DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=mime-version:date:message-id:subject:from:to:content-type;
        b=MDYKmkCEq5AVzLvoBYFsTK/naeOwiL5eTkJyzgC+SLZuVB8LrKPxwxIZijigldw6lB
         vJ7bkkydKSj6feS0juqR5McTegSa5/2x18/V8e9zFwR4VsvcjOcD2FQGHBDYGd3j/Xa0
         XZ33xfUwyyNgODyFGsDtL9hbMkqr7WGFweTpQ=
MIME-Version: 1.0
Date: Mon, 10 Jan 2011 14:22:23 +0000
Message-ID: <AANLkTinHC6VEy9e4OofWJRGxsiDRhhUe7xP=D-Gfx8Pf@mail.gmail.com>
From: Andy Talbot <andy.g4jnt@gmail.com>
To: rsgb_lf_group@blacksheep.org
DomainKey-Status: good (testing) 
Subject: LF: Capacitor warning and First 500kHz Class E breadboard
Content-Type: multipart/alternative; boundary=90e6ba4fc4c658a29304997eb28f
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org


--90e6ba4fc4c658a29304997eb28f
Content-Type: text/plain; charset=ISO-8859-1

A warning to anyone contemplating making power amps - especially class E
types.
Don't even think about using the small high voltage ceramic caps availalble
at low cost from the suppliers - and sold as "suitable for SMPSUs etc"  in
resonant circuits
They are appalling, terrible, revolting things.  You only have to breath on
them and they change their value.wildly

These things    http://www.g4jnt.com/Download/UselessCaps.jpg

I've just  made up a first version of a breadboard for a 500kHz Class E
amplifier, (target design 500W from 50 V supply),  calculated the L/C values
as per Alan's spreadsheet, combined with an L matched output for 50 ohm.
For the capacitors I aimed to get within 5% of the calculated values using
many of these small ceramic objects in parallel - lots in || should share
the current, shouldn't they, which would be good?

So I soldered them up - used the G4HUP L/C meter to measure the values - at
which point I should have been a bit suspicious as the measured C was
sometimes a fair bit lower than it should have been.  At the time I just put
it down to "finger trouble" - one does tend to ignore little things like
that sometimes.
Anyway, made up the amp, used a 12V supply rail and monitored the switching
waveforms, expecting to see something vaguely approaching the half sine on
drains etc.

Results, terrible, nowhere near what they should have been - clearly summat
was very wrong  I remeasured the || combination of the caps - after
unsoldering they were sometimes as much as half the value they should have
been  - until they cooled right down and approached room temp. Occasionally
a cap value approaching correct would show up on the meter once in a
while - but that was was probably just coincidence :-)
And yes, I did double, triple and quadruple check the meter against good
caps - it was OK every time.

These little blue things appear to have a huge temperature coefficent - are
completely useless, designed to do nothing of value, and all have now gone
into the waste bin.
In fact, I was wondering if they were random-capacitors.

I did have a just two trusted 22nF caps in the junk box that were within 20
- 30% of the right values for the amp and an awful lot of 3.3nF 1700V things
of a more respectable heritage, so just put these into the circuit, tested
at 12V and got waveforms almost correct.   Increased drive freq to 515kHz
(no real scope for changing cap values at this stage, so freq change would
be easier to try with) where the waveform looked best, and increased supply
volts.      Got to 30V and it was delivering 150 Watts; 40V and was getting
around 350 Watts at which point I assumed it was going to work, so switched
off for a break.   Wasn't going to tempt fate by gong to the full 50V having
not even bothered to tune things up properly
A picture of that breadboard can be seen at
http://www.g4jnt.com/Download/ClassE_BrdBd01.jpg

The yellow cap shunting the device was running quite warm - which is hardly
surprising just a single poly[whatever] carrying all those amps - and the
one in series with the tank was slightly warm.    So, now need to browse the
Farnell catalogue (the cubic printed paper version is easier) for some caps
that will do the job properly - like more of those yellow ones.  Actually,
may have enough of the 3.3nF ones, but that would end up with lots in
parallel

So, for anyone who is contemplating a class E amplifier design:

Get close to the calculated component values and it will probably
mostly-work immediately.
Play with the drive frequency while monitoring waveforms  to see which way
tuning needs to go
Don't use cheap high voltage ceramic capacitors
 Don't use cheap high voltage ceramic capacitors

They would work well in a random number generator tho.....

Andy
www.g4jnt.com

--90e6ba4fc4c658a29304997eb28f
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div>A warning to anyone contemplating making power amps - especially clas=
s E types. </div>
<div>Don&#39;t even think about using the small high voltage ceramic caps=
 availalble at low cost from the suppliers - and sold as &quot;suitable fo=
r SMPSUs etc&quot;=A0=A0in resonant circuits</div>
<div>They are appalling, terrible, revolting things.=A0 You only have to=
 breath on them and they change their value.wildly</div>
<div>=A0</div>
<div>These=A0things=A0 =A0 <a href=3D"http://www.g4jnt.com/Download/Useles=
sCaps.jpg">http://www.g4jnt.com/Download/UselessCaps.jpg</a></div>
<div>=A0</div>
<div>I&#39;ve just =A0made up a first version of a breadboard for a 500kHz=
 Class E amplifier, (target design=A0500W from 50 V supply),=A0 calculated=
 the L/C values as per Alan&#39;s spreadsheet,=A0combined with an L matche=
d output for 50 ohm.=A0 For the capacitors I aimed to get within 5% of the=
 calculated values using many of these small ceramic objects=A0in parallel=
 - lots in || should share the current, shouldn&#39;t they, which would be=
 good?=A0=A0</div>

<div>=A0</div>
<div>So I soldered them up - used the G4HUP L/C meter to measure the value=
s -=A0at which point I=A0should have been a bit suspicious as the measured=
 C was sometimes a fair bit lower than it should have been.=A0 At the time=
 I just put it down to &quot;finger trouble&quot; - one does tend to ignor=
e little things like that sometimes.=A0=A0=A0 </div>

<div>Anyway, made up the amp, used a 12V supply rail and monitored the swi=
tching waveforms, expecting to see something vaguely approaching the half=
 sine on drains etc.</div>
<div>=A0</div>
<div>Results, terrible, nowhere near what they should have been - clearly=
 summat was very wrong=A0 I remeasured the || combination of the caps - af=
ter unsoldering=A0they were sometimes as much as half the value they shoul=
d have been=A0 - until they cooled right down and approached room temp. Oc=
casionally a cap value approaching correct would show up on the meter once=
 in a while=A0-=A0but that was=A0was probably just coincidence :-)=A0 </di=
v>

<div>And yes, I did double, triple and quadruple check the meter against=
 good caps - it was OK every time.</div>
<div>=A0</div>
<div>These little blue things appear to have a huge temperature coefficent=
 - are completely useless, designed to do nothing of value, and all have=
 now gone into the waste bin.</div>
<div>In fact, I was wondering if they were random-capacitors.=A0=A0</div>
<div>=A0</div>
<div>I did have a=A0just two=A0trusted 22nF caps in the junk box that were=
 within 20 - 30% of the right values for the amp and an awful lot of 3.3nF=
 1700V things of a more respectable heritage, so just put these into the=
 circuit, tested at 12V and got waveforms almost correct.=A0=A0 Increased=
 drive freq to 515kHz (no real scope for changing cap values at this stage=
, so freq change would be easier to try with) where the waveform looked be=
st, and=A0increased supply volts.=A0=A0=A0=A0=A0 Got to 30V and it was del=
ivering 150 Watts; 40V and was getting around 350 Watts at which point=A0I=
=A0assumed it was going to work,=A0so switched off for a break.=A0=A0=A0Wa=
sn&#39;t going to tempt fate by gong to the full 50V having not even bothe=
red to tune things up properly</div>

<div>A picture of that breadboard can be seen at=A0=A0 <a href=3D"http://w=
ww.g4jnt.com/Download/ClassE_BrdBd01.jpg">http://www.g4jnt.com/Download/Cl=
assE_BrdBd01.jpg</a>=A0 </div>
<div>=A0</div>
<div>The yellow cap shunting the device was running=A0quite warm - which=
 is hardly surprising just a single poly[whatever] carrying all those amps=
 - and the one in series with the tank was slightly warm.=A0=A0=A0 So, now=
 need to browse the Farnell catalogue (the cubic printed paper=A0version=
 is easier) for some caps that will do the job properly - like more of tho=
se yellow ones.=A0 Actually, may have enough of the 3.3nF ones, but that=
 would end up with lots in parallel</div>

<div>=A0</div>
<div>So, for anyone who is contemplating a class E amplifier design:</div>
<div>=A0</div>
<div>Get close to the calculated component values and it will probably mos=
tly-work immediately.</div>
<div>Play with the drive frequency while monitoring waveforms =A0to see wh=
ich way tuning needs to go</div>
<div>Don&#39;t use cheap high voltage ceramic capacitors</div>
<div>
<div>Don&#39;t use cheap high voltage ceramic capacitors</div>
<div>=A0</div>
<div>They would work well in a random number generator tho.....</div></div=
>
<div>=A0</div>
<div>Andy</div>
<div><a href=3D"http://www.g4jnt.com/">www.g4jnt.com</a></div>
<div>=A0</div>
<div>=A0</div>
<div>=A0</div>

--90e6ba4fc4c658a29304997eb28f--