Re: EZ81 - series resistor - first cap in CLCLC filters
Hi Pierre,
?
- Yes please put a resistor in place as per data sheet advice as this will limit surge/peak current and extend the life of the tube cathode
- I'd use 47uF if I was just using a C filter or CRC as that will be the optimum value. However... as you are using CLCLC, it's a specific circumstance that might benefit from a smaller value. If you have C1 L1 C2 L2 C3, it's good for C1 < C2 and C2 < C3 as this will help to prevent ringing in the system
Some tips that might help you with all of this:
?
- You can simulate the anode resistor by working out the source impedance using one of the transformer calculators, then simply add the value of a single anode resistor to this value. For example if you have calculated 110 ohms source impedance and have a 230 ohm protection resistor for each anode, use 110+230 = 340 ohms
- Check out the tube data as advised by others, my TDSL site has links to data sheets which you may find useful
- Check the peak current in the rectifier and compare with the data sheet. Try different capacitor values and you can see the effect of changing the value. Lower peaks are likely to translate into longer tube life
- Experiment with a smaller capacitor value at the start of the filter chain, it's really easy to try different ideas with PSUD to get to where you need quickly, look at the effects on filter ringing
- If your filter is left with a small amount of ringing, use the soft start feature in PSUD to simulate that the tube conduction will slowly rise; that should make it go away
- Sonic differences will always be subjective, and for that reason it's a topic I generally avoid, however you may want to consider a couple of items:
- Lower ripple is generally better. Single ended amps will end up with ripple in the output so less ripple will translate to less hum out of the speaker. Push-pull amps should cancel the hum, however they will be affected by intermodulation distortion so they still benefit from lower ripple. As your target is a line stage, this would be treated the same as single ended
- Power supply impedance - there are two things to consider, one is "sag" from the power supply which is where it runs out of steam at high load and the voltage will reduce, potentially resulting in quieter periods after transients. The other is the impedance to AC variations in the load which can affect the sound output at different frequencies. PSUD will check for sag; the stepped current load is great for assisting with this. It doesn't currently allow for output impedance vs. frequency to be plotted
Hope the above is of some help to you.
?
Regards,
Duncan
|
Re: EZ81 - series resistor - first cap in CLCLC filters
Have you looked at the EZ81 spec sheet and understood it?
Craig
------ Original Message ------
Date 04/04/2025 17:24:37
Subject [duncanampspsud] EZ81 - series resistor - first cap in CLCLC filters
Dear All
?
I'm going to do my first PSU... tnx for this amazing tool that is PSUD. I am implementing a PSU for an high quality line stage. The load on the PSU will not be more than 40ma and the final B+ should be about 250V. I have several EZ81 around, so that will me my rectifier. The PSU will be a CLCLC. I hope somebody can help with this:
?
1. I read that it's better to put a resistor in series with the EZ81 between the secondary and the tube, what do you think? Any experience with that?
?
2.? I was told that it's best to have the first cap really small, something below 15uF. I saw cases where people use a first cap of 47uF and cases where people use 1uF... what's the sonic difference?? I tried to simulate both, of course I get different ripple and mostly different rise time.?
?
I'm interested in sound, do these choice impact sound significantly??
?
I hope somebody is willing to help
?
Best Wishes to all
Pierre
?
|
Re: EZ81 - series resistor - first cap in CLCLC filters
Pierre,
Mullard specifies a maximum capacitor value of 50uF, and a minimum 230R limiting resistance in series with each anode (which includes the winding resistance). Bear in mind that this is specified at maximum voltage and current, which you are nowhere near. The usual thing is to use a 47uF capacitor and add whatever resistance (if any) you need to give the voltage you want.
|
EZ81 - series resistor - first cap in CLCLC filters
Dear All
?
I'm going to do my first PSU... tnx for this amazing tool that is PSUD. I am implementing a PSU for an high quality line stage. The load on the PSU will not be more than 40ma and the final B+ should be about 250V. I have several EZ81 around, so that will me my rectifier. The PSU will be a CLCLC. I hope somebody can help with this:
?
1. I read that it's better to put a resistor in series with the EZ81 between the secondary and the tube, what do you think? Any experience with that?
?
2.? I was told that it's best to have the first cap really small, something below 15uF. I saw cases where people use a first cap of 47uF and cases where people use 1uF... what's the sonic difference?? I tried to simulate both, of course I get different ripple and mostly different rise time.?
?
I'm interested in sound, do these choice impact sound significantly??
?
I hope somebody is willing to help
?
Best Wishes to all
Pierre
?
|
Missed note.? ?My new? 0-60 vdc @ 10 amp current limited labs supply? ?will not go down to 0 vdc.? ? It will? only go as low as .8 vdc which resulted in 7.2 amps? ?.8 / 7.2 = .11111 ohms? ?across the pri.?
IF the vdc could go lower, less current will be drawn.? ?My other lab supply is just 0-30 vdc? @ 3 amps also current limited.? The smaller supply will go down to zero vdc.? ?With .2 vdc it's just 1.8 amps = same .1111 ohms.?
|
OK, while on a rant,? (and misplacing my PSUD-3..........'wish list' from years ago)? ? I just remembered? a useful item that would be? an excellent addition to PSUD-3.?
And that's the option of using a resonant? choke filter.? ?On any resonant choke filter, the choke is right after the FWB assy, followed by the usual filter cap.? ?A? 2nd cap is wired directly in parallel? with the choke.?
The choke plus the cap across it, parallel resonates at the ripple freq (120 hz, or? 2F if using 60 hz power).? ?100 hz, or 2F if using 50 hz power.? ?That filter config works superb for a lot of applications, like tightly regulated screen supplies.? ?Collins used a resonant choke setup for both the B+ supply...and the? screen supply in it's 30-S1 linear amplifier.? ?Henry radio also used the resonant choke setup? for it's? B+ supplies.? The resonant choke forms a notch filter at 120 hz (2F).?
On some smaller 5 kw, single phase FM broadcast B+ supplies, the resonant choke setup was also used....and then followed up with a conventional LC filter. In that case, the resonant choke? forms a notch filter at 120 hz...... but was not enough suppression at harmonics of 120 hz, hence the conventional LC filter? was added? on.?
Dunno if a resonant choke? can be added to PSUD-3? or not.? It would be a nice addition.?
Jim? VE7RF
|
OK, here is what I did.? ?I used my newer? 0-60 vdc? @ 10 amp lab supply.? ( has adjustable vdc, and also adjustable current limiting).? Lab supply has built in digital? voltmeter and ammeter.
Short across secondary winding.....( turns out not required).? ?Tweaked lab supply so it was exactly 1.0 vdc.? Resulting current was exactly? 9 amps.? ? ?1 / 9 =? .1111 ohms.? ?.1111 ohms is a LOT lower vs the .5 ohm dcr I measured with my fluke 87 dvm.?
While at it, I used the same procedure to measure sec DCR.? 6.0 vdc applied... and current was 1.0 amps.? ? 6/1 =? 6 ohms.? ?6 ohms was exactly what I got with the fluke 87.?
Short of it is, precise measurement of? both the pri and sec DCR is essential, when plugging both into the source resistance calculator.? Fluke and other DVM's? are not accurate when measuring dcr's < 1-2 ohms.?
Ok, now source? resistance is? 61.43 ohms.? ?Before, (using the incorrect .5 ohm dcr on the pri),? source resistance was 255.4 ohms? ?Well that makes one helluva huge difference to the results.? Loaded? B+ is a lot higher.?
Turns ratio is? 5631 vac? / 230 vac =? 24.48 :1? ? ?6.0 ohms sec DCR / 24.48 =? .245 ohm pri DCR.? ?That rough calc is better then the fluke DVM's? .5 ohm dcr.?
I believe the reason the actual .1111 ohm dcr is lower than the? above rough .245 ohm calc is? because the primary winding on my 15 kva xfmr is essentially wound with flat magnet wire, it's huge material, so the DCR is very low....and the? various pri taps come out as buss bar. and terminated on the underside of the 3/8"? bolts through the red HV micarta.? ? The sec HV side is rated for 3 amps? CCS..... and presumably smaller ga wire used.
I think it was GM3SEK, eons ago, pointed out that if you knew the distance from your main electrical panel...to the xfmr on the ole out on the street, and knew what ga? aluminum was used, a calc could be done to approximate the total DCR loop resistance, external to the home.? ?That DCR could? then be added to the source resistance, to get a very realistic? loaded B+ value.?
On a similar note,? ?I took the (loaded) 5124 vac? RMS xfmr sec voltage, and multiplied it by the (loaded)? ?xfmr sec 5.02 RMS current and it's wicked.....at a whopping 25.722 KVA.?
Meanwhile, on the loaded B+ side, it's? just 6761 vdc and? 2.5 amps DC..... = 16.903 kw dc input.?
16.903 KV? dc input / 25.722 KVA =? .657? apparent power factor.? ?We only pay for actual? DC input power, not kva power.? ?Still, with the lousy .657 apparent power factor, the RMS current on the incoming 230 vac line is wicked....at 111.83 amps.?
I call it......'apparent power'? since it's NOT due to any leading / lagging phase shift.? ?On the graph, the 12.792 amp? peak current pulse (every 8.3 msec) is exactly dead center on the flattened voltage waveform.?
One other item of note. The AC ripple current is a whopping 4.14 amps rms.? The DC load? is 2.5 amps.? ?AC ripple current is 4.14 / 2.5 =? 65.6%? greater than the actual? DC load current.? ?Depending on config, it can be a lot higher than 65.6%.? ?Like on a FWD config.? AC ripple current can easily be 2.56 X greater than the DC load current.? This is why I use 2 x banks of series lytics ( in parallel) on the +2650 vdc supplies on my drake amps.
Anyway, that's my sob story for today.
Jim? VE7RF
|
I laid a dead short across the sec..... then put the fluke 87 DVM in manual ( 40 ohm)? range....instead of auto ranging.?
I will try again, but using my? B+K 875A? ?digital LCR meter, in the lowest resistance range.? ?It measures DC resistance.
I also have a newer? B+K? 875B,? BUT it measures? AC? resistance...and using a 1 khz square wave .... which does not work? on a plate xfmr.?
I will try again,? but using my adjustable? ?DC lab supplies.? ? ?#1 is? 0-30 vdc @ 3 amps? CCS.? ? ?#2 is? 0-60 vdc? @ 10 amps? CCS.? ? ?Both lab supplies have adjustable current limiting.? Both have digital readout of vdc and current.?
Temp short across the entire secondary winding.?
Then apply say? 1-3 vdc across the primary winding....and measure the resulting? ?DC current.? Then it should just be? applied Vdc / DC current =? DCR.?
With a 5361 sec and a 230 pri, turns ratio = 23.3 : 1? ? ?Sec DC resistance is 6.0 ohms.? 6.0 ohms / 23.3 =? .257 ohm? primary resistance.? ? That's what it should be aprx.??
Stay tuned. I will run some more tests and try my lab supplies.
Jim? VE7RF
|
On 3/25/2025 2:11 PM, Duncan Munro via groups.io wrote: some DMMs do a really poor job of measuring the resistance of inductive items It may help to short the other winding to suppress inductive spikes from meter auto-ranging and general digital sin.
|
Hi Jim,
?
Thanks for the detailed information - it's really handy. Pretty much everything looks good in the foregoing, except the primary resistance - that's the one that's jumping out at me.
?
Here's why...
?
I set up a .psu exactly as you described and got the same numbers. So far so good. I then checked the peak current in the secondary of the transformer, this was topping out at around 7.9A. Given the turns ratio is approx 22.34, that would suggest 22.34 x 7.9A roughly 176.5A peak in the primary.
?
Multiply 176.5A by the 0.5 ohm primary and you have resistive losses in the primary on a grand scale - that's the bit that doesn't add up for me. It's 88V loss on 339V mains peaks which means you have a 26% power loss in the primary, or it suggests a much lower value than 0.5 ohms in reality.
?
The other thing to bear in mind, based on my own experience, is that some DMMs do a really poor job of measuring the resistance of inductive items, especially when the resistance is low - don't ask me why! I'd expect your transformer to have a primary DCR somewhere in the region of 6.0 / (22.34 ^ 2) ohms. Give or take a bit, but in that ball park. To that you would have to add the incoming wiring all the way back to the power source, contactors, isolators, trips, etc.
?
I'd be tempted to do something "old school": Isolate the transformer, short across the secondary and then wire the primary from a car battery through to a headlight or something similar. Add some metering in place to measure the through current and the voltage across the winding. I have a feeling that will present a lower number than 0.5, but understandably you may not wish to screw around with the supply if you've already built it.?
?
Regards,
Duncan
?
|
Forget the 3 phase power option..... that was more just out of interest on my part.?
I use the measured? off load voltage? and pri/sec? DC resistance values? method on PSUD-2? to? arrive at the source resistance.??
IE: 5361 Vac no load secondary.? .5 ohm? DCR primary.......and a 6.0 ohm DCR secondary DCR.? ? Source resistance per PSUD-2? =? 255.4 ohms?
FWB rectifier assy? ( 16 x 1N5408's used per the drop down rectifier option).? ?In actuality, it's 24 x 6A10's for each of the 4 x legs of the FWB assy.? ?I then paralleled 2 x identical FWB assy's.
800uf filter cap assy.? ?25 x 10,000 uf lytics in series.? Each cap has .0159 ohm ESR.... = . 3975 ohms? ESR.? Total uf is? 10,000uf / 25 = 400 uf.?
2? identical banks of series caps? are wired in parallel.? ?Results in a total of .19875 ohms? ESR.? ?Total uf = 2 x 400uf? = 800uf.?
Loaded to 2.5 amps,? PSUD-2? results in a loaded B+ of just 5549.4 vdc.? ? ?P-P ripple is just 14.914 volts.?
(unloaded, with just 6 ma of EQ current, 100k resistor across each lytic, it's? 7543 vdc.?
The loaded B+ results on PSUD-2? are WAY too low? vs measured results.??
This is using a dahl 253 lb? hypersil C core? plate xfmr,? rated for 16 kva? CCS.?
But by reducing the pri DCR? down somewhat....( and keeping the sec? DCR fixed at it's measured 6.0 ohms), the resulting? source resistance is greatly reduced.......... which then results in a loaded B+? that is the same as what was measured. (6400 vdc).?
|
Hi Jim,
?
When is? PSUD-3? ? coming out ?? ? Will? it? do 3 phase power ??
PSUD3 has been out in Beta already but suspended as it still needs a little more work, there's no set date for the next release. I can confirm there is no plans to do three phase and the reasons are:
?
- PSUD2 / PSUD3 are both aimed at Hobbyists
- Likely adopters are experimenters in vacuum tube audio, ham radio, etc.
- It's not intended for industrial or commercial design, there are better tools for this such as SPICE and its variants
?
I'd like to know the P-P level? at? 240-360-480 etc.?
This is one of the features in PSUD3 is that it will export the raw data to a CSV file. Once in this format, it becomes easy using spreadsheet tools to do a harmonic analysis. In the interim, many SPICE products (I use LTSpice) can do FFT type analysis.
?
PSUD-2? works pretty good for the most part.? But it hinges on the source Z? of the xfmr.? ?That's where I have to get a little creative, and tweak a bit, so PSUD-2 results = measured results.
Are you using the two calculators built into PSUD2? These can work with either a nominal voltage and % regulation parameters, or measured off load voltage and resistance values. Both should give close results if source voltage, load and so on match up. What % out are you between what PSUD2 is saying and your measured values?
?
If you are doing some heavy duty installations where the resistance of the wire, contactors etc., becomes significant then this would need to be taken into account too.
?
Sorry if the above isn't 100% helpful, however I have a feeling you're trying to use PSUD2 for something it was never intended for, or going to be any good at. Again, there are much better tools out there for the task you have in hand.
?
73,
Duncan
?
|
When is? PSUD-3? ? coming out ?? ? Will? it? do 3 phase power ??
What would be helpful is being able to look at the harmonics, like for P-P ripple.? ?On? single phase, ripple is? 2F.? ? ?On 3 phase, ripple is 6F.?
On? PSUD-2, say for 60hz,? ?P-P ripple is done at 2F....or? 120 hz here in NA.? ?I'd like to know the P-P level? at? 240-360-480 etc.?
PSUD-2? works pretty good for the most part.? But it hinges on the source Z? of the xfmr.? ?That's where I have to get a little creative, and tweak a bit, so PSUD-2 results = measured results.
What else that would be nice is data results for the input side of the xfmr, like the? 240 vac mains from the primary side of the xfmr.?
On the sec side of the xfmr, feeding a FWB, then a cap,? and displaying? ? current pulses every 8.3 msecs...... and voltage,? ? it depicts? the current pulse, dead center in the now flattened V peak.? IE: there is no phase shift / leading / lagging.? ?Yet the? VA power on the primary side is greater than the actual rms watts consumed.?
? The indicated results on a simple? FWB + cap? setup on a hv supply? depict a lot higher? RMS current draw on the secondary of the plate xfmr? vs? ?actual? DC power drawn on the output of the supply.?
On a similar note,? on an? AL-1200? HF amp, running 1.5 kw PO,? the (measured) peak? current every 8.3 msecs? ?on the incoming 240 vac mains is a whopping? 61 amps !? ?That amp uses a simple? FWB + 25 uf? filter cap? setup. All that high peak current pulses on the pri side all have to pass through relay contacts / switches / contactors / wiring etc..... causing more V drops.
The other story I got was if trying to measure AC current on the mains side, if a true rms AC ammeter is not used ( or an iron vane style panel meter), the indicated AC current will read? 20-40% lower than it actually is, if a FWB + C filter used.?
|
Hi Jim,
?
is this difficult to add on ?? ? Or can it be part of the ....'PSUD-3'? ??
?
?
The good news is it's easy to put on, the bad news is that it will need a whole sequence of tests to be carried out before the software can be up-versioned and re-released. Given that PSUD2 is in "sunset" mode, it would represent a significant amount of work just for one usage case, sorry.
?
It was set at 10kV as I figured the highest voltage use case would be for vacuum tube linear amplifiers, anything else would be venturing into industrial / exotica so best served with one of the SPICE products as suggested by Paul.
?
73,
Duncan
?
|
Dunno if it would scale....or not.? ?How do I scale a huge plate xfmr ?? ? Do I multiply the source Z? by 10? as well? to simulate the 1/10 size plate xfmr ??
|
Everything scales. Pretend 1.0KVAC. Instead of 0.5Amp current, draw 0.05Amp. Result will be 1.414KV which you multiply by 10 to get actual voltage. R and C and %ripple values will be unchanged.
Or wrestle with the Spice, PSPICE, LTSPICE or TINA of your choice.
(I won't say 'sliderule'.)
|
Can the max vac on the sec of the plate xfmr be increased...... like up to 15 kvac ??
As is, it's maxed at 10 kvac. ( aprx? 14.14 kvdc? ?with a? FWB and? C input filter).?
I was trying to simulate a 21 kvdc? B+? supply...and can't.? ?That was for a water cooled transmit tube.?
In other cases, 30 kvdc is required for a hi-pot tester....( for testing insulation break down, or between elements in a tube).? ?Used for all sorts of various applications.?
is this difficult to add on ?? ? Or can it be part of the ....'PSUD-3'? ???
|
Back up! Thanks.
I also deleted all the PSUD registry data as I had some Beta 3
entries.
NOT recommended for everyone UNLESS you REALLY know what you are
doing.
- Ian
On 3/10/2025 00:52, Duncan Munro via
groups.io wrote:
toggle quoted message
Show quoted text
Hi Ian,
?
If you're getting the fullwave solid state diagram as above
then PSUD2 is initialising and then failing at the point at
which it tries to load the default PSU which is located
(typically) at:
?
? ?
C:\Users\<myname>\AppData\Roaming\psud\examples\default.psu
?
If this is failing, it's because the default.psu is missing
or corrupt (it should give a message), or the 1N4007 rectifier
profile has been renamed, removed or corrupted badly. It needs
the 1N4007 rectifier for the default PSU to load.
?
A couple of other items:
?
- I'm using exact same build of Windows 11 as yourself so I
don't think there's a Win version issue
- I would probably shy away from using a text editor to view
the PSU files as they are binary; I use HXD which is a great
tool for checking binary content. Only mentioning this as if
you try and load into a text editor and save again, the file
will become "unwell", but I guess you know that anyway.
Rectifier files are text/XML so they don't have any problems
in this regard
?
?So, how do we fix this? I'd recommend the following steps:
?
- Check C:\Users\<myname>\AppData\Roaming\psud\examples
and \rectifiers for custom rectifier and files you wish to
keep - save them somewhere else
- Deinstall PSUD2
- Using file explorer, wipe out
C:\Users\<myname>\AppData\Roaming\psud
- Reinstall PSUD2
?
Hopefully, this will take things back to a stable position,
let me know how you get on.
?
Regards,
Duncan
?
?
?
|
Hi Ian,
?
If you're getting the fullwave solid state diagram as above then PSUD2 is initialising and then failing at the point at which it tries to load the default PSU which is located (typically) at:
?
? ? C:\Users\<myname>\AppData\Roaming\psud\examples\default.psu
?
If this is failing, it's because the default.psu is missing or corrupt (it should give a message), or the 1N4007 rectifier profile has been renamed, removed or corrupted badly. It needs the 1N4007 rectifier for the default PSU to load.
?
A couple of other items:
?
- I'm using exact same build of Windows 11 as yourself so I don't think there's a Win version issue
- I would probably shy away from using a text editor to view the PSU files as they are binary; I use HXD which is a great tool for checking binary content. Only mentioning this as if you try and load into a text editor and save again, the file will become "unwell", but I guess you know that anyway. Rectifier files are text/XML so they don't have any problems in this regard
?
?So, how do we fix this? I'd recommend the following steps:
?
- Check C:\Users\<myname>\AppData\Roaming\psud\examples and \rectifiers for custom rectifier and files you wish to keep - save them somewhere else
- Deinstall PSUD2
- Using file explorer, wipe out C:\Users\<myname>\AppData\Roaming\psud
- Reinstall PSUD2
?
Hopefully, this will take things back to a stable position, let me know how you get on.
?
Regards,
Duncan
?
?
?
|
Today I tried to load PSU files and always get the same layout which
is the stock after install file.
Beta 78 & 79 behave the same.
PSU files I'm trying to load are all correct date and differ
considerably when viewed in a text editor.
Files date from two to four years ago.
PSUD uninstalled and Beta 79 installed again.
Note loaded file in bar:
 
Same cct.
--
- Ian
|
Paul Reid
Ian Eales