but think about it, as we enter saturation, we generate third harmonics, and the little loudspeaker is certainly capable of reproducing them. On the other hand, perhaps that will give an illusion of bass from shoebox loudspeakers.
I would certainly prefer to see a high pass filter so that the amplifier only has to cope with signals that it and the loudspeaker can handle, but that's probably heresy.
If I'd realised it was April 1st, I would have come up with something even more surprising. Still, look at it another way, largish reflex loudspeakers will work better than you would expect with a given primary inductance. I'd always felt that reflex loudspeakers suited valve amplifiers very well, and had previously assumed it was simply due to their highish output resistance, but it could also be the kinder loading at low frequencies.
From the practical side of it. a One Electron transformer of 4800k impedance using the same size laminationss has about 40H primary inductance. One has to draw the line between the benefits of having higher primary inductance and pumping up the leakage inductance too much, especially as the leakage also varies as T^2.
So far, I'm looking at buying strings of lams for EI75, EI100, EI112, and EI150. The smaller sizes would be used for applications like interstage transformers or load inductors, and the larger lams for brutal output iron. It all depends on if I can find a distributor willing to stock 6 mil lams in the sizes I need. It would have been a lot easier 10-20 years ago.
So far, I'm looking at buying strings of lams for EI75, EI100, EI112, and EI150. The smaller sizes would be used for applications like interstage transformers or load inductors, and the larger lams for brutal output iron. It all depends on if I can find a distributor willing to stock 6 mil lams in the sizes I need. It would have been a lot easier 10-20 years ago.
Just so folks remember, it's the excitation current that contributes to saturation in a transformer, not the load current, as the ampere-turns represented by the load are equal and opposite on primary and secondary to a fair degreee of precision (assuming the leakage induxctance is not way. way out of bounds). This is true for both gapped and non-gapped tramsformers, it's just that the energy storage and the excitation current is much larger with the single-ended transformer due to the presence of the gap.
I've been doing preliminary estimates for turns and gap by reconciling the two equations B = (0.4 pi X Np XIp)/Lg and
L = (0.4 pi X Np^2 X Ae X10^-8)/Lg. I use the first equation to get a figure based on peak excitation current and peak flux density (I'm using 15-16 kG, right before the hysteresis loop of silicon steel turns the corner toward saturation. This is then substituted into the second equation to get the required turns for a given inductance.
It's true that the quadrature/excitation current will increase the distortion for low frequency, high level operation, but that is the price that one pays for using a transformer in the first place. Most of the commercial desgns I've seen for SE transformers seem to peter out at about 50H, probably due to the tradeoff between turns/size and leakage inductance. I'll have to see if I can get some figures for some of the higher priced transformers out there, as the ones I have are for lower priced units like the One Electron and Hammond. Still, there's only so much one can do with copper and silicon steel, which has the highest flux capability of any of the common magnetic materials. The only way one could do better would be to somehow get hold of some supermendur cut cores and rob a bank to pay for them...
I've been doing preliminary estimates for turns and gap by reconciling the two equations B = (0.4 pi X Np XIp)/Lg and
L = (0.4 pi X Np^2 X Ae X10^-8)/Lg. I use the first equation to get a figure based on peak excitation current and peak flux density (I'm using 15-16 kG, right before the hysteresis loop of silicon steel turns the corner toward saturation. This is then substituted into the second equation to get the required turns for a given inductance.
It's true that the quadrature/excitation current will increase the distortion for low frequency, high level operation, but that is the price that one pays for using a transformer in the first place. Most of the commercial desgns I've seen for SE transformers seem to peter out at about 50H, probably due to the tradeoff between turns/size and leakage inductance. I'll have to see if I can get some figures for some of the higher priced transformers out there, as the ones I have are for lower priced units like the One Electron and Hammond. Still, there's only so much one can do with copper and silicon steel, which has the highest flux capability of any of the common magnetic materials. The only way one could do better would be to somehow get hold of some supermendur cut cores and rob a bank to pay for them...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.