It's time to go full Karna Kay

[Max N]

Well, It turns out that 2 x 68R series = 134R as a shared cathode resistor actually gets you slightly closer to Mullard's 250V / 34mA operating point than 270R parallelled (or in each leg). According to the model, that is.

It might be nice to build with the two 270R resistors with a switch to optionally connect them in parallel or leave them separate. Then you could do the experiment and see which sounds best.

[Thermionic Idler]

Well, It turns out that 2 x 68R series = 134R as a shared cathode resistor actually gets you slightly closer to Mullard's 250V / 34mA operating point than 270R parallelled (or in each leg). According to the model, that is.

It might be nice to build with the two 270R resistors with a switch to optionally connect them in parallel or leave them separate. Then you could do the experiment and see which sounds best.

Yep, I think you're right - a lot of the circuit is migrating onto PCB's as part of the rebuild process, so a simple jumper could probably do the job.

I think I'm also going to revert the interstage back to the LL1692A - the recommended PP to PP configuration steps up slightly instead of stepping down. The EL84's will cope with this easily and it'll give a bit more headroom to the driver stage which is one of the key aims of the design - the output stage needs to be well into grid-positive territory before the driver stage even thinks about starting to clip. Reducing the R between CCS and EL84 anodes to 100R helps with that as well - I honestly can't see any reason for them to be 1k and dropping 34 volts of headroom, as long as the CCS heatsinks are big enough.

[Nick]

No neither can I. There is some merit in a 100R or so resistor in series with the CCS, partly its a useful place to measure current and it also helps to provide some resistance to VHF that will find its way past the CCS due to capacitance effects.

[Thermionic Idler]

No neither can I. There is some merit in a 100R or so resistor in series with the CCS, partly its a useful place to measure current and it also helps to provide some resistance to VHF that will find its way past the CCS due to capacitance effects.

Yep. All that happens when I change the range on my digital multimeter is the decimal point moves one position to the left, so it's not like it's any less accurate a measure of current.

[Max N]

TI, I can’t remember if you have a scope or measurement interface. If you do, then you could also use the voltage across one of the cathode resistors to see the behaviour with the switch in both positions, whether any signal current flows through the valves in each position, etc

[Thermionic Idler]

Now then, power supplies. Here we address power supply 1 which exclusively feeds the 300B output stage.

Going to solid state, I didn't just want to blindly copy what Lynn had specced, as I now had the freedom to make cap 1 of the CLC filter somewhat bigger, to increase the DC voltage obtainable from the secondary.

For the mains transformer, a good value off-the-shelf and British-made option seems to be this from Primary Windings: https://primarywindings.com/product/pwh ... a-6-3v-3a/ - it should run nice and cool at 40% of its rating (128mA / 320mA) and I'm hoping I'll be able to get a B+ of ~428V out of it - there are lots of primary winding taps from 220 - 250V and I figure just pick the one that gives me the closest result.

We keep the existing Hammond choke which is rated 5H/200mA, now we just need to size the caps. I've bought the Merlin Blencowe book on power supply design and one thing he notes to watch out for with CLC supplies is the resonance you get with that combination, which with the values I've been playing with seem to correspond quite well to record warps.

So rather than do lots of hard maths, I've been playing with PSUD2 to solve the problem iteratively. PSUD includes a current draw step change facility, and I knew that the current draw on the output stage would change if the 300B's went grid positive. Replicating 200V pk/pk voltage drive to the 300B grids suggested the current draw would drop from 128mA to 108mA or so on music peaks (or movie explosions), so I input that as the step change in PSUD to see what would happen.

Well, we can see the result of poorly chosen values right here - hey 12H isn't too large, we don't need loads of filtering as it's push-pull so 2x20uF ought to do.... oh dear:



For shits and giggles I put in the values from the Mullard 5-20 power supply - hmm, still a bit wavy on the step change....



It's a push - pull output stage, so getting ripple down to microvolt levels isn't a priority here. The minimum value for C1 with the aim of getting within a few volts of the highest voltage from a given transformer seemed to be about 20uF. Taking this and using my existing 5H 65ohm choke, the second cap seemed to need to get to 130uF before the bounce could no longer be seen:



If you zoom in, there is a *tiny* overshoot of 0.1V on the step change which I think we can live with. The ripple is 0.2V, about a tenth of what I had on the 300B's before, and they were silent then.

The cap figures are based on what's available in the ClarityCap TC range, by the way.

So I'm happy with that. Next up, the input / driver stage power supply.

[Nick]

I am guessing you haven't checked the actual DCR of the caps or tried changing it in the sym?

[Morgan Jones]

That's the way to do it. But ESR of a plastic capacitor will be way less than the default 2 ohm value used in PSUD2; 10m ohm is typical. The good news is that it's the choke's DCR that is significant, that and the L/C ratio. Big capacitors will tame the Q.

[Thermionic Idler]

I am guessing you haven't checked the actual DCR of the caps or tried changing it in the sym?

Strangely enough no, because I dont have any of the caps in question yet so I have no idea what the actual numbers are. ClarityCap, bless them, don't publish DCR on their datasheets so... where am I supposed to get those numbers from? I have no choice but to go with PSUD's defaults as a 'rule of thumb' because I have no other data to go on. 2 ohms and hope for the best but I would be REALLY grateful for the accurate figures. Given what ClarityCap charge for their products, i really wish they would publish ALL the relevant fucking data that we as DIY-ers need to be able to make an informed decision.

And.... breathe.

I'm similarly frustrated with Primary Windings - zero data on their spec sheets about open circuit secondary voltage and DC resistance primary and secondary which would help me determine if I'd get the right B+. I pinged them asklng for those figures two daye ago and what do I get back? Nothing. Nada. So i have to either buy the bloody thing and cross my fingers and hope the DCR will be low enough to give me the B+ I need, or find an alternative, probably more expensive and imported (with added Trump tariff), which at least has the data published that'll give me more certainty. Thing is, I really want to support British manufacturers, but... yeah.

And.... breathe again.

[Thermionic Idler]

That's the way to do it. But ESR of a plastic capacitor will be way less than the default 2 ohm value used in PSUD2; 10m ohm is typical. The good news is that it's the choke's DCR that is significant, that and the L/C ratio. Big capacitors will tame the Q.

Holy moly, Mr Jones himself *bows* we are not worthy!

I think I was still formulating my rant whilst you were posting your reply. I will certainly retry the calculation with the 10m figure - interesting to see what difference it makes.

By the way, I have the latest version of "Valve Amplifiers" on Kindle and the earlier version in print, along with "Building Valve Amplifiers".

[Thermionic Idler]

By the way, sorry about my grumpy rant last night - it genuinely wasn't aimed at you Nick. I'd not had the best of days and I was already feeling frustrated at the lack of published data for the transformer I wanted to use (which looked perfect for the job otherwise), and the non=response from the manufacturer when seeking said data.

Given the thing operates mostly in deep Class A push pull, I'm wondering if I even need to bother with boutique caps in the power supply - what exactly is the TC series going to bring to the table that a Wima won't. Plus a wider range of values and... better data!

[Thermionic Idler]

I recently found the Audiogon post from the point they changed the driver over to the KT88 in the Blackbird. They also spilled the beans about the operating points (14 watts dissipation, about 55mA, so that's about 250V plate then).

I have to admit, from the description I am still sorely tempted. I've worked out a way I can make them fit, too.

https://forum.audiogon.com/discussions/ ... id=2616731

[izzy wizzy]

Given the thing operates mostly in deep Class A push pull, I'm wondering if I even need to bother with boutique caps in the power supply - what exactly is the TC series going to bring to the table that a Wima won't. Plus a wider range of values and... better data!

I guess coz most of my amplification is diff/push pull, power supply caps don't seem to make much difference compared to other things like topology. In the line stage, I couldn't reliably tell the difference between a wkz black gate and regular electrolytic. Apart from phono, single ended, all my power supply caps are electrolytics and are relatively big

Coupling caps are a different story completely but that could end up as a rant

I don't have those caps from ht to cathode nor any cathode decoupling. Couldn't reliably tell if they we're there or not. Ymmv of course.

[Thermionic Idler]

So I'm back to the power supply design again, because I'm an idiot.

I've been bumbling merrily along thinking that the 300B output stage in the Karna Kay is fed by a power supply consisting of 5uF -> 5H -> 40uF, and assessing ripple and resonance solely on that, treating the 300B output stage as a 120mA constant current drain.

Er.... no.

Let's have another look at the Karna Kay circuit shall we and see where I cocked up?



What I had entirely forgotten to take into consideration were the three caps in the 300B output stage - there's an 80uF connected between OPT centre-tap to ground (via interstage secondary), a 40uF (sometimes reduced to a 20uF) going from OPT centre tap to cathode (I think that's the Ultrapath thing that is supposed to inject power supply noise at the cathode and give some sort of cancellation), and 80uF cathode bypass. So the power supply is actually seeing 40uF + 80uF + (1 / (1 / 20uF) + (1 / 80uF) ) the other side of the choke, not just 40uF.

Which also points to the pigs ear I made of the second power supply for the input and driver stage on my original build, as I fed that with the same 5uF / 40uF chain, albeit with a 12H choke betwixt (introducing a bit of resonance to the party). Only 40uF of smoothing before the CCS's on the 6V6 drivers with no Ultrapath cancellation. No wonder the thing hummed if the 6V6's were not well matched - although with the JJ 6V6's it was completely silent even when powering up.

Oh well, it still sounded pretty good and ran reliably for years... who knew...

[Paul Barker]

Perhaps diminishing returns; I do ultrapath in se and clearly hear improvement, but for me se is easy to calculate the value of the ultrapath. Also in se when we originally found this we had loads of greait paper in oil caps and tweaked the value for best sound which you could hear.

But those schematics you are looking at I cant fathom. My cognitive abilities are throwing me out not letting me see through a complete maze.

Im having a tough enough job following Williamson. Im really greatful there are no cathode resistor bypass caps in Williamson. Soon I saw that, Im like ! Fantasidosey!

On my 300b SE because I am applying final stage feedback Im not playing with ultra-path. Anyway the power supply well filtered.

Have you thought of stripping the trickery out and hearing what happens? OTOH my formula says youre ultra-path cap vlaue is 16uF. You said some are using 20, if they only use mu and not mu+1, 22 is the optimum, most mu formulas use mu+1, which suggests 16.

But you probably wont hear the benefit : these blessings are needed in SE and clearly audible. But really? PP, can you really hear that tool?

Also why does youre amp need a bypass caps in when Williamson doesn’t? (Edit:Ive found out from a vintage radio forum “not bypassed in combined cathode resistor removes a CR resonance from the global feedback path so avoids instability.) hence why Williamson didn’t bypass.

There is a place in the Williamson cathode resistance where the resistors are not combined but that is low value and such that you can observe the current flow to ballance it. A small amout of unbypassed resistance isnt an issue. In se you can play with the cathode resistance path, dont bypass a portion to give a portion of feedback, but not so much to make the anode load insufficient.