Film Cathode Resistor Bypass Caps

[Deke609]

Following the lead of PB and Alonzo (https://bottlehead.com/smf/index.php?topic=11590.0) this morning I replaced the stock electrolytic cathode resistor bypass caps with film caps. I used Solen PA10000 100uF 250VDC.

There is a noticeable improvement in sound - lots more texture across the entire bandwidth (aka "detail" - but in a lively/robust, not sterile/analytical, sense) and the dynamics got even better. The amp now sounds at once more at ease and more decisive - if that makes sense.  I had intended to scope the amp today in preparation for playing around with the power supply, but liked the results of the new caps so much that I ended up listening to music all afternoon!

cheers, Derek

[Alonzo]

Great, glad it worked out for you!  I'm looking for a reason to get underneath and add some bypass caps or some Russian PIO or Obligatto film oilers but it sounds too good to mess with now.
Alonzo

[Deke609]

I'm looking for a reason to get underneath and add some bypass caps or some Russian PIO or Obligatto film oilers but it sounds too good to mess with now.

Based on my experience today, I don't think film caps in the Kaiju power supply make a big difference. I removed the 2 stock electrolytic pairs (2 x 220uF 350V) and hooked up 2 x 100uF 700VDC metalized polypro caps in their place. Does it make a difference?  Maybe, but I'm not sure. Perhaps a bit of added clarity in the upper end. But that could be imaginary. In contrast, replacing the electrolytic cathode resistor bypass caps with the film caps resulted in an immediate and obvious improvement.


I plan to keep the new big caps in the PSU, if only b/c they sit above (or below, depending on your perspective) the PEC pots and 4 pin sockets, freeing up a bunch of space around the power switch for power supply chokes.  We'll see whether the reduction in ripple has any noticeable positive effect. I may be at the point of vanishingly small returns with my tinkering. but it's fun ... and gives me a reason to figure out how to use a oscilloscope. 


cheers, Derek

[EricS]

Derek - where are you measuring ripple?  I'm curious to learn how much ripple you are seeing at the plate of your 300B tubes.  I am finding about 3mV ripple on my 300B plate voltage of ~340v and the same amount on my driver tube with a plate voltage of about 170vDC.    The 300B has a CLCL filter and the driver has a CLCRCR filter. 

My latest fiddlings with the 300B filament supply (now a CLRC filter) have reduced ripple to about 8-10mV.  I need to increase the resistance just a pinch to bring the filament voltage down.  I'm targeting 5.1vDC on 122.5vAC mains, they seem to wander from 120 to 125vAC on a daily basis.

[Deke609]

I'm curious to learn how much ripple you are seeing at the plate of your 300B tubes. 

I'm really curious too!

Confession: there is a particularly intense level of idiocy in my brain reserved specially for oscilloscopes.

I tried measuring ripple this am with both Fluke 117 DMM and a new Siglent o-scope.  Problem: both are picking varying levels of ambient AC signal that bounces around from 5 -30mV. I think it might be coming from my wireless router. To test this, I held the DMM leads close to the router and registered a bouncing ac signal approaching 100mV.  So I probably need to power off the wireless.

A potentially compounding factor: my mains voltage bounces around too. It's currently reading 117.2 - 117.8, and showing 100mV variations close to every second - which makes me think that smaller variations (e.g., 10-20mV) are happening much more frequently.

What I see on the scope when I adjust the horizontal and vertical divisions to look for 120Hz ripple: a big fat fuzzy block of noise.

I haven't yet figured out how to set the trigger conditions - I'm hoping that they can be set to pick up on changes in slope reoccurring at a certain frequency.

I'm sure this is all user error/ignorance. I just need to figure out how to properly operate a scope.

Any pointers?

cheers, Derek



[edited to remove all the double line spacing - I'm not hitting enter twice so I have no idea why it's there - if someone knows, please post]

[Paul Birkeland]

There's probably a math function, and you can set a low pass filter at 5kHz that should leave something more useful on the screen.

[Deke609]

Many thanks PB. Yeah, it's got a bunch of math function capabilities. I'll look into that later today.

[EricS]

I have to admit that I'm no pro with a 'scope.  I have an old analog scope with a CRT built into it that was given to me years ago.  Every now and then, I look at the new digital scopes, but I've decided that I like the price of mine better.  It seems to work well enough for my purposes and would probably be more useful if I had time to learn about all of its functions...  Most of the measurements I posted above are done with a DMM set to mV AC scale and see what comes up. 

I have no idea of what kinds of figures are "good" or what "normal" ranges might be, so I thought this might be an interesting comparison point to learn a little.

[Paul Birkeland]

The only number that really matters is what's seen at the speaker binding posts.  There is definitely a point of diminishing returns where when you get to under 1mV RMS, it gets hard to rationalize getting much quieter.

When you are looking at where noise is coming from in a directly heated single ended amp, it can be helpful to clip lead a big film cap in various places to break things up.  8-10mV in a DC heated 300B amp would have me looking for a floating driver heater winding, or some other fundamental issue beyond a lack of filtering (it sounds like you have way more than enough filtering).

[EricS]

I would agree- the binding posts are the most important place to start as it will suggest that there are potentially problems elsewhere that need to be addressed.  I was pleased to see 0.0mVAC at my speaker terminals, but continued to poke around at other places in the circuit merely out of curiosity with my first tube amp.

[Deke609]

Eureka! PB's FFT suggestion did the trick:I can use it to dial in the frequency I wan to look at. Setting it for 120Hz, I see approx. -28dB at the first ps cap (10uF) after the first ps choke (4H). At the second ps cap (200 uF) after the 2nd choke (6H), there's nothing at 120Hz - or whatever there is, is within or under the noise floor of approx. -80dB or so.

I'll have to do the dB to mV calculation (PB can probably do that in his head), but I'm guessing I have no discernible ripple.

Interesting thing: the scope suggests that the 120Hz ripple is actually at 120.5Hz - not sure whether this is a scope calibration issue or my mains is actually approx. 60.25Hz. Guessing it's the scope.

Not sure what to make of the MHz noise I see- is that just coupled-to/induced-in the probe leads, or is it actually in the amp (or both)? But I guess since it is way above the audible range it doesn't matter.

cheers, Derek

[grufti]

Your Mhz noise is more than likely coupled to the probe leads. There is a very good article online on the subject of measuring noise from switching power supplies. You can probably find it by searching for those terms. It discusses the issues related to Mhz noise measurements in great depth and reasonably understandable terms. I'll look for it as well and will post a link, if I find it again. One thing that I remember: the ground lead on your probe is a great antenna. The author suggest measuring with a stripped down probe and shows exactly why.

[Paul Birkeland]

For audio measurements with gear not containing switching supplies, it's often helpful to cut everything off over 1mHz.  There are some instances where you may want to hunt down oscillations, and that will require the high bandwidth, but for chasing audible noise it will make life more difficult.

[Deke609]

I cheated and used a dBV calculator instead of looking up the formula nd doing it myself: it looks like about 44mV at the first ps cap and 0.1mV at the second. I'm not sure the scope can be that accurate, so I'll just conclude that 120Hz ripple is insignificant after filtering.

@ grufti: thanks. I'll try to find that article. I read something similar recently by Rod Elliott (I think). I wonder whether shielding the probe leads with copper braid would work? And then maybe attach the drain wire to the ground tab at the front of the scope? Will look into this.

cheers and thanks, Derek

[Deke609]

Since I started this thread about the cathode resistor bypass caps, I should probably say something about my experience with the three mods discussed above. Did the following make a difference?

(a) film caps in place of electrolytic in the power supply: possibly, but any audible benefit is tiny, and quite possibly imaginary [Edit: but the result might have been very different had the amp been series feed (output transformer in series with output tube) rather than parafeed (ot in parallel)]

(b) combination of film caps and chokes (replacing resistors) in the power supply: I think so, but the effect, if there, is small [Edit: ditto]

(c) film cap in place of electrolytic capthode resistor bypass cap: definitely and plainly noticeable right away - and I think it's improved as the cap got some use.

So, if I had to put numbers to it all, I'd guess the combo of caps/chokes in the power supply is responsible for <20% and the film bypass caps are responsible for >80%.  The film bypass cap mod was also the cheapest and easiest: I think I paid, with shipping and taxes, about $50 CDN (approx $35 US) for each Solen cap, so $100 CDN ($70 US) in total. And they took all of 5-10 minutes to install - most which time was spent lifting and flipping the chassis of the beast and waiting for the soldering iron to get hot.

The power supply caps and chokes were about 3 X the price of the cathode bypass caps and the install was a lot more involved (drilling chassis, making cap retaining brackets, getting up the nerve to turn it on after it was all installed for fear of explosion).

All that said, the ps caps and chokes were the most fun 

Sadly, i think this means that the Kaiju is now done.  I've really enjoyed tinkering with it. It has so much free space (or, at least it did) that it just calls out for tinkering.

cheers, Derek