Mercury Power Supply Design - Checking for Efficacy

[Grainger49]

Alex,

This is a stereo amplifier.

There will be one power transformer for the B+ (500V-0-500V) feeding a single GZ33 (dual rectifier).  That will feed four filters, driver and output tubes, right and left channels dual mono after the rectifiers.  This is an old Bottlehead trick of PDMPS (pseudo dual mono power supply).

So when we talk about the 50uF that the rectifier "sees" it is the total of the four input caps.  Then each string will have a choke and a film cap to feed the voltage to the tubes.

Does this make sense???

 

[Paul Joppa]

Sorry I missed the switch - you are now looking an a CLC filter instead of LC or LCLC? That will increase the power supply voltage 35-40% - like, 600v or more from a 500v winding. That's a LOT, and way too much for 76s or 45s.

Am I missing something?

 

[ALEXZ]

Grainger,
Now I'm lost .... How rectifier " sees" four capacitors ? If rectifier is loaded by capacitor, it's one capacitor, never mind if it is made out of four parallel connected capacitors. If rectifier is loaded with LC, there are no capacitors at all, rectifier "sees" inductors first. Also, with regards to  dual mover supply (psudo or otherwise), I'm not clear why driver and output tube need to have separated filters.
Speaking of output voltage, "Typical operational conditions" for GZ33 shows with 500v AC and C load (from 8uF to 60uF) output voltage 479-493 v DC with 250 ma (!!!) current.
I not really comfortable making recommendations, i hope Paul will comment, but with 1000vCT transformer, the way to go is LCLC or LCRC for each side with additional R for bleeding.

 

[Grainger49]

Paul,

We modeled this on PSUD II with a GZ33 rectifier and the 500V transformer.  We didn't get that high a voltage out at all.  We had started with a 375V transformer and had to keep increasing the voltage.

I think it would be best if I can somehow post the circuit from PSUD II.

Alex,

The four capacitors are the first caps of the CLC filter.  They are all across the rectifier's output so they sum to 50uF.  If we model with a 15uF cap the rectifier has too high an inrush current.  So we were modeling with the 50uF as the input cap.  I made no mention of an LC filter above.

Here is a link for the PDMPS on VoltSecond's site to catch you up:

http://www.siteswithstyle.com/VoltSecond/PDMPS/pseudo_dual_mono_foreplay.html

I don't understand your emphasis on 250mA current, is it necessary to draw full current to get the 479-493V output?

There will be a ~0.5M ohm bleeder on each B+ leg.

 

[ALEXZ]

Not sure what  PSUD II shows. Did you design four separeted CLC filers with it ?  Anyway rectifier GZ33 with 50uf input and current much less then 250 ma will give you  from 1000vct transformer voutage higher then you need.  
250 ma is important to get voltage drop on internal resistance of rectifier. Lower current, smaller drop.
Put GZ33, 1000vct, 50uf and current of 2x76 +2x45 into PSUD and check what voltage you will get.
Bleeder resistir: 500k - it's  less  then 1 ma , won't help to much with rectifier/ filter load.

 

[Grainger49]

Since there are pairs of rails, one pair for drivers one pair for 45s, we only looked at two CLC filters, so we did cover all four.

So you are saying that we need more current drawn through the rectifier to get a lower voltage, right?

I'll try to post PSUD shots later today.

We are not looking for the bleeder to be part of the load, just a safety device.  Most Bottlehead equipment, maybe all, has a bleeder of 400k+ ohms.  It is simply for bleeding the PS dry when you turn it off.

 

[ALEXZ]

Grainger, just look at the first stage of the power supply. You have 2x500v AC, full wave rectifier loaded by 50uF capacitor. If you have no load and no losses in the rectifier, you will get ~700v DC on the capacitor. Now you're loading this PS with 4 tubes via LC filters. That's were current starts to play a role. All of the voltage drop will be on the rectifier (there are relatively  small voltage drop on the DC  impedance on a choke in CL filter, as opposite  to LC).
I downloaded this PSUD software. With 1000VCT/GZ33/47uF/3H/47uF/120ma load  PSUD shows 600v output.

 

[Grainger49]

Alex,

Thanks, I have decided, for some reason, PSUD is not computing properly on my computer.  When I "simulate" I can not read the scales, neither the left side or the bottom.  I see what you are saying.  The resulting voltage is too high.  And I'm with you, there is only a drop in voltage across the inductor if there is appreciable current.  Since there is not much current total here.  I would guess less than 100mA.  We wanted a Beefy power supply. 

Since I can't run PSUD well on my computer, what transformer voltage would you suggest for a 425V output?

Your trace is on the first cap and not after the inductor so I won't worry about the ripple.  How did it look on C2?  Was it smooth? 

BTW, how did you get a JPG of the screen???  I'm not much of a computer guy.  I'm old!

 

[ALEXZ]

Ok. Last thing first. Whatever screen is active, you press alt+contr+shift+print_screen and after that go to whatever image apps you like (Microsoft Paint will do) and paste into into apps (could use contr+v)
Control+shift+print screen will cupture entire screen, with ALT only wiindow "in-focus"

For 425vDC after gz33 with C input, gz33 datasheet shows 2x375v AC (750vCT).
I installed PSUD in the office. Will let you know what's data I can get from it tomorrow.

 

[Grainger49]

Thanks for it all!

 

[ALEXZ]

here we go ...2x475v AC

 

[Grainger49]

Beautiful!  It never reaches 500V, the rating for the inductor's insulation.  Looking at the RMS for R1 it is up at 379V.  Interesting, the graph shows different voltage than the RMS value in the chart next to it.  

This is part of my confusion about PSUD II.  The RMS voltage should be the steady state supply voltage.  That is why I'm so befuddled with PSUD.

 

[Grainger49]

Ok, I tried with all the same components you have and I got much different results.

WOW  That is tiny.  Click and it will open in a window.  As you can see I have ripple across R1.  I don't think any of the voltage and current values are the same as you have.


This is why I get frustrated with this program.

 

[Paul Joppa]

Grainger, your first capacitor is 50 farads, not microfarads. That will make a difference!

 

[Grainger49]

I'll look at it again, Thanks!

We are still wondering, looking at Alexz's example above, how do you start with a transformer that puts out 750V and R1, the load has a calculated RMS voltage of 379V?  The graph of Vc1 shows DC of 480-490V.

What am I missing?

 

[Grainger49]

Second try got almost the same numbers as Alex.

 

[Paul Joppa]

Oh yeah - I forgot to post this one. It's bit me in the backside more than a few times!

The average, rms, max, min, etc. cover the entire reported interval. If you don't use a delay, then it includes the initial start transient. A ramp from 0 to 500 volts is an average  of 250 volts. You have to use the "after an initial delay of..." if you want to know where the voltages etc. end up.

By the way, there's another hidden time bomb. If you use a constant current load instead of a resistor, it offers the option of changing the current after a delay time. Very good for approximating tube warmup - but in my experience this gets introduced without your asking for it, at seemingly random times. If it looks weird, re-check the current load specs!

Oh yeah, here's another one. The transformer is described by its UNLOADED voltage plus its DC resistance. Transformers are rated by their LOADED voltage, after the resistance has dropped the voltage due to the entire rated current. The difference is usually at least 5% - so if you use a generously over-rated transformer the voltage will be at least 5% higher than the rated voltage. Plus, most transformers today are still rated for 115 volts input, and most houses have 120 to 125 volts. There's another 5% to 10%. So a transformer rated 375v may have an unloaded voltage of as much as 430 volts. And by the way, the resistance is the secondary resistance PLUS the primary resistance multiplied by the turns ratio squared. There's a calculator hidden in PSUD, but a lot of people aren't aware they need to find it and use it. Bottom line, you need to buy the transformer, measure the heck out of it, and do some calculations before you can actually put it in PSUD and expect realistic results. You can't just use the specs in the catalog. Serious pain in the aforementioned backside.

It's a really cool simulator and very useful, but it does have a lot of traps for the unwary! You know how, in house painting, the surface preparation takes more time than the actual painting? Same thing with PSUD - getting all the numbers and getting them into the program without triggering any hidden "features" takes more time than using it to get useful results.

 

[Grainger49]

Paul, Thanks!  I am beginning to see the light at the end of the tunnel.  I added the delay and the voltage plot is a sine wave.  But the ripple goes from 476.58V to 476.56V, 20mV.  Additional resistance should bring that down, right?  Or is 20mV good?

PSUD is a free program and so is EAC.  Neither seem well documented.  But I got what I paid for.

If we need to shed voltage do you suggest a resistor before the tube Anodes, after rectified or in series with the inductor.  I am thinking in the AC legs will soften the blow on everything.