The AM Forum
December 02, 2024, 02:11:31 PM *
Welcome, Guest. Please login or register.

Login with username, password and session length
 
   Home   Help Calendar Links Staff List Gallery Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: calculating the output impedance of a PWM power supply  (Read 13452 times)
0 Members and 1 Guest are viewing this topic.
N4LTA
Contributing
Member
*
Offline Offline

Posts: 1070


« on: October 25, 2019, 08:48:11 PM »

I am working on an output filter for my PWM power supply. It is fairly easy to determine the input impedance of the class D transmitter by measuring the input voltage and current but determining the output impedance of the PWM supply is not discussed in any of the literature. I am trying to figure out how the Butterworth filters are calulated. I maybe able to guess by reverse calculations  but I would rather have a more accurate method.

Any help would be appreciated

Pat
N4LTA
Logged
vk3alk
Member

Offline Offline

Posts: 263


« Reply #1 on: October 25, 2019, 11:12:43 PM »

Thats all I do....the load the TX presents to the Modulator.....
I use http://tonnesoftware.com/svcfilter.html ......
Have used it so much that I felt a bit guilty so I purchased his Meter program.....

In the family area select Butterworth as the ripple content at the knee is lower then the other types....
I tend to use a reference frequency of 250Khz so using a 4 pole offers plenty of attentuation .....
Also use a cutoff frequency of appox 25Khz as well....

Maybe make the TX first and see what load it offers .....
The efficiency point of the transmitter might be different to what you expect as well ......


Wayne

Logged
KQ6F
Member

Offline Offline

Posts: 124


« Reply #2 on: October 26, 2019, 11:59:10 AM »

Proper design of the PWM low-pass filter involves a few tradeoffs and is not trivial.  I recommend you read this first:
http://www.tonnesoftware.com/appnotes/pwm/HamPWM.pdf.

As Wayne suggested, you'll need to build the RF stage and operate it at your intended carrier level in order to determine the filter load impedance (DC voltage divided by current).

Then use this program to design the filter http://www.tonnesoftware.com/elsie.html  The Student version is free and allows up to 7 network components which is sufficient for your design.  Like Wayne, I felt guilty and bought the Professional version after repeatedly going onto Jim's website.  I forget what it cost but it wasn't expensive.

In Elsie start with Inductor input lowpass topology Butterworth family.  For 3dB Bandwidth use your chosen cutoff frequency.  In general the higher you make it the lower the component values but that involves the first tradeoff.  You need to consider your switching frequency and to what extent you want the filter to suppress it.  In my case the switching frequency is 100kHz and I chose a 15kHz cufoff.

Next put in the number of inductors and caps under Order (mine is 4).  Finally put your load impedance in under Input termination and select the Schematic tab.  You'll see that your load impedance will appear both as the load and the source impedance.  The source impedance is not accurate since the modulator is essentially a voltage source and its output impedance is very low.  Use the Edit tab to change it to a low value.  I made mine 0.5 ohm.

Now select the Plot tab.  On the left side click on Input impedance;transmission and observe the filter response and input impedance curve.  You'll see that there is an unwanted bump up in the response at the cutoff frequency and probably a dip down in the input impedance.  At this point you'll need to use the Tune Part tab to play with values to both lower the bump up and smooth out the input impedance curve.  This tuning procedure may seem like cheating but it's what Jim recommends.

As a final check click on Analysis/Transient and under Waveform Type select Squarewave with a Squarewave frequency of say 1000 Hz.  If you have tuned everything properly you should see minimal ringing on the squarewave.  My actual filter squarewave response has just a bit of overshoot and exactly matched the simulation.

As a final tweak I added a capacitor in shunt with the second inductor to produce a deep notch at the switching frequency.  

This all sounds complicated but it was very educational and great fun at least for me.

Rod
Logged
vk3alk
Member

Offline Offline

Posts: 263


« Reply #3 on: October 26, 2019, 08:43:48 PM »

An excellent post Rod....worth printing out and pinning to the Frig...

Using a capacitor across the last inductor forming a shunt at the switching frequency is a good idea too.....
Particularly at your power levels while ours down here are so much lower doesn't really matter....
I use this WEB site to calculate https://goodcalculators.com/resonant-frequency-calculator/
There are many out where on the web...

Have played around with the switching frequencies too...
Found a big difference using complex audio like music at 100Khz and 250Khz but voice frequencies as expected no difference...

Just a question to you Rod.....why use such a low cutoff of 15Khz.....
I did experience alaising issues some years ago now on my 40M system and amongst other things decided to move away from my highest audio frequency and increased the cutoff to 25Khz and the problem went away.....

Also too I think you use H Bridge for your TXs.... what power levels are they running at and what efficiences etc:

Just thinking about the choice of transmitter design for Pat....


Wayne
Logged
N9NEO
Member

Offline Offline

Posts: 282


« Reply #4 on: October 27, 2019, 05:17:30 AM »

A series LC trap across the output capacitor will do a great job of squelching the switching harmonics of modulator.  I have a very old design paper around here if you are interested I might be able to find.  Modulator I did was at 360kHz and so trap was small.  Todd down in SC came up with suggestion of trap.


The actual output impedance of the modulator is a function of frequency.  There is a feedback loop in there that is frequency dependant.  Not to mention the output LC filter within the loop.  In the simplest terms load modulator up and measure the change in (DC) output voltage as a function of the (small) change in output current.  Now sweep this measurement across the audio spectrum you expect to operate in and you will have output impedance as a function of frequency.  A lot of work that may not be necessary to dial in your design.  Don't over-design the filter and save your self some grief.  Just need to keep the switching sidebands under control to keep the bands clean and your radio neighbors happy.

Fun stuff you're working on.

N9NEO
 
Logged

I'd rather be anatomically correct than politically correct!
N4LTA
Contributing
Member
*
Offline Offline

Posts: 1070


« Reply #5 on: October 28, 2019, 07:55:15 PM »

Thanks everyone for the info. I really like Jim Tonnes paper. Very informative. I also like the idea of a trap.

I hope to get the RF Deck ready to go soon. I am building W1VD's VFO now .


To initially test the RF Deck - I assume that I can hook up my 35 volt current limited supply and a signal source and generate a carrier with some amount of safety. Trying to figure the safest way to minimize smoke. I also have a big 0-75 volt power supply rated at 1000 watts or so that I used  with my experimental class D transmitters on 500, 137 and 60 KHz.  IT has a big variac in the front end but is still a "welder" as it has no current limiting.

Thanks again,

Pat
N4LTA
Logged
KQ6F
Member

Offline Offline

Posts: 124


« Reply #6 on: October 28, 2019, 08:46:57 PM »

Bringing up your RF deck with a current-limited power supply is a very good idea.  Watch your gate and drain waveforms carefully with a scope.  Look for evidence of parasitics.  You can get a rough idea of the modulator load impedance at this point but it will change some as you go up toward your final desired carrier output.  So wait until you get to final carrier level before measuring.

If things look good then you probably can proceed to use your non-current limited power supply to go the rest of the way.  Just carefully watch waveforms as you progress.

Rod
Logged
N4LTA
Contributing
Member
*
Offline Offline

Posts: 1070


« Reply #7 on: October 28, 2019, 09:51:53 PM »

Thanks Rod,

I appreciate your assistance and I will probably ask for more.

I have a board with SMT drivers installed and made for SiC FETs. I have not decided on what type of output transformer I will attempt but current mode class D seems like a good try. W1VD
has examples using a transmission  line output transformer - I feel more comfortable with a toroidal transformer but do not have a good example.

I am also intrigued by the H bridge circuit described by WO1U in the transmitter forum using modern SMT drivers. The output circuit looks straightforward. Not sure exactly how to impliment a SMT driver without having floating common buses? i think Wayne has used an H bridge format.

Thanks

Pat
Logged
vk3alk
Member

Offline Offline

Posts: 263


« Reply #8 on: October 29, 2019, 08:12:06 AM »

As far as the TX goes I would go with what you feel like experimenting with.....
Giving Nigel's TX a go first perhaps which is VMCD.....
To be honest VMCD is my least favourite but Nigel has been totally successful so I cannot say anymore.....
I did build up that CMCD with 2 FETs and posted a photo of it somewhere and that went well really....and cannot see why using 4 you could not get similar results with more power....

But like Rod the H Bridge design is now my  favourite RF module and have uploaded some pictures.....
Instead of just one module I use 2 and feed them into a combiner......it also adds another level of protection as if one module fails the resistor in the combiner takes the load and protects the PWM and power supply perhaps...

The 80M TX is finished now .... the other is of modules for my new 160M TX that have been tested and now are building the box etc:

Its good you have popped in too Rod.....



Wayne





* 80M nearly finished.JPG (468.87 KB, 1600x1200 - viewed 993 times.)

* 160M TX tested.JPG (467.16 KB, 1600x1200 - viewed 1035 times.)
Logged
KQ6F
Member

Offline Offline

Posts: 124


« Reply #9 on: October 29, 2019, 02:30:07 PM »

A few years ago I experimented with CMCD and built a H-bridge RF deck using SiC FETs.

http://amfone.net/Amforum/index.php?topic=38429.0

I never could achieve better than about 83% drain efficiency and finally gave up.  I figure part of the problem was shoot-through but could not prove it since I had no means of observing the current waveforms.

I took the same FETs and built what I have now which is a hybrid configuration somewhere between Class D and Class E topology.  It has large shunt caps from drains to ground (like Class E) but has an output filter like Class D.  I wanted to avoid pure Class E topology because of its large series-tuned components in its output filter.

Unless you're in a hurry you should probably take the time to try different topolgies.  The learning experience is very worthwhile and is great fun...at least it was for me.  You might want to take a look at what K7DYY has done with his popular Super Senior transmitters.  Schematics for both his VMCD and CMCD units are on his website.

http://www.k7dyy.com/

Rod
Logged
N4LTA
Contributing
Member
*
Offline Offline

Posts: 1070


« Reply #10 on: October 29, 2019, 05:17:53 PM »

Real good information. Thanks for all the assistance.

How do you drive your H bridge? Do you use drivers that are high side, low side combination drivers? I have seen them driven with transformer and then I see some  high side drivers than use the same power supply as a the low side driver or combo high side/low side drivers - but have not seen them used. Do you recommend the gate transformers driven by a heavy duty driver.
I am tending toward trying an H bridge foir a first try.

I am working on a VFO much like W1VD uses to get a signal source. I have several signal generators but I'd rather test with the same signal source that I will use with the finished project.

Waiting on a new power transformer due tomorrow and will be out of town for a long weekend, so things will slow down a bit.

Pat
N4LTA
Logged
KQ6F
Member

Offline Offline

Posts: 124


« Reply #11 on: October 29, 2019, 06:10:13 PM »

I used transformers.  See schematic attached of a 40M deck I once built.  It's a H-bridge configuration although not exactly drawn that way.

The gate capacitance is usually pretty high.  If driven directly the driver ICs will frequently run hot owing to large current spikes.  But if instead transformers are used and a small inductor of the right value is placed between the driver output and the primary, a serendipitous effect results.  The gate C will resonate with the circuit inductance and create a nearly-perfect sine waveform at the gate.  This has two benefits.  Sine wave drive automatically prevents overlaps that square wave drive might otherwise cause due to uneven turn-on/turn-off times in the FETs.  By correctly setting the value of resistors in series with the gates to dampen the circuit, the gate voltages can be controlled to say 25V peak-to-peak which ensures solid turn on/turn off performance.

Rod

Edit:  wrong schematic was posted


* ClassD_PA_40M.jpg (314.92 KB, 2040x1540 - viewed 1120 times.)
Logged
vk3alk
Member

Offline Offline

Posts: 263


« Reply #12 on: October 29, 2019, 10:16:40 PM »

Playing around builds up knowledge over a period of time..... Shocked
Have done the same here a little while ago now......

The photo shows a H Bridge on 40M that now has good efficiency in the low 90%.....
The input too has been made resonant resulting in the waveform shown on the CRO....
See the little coil on the input side ( appox 4 Turns )....
Actually made it resonant on appox 8.5Mhz so the TX also works on 30M band as well....

The whole thing has promise really and has been pushed aside just for the moment....for another day....


Wayne





* 40M H Bridge.JPG (470.14 KB, 1600x1200 - viewed 1030 times.)
Logged
Pages: [1]   Go Up
  Print  
 
Jump to:  

AMfone - Dedicated to Amplitude Modulation on the Amateur Radio Bands
 AMfone © 2001-2015
Powered by SMF 1.1.21 | SMF © 2015, Simple Machines
Page created in 0.085 seconds with 19 queries.