[coinmaster]

After looking into proper mosfet biasing and source follower configurations I learned that my above example of the grid drive mod would work but it is far from optimal.
 
This is a better way to do it.

M10 is the source follower mosfet
M11 is another source follower
M12 is a constant current source (maintains a constant current)
The zener diode D3 puts the gate of M11 at a set voltage above the output, in this case 27v.
This makes M11 produce 20v at its source (depending on the mosfet and its bias). So the voltage difference between the drain and the source of M10 is always going to be 20v at DC.
The capacitor C12 feeds the output signal to the gate of M11 so the voltage difference between drain and source of M10 will be the same at AC as well. 
The end result is that as far as M10 is concerned, it is sitting at a stable 20v even though the the source and drain voltages are all over the place.
 
M12 feeds M10 a constant current.
As current is drawn through R31 a voltage drop is produced which after a certain voltage thresh hold turns on the transistors Q12 and Q13 which brings the voltage on the gate of M12 to ground level.
After the the current decreases from this the transistors turn off, the end result being an equalibrium of current is maintained.
 
Typically transistors turn on at about .7v differential between the emitter (pointy end) and the base but if you add more like I did in the above example it makes them more sensitive and they turn on at a lower voltage which means they react faster and a more stable current is maintained.
 
So what does it all mean in the end?
Well the current source and M11 maintain a constant voltage and current across the source follower (M10) which means it is locked in place on the load line, no change in transconductance or linearity can occur, this also reduces the effective input capacitance since input capacitance is the sum of the capacitance between gate-source and gate-drain.
A source follower has pretty much no difference between gate and source voltage so there is no need to charge/discharge the capacitance between them and the voltage difference between gate and drain stays stable also meaning there is no need to charge/discharge the capacitance between them.
 
In the real world there will be small variations but this way the overall distortions remain minimal.
 
This means you should still probably choose the most linear part of the load line for the mosfet just to eek out every last bit of linearity.
Meaning you should probably have M10 and M11 be the same mosfet since you are feeding them the same current and preferably the same voltage so they will act the same which will prevent phase shifts,
 
I'm choosing the BUZ901 because I don't know the meaning of restraint and it is probably the most linear mosfet in existence and I'm going to feed it 3.5 amps @ 20v which gives an insanely linear performance even if the voltage/current wasn't locked.
 
I've learned a lot from this source follower design and I've already been able to greatly improve my upcoming designs with the technique.
However it is not a simple mod and it will not fit in case (especially if you want to run BUZ901s @ 3.5 amps).
This design can actually be used as the output stage itself and could drive speakers, which I fully intend to try next month. Paypal Credit, what would I do without you.

 

[Maxx134]

Coin,
Have you looked into balanced designs and topologies?
Now that you have your CCS I believe you should bring your attention to possible clever ways to achieve the effects a balanced topology does

 

[coinmaster]

Not sure I catch your drift. As far as I know the main advantage of balanced designs is EMI rejection.
Technically I am only looking into balanced designs because I see no reason not to and its better to be safe then sorry since balanced is often claimed better sounding.
 
Not sure where the CCS falls into that or what you mean by "effects" of balanced design.

 

[Maxx134]

Not sure what you mean by not sure.

I barely have time to push out my latest mods and now I have to explain things..

 

[coinmaster]

A balanced design uses anti phase signals on each side of the load, I''m not sure what a CCS has to do with it.

 

[Maxx134]

Exactly the point, now you have to address how will you maintain the force balace of the two triodes.
When one triode changes gain due to age, we need the other to adjust accordingly in relation to the other side.
Maintaining CCS on each side does not address this..
Edit:
Maybe a regulator to force both ccs to be fed equally.

 

[coinmaster]

A CCS does maintain balance with age. If the tube ages and its specs change the CCS changes with it. That's why it's called a constant current source. You can do whatever you want with the tube specs and it will maintain the same operating point on the load line assuming the voltage is the same.
 
But, for a WCF the top triode will not maintain a constant current unless you use a gyrator or inductor load. Just verified that in spice. But it also seems to cancel out the signal so there may not be a way to maintain top triode balance without a change in design, at least not with a CCS. 
The servo maintains the impedance of the top triode over its lifetime in order to keep a 0vDC output so that helps.
Also the auto biasing on the bottom triode helps counteract aging as well. 
I think that's about as good as you are going to get.
 
Since you mentioned gain I assume you are mainly referring to the input stage, in which case a CCS will maintain a balance between the two regardless of age/tube type/etc.
 
For example, lets say one tube is half dead and one tube is new.
The new tube requires the cathode to be 2V more positive then the grid and the half dead tube requires the cathode to be 3v more positive then the grid in order to maintain a 10ma current.
 
The current sources will put 2v and 3v on the triodes cathodes and 10ma will be maintained through both.

 

[Maxx134]

OK I see,
So you will need at least this CCS to stabilize the triodes, in order to implement the cathode follower / grid drive mod...

Now count the amount of components (maybe custom chip, or mosphets, transistors, diodes,caps , and their separate PSU stage) in order to implement this.

While I am not going to say anything until I try for myself, and in fact once I extend my bottom case, I would be open to try, especially once you do it ha.

But...

Remember this is all to replace one coupling cap and one cathode resistor in our current amp..

Now our coupling caps are large enough, and of highest order (copper foils),
as is anode/cathode resistors in driver stage (zfoil resistors) .

Now this is where design choices are made.

Wich path will we decide to choose?.
And will we gain a sonic benifit specifically where implemented,
Not in a general sense .

Will our driver/power tubes benifit from these changes?
As they can handle such a wide range of variables.
And which tubes will benifit most?
And which tubes will sound best regardless of these changes?

This is the whole problem at hand that we are dealt with here, wich will take some more time to sort out.

I am currently in a different project to test all options of tube types in my amp.
I almost covered testing most all 9pins and 8pin types from both MK6&8, in my amp,
In various configurations in both driver and power areas.

My current goal is to see for myself what types works best in current design configurations.

So coinmaster you must have even more choices to make because you has more options on tube types to use.
I assumming basically you want that sweet spot that you heard in past.

 

[coinmaster]

So you will need at least this CCS to stabilize the triodes, in order to implement the cathode follower / grid drive mod.

Grid drive mod doesn't need it, using it in conjunction with the "bootstrap" source follower just locks its load line into a single point on the load line so the overall distortions are minimal if any.
 

Now count the amount of components (maybe custom chip, or mosphets, transistors, diodes,caps , and their separate PSU stage) in order to implement this.

Component count is probably the biggest problem other then size, mainly because you need to have a decent electronics setup in order to build and troubleshoot the circuits which from my own experience can be a huge pain.
The best way to do it is simply by ordering PCBs which saves space, time, and probably money, and its safer.
I have the PCB files mostly done already other then labeling the values on the board but I still need to test them to make sure they are public ready first so unless someone else wants to take that job they will have to wait until I have the time and money plus the 2-3 weeks wait for shipping.
 

Remember this is all to replace one coupling cap and one cathode resistor in our current amp

The coupling cap doesn't go anywhere, assuming you are talking about the grid drive mod.
You still need to separate B+ and the servo voltage.
The grid drive mod just turns the high impedance output of the input stage into a low impedance one so it can quickly and easily overcome the capacitances of the output tube which should increase dynamics.
 
I tried simulating this but I'm not sure the proper way to go about it.
Instead I simulated the input stage feeding a 6080 cathode directly and nothing really worked, the 6sn7 with a resistor on anode/cathode produced a few ma into the 6080 but the CCS/gyrator did not.
So I cannot say what will happen if I directly couple the input and output stage, perhaps at a minimal it will sound as good as the duelunds but in order to directly couple the stages the input stage plate voltage needs to be at the grid voltage level meaning the servo would have to adjust the B+ of the input stage instead of just the grid voltage, but that also means that the grid of the input stage triodes needs to be below ground level so unless you float your entire dac to that level or unless your dac has an output capacitor it won't work.
My dac is direct coupled so I would need to float the entire thing to whatever voltage I need on the grid which is project in itself (and one I'm not sure I'm willing to try).
 
I do have an idea though, if a low output impedance is needed to overcome grid capacitances of the output tubes then why not use a Jfet in cascode with a low RP tube like a 6080/6as7 or more preferably the 300b.
 
Jfets are commonly used in input stages because they are so quiet and they are very tube-like in sound and operation. Normally they are difficult to use because of low voltage requirements but they would work for this purpose.
The problem with this idea is that a jfet/tube cascode would have so much gain you would never be able to use it, but then again I don't know much about Jfet biasing, perhaps there is a way to reduce the gain.
 
Such a mod would theoretically be relatively easy to implement.

 

 
And which tubes will benifit most?
And which tubes will sound best regardless of these changes?

That's a can of worms I don't intend to open, you can break the bank and use up all your time trying to find out. There's a lot of tubes out there.
 

Will our driver/power tubes benifit from these changes?

The driver tubes are by far the weakest link in the amp, even as skeptical as I am I can't ignore the fact that at the very least the technical improvement would be huge with the mods.
 

This is the whole problem at hand that we are dealt with here, wich will take some more time to sort out.

Yeah unfortunately, I already have the materials needed to do all of the testing but I need to wait to order a few more odds and ends so I can properly finish up my diy build which is modular and will allow me to switch between different stage designs fairly easily but then again while I do intend to test different input stage designs I want to leave my MK6 output stage alone for now.
I haven't actually listened to the modded MK6 since early september or something so even if I do test the mods I can't speak too much about how they improve over the original.
 
I'm secretly hoping someone else will take on some of the load because at the end of the day I just want to listen to my music 

 I suspect my drive to design and innovate will go down once I finally have an amp to use.
But I do have 3 more output stages to test, a solid state and 2 hybrids. One is a modified version of what MrCurwen recommended me a while back, one is a modified version of the WCF that always maintains equal current balance between the triodes at all frequencies, something the impedance mod cannot do, and one is push pull version of the "common cathode" input stage (real common cathode, not the one we have which isn't really common cathode) which seems to be the universally accepted "best" input stage topology since it cancels out all distortion and gets heavenly claims from everybody.
The push pull "output stage" version of this that I have been working on does the same thing but is able to be used as a output stage for driving headphones however it is insanely complex, so far it needs 20 opamps per channel to bias 4 tubes. It works well in spice but can still use some optimizations, I would like to get some second hand advice on it but I am banned from DIY-audio website which is sadly the only place where people have the experience to comment on such a design.
 

So coinmaster you must have even more choices to make because you has more options on tube types to use.

Sadly if I want to keep an OTL/OCL output stage my tube choices are very limited, I was thinking of using 300b tubes which would be very complicated to make into OTL and may or may not be better given the fact they have more then twice the output impedance of our tubes.
 
As for input tubes I do have a lot of choice but my Tungsol RP 6SN7 "holy grail" tubes blew away the competition I tried so I will probably stick to them for now although I would like to try 300b at some point.
I am for sure going to try the "common cathode" input stage since it is so often considered to be unspeakably good due to its distortion canceling capabilities but since it is  distortion canceling it may not matter what type or quality of tube I use.
 

 

I assumming basically you want that sweet spot that you heard in past.
 

I'm going on the assumption that what I heard in the past can be improved on, otherwise I would be listening to an MK6 right now. It really was that perfect sound that I didn't think actually existed.
 I truly cannot imagine it sounding better then it did. This was before the WCF cap mod was introduced which is almost on the level of the coupling cap mod right? Wow.
 
I have a sub goal of creating a new product, or at least something that I can recoup the thousands (around $7,000 so far, ouch) I've spent on materials for R&D so far.
I've already helped create a new output stage which I was praised for by broskie, the tubecad guy, which I've made further improvements on and I'm hot on the trail of another design that seems to be really good so I have some material to work with, it all depends on how they actually sound in the end. I want to avoid using known designs.

 

[Maxx134]

I am for sure going to try the "common cathode" input stage since it is so often considered to be unspeakably good due to its distortion canceling capabilities but since it is distortion canceling it may not matter what type or quality of tube I use.

This is exactly the point I am trying to make about the distortion cancelling abilities of the common cathode design.

Which takes on more importance and effect in the low level signal driver stage.

As far as I have checked,
this driver stage implement almost textbook spec.

Yet you and MrCurwen kept pointing out amp by stating the driver stage is poor and non optimal.

In reality nothing is optimal.
We have to deal with what we have.

So our amp driver stage may not be perfect, but I have not notice anything poor in driver circuit except the tubes themselves.

Many preamp circuits run tubes at lower power than optimal so running at 250v or less is not uncommon.

So hopefully we can look over this driver stage to see how we can fine tune and optimize it's performance further.

Edit:
Looking over the values it does seem this circuit is more for triode balance than the "common cathode" benifits of noise cancelling.
Edit2
I changed some words to be more polite so not take me wrong way.
I do realize We need to expose weaknesses .

 

[coinmaster]

This is exactly the point I am trying to make about the distortion cancelling abilities of the common cathode design.

Which takes on more importance and effect in the low level signal driver stage.

As far as I have checked,
this driver stage implement almost textbook spec.

Yet you and MrCurwen kept crapping amp by stating the driver stage is poor and non optimal.

We do not have a common cathode input stage(this has been said multiple times including once in my above post). A common cathode input stage is a cathode follower feeding a grounded grid amplifier, the cathode follower distorts one way and the grounded grid distorts the opposite way, the end result being the distortion is canceled out.
 
It would require 2 dual triode input tubes per channel to do this with a balanced amp.
 
The purpose of the two triodes in our amp is to amplify a pair of anti phase signals because the output requires the signal to be anti phase on each side of the load in order for it to be a balanced design.
It is basically two SE amplifier tubes that are amplifying two different signals, this is why balanced designs require two sets of dacs and amps because they have an in phase and an out of phase signal (opposites).
 
 

In reality nothing is optimal.
We have to deal with what we have.

So our amp driver stage may not be perfect, but I have not notice anything poor in driver circuit except the tubes themselves.

Many preamp circuits run tubes at lower power than optimal so running at 250v or less is not uncommon.

So hopefully we can look over this driver stage to see how we can fine tune and optimize it's performance further.

The technical improvements that can be had to the input stage are huge, you can use word play all you want but that is still the truth.
The input stage is, from a technical point of view, garbage.
Will it sound better by "improving" it? I'd be surprised if it didn't. The only way I see otherwise is if the suboptimal implementation is creating "good" distortions. Everything about the input stage is poorly implemented, obviously to save on costs.
 
If you have an extension box then you can optimize the input stage without too much trouble, the problem is creating the CCS/gyrator/power supplies.
Like I said I'm already ordering PCBs for the CCS/gyrator so you can just leech off my design for that and order your own which makes things a whole lot easier.
The problem is the power supply for the B+ and CCS.
It may be worth it just to order a Salas shunt regulator for $60 since they are the bees knees in the diy community and they come in kits. This shunt regulator can also improve on the output stage power supply and will no doubt improve the sound due to lower output impedance.
As for the CCS power supply, I may have to create a PCB design for that unless people want to make it themselves. It would be a -15v supply to have enough headroom for the CCS to regulate.

 

[Maxx134]

We do not have a common cathode input stage(this has been said multiple times including once in my above post). A common cathode input stage is a cathode follower feeding a grounded grid amplifier, the cathode follower distorts one way and the grounded grid distorts the opposite way, the end result being the distortion is canceled out.

Depends which version your looking at..
The actuall name is different here but the objective and schematic is exact same as in our amp.







You say input stage is, from a technical point of view, garbage..


Why does simplicity lose out over complexity with you?



I believe this is what makes the tube rolling more noticeable. 

 

[coinmaster]

It's a fancy way of saying two SE amplifier tubes that amplify opposing signals with the downside that if they are not perfectly matched they will draw uneven current.
There is no distortion cancelation in the design, it is not a common cathode input stage.
 
This is a common cathode input stage

The right triode counteracts the distortion of the left triode.
 
It doesn't "depend" on which version I'm looking at. 
You don't take the words "balanced" and "common cathode" literally when discussing topologies. 
 
Here is a read on common cathode stages http://www.tubecad.com/2005/April/blog0042.htm
 
 
Quote:

You say input stage is, from a technical point of view, garbage..


Why does simplicity lose out over complexity with you?

I believe this is what makes the tube rolling more noticeable. 

Has nothing to do with simplicity or complexity.
 
The input stage power supply is unregulated.
The resistor values are poor
The bias is poor, depending on the tube you use
The balance between the tubes is poor
Resistor loading in general is poor due to the impedances and the way the signal travels along the load line.
 
The ideal load for a triode in order to get the most linear impedance and transfer characteristics is an infinite impedance and resistance, something a resistor can not do and still have current running though it unless you want to run a 10KV power supply.
That's why CCS/gyrator is good because a CCS is the closest you can get to such a situation and puts an effectively infinite impedance on the cathode and the gyrator puts an effectively infinite impedance on the anode while maintaining a constant voltage.

 

[coinmaster]

Edited the above post to be a little more clear on common cathode.
 
Did anyone ever find out if changing the pot made a difference in sound? People kept talking about it a while back but I don't think anyone ever reported on the findings.
 
I have a 32 bit dac so I should have a good range of digital attenuation via the PC before I get into any bit compression.
So I don't intend to use any pot but I am curious to know whether it makes a noticeable difference either way.

 

[Maxx134]

I am thinking the values on the driver resistors to was set that way because of the Limited PSU so I will look into that

Edit
I was told that using a pot was a bad idea