Strange high frequency push pull transformer action

[mook johnson]

As mentioned before I'm working with a design that has a 10MHz push-pull
transmitter to drive a 100 ohm cat6 cable.

The cable is terminated with 100 ohms on both ends and the length is
~200 ft. This is a multi-drop implementation much like RS485 but
needed custom drivers due to the environment.

The driver is configured as a push pull with 5V on the center tap. The
2n7000 transistors are on each leg of the driver and are driven by high
speed logic 5V chips. What I expected to see during transmission on
the centertapped "driver" side of the transformer is a waveform like
this when measured across Q2.

+ _________
+ | | 2X Vin
+ | |
+ | |
+ | Q1 |
+ | on |
+ | |
+ | |
+----------| |------------| |----------- Vin
+ | |
+ | |
+ | |
+ | Q2 |
+ | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================

instead it looks like this where the drain voltage of Q2 drops to ~
1/2Vin when Q1 turns on.
+
+
+
+
+
+----------| |------------| |----------- Vin
+ | | | |
+ | | | |
+ | Q1 | | |
+ | on | | Q2 |
+ |_______| ~1/2 Vin | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================


This of course reduces the amplitude of the "bus" side signal since it
is the difference between Q1 drain and Q2 drain.

Pulses are 30 - 60nS wide.

So instead of the expected +/- 8 - 10V on the bus I am getting +/- ~2

What I've already looked at.

1) the signals look like a current driven push pull push pull power
supply. So I removed the transformer and shorted the two drain
connections together and measured inductance. (singe nH range. < 1 ohm
XL). I measure the CT voltage with a scope and it is rock steady, 10uF
ceramic directly from there to ground to insure that.


2) Changed the frequency down the 5MHz to see if it is a reflection of
the pulse. Signals look identical just 2x wide was expected.

3) checked insertion loss, with an impedance analyzer and it looked
good. < 1dB


4) Q1 was removed and replaced with a ohm ranger (semi flyback style).
and the lower the resistance got, the higher the amplitude seen on the
bus side. until transformer saturation.


There is something going on with that transformer I'm sure of it. It
was designed by a 3rd party that has RF experience and they are
convinced that it is some kind of RF/transmission line matching issue.
I tend to think its more simple than that.



Any thoughts?

 

[Jeroen]

As mentioned before I'm working with a design that has a 10MHz push-pull
transmitter to drive a 100 ohm cat6 cable.

The cable is terminated with 100 ohms on both ends and the length is
~200 ft. This is a multi-drop implementation much like RS485 but
needed custom drivers due to the environment.

The driver is configured as a push pull with 5V on the center tap. The
2n7000 transistors are on each leg of the driver and are driven by high
speed logic 5V chips. [...]
instead it looks like this where the drain voltage of Q2 drops to ~
1/2Vin when Q1 turns on.
+
+
+
+
+
+----------| |------------| |----------- Vin
+ | | | |
+ | | | |
+ | Q1 | | |
+ | on | | Q2 |
+ |_______| ~1/2 Vin | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================

It looks as if the transformer connections got mixed up, with 5V
applied to one end of the primary and the transistor drains on
the center tap and the other end of the primary.

Jeroen Belleman

 

[Spehro Pefhany]

As mentioned before I'm working with a design that has a 10MHz push-pull
transmitter to drive a 100 ohm cat6 cable.

The cable is terminated with 100 ohms on both ends and the length is
~200 ft. This is a multi-drop implementation much like RS485 but
needed custom drivers due to the environment.

The driver is configured as a push pull with 5V on the center tap. The
2n7000 transistors are on each leg of the driver and are driven by high
speed logic 5V chips. What I expected to see during transmission on
the centertapped "driver" side of the transformer is a waveform like
this when measured across Q2.

+ _________
+ | | 2X Vin
+ | |
+ | |
+ | Q1 |
+ | on |
+ | |
+ | |
+----------| |------------| |----------- Vin
+ | |
+ | |
+ | |
+ | Q2 |
+ | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================

instead it looks like this where the drain voltage of Q2 drops to ~
1/2Vin when Q1 turns on.
+
+
+
+
+
+----------| |------------| |----------- Vin
+ | | | |
+ | | | |
+ | Q1 | | |
+ | on | | Q2 |
+ |_______| ~1/2 Vin | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================


This of course reduces the amplitude of the "bus" side signal since it
is the difference between Q1 drain and Q2 drain.

Pulses are 30 - 60nS wide.

So instead of the expected +/- 8 - 10V on the bus I am getting +/- ~2

What I've already looked at.

1) the signals look like a current driven push pull push pull power
supply. So I removed the transformer and shorted the two drain
connections together and measured inductance. (singe nH range. < 1 ohm
XL). I measure the CT voltage with a scope and it is rock steady, 10uF
ceramic directly from there to ground to insure that.


2) Changed the frequency down the 5MHz to see if it is a reflection of
the pulse. Signals look identical just 2x wide was expected.

3) checked insertion loss, with an impedance analyzer and it looked
good. < 1dB


4) Q1 was removed and replaced with a ohm ranger (semi flyback style).
and the lower the resistance got, the higher the amplitude seen on the
bus side. until transformer saturation.


There is something going on with that transformer I'm sure of it. It
was designed by a 3rd party that has RF experience and they are
convinced that it is some kind of RF/transmission line matching issue.
I tend to think its more simple than that.



Any thoughts?

What's the open-circuit primary inductance from either side to CT AND
across both primary windings? (3 numbers).



Best regards,
Spehro Pefhany

 

[Joerg]

As mentioned before I'm working with a design that has a 10MHz push-pull
transmitter to drive a 100 ohm cat6 cable.

The cable is terminated with 100 ohms on both ends and the length is
~200 ft. This is a multi-drop implementation much like RS485 but
needed custom drivers due to the environment.

The driver is configured as a push pull with 5V on the center tap. The
2n7000 transistors are on each leg of the driver and are driven by high
speed logic 5V chips. What I expected to see during transmission on
the centertapped "driver" side of the transformer is a waveform like
this when measured across Q2.

+ _________
+ | | 2X Vin
+ | |
+ | |
+ | Q1 |
+ | on |
+ | |
+ | |
+----------| |------------| |----------- Vin
+ | |
+ | |
+ | |
+ | Q2 |
+ | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================

instead it looks like this where the drain voltage of Q2 drops to ~
1/2Vin when Q1 turns on.
+
+
+
+
+
+----------| |------------| |----------- Vin
+ | | | |
+ | | | |
+ | Q1 | | |
+ | on | | Q2 |
+ |_______| ~1/2 Vin | on |
+ | |
+ | |
+ | |
+ |--------|
+====================================================


This of course reduces the amplitude of the "bus" side signal since it
is the difference between Q1 drain and Q2 drain.

Pulses are 30 - 60nS wide.

So instead of the expected +/- 8 - 10V on the bus I am getting +/- ~2

What I've already looked at.

1) the signals look like a current driven push pull push pull power
supply. So I removed the transformer and shorted the two drain
connections together and measured inductance. (singe nH range. < 1 ohm
XL). I measure the CT voltage with a scope and it is rock steady, 10uF
ceramic directly from there to ground to insure that.


2) Changed the frequency down the 5MHz to see if it is a reflection of
the pulse. Signals look identical just 2x wide was expected.

3) checked insertion loss, with an impedance analyzer and it looked
good. < 1dB


4) Q1 was removed and replaced with a ohm ranger (semi flyback style).
and the lower the resistance got, the higher the amplitude seen on the
bus side. until transformer saturation.


There is something going on with that transformer I'm sure of it. It
was designed by a 3rd party that has RF experience and they are
convinced that it is some kind of RF/transmission line matching issue. I
tend to think its more simple than that.



Any thoughts?

Like Jeroen I believe the most likely cause is that one of the primary
windings is flipped. IOW, instead of connecting a dot side with non-dot
they may have connected two dot sides or two non-dot sides.

Hang a 100ohms resistive load to the output. If still looking as in you
2nd pic I could almost bet it's indeed a half primary flip. As Forrest
Gump said, ..it happens

 

[mook johnson]

What's the open-circuit primary inductance from either side to CT AND
across both primary windings? (3 numbers).



Best regards,
Spehro Pefhany

Its a 1:1 transformer with a CT on the driver side.

Bus side with primary open: 200uH
CT to leg with bus side open: 50uH (either leg)

CT to both legs tied together with bus side open: ~ 5nH

Nothing obviously wrong. Any fancy RF/transmission line stuff could be
taking place here?

 

[mook johnson]

Like Jeroen I believe the most likely cause is that one of the primary
windings is flipped. IOW, instead of connecting a dot side with non-dot
they may have connected two dot sides or two non-dot sides.

Hang a 100ohms resistive load to the output. If still looking as in you
2nd pic I could almost bet it's indeed a half primary flip. As Forrest
Gump said, ..it happens

Ahh that was my first though as well. Forgot the mention I checked for
that.

I guess i should have added more waveform.

If I probe Q1 drain instead of Q2, I get the same result. where Q1 goes
from Vin to ground and goes from Vin to ~ 1/2Vin when Q2 turns on.

Could it be some strange kind of saturation? I don't think so since it
the waveforms immediately look that way. Typically when i have
saturated a core the first part of the waveform looks right then when
the inductance falls the waveform drops off at the end. This one doesn't
do that unless it is at the very start and I can see it.

The wave shape looks pretty good with sharp edges and flat tops so it
doesn't look too bad but your never know.

 

[Spehro Pefhany]

Its a 1:1 transformer with a CT on the driver side.

Bus side with primary open: 200uH
CT to leg with bus side open: 50uH (either leg)

CT to both legs tied together with bus side open: ~ 5nH

Nothing obviously wrong. Any fancy RF/transmission line stuff could be
taking place here?

Schematic of your source termination?



Best regards,
Spehro Pefhany

 

[Joerg]

How did you check? Did you swap connections to one of the primaries and
get the same waveform?

Did you measure inductances with the drains shorted, or the drain
connections on the transformer shorted? If the drain connections on the
transformer are connected together I would expect to only see negligible
leakage inductance!

Yup, I'd double, triple and quadruple-check that. One method is to
ground the center tap, feed a square wave from a function generator into
the secondary and see if opposite polarities come out the primaries.

The very fact that the waveforms just look expanded at 5MHz suggests
there are no saturation or other artefacts.


My money is still on the centre tap isn't really the centre!! Measure
the inductance of each primary, then across the drain connections and
come back here with the result. The inductance across the drain
connections should be 4 x each winding.

If it was saturation the current intake would be high and the 2N7002
FETs would attempt to unsolder themselves.

 

[Jeroen Belleman]

Yup, I'd double, triple and quadruple-check that. One method is to
ground the center tap, feed a square wave from a function generator into
the secondary and see if opposite polarities come out the primaries.



If it was saturation the current intake would be high and the 2N7002
FETs would attempt to unsolder themselves.

Mmmh. There are several things that do not fit my first thought:

One is that if indeed one of the FETs connects to the CT instead
of either end, the remaining end goes below GND, which will
forward-bias the other FET's body diode. He wouldn't have missed
that, I presume.

Second, his inductance measurements appear to make good sense
for a correctly connected transformer. (50uH from CT to either
end and 200uH on the secondary with primary open. 5nH with
both ends of the primary shorted together. Although that would
imply a coupling factor of .9999, which is suspiciously
good.)

Third, in the first waveform diagram, one cannot possibly have
pulse widths of 60ns at a repetition rate of 10MHz. 30ns would
be OK though.

Maybe it's one of those cases where what you think you see is
not actually what is there.

You mentioned your twisted pair cable is doubly terminated.
Do you actually see about 50 Ohms between the wires?

Jeroen Belleman

 

[Joerg]

The amazing thing is that they sometimes succeed, and still survive.

You are old enough to remember the 2N3055, some of them had their TO3
cans adopt a blue tint from gross overheating, like the exhaust pipes of
a well-seasoned Harley-Davidson. Yet many of those were still good for
more rock sessions.

 

[Joerg]

Mmmh. There are several things that do not fit my first thought:

One is that if indeed one of the FETs connects to the CT instead
of either end, the remaining end goes below GND, which will
forward-bias the other FET's body diode. He wouldn't have missed
that, I presume.

Second, his inductance measurements appear to make good sense
for a correctly connected transformer. (50uH from CT to either
end and 200uH on the secondary with primary open. ...

That can also be the case if one of the primary halves is flipped.

... 5nH with
both ends of the primary shorted together. Although that would
imply a coupling factor of .9999, which is suspiciously
good.)

That doesn't sound possible. If it is true then we have a viable Nobel
prize candidate in our midst

Third, in the first waveform diagram, one cannot possibly have
pulse widths of 60ns at a repetition rate of 10MHz. 30ns would
be OK though.

That might have just been a notation error. The plot he described looks
suspiciously like flipped transformer connection. I've had that happen
on the bench when I whipped up a transformer by hand a bit too fast, or
the phone rang while wiring it up. So now I alwats keep a Sharpie in the
bench drawer for marking, same as the Edding pens in Europe.

[...]

 

[Jeroen]

Snip!



That can also be the case if one of the primary halves is flipped.



That doesn't sound possible. If it is true then we have a viable Nobel
prize candidate in our midst

It's not impossible. I have this 1:4 impedance ratio transformer
with a 6-decade bandwidth, which corresponds to a coupling factor
of some 1e-6 below unity. Granted, it's a Guanella transmission
line affair, using ferrite and Vitrovac loaded UT47 coax; not a
wire-wound thing like I suppose we have here.

Jeroen Belleman

 

[[email protected]]

Ahh that was my first though as well.  Forgot the mention I checked for
that.

I guess i should have added more waveform.

If I probe Q1 drain instead of Q2, I get the same result. where Q1 goes
from Vin to ground and goes from Vin to ~ 1/2Vin when Q2 turns on.

.------------->
| .--->
| |
'-'-'-'-'-'
=============
(A) .-.-.-.-.-.-. (B)
| | |
||--' +5v '--||
||<-. .->|| Q2
_||--+ Q1 +--||__
| |
=== ===

Not possible. If the transformer's right, the drain of the "off" FET
has to rise when the other turns on, there's simply no other choice.

If the "off" FET's drain didn't rise to 10v, or spiked then decayed,
etc.--that would be different. For Vds not to rise at all, but fall
instead--that's not possible.

The transformer's wrong. Has to be.

 

[[email protected]]

Mook seems to have gone very quiet.  It would be good to hear the
outcome even if he has to swallow a little humble pie.

It's pretty safe to assume Mook's off trying to untangle those darn
primary connections. Happens to the best of us.

 

[Jamie]

Yup, I'd double, triple and quadruple-check that. One method is to
ground the center tap, feed a square wave from a function generator into
the secondary and see if opposite polarities come out the primaries.





If it was saturation the current intake would be high and the 2N7002
FETs would attempt to unsolder themselves.

yeah well, that would help out in the replacement process

Jamie

 

[[email protected]]

You are old enough to remember the 2N3055, some of them had their TO3
cans adopt a blue tint from gross overheating, like the exhaust pipes of
a well-seasoned Harley-Davidson. Yet many of those were still good for
more rock sessions.

keep seeing those old schematics for transistorized ignition using a
2n3055 pop up now and then, surprised any of them work since it is
only rather for 60V

-Lasse

 

[[email protected]]

It's pretty safe to assume Mook's off trying to untangle those darn
primary connections.  Happens to the best of us.

Betcha he finds this:

.------------->
| .--->
| |
'-'-'-'-'-'
=============
(A) .-.-.-.-.-.-. (B)
| | |
+5v-' ||--' '--||
||<-. .->|| Q2
__||--+ Q1 +--||__
| |
=== ===

That would explain the first waveforms, anyhow.

 

[Joerg]

yeah well, that would help out in the replacement process

Not always: One of the guys in our army unit was building a hobby
electronics project. He sat at the desk in his socks. Something
unsoldered itself, fell down, melted through a sock, lodged itself
between his toes ... phsssss ...

 

[Jamie]

Its a 1:1 transformer with a CT on the driver side.

Bus side with primary open: 200uH
CT to leg with bus side open: 50uH (either leg)

CT to both legs tied together with bus side open: ~ 5nH

Nothing obviously wrong. Any fancy RF/transmission line stuff could be
taking place here?

Shot in the dark:

It could be miller effect due to the capacitance
between the drain and gate..
The opposite side is generating a pulse in an instance and thus the
cap located there is pushing a pulse over to the gate and forces it to
bias on. Of course this is short and shouldn't last long however, at the
frequency you are operating at, it maybe showing the effects if you are
not driving the gate low and hard enough in the off state.

You should probe at the gate of Q1 when Q2 turns on.

Jamie

 

[Joerg]

Betcha he finds this:

.------------->
| .--->
| |
'-'-'-'-'-'
=============
(A) .-.-.-.-.-.-. (B)
| | |
+5v-' ||--' '--||
||<-. .->|| Q2
__||--+ Q1 +--||__
| |
=== ===

That would explain the first waveforms, anyhow.

My bet would be this (asterisk is winding orientation dot):

.------------->
| .--->
|* |
'-'-'-'-'-'
=============
(A) .-.-.-.-.-.-. (B)
|* | *|
| | |
||--' +5v '--||
||<-. .->|| Q2
__||--+ Q1 +--||__
| |
=== ===