Break Through - How to make the best OPT

조회 수: 737, 2015-04-03 23:45:24(2015-01-19)

Winding related "bad stuffs" in the Out Put transformer are leakage fluxes and stray capacitances. 

If there is no leakage fluxes in the OPT (Ideal transformer), interleaving or split windings are not needed.

Modified R-core structure (minmized flux path length and maximized diameter with 2 balanced bobbins)  

puts out about 1/10  of the leakage fluxes that of an EI structure of the same power rating . 

Identical twin bobbin plays an important role for the leakage flux reduction.

Next the stray capacitances have to be reduced.

Nothing much you can do with the stray capacitace in the winding itself.

However, the capacitance between primary to secondary winding is lowest without interleaving. 

Thus the best performance is -->  no interleaving.

By reducing   the capacitance between the primary and secondary winding,  the valley and peak in the high end of the OPT frequency response  will be pushed up and often  out of the  F-bandwidth.

In the actual winding, wind 2ndary first to minimize DCR.

On top of this winding, place reasonably thick insulating layer to reduce the capacitance between the windings.

KYJ:

The frequency response of an audio output transformer is determined largely by two factors: primary inductance and product of leakage inductance (L’) and stray capacitance (C’). Increasing the primary inductance would extend low-end frequency response. Reducing the L’C’ product would extend high frequency response of the transformer.

The difficulties in designing good output transformer stem from the fact that any measures that improve one factor of performance leads to detrimental effects on other factors of the performance. To extend low-end frequency response, for example, one must increase the primary inductance. This requires increasing the number of turns as much as possible. However, this leads to increase of both L’ and C’, thereby limiting high frequency response. Also, some efforts to reduce L’ would leads to increasing C’ and vice versa.

Once a designer sets the required low-end frequency response of the transformer, the required primary inductance would be determined. Then rest of the design procedure will be devoted to deducing the L’C’ product to extend high frequency response as much as possible.

As mentioned above, most of the measures to reduce one factor, either L’ or C’, it causes to increase of another factor. For example, if one reduce the number of turns per layer in coil winding, then he must increase the number of layer to accommodate the required number of turns. This would reduce C’ but would increase L’ thereby maintaining the L’C’ product largely unchanged. In a rule of thumb, the L’C’ product depends on the surface area of the winding. More specifically, if one can reduce the length of an average turn, one can reduce both L’ and C’ at the same time.

Here comes the advantage of employing R core: the core with round cross section. Since the cross section of R core is round, the length of average turn becomes minimum. If all other factors remain the same, for example, the same cross sectional area, the same magnetic properties, etc, one can reduce almost 12% of L’C’ product compare with a core with square cross section, regardless of the core is either EI core or C core. If the cross section of the core is the one with 2:1 rectangular shape, the reduction could be as much as 30%. Also, due to absence of sharp edges in its overall core shape, it appears that R core reduces the leakage inductance caused by fringing effect at sharp edges in the core such as EI core.

Reduction of the length of an averge turn would also reduce the DC resistance. Another advantage offered by a R core!

100KHz 방형파 응답 특성이 웬만한 싱글 암프의 10KHz 응답특성과 맞먹네요. 참 대단한 특성입니다. 물론 상당량의 NFB를 걸어주어 나온 특성이기는 하나 출력트랜스가 좋지 않으면 NFB를 걸어주어도 좋은 특성이 나오지 않습니다. 무궤환의 싱글 암프 대다수의 10KHz 방형파 응답 특성이 이 암프의 100KHz 방형파 특성에 못미칠 것 같습니다.

--------

예, 처음 특성쟀슬때 X-time scale 이 잘못 됐나 확인해 보았습니다.
Core 자로 단축하고 직경 늘려서 권선수를 반이상 감소해도 충분한 저역특성이 얻어 집니다. 

KYJ 님 저역기준은 30 Hz 에서 1-2 db 내려가는 정도 - 권선수를 이 이상으로 하면 Amp x turn 이 커저서 포화가 속히옵니다. 많이 감으면 저역 Bandwidth 는 내려가지만 최대 출력이 줄고 Speaker 특성이 저역 재생을 못 하면 별 의미가 없습니다. Speaker 를 울릴려고 Amp 를 만드는 것입니다.