Date: Mon, 29 Feb 2016 18:02:50 -0500
From: Markus Vester <markusvester@aol.com>
To: rsgb_lf_group@blacksheep.org
Message-Id: <1532f45159f-67d8-9286@webprd-a39.mail.aol.com>
In-Reply-To: <56D3875A.4060405@abelian.org>
MIME-Version: 1.0
Subject: Re: LF: MF: EbNaut
Content-Type: multipart/alternative; 
	boundary="----=_Part_47631_1717867925.1456786970013"
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Status: O

------=_Part_47631_1717867925.1456786970013
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable


Hi Paul, your explanation is interesting - I've never attempted to look at =
this topic from a time-domain / traveling wave point of view.

I think the usual perception is that the center tap may exhibit voltage at =
even harmonics (mostly 2 f), which needs to be short-circuited to prevent a=
dditional unnecessary drain voltage.  However (especially in broadband HF P=
As), at some frequencies there may be a 2f parallel resonance between the s=
tray inductance and the drain capacitances, leading to a higher drain volta=
ge when the blocking cap (C1) is present.

Another aspect of looking at this is overdriving the PA into switched-mode =
class-D operation. This may either be voltage-mode class D when C1 is prese=
nt (rectangular drain voltages, half-sinusoidal currents), or current mode =
class D with only a large choke L1 (rectangular currents, sinusoidal voltag=
e). The voltage mode version needs to be connected to the load through a se=
ries resonant filter, providing high impedance for odd harmonics, whereas t=
he current-mode class D needs a parallel tank circuit to shunt harmonics.=
=20

Clemens, I think that saturation from DC-biasing would only be a problem if=
 the two transformer primaries (LP1 and LP2) were wound on separate cores. =
On a single toroid, the flux from the two DC currents should cancel.

All the best,
Markus (DF6NM)


-----Urspr=C3=BCngliche Mitteilung-----=20
Von: Paul Nicholson <vlf0403@abelian.org>
An: rsgb_lf_group <rsgb_lf_group@blacksheep.org>
Verschickt: Mo, 29 Feb 2016 12:52 am
Betreff: Re: LF: MF: EbNaut


Attached a circuit of part of the push-pull PA.

The center tap of the primary is at AC ground with the
DC supply coming in via a choke.

I've been getting a lot of ringing on the drains, as is
fairly normal I guess. Even with very low leakage inductance
there are large spikes appearing on the drain when the device
is off.

I took a close look at these spikes and realised they were
transients running down the primary and reflecting off the
AC grounded mid-point. For example there's a spike 40nS wide
occurring 350nS after the FET switches off, then another after
a further 350nS, and then a third, each with diminishing
amplitude. This is clearly a transient propagating back and
forth along the primary. My attempts to suppress this spike
with the usual snubber led to some hot resistors but not much
attenuation.

So over a cup of tea I wondered what would happen if I let
the transient go the full distance of the double primary.

I removed C1 to let the center float at AC.

Now the round trip time of the transient has increased a lot
and it's just arriving back at the drain when the FET switches
back on again. Result: no transient. An unexpected side
effect is that the transformer windings are now cool.

I'm not quite sure how the circuit is working without an AC
ground through C1 but the efficiency has improved.

Has anyone else had trouble with such primary transients?
On the scope they look quite different to the decaying
sinusoid of leakage inductance ringing with drain capacitance,
they appear as a succession of spikes with long intervals
between.
--
Paul Nicholson
--
------=_Part_47631_1717867925.1456786970013
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable

<div style=3D"font-size:10pt;font-family:arial,helvetica,sans-serif;color:b=
lack;"><div id=3D"AOLMsgPart_2_84237fec-12e3-4054-8ddc-f4e407d8cfd2">
<div style=3D"color: black; font-family: arial,helvetica,sans-serif; font-s=
ize: 10pt;"><div style=3D"color: black; font-family: arial,helvetica,sans-s=
erif; font-size: 10pt;">Hi Paul, your explanation is interesting - I've nev=
er attempted to look at this topic from a time-domain / traveling wave poin=
t of view.<br><br>I think the&nbsp;usual perception is that the center tap&=
nbsp;may exhibit voltage at even harmonics (mostly 2 f), which needs to be =
short-circuited to prevent additional unnecessary drain voltage.&nbsp;&nbsp=
;However (especially in broadband HF PAs), at some frequencies there may be=
 a 2f parallel resonance&nbsp;between the stray inductance and the drain ca=
pacitances, leading to a higher drain voltage when the blocking cap (C1) is=
 present.<br><br>Another aspect of looking at this is overdriving the PA in=
to switched-mode class-D operation. This may either be voltage-mode class D=
 when C1 is&nbsp;present (rectangular drain voltages, half-sinusoidal curre=
nts), or current mode class D with only a large choke L1&nbsp;(rectangular =
currents, sinusoidal voltage). The voltage mode version needs to be connect=
ed to the load through&nbsp;a series resonant filter, providing high impeda=
nce for odd harmonics, whereas the current-mode class D needs a parallel ta=
nk circuit to&nbsp;shunt harmonics. <br><br>Clemens, I think that saturatio=
n from DC-biasing would only be a problem&nbsp;if the two transformer prima=
ries (LP1 and LP2) were wound on separate cores. On a single toroid, the&nb=
sp;flux from the two DC currents&nbsp;should cancel.<br><br>All the best,<b=
r>Markus (DF6NM)<br><br><br>-----Urspr=C3=BCngliche Mitteilung----- <br>Von=
: Paul Nicholson &lt;vlf0403@abelian.org&gt;<br>An: rsgb_lf_group &lt;rsgb_=
lf_group@blacksheep.org&gt;<br>Verschickt: Mo, 29 Feb 2016 12:52 am<br>Betr=
eff: Re: LF: MF: EbNaut<br><br></div><br>Attached a circuit of part of the =
push-pull PA.<br><br>The center tap of the primary is at AC ground with the=
<br>DC supply coming in via a choke.<br><br>I've been getting a lot of ring=
ing on the drains, as is<br>fairly normal I guess.  Even with very low leak=
age inductance<br>there are large spikes appearing on the drain when the de=
vice<br>is off.<br><br>I took a close look at these spikes and realised the=
y were<br>transients running down the primary and reflecting off the<br>AC =
grounded mid-point.  For example there's a spike 40nS wide<br>occurring 350=
nS after the FET switches off, then another after<br>a further 350nS, and t=
hen a third, each with diminishing<br>amplitude.  This is clearly a transie=
nt propagating back and<br>forth along the primary.   My attempts to suppre=
ss this spike<br>with the usual snubber led to some hot resistors but not m=
uch<br>attenuation.<br><br>So over a cup of tea I wondered what would happe=
n if I let<br>the transient go the full distance of the double primary.<br>=
<br>I removed C1 to let the center float at AC.<br><br>Now the round trip t=
ime of the transient has increased a lot<br>and it's just arriving back at =
the drain when the FET switches<br>back on again.  Result: no transient.  A=
n unexpected side<br>effect is that the transformer windings are now cool.<=
br><br>I'm not quite sure how the circuit is working without an AC<br>groun=
d through C1 but the efficiency has improved.<br><br>Has anyone else had tr=
ouble with such primary transients?<br>On the scope they look quite differe=
nt to the decaying<br>sinusoid of leakage inductance ringing with drain cap=
acitance,<br>they appear as a succession of spikes with long intervals<br>b=
etween.<br>--<br>Paul Nicholson<br>--<br><br><br>
<br></div>
</div></div>
------=_Part_47631_1717867925.1456786970013--