I'm currently in love with applying a recent technique, which I called "self-screening" secondary technique. I'll try to explain it as simply as possible.
Some of you that are familiar with transformer design probably know that connecting secondary in series for most cases, aggravates the high frequency behavior of transformers. By trial, error and analysis, I've come to the conclusion that capacitance difference from primary to secondary layers causes this, where the higher the capacitance between the series connected secondary is, the higher the Q factor of the resonance and it is usually a dip resonance. Another factors come into significant play, such as the physical distance between these secondary layers, the turn number of the primary sections the secondaries are coupled to, etc.
So in order to have a multisecondary transformer with an uniform and nice high frequency behavior, years ago I began experimenting with shunting screens to shield from that capacitive difference. Now, in some interleaving configurations, it is possible to do so without the screens, but to employ the constantly paralleled secondary layers as screens.
The conditions to do so are to have at least three internal secondary layers and two outer secondary layers. For example, in this project, we have an interleaving pattern of.
SS-P8-SSSS-P4
In order to make into self-screening, we transform it into,
sS-P8-SsSS-P4, where "s" stands for secondary suitable for series connections and "S" is about always parallel.
Another configuration on my 300B OPTs is:
P3-SsS-P6-SsS-P3, where again, the "s" layers are meant for series connections and they represent 1R impedance windings, whereas the outer capital ones represent 4R layers. This gives us the freedom to have 4R, 9R and 16R secondary connections with uniform frequency response.
Here you can compare the FR behavior of identical core, interleaving pattern and almost the same primary impedance (4.6k vs 4.9k) of classic secondary pattern vs self-shielding secondary pattern. Note the dip resonance shifting from 90kHz to 190kHz
A few final words - this technique is even more handy for interstage transformers, where inevitably most secondaries are in series.
Some of you that are familiar with transformer design probably know that connecting secondary in series for most cases, aggravates the high frequency behavior of transformers. By trial, error and analysis, I've come to the conclusion that capacitance difference from primary to secondary layers causes this, where the higher the capacitance between the series connected secondary is, the higher the Q factor of the resonance and it is usually a dip resonance. Another factors come into significant play, such as the physical distance between these secondary layers, the turn number of the primary sections the secondaries are coupled to, etc.
So in order to have a multisecondary transformer with an uniform and nice high frequency behavior, years ago I began experimenting with shunting screens to shield from that capacitive difference. Now, in some interleaving configurations, it is possible to do so without the screens, but to employ the constantly paralleled secondary layers as screens.
The conditions to do so are to have at least three internal secondary layers and two outer secondary layers. For example, in this project, we have an interleaving pattern of.
SS-P8-SSSS-P4
In order to make into self-screening, we transform it into,
sS-P8-SsSS-P4, where "s" stands for secondary suitable for series connections and "S" is about always parallel.
Another configuration on my 300B OPTs is:
P3-SsS-P6-SsS-P3, where again, the "s" layers are meant for series connections and they represent 1R impedance windings, whereas the outer capital ones represent 4R layers. This gives us the freedom to have 4R, 9R and 16R secondary connections with uniform frequency response.
Here you can compare the FR behavior of identical core, interleaving pattern and almost the same primary impedance (4.6k vs 4.9k) of classic secondary pattern vs self-shielding secondary pattern. Note the dip resonance shifting from 90kHz to 190kHz
A few final words - this technique is even more handy for interstage transformers, where inevitably most secondaries are in series.
do you think to have only ine secondary choice ( in my case all are 5 ohm set) , tendentially low Z and leaving the nominal primary Z as usual ( p.e. 3k5 for 300) is good or not?
It will be less complicated and , at the end of the story, maybe you loose some power but in the region where the energy is high ( bass, mid bass) and the modulus of loudspeaker go up but quickly down you can delivery more dynamic power with a better DF
Walter
It is definitely the simplest and easiest form. Having additional secondaries also adds the complexity of commutating them, and I'm not a fan of tapped secondaries, mainly because there is always one inactive winding that wastes space, and I like squeezing available options to the max.
The classic way of doing tapped winding is running two consecutive layers, one for 4R and then one for 0.686R. The sum series of these, ( SQRT(4) + SQRT(0.686) )^2 = 8R. However, when using the 4R only, there is the inactive 0.686R winding.
As for capacitance distribution and parallel connection, classically, both the 4R and the 0.686R windings get paralleled, so you still get that smooth resonance behavior.
In my case, I have both parallel and series windings, 1R and 4R across the coils. That changes the interleaving rules a bit.
In your transformers, if you'd like to optimize the resonance , increasing it further into frequency, you could try capacitance dumping and capacitance equalization.
1. Capacitance dumping is the redirection of high voltage regions more concentrated into the primary to primary layers than primary to secondary.
2. Capacitance equalization means attempting to achieve the smallest capacitance difference between primary and secondary layers. It is arithmetically impossible for a SE transformer on the contrary to a double coil PP transformer. But you can achieve that for a SE transformer to a practically close to useful way, by splitting huge primary packages into smaller effective ones.
The classic way of doing tapped winding is running two consecutive layers, one for 4R and then one for 0.686R. The sum series of these, ( SQRT(4) + SQRT(0.686) )^2 = 8R. However, when using the 4R only, there is the inactive 0.686R winding.
As for capacitance distribution and parallel connection, classically, both the 4R and the 0.686R windings get paralleled, so you still get that smooth resonance behavior.
In my case, I have both parallel and series windings, 1R and 4R across the coils. That changes the interleaving rules a bit.
In your transformers, if you'd like to optimize the resonance , increasing it further into frequency, you could try capacitance dumping and capacitance equalization.
1. Capacitance dumping is the redirection of high voltage regions more concentrated into the primary to primary layers than primary to secondary.
2. Capacitance equalization means attempting to achieve the smallest capacitance difference between primary and secondary layers. It is arithmetically impossible for a SE transformer on the contrary to a double coil PP transformer. But you can achieve that for a SE transformer to a practically close to useful way, by splitting huge primary packages into smaller effective ones.
Uh, sort of, but not exactly, because it is passive in a transformer, whereas the screen is active in the tetr/pent/ode.
I've done Interstage transformers with selected voltage potential screen layers which turned out successful. That is, adding an additional layer between secondary and primary, depending of the goal, biased from another active layer.
Inside a transformer, we cannot entirely remove capacitance, but reduce it and distribute it into domains less offensive for the high frequency response.
In most output transformers, usually 80 to 90% of the capacitance is dominating between the primary and secondary. In my transformers, I like distributing it close to a 1:1 ratio, primary to primary vs primary to secondary.
In high leakage transformers, which can be inevitable for interstage ones, it's usually best to maintain primary to secondary capacitance at zero value. Which is hard to achieve for multi setting winding ratios.
I've done Interstage transformers with selected voltage potential screen layers which turned out successful. That is, adding an additional layer between secondary and primary, depending of the goal, biased from another active layer.
Inside a transformer, we cannot entirely remove capacitance, but reduce it and distribute it into domains less offensive for the high frequency response.
In most output transformers, usually 80 to 90% of the capacitance is dominating between the primary and secondary. In my transformers, I like distributing it close to a 1:1 ratio, primary to primary vs primary to secondary.
In high leakage transformers, which can be inevitable for interstage ones, it's usually best to maintain primary to secondary capacitance at zero value. Which is hard to achieve for multi setting winding ratios.
Last edited:
A lot of overthinking and analysis definitely.
For the frequency response test, a bode plot, 10 and 20kHz square waves are the way to go for me.
To measure effective capacitance is a bit harder and requires a multiple set of tests. First a couple of bode plots to identify the resonant frequencies. Then I execute another set of unloaded secondary bode plots under different source resistances to identify shunt capacitance RC -3dB roll-off. It's important to deliberately aim the roll-offs far away from the identified resonant frequencies, otherwise they change the slope and would bring wrong cutoff frequencies.
For the frequency response test, a bode plot, 10 and 20kHz square waves are the way to go for me.
To measure effective capacitance is a bit harder and requires a multiple set of tests. First a couple of bode plots to identify the resonant frequencies. Then I execute another set of unloaded secondary bode plots under different source resistances to identify shunt capacitance RC -3dB roll-off. It's important to deliberately aim the roll-offs far away from the identified resonant frequencies, otherwise they change the slope and would bring wrong cutoff frequencies.
In my last Projet : 6P36C in Circlotron circuit, I use a usual secondary arrangement,
The résults obtained are good but could be improved,
If I understand well, it‘d be better to connect secondary that are winding togather : bi-wiring in my case
In my schematic, in your maner, it 'll be
4 Ohm :1 and 2 in serie // 3 and 4 // 9 and 10 // 11 and 12
Is it a wong way ?
Yan
The résults obtained are good but could be improved,
If I understand well, it‘d be better to connect secondary that are winding togather : bi-wiring in my case
In my schematic, in your maner, it 'll be
4 Ohm :1 and 2 in serie // 3 and 4 // 9 and 10 // 11 and 12
Is it a wong way ?
Yan
Attachments
Hi Yan,
The 10kHz looks pretty fine to me for this OPT.
Your schematic doesn't look to have more than two internal layers and one external layers, so it is not suitable for the self-screening technique.
However, things are looking pretty good. If your primary Anode potentials start at the outer primary regions (1st and 3rd primary packages) and B+ ends at the 2nd primary package, the connections look pretty fine, as the 4R windings are shunting the highest capacitance regions. Then, the series connected 5,6,7,8 layers are located into less "offensive" regions.
Also, a PP OPT is much more forgiving, when it comes to capacitance difference, compared to a SE OPT.
In this case, I'd leave it as it is.
The 10kHz looks pretty fine to me for this OPT.
Your schematic doesn't look to have more than two internal layers and one external layers, so it is not suitable for the self-screening technique.
However, things are looking pretty good. If your primary Anode potentials start at the outer primary regions (1st and 3rd primary packages) and B+ ends at the 2nd primary package, the connections look pretty fine, as the 4R windings are shunting the highest capacitance regions. Then, the series connected 5,6,7,8 layers are located into less "offensive" regions.
Also, a PP OPT is much more forgiving, when it comes to capacitance difference, compared to a SE OPT.
In this case, I'd leave it as it is.
Did you calculate theoretical Ls and overall shunt capacitance of your transformer. To do that, you'll need.
1. Mean turn Length, preferably each MLT per primary/secondary interface.
2. Layer length of primary and secondary.
3. Turn count
4. Wire diameters.
5. Dielectric thickness and epsilon for each layer.
6. Primaries connection order.
7. Overall number of primary layers.
1. Mean turn Length, preferably each MLT per primary/secondary interface.
2. Layer length of primary and secondary.
3. Turn count
4. Wire diameters.
5. Dielectric thickness and epsilon for each layer.
6. Primaries connection order.
7. Overall number of primary layers.
We built an OT for an SE amplifier with OT100 class A2 impedance Z=6.6K/8 ohm 6P+4S plus a KFB winding of 5% and 10%, frequency band at -1dB 15Hz-42 Khz, primary inductance L=24.6H input 0.15mm, loss inductance 6.3mH R=235 ohm EI96 core section of 20cm* 1.7Tesla, current through primary winding 125mA
Attachments
In such configuration, you already have equal capacitance across all P/S region, so the OPT is good enough. This is inherent for correctly executed PP OPTs. I wouldn't add anything to the interleaving at first sight. Well done, also for the cross-coupling.
Is the resonance a dip or a peak one?
Hi
I's a dip peack
I've made 3 itérations with différent isolation thickness avarage of 0.35 to avarage to 0.5.
Not real improvement.
A Excel sheet for RLC serie calculation give me critical R about 600Ohm (Lf 2mH, Cs 500pF)
For critical R about 100 Ohm I 'd have less than 0.4mH and 2000pF but i dont find the same with OPT DA (Yves Monmagnon) ????
It 's why I'm thinking about 5P- 4S arrangement
Yan
I's a dip peack
I've made 3 itérations with différent isolation thickness avarage of 0.35 to avarage to 0.5.
Not real improvement.
A Excel sheet for RLC serie calculation give me critical R about 600Ohm (Lf 2mH, Cs 500pF)
For critical R about 100 Ohm I 'd have less than 0.4mH and 2000pF but i dont find the same with OPT DA (Yves Monmagnon) ????
It 's why I'm thinking about 5P- 4S arrangement
Yan
It's worth trying to redirect some of this capacitance into the primary region. This will shift the dip in frequency, the tradeoff is adding some series (peaking) resonance. However in an OPT, the odds are in favor of fighting the Q of the peaking resonance, so you have plenty of room for primary to primary capacitance. Here's how to do it.
You take your primary package of 5 layers.
You split it into three parts.
At first, a single layer of primary. Normal winding direction. Right to left
Then, three layers of primary, again normal winding direction. Start from left, three layers, end on the right
Finally, a final single layer of primary, normal winding direction. Left to right
Basically, a P5 package becomes a P1+P3+P1.
Then you connect as follows.
-----------------------------------------------------------------
d---------P5------e
b---------P3------c
b---------P2------a
Anode---P1------a
d---------P4------c
--------------------
SECONDARY
-----------------------------------------------------------------
Then continue to the second primary package the same way, but keep the third primary package continuously wound, as the P/S capacitance there is already very small.
SECONDARY
--------------------
n---------P15-----B+
n---------P14------m
l----------P13------m
l----------P12------k
j----------P11------k
--------------------
SECONDARY
SECONDARY
--------------------
j----------P10-----i
h---------P8------g
f----------P7------g
f----------P6------e
h---------P9------i
--------------------
SECONDARY
SECONDARY
--------------------
d---------P5------e
b---------P3------c
b---------P2------a
Anode---P1------a
d---------P4------c
--------------------
SECONDARY
What this will do is
a) Takes away some of that P/S capacitance that worsens the dip resonance.
b) Redirects this capacitance a little to this primary package, where it is less offensive and transforms itself into peak resonance.
c) Splits the big primary chunk into smaller ones, so less primary package leakage inductance interacts with the P/S interface, resulting in higher frequency resonance.
d) Less capacitive difference between top and bottom primary layers.
Note that capacitance between P4 and P1 , as well as P9 and P6 will increase 9 times, so if necessary, compensate with a thicker dielectric and/or lower epsilon. Capacitance between P3 and P5, as well as P8 and P10 will increase 4 times.
However in this configuration Cps1 decreases from factor 0,934 to 0,589, Cps2 = 0,49 (no change) ; Cps3 decreases from 0,4 to 0,188 ; Cps4 = 0,134 (no change), Cps5 = 0,09 (no change), Cps6 = neg.
An even better solution, but bothersome to wind is the primary snail winding technique. I made that name inspired from a snail shell. The anode connection starts in the middle of the package, then progressively increases. The tradeoff is that this increases each primary to primary capacitance interface by 4 times, however in a S-P-S interleaving configuration, this leaves almost zero P/S capacitance. Here's an example for a 5 layer primary package.
d---------P5------e
b---------P3------c
Anode---P1------a
b---------P2------a
d---------P4------c
You take your primary package of 5 layers.
You split it into three parts.
At first, a single layer of primary. Normal winding direction. Right to left
Then, three layers of primary, again normal winding direction. Start from left, three layers, end on the right
Finally, a final single layer of primary, normal winding direction. Left to right
Basically, a P5 package becomes a P1+P3+P1.
Then you connect as follows.
-----------------------------------------------------------------
d---------P5------e
b---------P3------c
b---------P2------a
Anode---P1------a
d---------P4------c
--------------------
SECONDARY
-----------------------------------------------------------------
Then continue to the second primary package the same way, but keep the third primary package continuously wound, as the P/S capacitance there is already very small.
SECONDARY
--------------------
n---------P15-----B+
n---------P14------m
l----------P13------m
l----------P12------k
j----------P11------k
--------------------
SECONDARY
SECONDARY
--------------------
j----------P10-----i
h---------P8------g
f----------P7------g
f----------P6------e
h---------P9------i
--------------------
SECONDARY
SECONDARY
--------------------
d---------P5------e
b---------P3------c
b---------P2------a
Anode---P1------a
d---------P4------c
--------------------
SECONDARY
What this will do is
a) Takes away some of that P/S capacitance that worsens the dip resonance.
b) Redirects this capacitance a little to this primary package, where it is less offensive and transforms itself into peak resonance.
c) Splits the big primary chunk into smaller ones, so less primary package leakage inductance interacts with the P/S interface, resulting in higher frequency resonance.
d) Less capacitive difference between top and bottom primary layers.
Note that capacitance between P4 and P1 , as well as P9 and P6 will increase 9 times, so if necessary, compensate with a thicker dielectric and/or lower epsilon. Capacitance between P3 and P5, as well as P8 and P10 will increase 4 times.
However in this configuration Cps1 decreases from factor 0,934 to 0,589, Cps2 = 0,49 (no change) ; Cps3 decreases from 0,4 to 0,188 ; Cps4 = 0,134 (no change), Cps5 = 0,09 (no change), Cps6 = neg.
An even better solution, but bothersome to wind is the primary snail winding technique. I made that name inspired from a snail shell. The anode connection starts in the middle of the package, then progressively increases. The tradeoff is that this increases each primary to primary capacitance interface by 4 times, however in a S-P-S interleaving configuration, this leaves almost zero P/S capacitance. Here's an example for a 5 layer primary package.
d---------P5------e
b---------P3------c
Anode---P1------a
b---------P2------a
d---------P4------c
Last edited:
For an overkill primary capacitance dumping, you can try continuing the high potential Anode layers towards the inners of the second primary package, as follows. Here's the idea:
--------------------
SECONDARY
SECONDARY
--------------------
j----------P10-----i
f----------P6------e
d---------P5------e
d---------P4------c
h---------P9------i
--------------------
SECONDARY
SECONDARY
--------------------
h---------P8------g
b---------P3------c
b---------P2------a
Anode---P1------a
f---------P7------g
--------------------
SECONDARY
The good news is that winding starts and directions are the same, so you could try both commutations. In this setting, Cps1 decreases from 0,934 to 0,32. Cps2 goes from 0,49 to 0,25. Cps3 and Cps4 remain the same as the previous arrangement. You have further equilibrium between all Cps factors.
However, capacitance between P7 and P1 becomes at a factor of 0.16, a 36 times increase, compared to a 0.00444 factor from continuously wound primary chunk. Between P3 and P8 = factor of 0.097. Between P4 and P9 = same. Between P6 and P10 = 0.0625.
So it's a good idea to add dielectric thickness / lower epsilon there.
--------------------
SECONDARY
SECONDARY
--------------------
j----------P10-----i
f----------P6------e
d---------P5------e
d---------P4------c
h---------P9------i
--------------------
SECONDARY
SECONDARY
--------------------
h---------P8------g
b---------P3------c
b---------P2------a
Anode---P1------a
f---------P7------g
--------------------
SECONDARY
The good news is that winding starts and directions are the same, so you could try both commutations. In this setting, Cps1 decreases from 0,934 to 0,32. Cps2 goes from 0,49 to 0,25. Cps3 and Cps4 remain the same as the previous arrangement. You have further equilibrium between all Cps factors.
However, capacitance between P7 and P1 becomes at a factor of 0.16, a 36 times increase, compared to a 0.00444 factor from continuously wound primary chunk. Between P3 and P8 = factor of 0.097. Between P4 and P9 = same. Between P6 and P10 = 0.0625.
So it's a good idea to add dielectric thickness / lower epsilon there.