Re: Re: R: Re: LF: Re: I: Fw: For today the FETs survived...

[Andy Talbot <andy.g4jnt@gmail.com>]

There is a famous quote by Sherlock Holmes that goes "Once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth."

You have shown by your substitutions that the drop-off in power cannot be the ferrite core or the the capacitors.  So they now become "the impossible".

What remains, however improbable, are the FETs, and something else.  The FETs  are well within their ratings and at low power are giving the expected output, so it is more than likely they are behaving properly so look for other improbables first

There is something else that could become an "improbable".   What are you using to measure the power output?  And is it your ONLY means of measuring the RF level.   Could this be failing at the higher RF?

I have seen a diode detector do something similar as the reverse voltage across the diode reaches breakdown.    Some small Schottky diodes can do this at surprisingly low voltages, so if you haven't checked their specification before use...

If you use a power meter, is it designed for use this low in frequency?   If not, the LF could be heating something.

Check and see if the DC changes with the power fall-off.   If DC power stays the same and indicated RF falls, that points very definitely to the sensor / detector.    IF DC falls with the RF, it suggests FETS or something early on.   

If not FETS and not the RF measurement - there is even less left to look at - so consider other "improbables".   There won't be many left by now.    Could there be some kind of feedback into the driver - unlikely but you're eliminating the impossibles one-by-one

Andy  G4JNT

On 3 June 2017 at 15:44, marcocadeddu@tin.it <marcocadeddu@tin.it> wrote:

Hi Stefan,

and thank you :-) so you have too something to mumble while sending
VLF SMS ;-)
There is nothing really new.. I started from an avalable
chassis with on board a PS giving 180Vdc (1200W) and assembled the
Andy's half bridge 700W switching PA and before give it the full power
I'm checking with lower DC supply. Attached you see the schematic with
some change at the moment..
The output xfmr has 7T/19T, the coil of the guard circuit is not
connected (now is in serie with L2) and the resonating caps now are
5x1000pF in parallel.
The aim is finally with 180Vdc of supply have an output of 0,5÷1kW
but.... as you probably red in the previous messages, from 10 to 30Vdc
the PA runs, at low power of course (up to 15W), and complies the
calculations; with supply between 30 and 50 Vdc the output go down to a
couple of W.
I made several trials it the last days, changing the core of the xfmr,
the turn ratio, the kind of capacitators but nothing changes (or at
least the change are within a fraction of dB).

that's the sad history Stefan...
All thoughts and suggestions are welcome! (including go to fish ;-)) )
Marco

----Messaggio originale----
Da: selberdenken@posteo.de
Data: 3-giu-2017 16.19
A: <rsgb_lf_group@blacksheep.org>
Ogg: Re: R: Re: LF: Re: I: Fw: For today the FETs survived...

Hi Marco,

If you like, i can help you. Just need a schematic to see what you
want
to do. Targer power, voltage, etc...

73, Stefan

Am 03.06.2017 16:02, schrieb marcocadeddu@tin.it:
> Hi Alan, Andy, Chris,
>
> I replaced the caps in the output LC with WIMA FKP1, the 2 caps wich
> replaces the missing half of the bridge are still Cornell Dubillier
> 940C serie (rated at 9A @100kHz), so now all the caps are pulse rated
> but...
> I regret to admit it, nothing changed :-((
> Power out increases from 10 to 30Vdc and at 40Vdc after a first pulse
> the power slowly decreases till a couple of watts..
>
> Just to check another thing: being the coil of the guard circuit
> disconnected, I jointed the 2 coils together (they are wound on the
> same support) increasing so the inductance of the L (25T more). Of
> course I had to retune and now I need C in the range of 5000pF but
this
> narrowed the 3dB bandwidth of the LC from abt 15kHz to abt 8kHz so is
> the Q increased!
>
> ... I should visit Decathlon and look for a fishing rod!
>
> your sincerely depressed IK1HSS
>
> ----Messaggio originale----
> Da: alan.melia@btinternet.com
> Data: 2-giu-2017 15.58
> A:<rsgb_lf_group@blacksheep.org>
> Ogg: Re: LF: Re: I: Fw: For today the FETs survived...
>
> Hi Marco, I may well be wrong I was experimenting with 150W at 35V I
> measured 12 to 15A through the capacitors (from memory) .......the
> confirmation is that the change is permanent. I dont remember seeing
> any
> heating effects. Single ended Class E may be considerably different
to
> H-bridge. The currents a lower power will be less,  about 1.5A or so
> (??) I
> am not sure it that would produce damage, but you would certainly see
> it as
> the power increased.
>
> Good Luck with it
> Alan
> G3NYK
>
> ----- Original Message -----
> From:<marcocadeddu@tin.it>
> To:<rsgb_lf_group@blacksheep.org>
> Sent: Friday, June 02, 2017 2:00 PM
> Subject: R: LF: Re: I: Fw: For today the FETs survived...
>
>
> Hi Alan,
>
> and thanks for joining :-)
> it's hard to believe at this stage that is correct to speak of "high
> currents"..
> The PA is an half bridge like
> Andy's 700W but now I'm testing at low voltage: I see these problem
> trepassing the 30Vdc supply level (and power is about 15W)..
> Anyway will try with the "pulse rated caps" you suggested just in
case
> my "low current" "low tension" ideas are wrong.
> 73 Marco
> ----Messaggio originale----
> Da: alan.melia@btinternet.com
> Data: 2-giu-2017 14.36
> A:<rsgb_lf_group@blacksheep.org>
> Ogg: LF: Re: I: Fw: For today the FETs survived...
>
> Hi Marco I dont know where you are putting the caps but I am assuming
> they
> are passing a high RF corrent. I found that the old style (valve)
high
> voltage caps in my Class E test rig changed capacitance permanently
> during
> and after a run (capacitance reduced). I assumed that the current was
> fusing
> the foil connection to the lead-out wires. When using "pulse rated"
> capacitors as used in SMPSUs I had no further capacitance change
> problems.
> The ones I have are RIFA PHE 428 2000v from RS Components. I think
> Farnell
> do a WIMA equivalent.
>
> Best Wishes
> Alan
> G3NYK
>
> ----- Original Message -----
> From:<marcocadeddu@tin.it>
> To:<rsgb_lf_group@blacksheep.org>
> Sent: Friday, June 02, 2017 1:06 PM
> Subject: R: Re: Re: Re: LF: Re: I: Fw: For today the FETs survived...
>
>
> now my thoughts are even more confused....
> Hi Andy, me again....
> after a long night of meditation, I remembered I had a bounch of old
> style mica capacitators 1000pF 1000V..
> I put
> 9 of them in parallel to get the closer value to the 8800pF I had
> before.
> Of course the resonance moved a bit and now is on 134.5 kHz with 3dB
> bandwidth of 9 kHz (Q=15 I know is a bit to high..) so carried out
some
> tests on 134 kHz (of course on the dummyload) all without guard
circuit
> 1st test xfmr 7T/19T (Ae197mm² R50 mat T38): 2Wout @10Vdc; 6,5
> Wout@20Vdc; 13,7Wout@30Vdc stable.. @40Vdc after a quick peak the
power
> slowly goes down till below 1W!
> 2nd test I had still on hands the previous xfmr 5T/12T (Ae197mm² same
> core as 1st test) so was worth to make a trial... same trend :-( for
> Vdc>  30V after a first peak... it goes down.
> I had another core available with different material (N30, Ae 154mm²
> R58) so I prepared a new xfmr with this core 7T/14T:
> 1,4Wout@10Vdc; 4,6Wout@20Vdc; 9,2Wout@30Vdc .... at 40Vdc a short
burst
> with almost 15Wout and then down down down... :-((((((
>
> I noticed a difference in comparison to your project: you connected
the
> + and - rails to ground via 10nF and here I used 100nF caps could it
> explain this behaviour?
>
> 73 Marco
>
> ----Messaggio originale----
> Da: marcocadeddu@tin.it
> Data: 1-giu-2017 22.20
> A:<rsgb_lf_group@blacksheep.org>
> Ogg: R: Re: Re: Re: LF: Re: I: Fw: For today the FETs survived...
>
> that was also my thought.. but they run at room temperature: the
> resonance is obtained with 4 polyesther 2200 pF 2000 Vdc caps. the
> classic boxes 25x15x5mm
>
> Marco
>
> ----
> Messaggio originale----
> Da: andy.g4jnt@gmail.com
> Data: 1-giu-2017 22.03
> A:<rsgb_lf_group@blacksheep.org>
> Ogg: Re: Re: Re: LF: Re: I: Fw: For today the FETs survived...
>
> If the conditions change as the PA is operating, it looks like
> capacitors
> heating up and changing its value.   I can't think of any other
> component
> that will change with dissipation / heating.  What type of capacitor
> are
> you using in the tank?
>
> Andy
>
>
> On 1 June 2017 at 19:59, marcocadeddu@tin.it<marcocadeddu@tin.it>
> wrote:
>
>
>> anybody has good ideas for a replacement hobby?? fishing? growing
>> flowers?
>>
>> Andy: the suggestion of try without the guard circuit revealed that
>> some effects in this area are present.
>>
>> I left the guard coil in place (I'm a bit lazy...) and disconnected
>> simply the 2 wires from the rectifier bridge.
>> The output improve of about 4dB with Vdd from 10 to 30V, the output
>>
> is
>
>> stable up to 20Vdd, on 30V it show a peak of 16W, then the output
>> starts to decrease till a couple of watts :-( this happens also at
40
>> and 50V.
>> The resonance of the output LC (with the guard coil open) shifts
from
>> 137 to 140 kHz and the bandwidth decreases from 20kHz to 10 kHz.
>>
>> I need a long weekend of meditation...
>> 73 Marco IK1HSS
>>
>>
>> ----Messaggio originale----
>> Da: marcocadeddu@tin.it
>> Data: 31-mag-2017 21.05
>> A:<rsgb_lf_group@blacksheep.org>
>> Ogg: R: Re: Re: LF: Re: I: Fw: For today the FETs survived...
>>
>> Andy.. you are almost better than online help desks :-D
>>
>> yes the guard circuit is on place but no current is flowing toward
>>
> the
>
>> PA, testing disconnetting it needs just to warm up the iron ;-)
>>
>> the PS should
>> provide enough juice for 1200W input and the IRF460A are rated for
>>
> 20A
>
>> @ 25°C (13A @ 100°C) so.. I admit it would like to give a try ;-)
>>
>> I don't guess the core is saturating specially at this power level
>> where rms is only 22V, the core is 50mm OD and has 195,7 mm²Ae: if I
>> am
>> not totally wrong B should be<  0,03T @50V with 7 turns on the
>>
> primary
>
>> Will tru to disconnect the guard circuit just in case the squirrel
is
>> running in its cage ;-)
>>
>> Thanks again for assistence
>>
>> Marco IK1HSS
>>
>>
>>
>> ----Messaggio originale----
>> Da: andy.g4jnt@gmail.com
>> Data: 31-mag-2017 20.39
>> A:<rsgb_lf_group@blacksheep.org>
>> Cc:<dead.fets@gmail.com>
>> Ogg: Re: Re: LF: Re: I: Fw: For today the FETs survived...
>>
>> 7:19 turns (assuming 50R output)  means you have a load resistance
of
>> 6.8
>> ohms which for 180V DC (81 V RSM fund sine) is nearly 1kW  I don't
>> think
>> you really mean to go that extreme do you?    13 ohms is more
>> realistic.
>>
>> As for the tank resonance changing as power increases, that is very
>> wrong.
>> I wonder if the transformer is saturating.   Not sure of your core
>>
> Ae,
>
>> but
>> lets assume 200mm square, a core of about 16mm diameter.
>>
>> V = 4.44.F.N.A.B    Plugging in 137kHz 7 turns, 200 mm^2 and a Bmax
>>
> of
>
>> 0.1
>> that suggests 85V RMS.
>> Which is exactlyly what you have.  I suggest more primary turns .
>> Before
>> a transformer ratio of 1:2 was suggested, for Rload = 13 ohms
>>
>> Is the guard circuit in place ?   Don't forget, it has to be
>> customised
>> to
>> you exact currents and coil Q.  Get teh PA operating to its proper
>> settings
>> foirst - that you can do at low voltage power, it scales perfectly.
>> Only
>> when it it working properly can you add and set up the guard
circuit.
>>
>> When I did teh 700W PA, I had a complete workign (albeit unreliable)
>> unit
>> before even thinking of teh guard circuitry.
>>
>> Andy
>>
>>
>>
>>
>>
>> On 31 May 2017 at 18:50, marcocadeddu@tin.it<marcocadeddu@tin.it>
>> wrote:
>>
>>
>>> Hi Andy... me again...
>>>
>>> I was so curious to see what could happen thatI had a very quick
>>> dinner and connected all, but...
>>>
>>> now the output xfmr has 7T/19T here my
>>> readings/calculations:
>>> (see attached picture)
>>> again the power increase from 10 to 30Vcc then from 30 to 50Vcc
>>>
> after
>
>>> an initial burst it start to fall down..
>>> I checked also the resonance of the LC: till 30Vcc is tuned on 137
>>>
>> kHz
>>
>>> with a 3dB bandwidth of 20 kHz, when I move to 40 and 50Vcc the
>>> "maximum" output shifts to 165 kHz...
>>>
>>> mumble mumble
>>>
>>> I tempted to have roasted FETs for dessert and see what happens at
>>> 180V!
>>>
>>> Marco, IK1HSS
>>>
>>>
>>> ----Messaggio originale----
>>> Da: andy.g4jnt@gmail.com
>>> Data: 30-mag-2017 23.50
>>> A:<rsgb_lf_group@blacksheep.org>
>>> Cc:<dead.fets@gmail.com>
>>> Ogg: Re: LF: Re: I: Fw: For today the FETs survived...
>>>
>>> I've just looked again at the circuit diagram you sent - on there
>>>
> the
>
>>> values are different from your statement in the email.  It shows
>>> primary 5
>>> turns, secondary 12 turns so a load resistance in the order of 9
>>>
> ohms
>
>>> which
>>> is rather low if you are intending a Vdd of 180V - but closer to
>>>
> the
>
>>> ideal
>>> Rl
>>>
>>> The tank components have a reactance of 130 ohms which is too high
>>>
> a
>
>> Q
>>
>>> is
>>> used with that 9 ohms Rload, You should be aiming for a Q in the
>>>
>> region
>>
>>> of
>>> 6.
>>>
>>> Even with the optimum load R of 13 ohms described last time for 500
>>> Watts
>>> from 180V rail the resulting Q of 10 is a bit too high - you will
>>>
> end
>
>>> up
>>> with high voltage and critical tuning
>>>
>>> Andy  G4JNT
>>>
>>> On 29 May 2017 at 19:07, Andy Talbot<andy.g4jnt@gmail.com>  wrote:
>>>
>>>
>>>> Yes.
>>>> As you'll see in my original write up, I originally forgot that
>>>>
> the
>
>>> peak
>>>
>>>> of the fundamental sine component of a square wave is GREATER
>>>>
> than
>
>>> the peak
>>>
>>>> by a factor of 4 / pi and initially my PA delivered a lot more
>>>>
>> power
>>
>>> (1.6
>>>
>>>> times) than it was supposed to.
>>>>
>>>> So if the square wave has a peak value of 1, its fundamental sine
>>>> component has a peak value of 4/pi or around 1.27.  The RMS of
>>>>
> the
>
>>>> resulting sine  is SQRT(2) less than this giving a Peak square to
>>>>
>> RMS-
>>
>>> sine
>>>
>>>> ratio of  0.9..   If you specifye peak-peak of the square wave, a
>>>>
>>> further
>>>
>>>> factor of 2 applies, leading to the 0.45 ratio described before.
>>>>
>>>> Incidentally, this same ratio appears in that equation for  flux
>>>>
>> in
>> a
>>
>>>> magnetic code,   V = 4.44.F.N.A.B
>>>> The magic number 4.44  is actually SQRT(2) * pi     and comes
>>>>
> about
>
>>> from
>>>
>>>> the same sort of sine to square transform.
>>>>
>>>> Andy
>>>>
>>>> On 29 May 2017 at 18:48, marcocadeddu@tin.it<marcocadeddu@tin.
>>>>
> it>
>
>>> wrote:
>>>
>>>>
>>>>> uhuh... a slightly silly misleading assumption... Vdc are the
>>>>>
> same
>
>>> of
>>>
>>>>> Vrms before FETs make their work!
>>>>>
>>>>> Thank you Andy for pointing out it!!
>>>>> With this approach calculation changes a bit and probably with
>>>>>
> the
>
>>>>> right Xfmr  the PA can give higher satisfaction :-)
>>>>>
>>>>> Hopefully the FETs will survive and this time I'm ready to
>>>>>
> burnout
>
>>> the
>>>
>>>>> antenna hi
>>>>>
>>>>> Will keep you both updated, thank you once more Andy
>>>>>
>>>>> 73 Marco, IK1HSS
>>>>> ----Messaggio originale----
>>>>> Da: andy.g4jnt@gmail.com
>>>>> Data: 28-mag-2017 21.18
>>>>> A: "marcocadeddu@tin.it"<marcocadeddu@tin.it>,
>>>>> <rsgb_lf_group@blacksheep.org>
>>>>> Cc:<dead.fets@gmail.com>
>>>>> Ogg: LF: Re: I: Fw: For today the FETs survived...
>>>>>
>>>>> First thing I noticed is that your turns ratio on the output
>>>>> transformer
>>>>> doesn't look right.
>>>>> You quote "* ... with primary winding of 15 turns and secondary
>>>>>
> of
>
>>> 12
>>>
>>>>> turns...*"
>>>>>
>>>>> 180V DC in a half bridge is 180V peak-peak square wave.
>>>>> The fundamental sine part of that is  4/pi * 180 = 229V pk-pk
>>>>> so is 229V /[2.SQRT(2)] = 81V RMS
>>>>>
>>>>> To a good approximation RMS(fund) from a half bridge is Vrms
>>>>>
>> (fund)
>> =
>>
>>>>> 0.45VDC
>>>>>
>>>>> For 500 Watts out, Rload =  81 ^ 2 / 500 =  13 ohms
>>>>>
>>>>> So to match to 50 ohms you need a turns ratio of SQRT(50/13) =
>>>>>
>> 1.9:
>>
>>>>> 1     so
>>>>> call it 2:1  Keeping 12 turns on the  secondary means you need 6
>>>>>
>>> turns
>>>
>>>>> on
>>>>> the primary
>>>>>
>>>>> When operating at reduced voltage, the power out will vary
>>>>>
> exactly
>
>>> as
>>>
>>>>> the
>>>>> square of the voltage.
>>>>> Recalculating from first principles for a 12V supply:
>>>>>
>>>>> 12V  DC = 12V pk-pk = 12 / [2.SQRT(2)] * 4/pi = 5.4V RMS
>>>>>
>>> (fundamental)
>>>
>>>>> in 13 ohms should give 5.4^2/13 = 2.2 Watts
>>>>>
>>>>> check using ratio of voltages, squared :
>>>>>
>>>>> (12V/180V) ^ 2 * 500W = 2.2 Watts which is the same as above.
>>>>> QED
>>>>>
>>>>> Your 15:12 ratio result sin a load impedance of (15/12)^2 * 50 =
>>>>>
>> 78
>>
>>>>> ohms
>>>>>
>>>>> At 40V DC == 18V RMS(fund) that will give 18^2/78 = 4.1 watts
>>>>>
>> which
>>
>>> is
>>>
>>>>> actually LESS that you are seeing - the 2* discrepancy is odd,
>>>>>
> but
>
>>> the
>>>
>>>>> low
>>>>> power is in the area of what you measured..
>>>>>
>>>>> Andy  G4JNT
>>>>>
>>>>>
>>>>>
>>>>> On 28 May 2017 at 19:34, marcocadeddu@tin.it<marcocadeddu@tin.
>>>>>
> it>
>
>>>>> wrote:
>>>>>
>>>>>
>>>>>> Hi Chris,
>>>>>>
>>>>>> I tried to post this message on the reflector but apparently I
>>>>>>
>> had
>>
>>> no
>>>
>>>>>> success..
>>>>>> As promised I keep you updated but as you can read in the
>>>>>> attachment the first trials were not enocouraging...
>>>>>> Andy, may I ask you to read my report? your interpretation and
>>>>>> suggestion are welcome!
>>>>>>
>>>>>> 73, Marco IK1HSS
>>>>>>
>>>>>>
>>>>>> -----Original message-----
>>>>>>
>>>>>> From: "marcocadeddu@tin.it" marcocadeddu@tin.it
>>>>>> Date: Sun, 28 May 2017 17:01:33 +0200
>>>>>> To: rsgb_lf_group@blacksheep.org
>>>>>> Subject: For today the FETs survived...
>>>>>>
>>>>>> Hi LF,
>>>>>>
>>>>>> hope that also the toroids of Chris survived!
>>>>>> My FETs survived, but they are not working as expected :-(
>>>>>> Attached the report on my attempt to duplicate the half bridge
>>>>>>
>> of
>>
>>>>>> Andy..
>>>>>> Has anyone suggestions before I try to cook all connecting to
>>>>>>
>> the
>>
>>>>>> 180Vdc supply?
>>>>>>
>>>>>> Thank you
>>>>>> 73 Marco IK1HSS
>>>>>>
>>>>>>
>>>>>> --
>>>>>> This message has been scanned by E.F.A. Project and is
>>>>>>
> believed
>
>> to
>>
>>> be
>>>
>>>>>> clean.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> ---------- Forwarded message ----------
>>>>>> From: "marcocadeddu@tin.it"<marcocadeddu@tin.it>
>>>>>> To:<rsgb_lf_group@blacksheep.org>
>>>>>> Cc:
>>>>>> Bcc:
>>>>>> Date: Sun, 28 May 2017 17:01:33 +0200 (CEST)
>>>>>> Subject: For today the FETs survived...
>>>>>> Hi LF,
>>>>>>
>>>>>> hope that also the toroids of Chris survived!
>>>>>> My FETs survived, but they are not working as expected :-(
>>>>>> Attached the report on my attempt to duplicate the half bridge
>>>>>>
>> of
>>
>>>>>> Andy..
>>>>>> Has anyone suggestions before I try to cook all connecting to
>>>>>>
>> the
>>
>>>>>> 180Vdc supply?
>>>>>>
>>>>>> Thank you
>>>>>> 73 Marco IK1HSS
>>>>>>
>>>>>>
>>>>>> --
>>>>>> This message has been scanned by E.F.A. Project and is
>>>>>>
> believed
>
>> to
>>
>>> be
>>>
>>>>>> clean.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>