Need Help with First Tube Build - WE91 300B Parafeed Derivative

[EricS]

Ah, I think I'm beginning to understand a little better now.  Given the relatively large size of the bypass cap and the very small mA draw of the tube, I think I was adjusting too quickly to see the effect that I was looking for.  I'll experiment again tomorrow and be a little more patient as I adjust things.  I'll be sure to get a meter on the plate as well.  Somehow, I think I may have missed the obvious  :-[

 

[Paul Joppa]

If you're going to put a pot in for the cathode resistor, you will want to adjust it very, very slowly to get the target plate voltage.  You are looking to be very, very close to your original value.  If you're off by more than a couple hundred ohms in one direction or the either, you will fall off a cliff so to speak in terms of operation.

I've been out of town for the holiday. Fortunately, this post by PB covers what I was going to say.

 

[EricS]

The 6SJ7 does seem to be out of spec.

You nailed it, Paul!  The tube I was using sure is out of spec!  Like most problems in life, this one is another a case of user error  :-[.  While I was adjusting R6, I was also measuring voltage drop across R6 expecting to see a much larger voltage change.  I had assumed voltage drop across R9 and R6 would somehow be constant and would co-vary, so I only put a meter across R6, not R9 that leads to the plate - duh...

So, while measuring the PLATE voltage at pin 8 (thanks for the kick, PB), adjusting R6 actually makes a big difference!  It was easy to dial in 170vDC at the plate once I started measuring correctly!  The problem is that until I was satisfied that the amp was working properly, I was using my most-used-looking driver tube and it seems that it is furthest out of spec - it takes a value of 873R to hit 170v on the driver plate.  The remaining 10 tubes are all clustered between 1200R and 1500R (average of 1327R) in order to hit a plate target of 170v.  Looks like individual tubes are just as variable as solid state transistors... 

I'll re-run my gain and bandwidth measurements and see where things stand next.  I'll also post an updated voltage map of the amp shortly.

Very happy now!    Many thanks for the help and support, guys!  I hope everyone had a nice holiday weekend!

Question: Is there any value to placing a pot in series with R6 so that adjustments can be made as any individual tube ages? 

 

[Paul Birkeland]

Pentodes are a fair bit touchier than triodes are.  You can expect the plate voltage to be a little different for each tube you use, there's nothing wrong with that.  I think you could stick with 1.4K cathode resistors and call it good. 

I wouldn't bother with the pot, good working tubes should all bias up to an acceptable plate voltage.

 

[EricS]

Now that I've got a better handle on the driver plate voltage and it is dialed in, it dragged down overall B+ by a few volts, which also reduced the output of the 6.3v secondary winding for the tube filaments.  Adjusting the driver tube back to ~6.3v was cake.  Adjusting the 300B back to 5.0v-5.2v range that Jac recommended is taking some more experimentation.

The original configuration was diodes --> 154B chokes (3mH, 0R15) --> 10,000uF cap --> 300B filament, but now my voltage is too low.  I see two paths and was hoping someone could provide a sanity check for me:

1) Put a smaller cap (1,000uF 35v) between the diodes and chokes to make a CLC filter.  This worked well and I was able to hit a filament voltage of 5.1vDC with 122.5v (pretty typical) coming out of the wall.  Using my variac, I was able to experiment with the impact of different sized caps and get an acceptable window of filament voltages over the typical range of my AC mains voltage.  I was pleased with this result until I measured ripple on the 1,000uF cap at 1.5vAC.  With an ESR of 0R016, this seems like a ripple current of 1.5v/0.016R = 93 AMPS ?!?!?  Is this calculation correct?  If it is correct, this poor little cap surely won't last very long before it pops....

2) Use two equal size caps (6,800uF ?) in the CLC (which results in filament voltage closer to 6.0v) then add a power resistor on each leg of the filament supply (because of the center tap hum pot) to drop the filament voltage down to the desired target.  This seems like a more stable approach, but I also want to do some additional measurements to make sure that it doesn't delay warming the filament relative to when the B+ arrives at the tube.

Any thoughts or reactions?

 

[Paul Birkeland]

1) Put a smaller cap (1,000uF 35v) between the diodes and chokes to make a CLC filter.  This worked well and I was able to hit a filament voltage of 5.1vDC with 122.5v (pretty typical) coming out of the wall.  Using my variac, I was able to experiment with the impact of different sized caps and get an acceptable window of filament voltages over the typical range of my AC mains voltage.  I was pleased with this result until I measured ripple on the 1,000uF cap at 1.5vAC.  With an ESR of 0R016, this seems like a ripple current of 1.5v/0.016R = 93 AMPS ?!?!?  Is this calculation correct?  If it is correct, this poor little cap surely won't last very long before it pops....

The cap input filter is what should be done to bring the voltage up.  Depending on how much additional voltage you need and what rectifier diodes you are using, you could also use more efficient rectifier diodes to bring the voltage up.  While the 1,000uF cap has an ESR of 0.016 ohms, it's capacitive reactance at 120Hz is 1.3 ohms.  Still, for this reason, you'll commonly see a few caps in parallel to share this abuse. 

2) Use two equal size caps (6,800uF ?) in the CLC (which results in filament voltage closer to 6.0v) then add a power resistor on each leg of the filament supply (because of the center tap hum pot) to drop the filament voltage down to the desired target.  This seems like a more stable approach, but I also want to do some additional measurements to make sure that it doesn't delay warming the filament relative to when the B+ arrives at the tube.

That will work and is an approach we use.

You can also rectify the 6.3V winding on the PGP 8.1 with a choke input filter and a big cap.  1N5820 diodes and the Hammond 155B got me to 4.95V in my own 300B project. 

 

[Deke609]

  While the 1,000uF cap has an ESR of 0.016 ohms, it's capacitive reactance at 120Hz is 1.3 ohms. 

I don't mean to derail the thread, but for those of us newbies trying to follow along, does this mean that what we need to look at to avoid filter capacitor "abuse" is the ratio of ripple to capacitor impedance (reactance + equivalent series resistance)? And that, in the case of very low ESR caps acting as reservoir or filtering caps in a power supply, capacitive reactance dominates and is all we really need to look at for practical purposes?


many thanks, Derek

 

[EricS]

Thanks for the confirmation, Paul!

You can also rectify the 6.3V winding on the PGP 8.1 with a choke input filter and a big cap.  1N5820 diodes and the Hammond 155B got me to 4.95V in my own 300B project.

This is where I was until I adjusted the plate voltage on the driver tube.  Then my 300B filament fell from 4.95v to 4.50v.  Moving to the 154B chokes only got me back to ~4.75v.  Still too low.

I'll continue to experiment with the CLC approach with similar sized caps and a few dropping resistors.  I was hoping to avoid using any electrolytics at all - oh well...

Derek: I'm not quite sure how/where Paul determined capacitive reactance at 120Hz to be 1.3 ohms.  I don't fully understand the implications of this measure.

 

[Paul Birkeland]

Reactance is AC impedance.  Capacitors and inductors have impedance that varies with frequency.  The impedance of a capacitor is usually (hopefully) higher than its ESR.

I'm surprised you didn't get more voltage from the 6.3V winding, especially if you are using a 154B.  You are using 1N5820 diodes?  (and your bog standard testing 300B?)

 

[Deke609]

Derek: I'm not quite sure how/where Paul determined capacitive reactance at 120Hz to be 1.3 ohms. 

I think it's the textbook equation for cap reactance (that I've read a million times but never used): Xc = 1 / (2*pi*f*C) = 1 /(6.28*120Hz*0.001F) = 1.32Ohms

Edit: And in quoting the formula for cap reactance, I think I've answered my own question above about relative importance of reactance versus ESR. Since reactance varies inversely with capacitance, the smaller the capacitance the higher the reactance. This suggests to me that for low ESR caps, reactance will dominate unless the cap is insanely large - e.g., tenths of a Farad.

 

[EricS]

Ah - I've seen that formula before - I'm still learning how to wield each of these great tools!  I'll get there eventually  .

I was surprised by the low rectified voltage from the 6.3v winding as well.  I was expecting something north of 5.0vDC with a choke input.  I'm using 80SQ045NG diodes (https://www.mouser.com/ProductDetail/ON-Semiconductor/80SQ045NG?qs=%2Fha2pyFadujIGlmyYa9M8%252BWQ31K%252Bb%252BCtAaPPtZbg%252Bfg%3D) that have a 0.55v forward voltage drop.  I also tried using IXYS DSS 10-0045B diodes (https://www.mouser.com/ProductDetail/IXYS/DSS10-0045B?qs=%2Fha2pyFadug59d17CRjWjt%252BNDUp2lITVLL7Zh40I8PUxMJTDr5BNUA%3D%3D), but the final filament voltage was identical to the previous.  I'll add some 1N5820's to my next order - the curves on the data sheet look encouraging.

The EML mesh tube draws 1.5A for the filament, so this is dragging things down a bit.  There is a clear difference in voltage level between my PSVane tube and the EML tube.  All of my notes are at home right now, so some of this is from memory.

 

[Deke609]

The EML mesh tube draws 1.5A for the filament, so this is dragging things down a bit.

That may be your culprit. I use EML tubes in all my BH amps and the additional filament current draw always drops filament voltage below Jac's targets and requires tweaking (which is pretty easy in BH amps b/c they employ dropping resistors at the filament that can be decreased). I just did some really quick and rough PSUD sims of a 6.3V filament power supply and switching from normal 300B current draw of 1.2A to 1.5A dropped voltage by approx. 250mV. 


cheers, Derek

 

[EricS]

Yep, I can confirm that missing 0.25v!  I was at 4.98v on the filament and doing my happy dance.  Then I adjusted the plate voltage back to target and switched to the EML tube and was suddenly the party was over.

I'll just use a larger first cap for the CLC to reduce the ripple current to sane levels and then insert some resistors to hit my target.

Caps on the filament PSU definitely reduce the audible hum from the speaker, even when the hum pot is dialed in.  At first power up, I had 5mVAC on the speaker terminal.  I could get it down to 0.0mV by adjusting the pot, but I could still hear it (I have no idea what the sensitivity my old test speaker is).  Adding a cap cured this by reducing ripple on the filament from 1v5 to 80mV.  Adjusting the hum pot now produces a very subtle difference because the hum is so very low.  These tests need to wait until the kids are in bed so the house is quiet.  The CLC takes it down to something closer to 35mV at the filament.

 

[Paul Birkeland]

Yeah, 1.5A is no longer a 300B, and you'll lose a lot of filament voltage because of that.  The Hammond 156B might be a better choice, or more likely a different set of tubes.  If you intend to design and amp to use both, aim to be slightly low on the EMLs and slightly high on a 300B.

 

[Deke609]

How about one of these Hammond 8.5V CT @ 3A filament PT's for each EML 300B mesh? https://www.mouser.com/ProductDetail/Hammond-Manufacturing/166M8?qs=sGAEpiMZZMvwUzoUXIIvyQPvPmwnNFGywL5FidNAeTk%3D

There's also a 2A version for about 4 bucks less.

You'd have some extra voltage to drop, but that would give you lots of wiggle room for filter stages to get ripple super low.

cheers, Derek

 

[Paul Birkeland]

It's easier to go down. The choke after the bridge just needs more DCR.

 

[EricS]

I did contemplate a separate PS for the filaments. 

Then I started thinking about the quantity of large parts that already need to go under the hood...  And that it would be one more part to spew an electromagnetic field...  and how that might just create a bigger problem than I already had on my hands...  For now, I think I'll stick with some additional caps and resistors.

Darn that "no free lunch" thing  ;D

 

[EricS]

While the 1,000uF cap has an ESR of 0.016 ohms, it's capacitive reactance at 120Hz is 1.3 ohms.

Now that I have some time to think about this, does this mean that my previous calculation of ripple current was way off?  Is a more accurate calculation based on frequency specific reactance of 1v5/1R3 =  1.15A of ripple current?

 

[Paul Birkeland]

If I had concerns about ripple current, I would model the supply in PSUD, then insert an AC ammeter between ground and the negative lead of that cap and actually measure the current.  You'll need a meter with a current shunt that's got very low DCR though.  You could also put a 0.1 ohm resistor between ground and the negative terminal of the cap in the absence of a low resistance current shunt in your meter, then measure the AC voltage across this.  You can also try a 0.01 ohm resistor to get even closer to the actual value.

Doing this will eliminate any unpredictable behavior from your meter attempting to resolve AC voltage riding on DC voltage. 

 

[Deke609]

Just doing some "just in case" thinking here.


@PB - Could the Kaiju DCF handle a 1.5A draw, maybe with heavier duty heatsinks?


It looks like the board could be scored and split into two separate reg boards.  Since the DCF was designed to provide about 5V DCV after approx 1.2V filament resistor drop, it might offer a solution to the voltage sag issue with the EML tubes.

cheers, Derek