[KimLaroux]

Uhm honestly, the first time I tried the circuit I used cheap 12AX7 and It worked fine. The problem only shows up with the 12AU7.

 

[holland]

OK, I have a bit more time now.  I'll try to explain.  I'm not an expert by any means so my understanding may be not completely correct, but not incorrect neither.
 
CCS holds the current constant.  Vacuum tubes need current to flow from the cathode up through the plate.  You wired 2 tubes in grounded cathode mode.  This means you have 1 input, the grid.   You also have 2 outputs, the cathode and the plate.  As the signal goes into the grid, it wants to get current to flow.  Since the cathodes are coupled, and to a CCS, the outputs are colliding and no current is able to flow.  With a resistor to ground, the current fluctuates and the cathode voltage changes.  This is why you get no signal out of the 2nd output, the plate output.  One output is stuffed up, pushing against a wall.
 
Why the SOHA 2 works.
 
The triodes are configured in 2 different topologies.
 
The first triode is a cathode follower, it is grounded cathode mode.  That is the input is the grid, and 2 outputs.  The plate and the cathode, just as above.  The difference is that the 2nd triode is configured in grounded grid mode.  This means that the input is the cathode, and the grid, and the output is the plate.  The output of the first triode flows into the input of the second triode.  The CCS presents a wall and makes the flow "perfect", holding the cathode bias constant and a constant current flow up.  This is the basic difference.  Output flows to input.  The outputs are not colliding with nowhere to go.

 

[holland]

Quote:

Uhm honestly, the first time I tried the circuit I used cheap 12AX7 and It worked fine. The problem only shows up with the 12AU7.

 
Yeah, and from the sim it looks like the CCS is not working there.  I should say, partially working.  The amplification is no where near what it would be.  It is extremely low for a 12AX7.
 
Edit: because it is not working properly, for whatever reason, AC is seeping to ground, and current flows through the triode.  It is as if it were a resistor of some sort.  A functioning CCS will not have this.

 

[holland]

When the cathode voltage drops below 1.2V the top transistor in the ring of two doesn't turn on, and you get AC signal output.  The 12AX7 runs the cathode, at the points you have "selected" with a 0.7V cathode.
 
I wouldn't call that working.  That's broken in a different way.

 

[wakibaki]

Quote:

 
I noticed with the 12AX7 it shows gain.  With the 12AU7, it does not.

 
Well, I'm just looking at this in a very superficial manner, I didn't intend to get into a slog of work analyzing the circuit, or I'd have figured the DC conditions for the CCS instead of just letting the sim deal with it. There's no doubt that in my sim the CCS is not functioning as it should, but I just looked at the 25V that it's connected to and thought, 'Oh, I can't be bothered figuring that out'. The milliamp of current that is running through the tubes, is, however, entirely sufficient to see some gain with a 7k5 load on a 12AX7. You'd normally expect to see a much bigger resistance there with a 12AX7 though...
 
Typically no gain shows a tube that is cut off (or saturated), and the bias conditions from one tube to another might mean that that is the case with the 12AU7. That's why I suggested that Kim should run the sim himself, because in my experience it's a lot more revealing than drawing load lines on a graph, as long as the sim models are not too far out. I didn't realise that he doesn't have a Windoze box, but like you say, there's always Wine. I've used these models quite a bit and I think they're reasonably trustworthy if used with discretion. You can turn the voltage up and up and they won't blow up, but, like I say, you have to exercise a bit of discretion.
 
w

 

[holland]

Quote:

 
Well, I'm just looking at this in a very superficial manner, I didn't intend to get into a slog of work analyzing the circuit, or I'd have figured the DC conditions for the CCS instead of just letting the sim deal with it. There's no doubt that in my sim the CCS is not functioning as it should, but I just looked at the 25V that it's connected to and thought, 'Oh, I can't be bothered figuring that out'. The milliamp of current that is running through the tubes, is, however, entirely sufficient to see some gain with a 7k5 load on a 12AX7. You'd normally expect to see a much bigger resistance there with a 12AX7 though...
 
Typically no gain shows a tube that is cut off (or saturated), and the bias conditions from one tube to another might mean that that is the case with the 12AU7. That's why I suggested that Kim should run the sim himself, because in my experience it's a lot more revealing than drawing load lines on a graph, as long as the sim models are not too far out. I didn't realise that he doesn't have a Windoze box, but like you say, there's always Wine. I've used these models quite a bit and I think they're reasonably trustworthy if used with discretion. You can turn the voltage up and up and they won't blow up, but, like I say, you have to exercise a bit of discretion.
 
w

 
Yeah, me neither.  It was just interesting to try and find out why your sim worked.  I didn't have LTSpice installed and it totally went against what I thought and knew.  Unfortunately I didn't connect the dots with your other data point about 1mA going through the triode.
 
FWIW, I do know that if you load a tube with a CCS on the tail of a single triode and try to get output from the plate with an output transformer, you won't.  There won't be current fluctuations from AC to activate the transformer (ideal CCS, practical might be a bit different).  I don't think the circuit will work without the transformer and just an output through a cap just because of where the CCS exists.  Knowing this, I was befuddled.  Mystery solved.
 
Well, at least now I have LTSpice installed.  I haven't used Spice since I left school 20 years ago, so I'll need to do a refresher course.  Hopefully Kim picked up a few things in my ramblings.

 

[nikongod]

Quote:

The SOHA doesn't have a voltage source there... unless I'm missing something. The cathodes DO sit at their expected voltage, after all.
I'm thinking a higher value plate resistor would improve PSRR, but then it would either drop too much voltage and/or limit the amount of current I can push trough the tube. My HV is only 160V after all. Any recommendations on what would be a better value?
 

 
 
The SOHA has a resistor bypassed with a cap which is conventional for a *common cathode* tube stage.  
The SOHA II uses a *long tail pair* tube stage which (basically) requires a CCS between the cathodes and ground.
 
since your design is screwed up (it happens) you have 2 choices to fix it. 
First is to accept that you have a common cathode and should pick between a resistor, resistor & cap, or an active voltage source to replace the CCS. 
 
The second (and my personal preference) is to keep going down the route of trying for a long-tail-pair front end. There are many compelling advantages to the LTP, and you already have all of the expensive parts accounted for... so change what you have now into this. 

 

[KimLaroux]

Quote:

 
 
The SOHA has a resistor bypassed with a cap which is conventional for a *common cathode* tube stage.  
The SOHA II uses a *long tail pair* tube stage which (basically) requires a CCS between the cathodes and ground.
 
since your design is screwed up (it happens) you have 2 choices to fix it. 
First is to accept that you have a common cathode and should pick between a resistor, resistor & cap, or an active voltage source to replace the CCS. 
 
The second (and my personal preference) is to keep going down the route of trying for a long-tail-pair front end. There are many compelling advantages to the LTP, and you already have all of the expensive parts accounted for... so change what you have now into this. 

 
Now I'm curious about using a linear reg on the tail... never seen that done, wonder why?
 
I'm starting to understand now... I think.
 
As for the long tail pair, is not the goal of such circuits to provide both an inverting and non-inverting output? I don't need those two signals for my output stage. So how would such a topology provide any benefits in my case?

 

[holland]

Quote:

 
Now I'm curious about using a linear reg on the tail... never seen that done, wonder why?
 
I'm starting to understand now... I think.
 
As for the long tail pair, is not the goal of such circuits to provide both an inverting and non-inverting output? I don't need those two signals for my output stage. So how would such a topology provide any benefits in my case?

 
Hmm, I've seen them around on the tail.  I don't think it favored well, but I never paid much attention.
 
As for the LTP, yes and no.  The current mirror is used to recombine the signals to SE.  The LTP is comprised of the first triode in cathode follower configuration (grounded cathode), and the 2nd triode in grounded grid.  The CCS guarantees the AC signals are "balanced" as well as possible between triodes that do not match.  Vacuum tubes don't really match between triodes.  The current mirror is used for 2 things.  To set voltages at the DC level, and to recombine the AC signals to single ended.  You don't need to do the recombination, and can pull from the plate of the 2nd triode.
 
The LTP, or more accurately referenced as a cathode-coupled amplifier for vacuum tubes, is really good.  The reason is that the first stage adds it's flavor of distortion.  The 2nd stage complements that (or reverses it) with it's flavor of distortion which is the inverse of the first stage's.  The reason why CCS works there is because current flows from the first stage to the 2nd stage of the triode.
 
The SOHA 2 input stage is used to drive a SE output.  A similar stage is used in the EHHA to drive a push/pull output stage.  If you use push/pull, then you will want the differential.  If you don't then you use the current mirror trick, or you just pull the output from the 2nd triode and put a cap across the plate resistor on the first triode.
 
Borbely used the same design as the SOHA 2, it predates the SOHA 2, with some differences.
 
FWIW, my own forthcoming DIY amp is using the cathode-coupled amplifier front end in a cross coupled differential amp (balanced).
 
Edit:  freakin' grammar and typos.

 

[holland]

http://www.cavalliaudio.com/diy/soha%20ii/main.php?page=design/ipdesign
 
This will help you understand the SOHA 2 input stage design.  However, I believe it has been proven (on diyAudio) that you don't need to go to that extent.  The AC will match amplification even though the DC is different, as long as you have the CCS on the tail.

 

[holland]

Here's more reading material which should explain everything by a true master.  John Broskie.
 
http://www.tubecad.com/2010/11/blog0194.htm

 

[KimLaroux]

Well, I ordered some BC560C too in my last Digikey shipment, so maybe I should just go ahead and build the rest of the SOHA II circuit... the current mirror. 
 
But since I'm running from 160 V and not 75 V, is there anything I should tweak? I'll probably be running the tubes much hotter than the SOHA II so I'm guessing some change in resistor value is necessary. I'm just not sure how to do the math with the current mirror up there.
 
Thanks everybody.

 

[holland]

Quote:

Well, I ordered some BC560C too in my last Digikey shipment, so maybe I should just go ahead and build the rest of the SOHA II circuit... the current mirror. 
 
But since I'm running from 160 V and not 75 V, is there anything I should tweak? I'll probably be running the tubes much hotter than the SOHA II so I'm guessing some change in resistor value is necessary. I'm just not sure how to do the math with the current mirror up there.
 
Thanks everybody.

 
Yes, that should be fine.  I actually buy BC550C and BC560C in lots of 100, so I have tons on hand, and they are good to just have around for doing things.  I fry them left and right, unfortunately.
 
Also, you should be able to pull in a negative rail.  The reason the negative rail is better is because it offers more headroom, like when the cathode voltage drops down to low voltages (as seen with the 12AX7 and you turning up the bias with low B+).  With a positive rail driving the CCS, you can't operate the CCS.  The negative rail offers much more headroom.
 
What I intend to do, is to take the heater voltage, rectify it and tag it as a negative voltage source.  -12.6V or -6.3V and drive the CCS.  The heater is going to be on a different transformer secondary (or different transformer entirely) than B+.  Everything is solved, but I am a bit concerned about heater noise.  I will see when I build it and measure it.
 
Also, if you're going to run 5mA or more through the tube, bring B+ up higher.  I used 245V on my SOHA 2, with an unregulated DC voltage of about 330V, plate is sitting at 170V with 5mA through each triode.
 
Another also, in the SOHA 2.  You can bypass the resistor on the plate of the 1st triode with a small cap.  Broskie talks about this.  Some of the SOHA 2 prototype team (I was on it) did it.  It was never documented, but the reasoning is explained in the tubecad articles.  It's just better, if you take the words of the tube god.  I never measured it.

 

[KimLaroux]

Well, I rebuilt the whole PSU in this last run of modifications, and don't really feel like going back and redoing it again. Besides, I'm running out of space in the enclosure. At this point I'm just trying to make the best of what I've got.
 
I'm not certain how a negative rail improves things... and if it's necessary in this design. I see it as relative voltage differences. In the SOHA II, I see the tail CCS is connected to - 15 V and then the grids are referenced to ground... so it's as if the grids are some 15 volts higher than the bottom end of the tail CCS. Is this necessary in order for the whole design to work? Can't I just raise the voltage on the second triode's grid instead?
 
But lemme read the tubecad article, it seems to explain all that...

 

[holland]

Quote:

I'm not certain how a negative rail improves things... and if it's necessary in this design. I see it as relative voltage differences. In the SOHA II, I see the tail CCS is connected to - 15 V and then the grids are referenced to ground... so it's as if the grids are some 15 volts higher than the bottom end of the tail CCS. Is this necessary in order for the whole design to work? Can't I just raise the voltage on the second triode's grid instead?
 
But lemme read the tubecad article, it seems to explain all that...

 
The tubecad article won't explain it.
 
The negative rail gives headroom if you plan to run cathode voltage at a low voltage.  If you intend to operate above 3V, then you should be fine with the 25V source.  It's only when you get low that the CCS stops working.  This is proven with your initial proto when you ran the bias up to about 5.5mA per triode (11mA through CCS) and the cathode voltage dropped down to 0.7V.  The first transistor stopped doing what it should do, and you get audio out because AC (audio) starting moving to ground and it became a simple resistor based triode with no CCS at all.