Need Help Designing PDM Switching Filter for 4X1 Rig

[K1JJ]

I am having trouble getting the Butterworth type PDM filter design correct for the new 4-1000A X 4-1000A PDM rig. I need a standard 4 pole filter, using L1, C2, L2, C2.  L1 must be bigger than L2.

The standard design software out there works great, [Bill GF's software] but puts the small inductor first. I need the big inductor FIRST in line.

If that's not enuff, I need to use the capacitor values below that I have available. The voltage will be about 7500VDC. 

Anyone have software or the ability to design a 4 pole filter with a cutoff of 10kc and an impdance of 3,000 ohms... that will let you use the cap values below [parallel caps are OK] and also make the bigger inductor first in line???


Caps available, actual measured values:

.001907 uf @ 10kv
.001905 uf @ 10kv
.001309     7.5kv
.001206     20kv
.001201    20kv
.00250    20kv
.00247    15kv

.000214    10kv
.000107    10kv

I can possibly get some .01 ufd caps to fill in.

Thanks.
Tom, K1JJ

[WA1GFZ]

Tom,
You can scale my filter to 3 k but who knows the right way to design the filter both ways work. I got my numbers from an old Don White filter book. This put the big inductor next to the switch. Heck that is how you do a filter in a switching regulator. In a power supply the big choke is first then a small smoothing choke. Some smart person who really knows please set us straight.

[Tom WA3KLR]

Tom,

I scaled the design I did for my upcoming Class E project.

What I came up with:

- the input inductor 62.06 mH.

- the first cap. to gnd.  0.01034 uF.  (You can get this w/ (4) 0.00250 in parallel.)

- the second inductor 62.06 mH.

- the final cap. to gnd.  0.00345 uF.  (You can get 0.00367u with 0.001201 and 0.00247 in parallel.)

I ran the values in my SwitchreCAD simulations along with my 7.5 Ohm filters. 

Looks good in ac sweep and transient analysis.

Do you have Linear Technology's SwitcherCADIII circuit simulator recommended by Young Bill?  I love it.
I could sent you the files and you can run them yourself.

[WA1GFZ]

second section has a higher corner frequency than first so may work fine.
I was taught a 3:1 ratio will keep you stable but know you can cheat a bit. Tom what was your drive source. I think you can select square wave with that program.

[Tom WA3KLR]

Hi Frank,

I just looked at the transient simulation and the voltage source is a PULSE. 
It is configured for a 132 Hz. square wave with about 67% duty cycle.

I tend to copy other circuit files as a starting point, as I did with this exercise.
I don't remember specifiying the generator.  Maybe the transient analysis did it itself.
Works well though.

Simulation runs for 1 millisecond.  The filter settles in about 0.2 milliseconds.

The filter is my own design; a Bohlander filter.  It evolves from transmission line/lumped LC theory.
Similar to Butterworth, but a little better for our application, I feel.

[K1JJ]

Outstanding, Tom!

Thanks very much, OM.

Those are some BIG coils, like 2-3 ft long with a single layer...

The cap values I gave you were only for quantities of one. So I cannot use those parallel values you suggested.  BUT, I found some more caps to parallel and was able to come up with the following values for C1 and C2:

C1= .010018
C2= .00379

Could you recalculate using these new values, please?

Also, I plan to put in a trap at the end, near the input to the final cathode to suck out any residual pulses. This worked FB on my class E rig.  My switching freq is about 112 kc. What values would you suggest there? It will be a series resonant trap to ground.  I'm trying to picture what to use for real world parts, cuz it will be 7.5kv
there. I suppose a high Q trap is needed???  The trap coil will be maybe 1/10th the size of the big filter coils, I'm hoping. Something like a plate choke maybe.

BTW, thanks for the offer on the software, but will pass. This is the only time I will be using it, and have much more designing to do in other areas - thanks anyway.

T

[K1JJ]

Tom and Frank,

I came up with the physical coil sizes for the 62 MH coils.

Using a 12" diameter PVC pipe, that is 15" long will require about 592 turns. This is single layer and #22 enameled wire.

That's not too bad and smaller than I expected for the pair. Once I get the exact PVC form I can recalculate, but that's a ballpark that can fit into the rack.

T

[Tom WA3KLR]

Tom,

No need to re-simulate.  Your new values are very close to the design values and to the capacitor values I thought you would use that I simulated with before.

How about using the empty Home Depot spools from all of that 500 and 1000 foot rolls you bought in the past.  I hope you saved them.  Maybe re-winding them with multi-layer solenoid winding is the way to go for the 62 mH. inductors?

As far as the additional trap, I think that a parallel trap in series with the LPF is really the way to go.  The series trap to ground has a bad interaction, I would think.

To realize the full performance of the filter, you will need to keep the 3 inductors from coupling to each other via spacing, orientations, and maybe individual metal boxes.  Maybe Steve QIX has some thoughts on that.  When you get far enough along in the construction, you can drive the filter with a signal generator and look at the output with a scope or receiver to optimize the coil positions and shielding.

Maybe I'll look at the trap design tomorrow.  It's getting late.  Time to shut down.

GL OM.

[Tom WA3KLR]

Tom,

What is the PWM switching frequency in this tube design?

[WA1GFZ]

Tom,
Can you run the simulation at 100 KHz or was it just a typo??
Does the response have any peaking near the corner?  fc

[WA1GFZ]

Tom vu insulated wire will take more space but make a better L with less C between turns. This will help reducing the switching leakage through the filter.

[K1JJ]

Tom vu insulated wire will take more space but make a better L with less C between turns. This will help reducing the switching leakage through the filter.

Hmmmm... not a bad idea. Thin insulated wire is easy enuff to get. It will stand up better than enamel too. Just larger coil. The old coil finally imploded on itself on key up. That was the final straw when I tore the rig down....   [layout issues, again]

Tom, OK on the layered coils. Steve's tube PDM paper says to use a single layer for the input coil, but layered is OK for the 2nd one. But, I've always had poor luck with layered coils blowing up. It's really an art to wind them that takes trial and error to get it right with min C, etc. I will stick to the simpler single layer coils and make them identical.

The switching feq is 112 kc. I am using the same PDM generator/driver from the class E rig here.

OK on the filter values. I will use what I have then. There is probably enuff stray capacitance from the fil transformer and surrounding stuff to make the caps values want to be slightly smaller anyway.

That parallel resonant trap is an interesting idea. Though Steve used a series trap to gnd on his old tube rig and it sucked out the images into the noise. I heard the A/B. It worked well also on my solid state rig.   The shunt acroos the filter will be minimal with the cap/coil resonant at 112kc, won't it?  Either way, I should axe Steve about the parallel trap, as it does seem like a better idea.

T

[WA1GFZ]

just busting your butt here mr vu..... but who changed the system spec by 7dB QRO after PDR.

[K1JJ]

just busting your butt here mr vu..... but who changed the system spec by 7dB QRO after PDR.

Was it 7db?  Oh well, you know the routine. Get the bugs out by ramping it up and blowing up parts. Then back it down into the safe zone once the weak hands are cleared out.

But in this case the ramping up was too aggressive....   It really was a FB rig for 6 months or so.  I remember Tron was here watching it work. He just stared at the 4X1's working without any mod iron and running 90% / 80% efficient -he had a longing, distant look in his eyes. He said something to the effect of that being the way to go...

T

[WA1GFZ]

I remember Tim and I watching the arc traveling up the feed line as it echoed through the wooks as a modulated spark. That was priceless.

Just think back when certain mistakes were made and avoid them so the rig doesn't become a dead hulk on the shelf.

[Tom WA3KLR]

Here is the plot of the ac sweep results.  I made an educated guess at losses in the inductors and caps for this simulation.

(What size wire are you planning to use Tom.  It doesn't have to be very big.  I guessed #14 for you.  Even smaller would be o.k.  But it's good to get as thick an insulation as possible.)

112 kHz. is 78 dB down.

0 dB loss in the filter is referenced to -6 dB on the chart.

[K1JJ]

Tom,

78 db down is pretty decent.  Hopefully in the real whirl with coil placement it can be achieved.   The trap WILL be needed if I am going to see 100db+ down.  The 112kc image works out to be down on 3778, near the DX window when operating on 3890. At -78db, a 60 over S9 signal would be about an S6 down there... that could be heard, thus the need for more attenuation.

Frank and I have been emailing other details to get a decent design. That's actually the most fun before the drilling and blasting starts.

I haven't gotten any coil wire yet, but if you look back at my recent post you will see I planned to use #22 wire. That's what Steve used for his big tube rig. With insulation on the wire, it will be a mighty big coil as it is. Going to bigger wire yet wud make it too large for a single layer coil to fit in the rack.

T

[Tom WA3KLR]

Hi again Tom,

I updated the filter simulation with your capacitor values, the #22 wire and a pararel-resonant tlap on the output of the LPF.  Real losses are estimated and modeled for the caps and inductors.

112 kHz. is 97 dB down.  Results is very dependent on the actual losses of the components and mutual coupling.

The tlap is 1 milliHenry and 2000 pf. for the simulation; but not necessary to copy these values.  I don't recommend an L-C with less than 2000 pf.  Re-calculate the L with one of your 0.0025u caps. or something greater than or equal to that.  Make this inductor with larger wire than #22; #14?

By the way, the other line on the graph is the phase response.  I don't see a way to get rid of it.  It makes the graph confusing to read.

[K1JJ]

Hey, that's great, Tom!

-97db is probably do-able.

I like the sharp dip at 112 kc the trap produces.

If I needed more attenuation later after the rig was running, could another section be added to the filter without changing the existing part values?  How much more attenuation cud be expected with an additonal L3 and C3?

So, it appears in the simulation that #22 wire is going to work, right?

Tnx again.

T

[WA1GFZ]

#22 should be fine. We did #26 on the last filter good for around 700 ma.

[Tom WA3KLR]

Tom,

Here is your filter with and without the real component losses.  You can zoom in on the notch area with the Adobe Acrobat reader.  Blue trace is the "ideal" components.

The simulator defaults to having 1 milliOhm series resistance with each capacitor and inductor.  That is why the notch depth is not infinite, but is aboot 25 dB greater than the realistic model.

[K1JJ]

OK, Tom -

Yes, that plot shows the importance of using low loss components!  You sure pick up some attenuation with Hi Q.

The bad news:
I have some #22 insulated wire. It fits 18 turns per inch.
 I calculated that 62 MH needs a 30"  long, 18" diameter form with 540 turns. That's 3,050' of wire for each coil!  Way too big.

Too bad cuz I have a lot of this wire and it wud make a hi Q coil, but would have to hang these from the ceiling to fit....

And #22 enamel wire results in a small coil, but is too close spaced.

I was thinking of using those cheap 12" and 14"? cardboard concrete pillar forms for the coil form. At Home Depot.

Will have to think about this more.

T

[w3jn]

Is there a reason you can't use toroids, Tom?  You're not running a shitload of current thru 'em, they're self-shielding, and MUCH more compact than a solenoid inductor.  More of a PITA to wind than a solenoid inductor though.

[WA1GFZ]

John,
I have 6- 2 1/2 inch toroids as a filter in my V2 the Ui=60 and still an air core inductor sounds better. They provide a shielded inductor but the C between turns can be an issue if you want to get any inductance. I have 200 turns of #26 to get 8 UH good for about 700 ma. Turns are side by side. I've had problems with higher perm material swinging widly as the modulation changes.

Also it is cool to watch an air core inductor do pullups as you talk. Tom had a big one hanging by ropes under a steel plate in his PW PDM rig.

[K1JJ]

Is there a reason you can't use toroids, Tom?  You're not running a shitload of current thru 'em, they're self-shielding, and MUCH more compact than a solenoid inductor.  More of a PITA to wind than a solenoid inductor though.

Yeah, sure wud be nice to have a tiny ferrite coil, John. But as Frank said, for a quality PDM waveform and minimal audio distortion, a big ass air coil needs to be used. At least for the first coil.

It's a pity #22 insulated wire wud be so big cuz I have maybe 4,000' of it here from Hoss flea mkts.  Was gonna use it as ant gnd radial wire, but opted for #14 instead.

The proposed coils, 30" long X 18" dia will also need to be placed where they can be at least a diameter away from other stuff, so needs big real estate to see the -100db system atten [with the trap] that Tom/KLR speced out.

I hate to resort to #22 enamel or a thinner insulated wire to make a smaller set of coils.

Remoting these coils sounds like a bad idea too....  I COULD hang them from the ceiling, horizontally, above the rack with maybe a plexiglass enclosure of some kind.Some shielding betwen them....

It's really the compromise of not having to haul in 300+ pounds of mod iron. And, a lot more efficient than class B modulators.  But the glow is still their with 4X1's in the final and modulator.

John/JN, do you still have spools of wire from that big score you made? Maybe a 5,000' spool of insulated #24, #26, etc would work - it would make a smaller coil...

T