[coinmaster]

Looks like MrCurwen is sidestepping any point that I made. Not even going to repeat myself.

 

[baronbeehive]

Maybe. I think it's more likely they wanted to 'trap' some guys into tube and cap rolling by making the amp susceptible to small variations in these components. Makes for a fun hobby for many people.

You said that.... I didn't want to! I also didn't want to mention Chinese tubes!!

My point was that their circuit design was done knowingly and no attempt was made to produce a perfect circuit, if you look at the nano technology adopted in headphone manufacture nowadays by the Chinese I know that they are well able to produce cutting edge design if they want to.

CCS tail is what is used to force balance.

Thanks, that's what I thought. This is probably the next mod then... a little bit of circuit design never comes amiss! The balance issue is important along with the PSU issue.

Are you actually using NOS never-been-plugged-in tubes? You guys must be rich.

Once a tube has had the cathode emitting some dozens of hours, there is no more burning in that can ever be done.

By the way; if you make the circuit not dependent on tube gm, burn in doesn't matter anymore.

I'm using NOS testing tubes, more by luck than by design. I don't mind trying my luck with tubes with no test data, they often turn out fine, but its a bit risky. The burn in is more for the Jupiter caps, I never bother to burn in tubes.

Rich, not a chance!!I wanted to try the Western Electric's, but chances are they will have to go at some point to pay for the car to be serviced unfortunately. That's why I am in fact quite happy with the RCA's, I've always said that I prefer the cheaper tubes on the whole, some of the expensive ones are ludicrously overpriced and under perform as well, yet they have a massive following. As an example I've always used the TS6SL7's, quite cheap when I bought mine but they sound exactly the same as the very expensive "holy grail" TS 6SN7 and the same goes for the extremely tight tolerance and expensive 6SU7GTY's.

Edit: Hmmm, that's a idea, sell the Western Electrics, and buy the parts for the LTP amp, and still have enough left over for another 3 amps, or an assortment of headphones, and speakers...

 

[MrCurwen]

Thanks, that's what I thought. This is probably the next mod then... a little bit of circuit design never comes amiss! The balance issue is important along with the PSU issue.

If you force balance, PSU disappears. You can have potatoes as filter caps after that.

Edit: Hmmm, that's a idea, sell the Western Electrics, and buy the parts for the LTP amp, and still have enough left over for another 3 amps, or an assortment of headphones, and speakers...

Unless you really have money to burn, spending big bucks on tubes is madness in my personal opinion. Just go Russian; they're usually fully NOS never seen a tube socket. Cheap as bread and perform very well. 6P31S is a gem.

Also EL81 is fantastic, and still affordable. That's my new favourite tube to use. I've grown tired of making filament supplies for 4P1L, even though it is the best of the best. The difference is fleeting, and probably nonexistent.

 

[MrCurwen]

6E5P is superbly linear as well, just as good as 4P1L, but it's a bit tricky. It tends to oscillate while still in the box. Too high gm even after degeneration, and too high µ. Really great linearity though, one of the best in the whole world.

My current breadboard workhorse has 6E5P finals. Great sound but there's some background noise. Too small to get me off my ass to really fix it, so it just keeps annoying me. That's why EL81 is my new favourite.

 

[SonicTrance]

28 Vpp is more than enough. gNFB influence is already present at the grid input signal, and WCF is let's say 0.8 gain, so with 28Vpp at tube anode you'll have 70 Vpp across the headphone! That's crazy loud, unuseable.

I found the gain formula for WCF:

Gain = mu² + murp/Ra
(mu² + mu + 1) + (mu+2)rp/Ra

If we use the 6AS7/6080 tube we get about 0.55 gain. But how do you calculate? How do you get 70 Vpp from 28 Vpp with 0.8 gain? Probably a stupid question but I don't get it, lol.

 

[baronbeehive]

If you force balance, PSU disappears. You can have potatoes as filter caps after that.

So.... cheap as chips..... sorry you probably didn't want to say that either!

I was wondering that about balancing and the PSU, I thought they were separate issues, so how does balancing the signal avoid the need for PSU regulation for example? Do any perturbations cancel out or what?

 

[MrCurwen]

If we use the 6AS7/6080 tube we get about 0.55 gain. But how do you calculate? How do you get 70 Vpp from 28 Vpp with 0.8 gain? Probably a stupid question but I don't get it, lol.

By calculating wrong, apparently. I guess my brain was still a bit foggy after lifting weights just before writing that...

0.55, that's lower than I recalled. Anyway you still shouldn't need that much voltage swing on the input section anodes. At the load you sum the individual swings together.

I was wondering that about balancing and the PSU, I thought they were separate issues, so how does balancing the signal avoid the need for PSU regulation for example? Do any perturbations cancel out or what?

For starters look at my previous few messages, especially the experiments.

Any signal that is present at the cathode node does not appear in the output. Also if current draw is constant, PSU ripple decreases significantly. PSU is played out of the equation.

For best results I would still strongly advocate using a simple 2 FET reg that I've posted. It's strong enough to provide a decent output impedance even for SE use, and simple enough to be cheap, easy to build and set up. Only use the least level of complexity required.

Reg also significantly improves stability. That's the #1 reason I use it.

Probably wouldn't fit in the chassis though.

 

[MrCurwen]

Differential means that the only signal that is taken into consideration is the DIFFERENCE between the phases. If something is shared, it's called common mode signal, and it doesn't appear in the differential output.

PSU is shared, it's distortions are shared, so they don't appear in the output.

 

[Maxx134]

Why?

As to the 33k; what amount of swing are you looking to get and why?

Why..

Because One jfet sounds better to me than one mosfet plus 2 transistors.


As for the 33k anode,
I already tested that setting , but with slightly higher B+ & cathode resistance.
That setting was too low voltage swing and gimped/neutered the dynamics.
The overall sound was still very euphoric and nice with the tubes in question (ecc35) so it may have not been noticed without comparison.

I will have to test the whole thing again with same cathode setting,
But the whole thing may be a waste if we use CCS anyways.

Also that low anode reaistor setting was much worse on other tubes so really depending on which tube.
Bottom line listening,
And I chose 68k.

Oh well the real topic is what choice of CCS is best for this admittedly compromised driver stage.

 

[MrCurwen]

Because One jfet sounds better to me than one mosfet plus 2 transistors.

I agree as a principle to go with the least complicated solution that gets the job done good enough. The CCS I and Sonic use is that; small component count, small loop (only one loop). Still insanely powerful.

That chip you are talking about has much more stuff inside it. Look at the datasheet page 16, there is a schematic of the insides.

 

[baronbeehive]

Differential means that the only signal that is taken into consideration is the DIFFERENCE between the phases. If something is shared, it's called common mode signal, and it doesn't appear in the differential output.

PSU is shared, it's distortions are shared, so they don't appear in the output.

Right so it looks like the CCS tail mod is most important, and moreover probably makes the other 4 mods I mentioned: Force balance, Source Follower, Regulated PSU, and Redge's decoupling mods, rather redundant. Regulated PSU becomes less important, Redge's decoupling mods no longer needed, and the Source Follower mod I'm assuming is no longer needed if we already have the impedance mod.

 

[MrCurwen]

Right so it looks like the CCS tail mod is most important, and moreover probably makes the other 4 mods I mentioned: Force balance, Source Follower, Regulated PSU, and Redge's decoupling mods, rather redundant. Regulated PSU becomes less important, Redge's decoupling mods no longer needed, and the Source Follower mod I'm assuming is no longer needed if we already have the impedance mod.

Pretty much yes, except for the source followers. If you mean the grid drivers by these, then there is nothing performing that particular function.

 

[Maxx134]

so it looks like the CCS tail mod is most important, and moreover probably makes the other 4 mods I mentioned: Force balance, Source Follower, Regulated PSU, and Redge's decoupling mods, rather redundant. Regulated PSU becomes less important, Redge's decoupling mods no longer needed, and the Source Follower mod I'm assuming is no longer needed if we already have the impedance mod.

The decoupling in the driver stage is actually not an optional "add-on", but instead a substitution.

Remember the tiny 33uf electrolytic caps we took out?
They were part of the driver stage PSU "RCRC" filter..

So you already replaced them with nice film caps,
which perform way better, so I would not say it is not needed..

Take it out and you loose the last part of your PSU RCRC filter(!).

Plus since they are larger, you can lower the 3k3 resistor to my suggested value of 1k (or lower like SonicT did),
To get a higher B+ out of the anemic driver stage PSU.

It is sorely needed, and depending on the tube used you will have a different B+ because of the last R volt drop of the PSU RCRC design, (since there is no SS regulator).

So for driver tubes,
you like 6SL7,
and I preferred the (similar) 6C8G,
And SonicT like the Mullard ecc35, right?

They all higher "mu" than the 6SN7,
but different plate resistance & different load lines.

I guess we should find some optimal bias settings for each of these tubes to list in first page what bias for which tube (a can of worms lol)

My situation is that I have not implemented the CCS yet.
So I will order parts now.

Notice that although I have MK8 (which use 9pin tubes),
I preferred the use of adapter for the 6pin tube types,
as they sounded best to me (on this same MK8&MK6 driver stage).


Another thing I noticed is that Both you and SonicT prefer the lower gain setting on MK6.

For my MK8, the higher gain setting worked better, and I believe it is because of the different output tubes used on MK8 (9pin 6h30 type but I currently using cheaper 6n6 which actually more realistic for me).
It funny that the MK8 cost more but use cheaper tubes.

 

[Maxx134]

Just look at the tube comparison PM's

Yes, it was massive improvement I don't think is stated enough.

You should post the stock settings on these tubes.
I remember you made them .
They were absolutely dismal compared to your final load lines.

Show a before and after so everyone can realize the HUGE improvement you made just on bias alone.

Also of interest would be noting any audible changes of your CCS implementation.

 

[Maxx134]

Here is an article showing how setting bias boils down to what sounds good, instead of what looks good:

http://www.diyparadise.com/variation.html