[dBs]

Nah, like the RL from here: http://www.freewebs.com/valvewizard1/mufollower.html
 
They use lower case L so it looks like a 1 kinda. Basically top tubes CCS load resistor.
 
EDIT: Ok, never mind! I just read that and saw it, lol. I google for a simple example picture and I find a good explanation. It's ok, I've made a fool of myself before and I am sure I will again, lol.

 

[dBs]

Ok, I'm going to have to explicitly ask for help this time. I have been toiling over this for the past month or so I would say. There is a constant struggle between finding a pentode for the CCS that can yield a good rp (not Rp), but also not exceed the screen dissipation. I found what I thought was a good combination but that screen dissipation keeps killing me. Both find mutual benefit (good rp and lower screen dissipation) when Ip is low. There is a bit of a trade off between lowering the Ip and consequently increasing the Rk+RL, but the Pg2=Ip^2*Rg2, so it makes more sense to favor lowering Ip than Rk+RL, especially when increasing Rk+RL benefits rp.
 
Now I am reading that all this struggle may have been for no reason. I have been toiling under the idea that keeping in the linear region of operation is as important for a pentode CCS as it is for the triode that it supplies. I ran across some information that I should have looked at before that made it sound like a constant current SINK (not source) doesn't need to be run in the linear region like a the triode would. In fact, it goes so far as to say that you should operate the pentode in a sink with significantly less Ip than where you would normally run it. I do not know if this applies to CC sources as well as sinks, and my gut tells me that it does not, but I very much hope so. If I can feel sonically safe running the pentode CC source in a much less linear, lower Ip area of the curves, then it would REALLY save me from bashing my head more than I already have.
 
I will finally have the money to order parts this Friday, but until I can get this blasted step figured out, I will be going nowhere. I can clarify if anything I said above doesn't make sense. I can also supply some visuals if that would help clarify things. I'm a very visual person so I've been using Photoshop to handle all my load lines and such.
 
Hahaha, I've run through the equations so many times that I finally just made an Excel spread sheet so I can just throw in my load line numbers and have instant answers. This is actually working REALLY well and I think I will be continuing to do this for all future designs. Essentially making a quick look up table to determine what tubes and operating conditions will and wont work with one another with almost no effort. I could then even put them online for other people who may not want to break out the really difficult math but still want to get a bit more involved than just doing a Lego set amp...Lego SET amp, interesting idea XD
 
I am also going to eventually throw together (probably as I go along AFTER this amp is done) an Excel sheet of as many tubes as I can with all their most important ranges and values stored so that they can be organized and browsed as quickly as possible to locate alternatives.
 
So, if there was a time I needed some insight, it would be now XD I am at your gentle mercies.

 

[Leny]

I can't help with pentodes 'cos I've only played with triodes.
 
However if you find yourself fully stuck then consider using a solid state upper element.
Nip over to DiyAudio.com, go to the Articles section, then to 'His Masters Noise' phono amp, then Page 10, then Figure 5, then steal the cascode of FETs !!!
Tweak the current set resistor to your liking, then take your signal from the source pin of the lower FET, and it's 'Job done'.
 
Just a thought.
Good luck.

 

[nikongod]

RJM twilight FTW!!!111

 

[dBs]

It looks like I didn't make much sense, so I will try to clarify. The design I would like is your basic mu follower with two triodse in parallel in the lower section and single pentode up top. That is starting to become rather difficult so a basic mu follower with a pentode top is probably what will really result.
 

 
Take the example I THOUGHT was going to be a real winner:

The blue line is where the load line for the lower triode would be if I can find a way to parallel two of them together in the lower stage. The red load lines are where I would like to operate if I am forced into a single lower triode. The currents are 7.5mA and 11mA respectively. The swing is 130V in both cases. Amplification factor is 20, and rp is the usual 6600 and 7700 Ohms.
 
Now the upper pentode:

 
This is the 6JC6A pentode. The lines correspond as before, mostly. The red/pink lines in this case, however, are lower than they should be, I started going lower to see if I might get it to work at a lower current. Neither way worked. The voltage swing of the blue line gets a bit closer to 130V than I would like, but I was willing to take the risk. Also note that the 400V on the pentode would not be possible, that was more for an experiments sake that I will explain bellow.
 
Now the problem arises in the combination of factors that are joined at the hip, as it usually the case, they conflict with one another. I have been laboring under the possibly false idea that pentodes acting as CCS's still need their linearity considered in order for them to sound their best. I have been trying to find pentodes that were the most linear in the current ranges I have been hoping to operate under, like the one above. The problem is that CCS's benefit from having the flattest load line possible as that will increase the cathode and load resistors (Rk2+RL) which will in turn increase the rp seen by the lower triode/s. rp=(Rk2+RL)/(1-Av'), where Av' is the cathode follower gain of Av/(Av+1). So, I have been working to find a tube that would allow me to have as flat of a load line as possible while offering the current range linearity I also wanted. When I speak of acceptable rp values, I am hoping for at least a megaohm, which is achievable, buuuuut...
 
Additionally, the screen power dissipation Pg2 benefits from having a low current. Pg2=Ig2^2*Rg2, where Ig2 is the screen current and Rg2 is the screen resistor. Now, lowering the current, would mean a flatter load line, which would increase Rg2, BUT also decrease Ig2 and because Ig2 is squared in that equation, that is the better trade off. The greatest struggle has been keeping this Pg2 down to the .7W that is specified by the pentode shown above. I could accomplish it, but the currents were so low as to make the triode very nonlinear, as well as the pentode itself. The voltage of the pentode could be increased to have the same effect, unfortunately, all pentodes seem to be basically voltage bound at or bellow 300V in these current ranges.
 
The idea then was to find a pentode with higher current capabilities, this got me researching again. I found a section in the Morgan Jones book talking about constant current SINKS (not source like I am trying to develop here). In the book he talks about using pentodes at unusually low currents when trying to use them as sinks. Well that would solve my problem perfectly! The question though is if linearity is a problem? It would appear that it is not much of a problem in sinks as he recommends it and even offers up pentode options that work well for specific current ranges. Now that has me wondering if the same is true for sources. I have looked far and wide for that information but I cannot find it. That probably means the answer is so obvious that I have overlooked it. From the schematic for a pentode sink given in the book, I would not think that the two are the same, so what may work for the sink may not for the source. This also has me wondering if maybe it would be possible to make a mu follower using a sink instead of a source. I do realize that I would lose the PSRR in that case so it may not be a true mu follower, but maybe a more simple topography to implement, while offering most of the same benefits. As I was planning to go parafeed anyway, the PSRR should be pretty good so this makes this option even more intriguing.
 

Vp 1	Vp 2	Vp max	Vg1	Ip:Ig2 ratio
7	125	300	2.75	3.81
Vg1-1	Vg1-2	Ip max	Ivg1
0	2.75	0.0176	0.0104
Av1
42.91
Av1'
0.98
Rk1+RL
17,045.45
rL2
748,450.41
Rk1
264.42
RL
16,781.03
Rg2
63,935.73
Ig2
0.00
Pg2
1.36

(Rg2 in the table is the screen resistor, not the grid to cathode resistor like in the image above)
 
Hopefully that clears things up a bit.
 
Now I MIGHT have found a pentode (Z759) that would work. It gives me an rp of 748kOhms with a Pg2 low enough to pass. It even seemed relatively linear, not as good as I would like, but better than I could with any other tube I could find. The values for this tube can be seen above. While the tube isn't terribly expensive, it is really difficult to find. It's the best option I have found so far though, unless I find out that I should be running a pentode source like a pentode sink.

 

[dsavitsk]

Leaving aside the mu follower stuff for now, a pentode as a CCS late load is going to work just like a depletion mode mosfet. You put a resistor under the cathode, and tie the grid (through a stopper) to the base of the resistor. It will self bias and act as a constant current source, or sink (only with BJTs does this distinction matter).  With a pendode, what you are looking for is flat curves -- MJ mentions a few. There are any number of ways to bias the screen grid, but a resistor from b+ and a cap from the SG to the cathode fine.
 
But, the important point is that, like any other CCS, the pentode will drop whatever voltage it needs to, and this will be controlled by the triode.  The pentode's linearity does not enter into it.  Does this change taking the output from the pentode's cathode instead of the triode's plate? Maybe, but I'd wager not very much. The pentode is running under 100% local negative feedback that should linearize it. That would be my guess anyway.
 
So, I don't see any inherent difficulty in running parallel triodes, other than the bigger power supply -- thought I am not sure I see much of a point in it, either. As for the pentode, if you need a ton of current use an EL34. That should handle it.   Seriously, you just need to find a pentode that can handle heat dissipation and current, and that has reasonably flat curves.

 

[JB197]

dBs,
There is a pentode mu-follower in post #5 here: http://www.diyaudio.com/forums/parts/79300-lundhal-line-input-transformer-question.html .
It shows a 6AC7... I have no idea if this will help you, but surely worth a peek...

 

[dBs]

Quote:

Leaving aside the mu follower stuff for now, a pentode as a CCS late load is going to work just like a depletion mode mosfet. You put a resistor under the cathode, and tie the grid (through a stopper) to the base of the resistor. It will self bias and act as a constant current source, or sink (only with BJTs does this distinction matter).  With a pendode, what you are looking for is flat curves -- MJ mentions a few. There are any number of ways to bias the screen grid, but a resistor from b+ and a cap from the SG to the cathode fine.
 
But, the important point is that, like any other CCS, the pentode will drop whatever voltage it needs to, and this will be controlled by the triode.  The pentode's linearity does not enter into it.  Does this change taking the output from the pentode's cathode instead of the triode's plate? Maybe, but I'd wager not very much. The pentode is running under 100% local negative feedback that should linearize it. That would be my guess anyway.
 
So, I don't see any inherent difficulty in running parallel triodes, other than the bigger power supply -- thought I am not sure I see much of a point in it, either. As for the pentode, if you need a ton of current use an EL34. That should handle it.   Seriously, you just need to find a pentode that can handle heat dissipation and current, and that has reasonably flat curves.

See, all this great new stuff I'm learning is why I wanted to do this from scratch rather than following some recipe =).
 
At 2am on Saturday night, I realized just how stupid I have been all this time. I realized why I've had half of the problems I've had. I'm guessing this is a problem that most either see as too obvious, or maybe it's just such a rare issue that it is overlooked. It is best shown by illustration:
 

 
I forget how I found this beauty of a tube but she is gorgeous! She solves the other half of the problem I've been having; screen dissipation. I've run the numbers and with paralleled triodes or single triode, I can operate well under. The revelation of my stupidity can be seen in the lines though. This whole time, whenever I have been calculating my rL2, I have been basing my Av1, Av1',... values from the intersection of the load line with the quiescent current and the intersection of the load line with Ec1=0V. Essentially the maximum swing possible. On a triode this makes perfect sense to do this. On a pentode, however, it makes no sense because the change in Ec1 is VERY fast near Ec1=0V, deceptively fast (at least deceptive mathematically). You only gain a couple of plate volts for LARGE changes in Ec1. This gives your voltage gains Av1, Av1',... values deceptively bad results. When you make the calculations from a more reasonable value Ec1>0V, then the real numbers become clear. I can't believe I didn't see this before. The same affect happens on a triode if you try to max its Ec value, but most people don't run them that way so this problem isn't often encountered.
 
Either way, this tube solves both the problem I just described, and my plate dissipation. So I can FINALLY move on to the other aspects of the amp that should HOPEFULLY prove easier.
 
As far as why I want to go parallel on the triodes, I would like the increased gain, especially since it will be mostly current gain so as to work better with low impedance headphones. Also, it will halve the rp value and output impedance which should mean a smaller output transformer is needed as well as improving the performance with lower impedance headphones.

 

[dBs]

Long break as I had family visiting for basically the past 3 weeks. In the last few days, I have made a lot of progress. All my math and component values are pretty much determined for the amp itself. The last part I'm on is the OPT. Transformers, despite their simple basic concepts (not the more involved emag concept), and simple math, they have always been an Achilles heal of mine. It seems as though the number of turns is of little consequence so long as the input impedance is correctly chosen to match well with the plate choke and para capacitor and the turns RATIO is sufficient for the desired Zo and plate resistance. I'm guessing that as far as turns ratio is concerned, I will probably be forced to compromise as having one custom made will not be cheap. I am reading that standard toroidal PSU transformers can work and work well so I am also investigating that path.
 
After this it's just the PSU and I am confident that that should not prove to be as time consuming as the amp itself has been, lol.
 
I just figured that I would give a quick update and that things do seem to be going well =D

 

[regal]

Electra-Print is the go to DIY budget custom headphone amp OPT supplier.  His transformers are very good but not real expensive.  If you want expensive transformers no holds barr transformers use O-netics (contact Bud at DIYaudio).

 

[dBs]

Quote:

Electra-Print is the go to DIY budget custom headphone amp OPT supplier.  His transformers are very good but not real expensive.  If you want expensive transformers no holds barr transformers use O-netics (contact Bud at DIYaudio).

Great info. I will absolutely look into both of them. Thank you.

 

[natrix]

dBs,
 
I've finished reading most of this thread - there is a lot of really good information here.
 
I plan on creating my own amplifier, although I am going the solid state route, with a +/- 15 volt supply (I currently am using the AMB sigma22 until I figure out a design of my own) which seems a whole lot easier than dealing with tubes. I've already purchased some decent opamps and some output buffers for prototyping - the LME49600 are about $10 a piece! However, they look great on paper, at 0.00003% THD+N with a 2000V/uS slew rate capable of delivering 250mW to each channel. I intend on driving HD650's in an unbalanced fashion, I will probably be using a Cirrus Logic DAC as a source with some kind of switch for bypassing the DAC/active filter and going right to the gain/buffer stage from an analog "line level" input.
 
I have already designed and tested an active lowpass filter with a gain of 1 and a cutoff frequency of 25kHz. It works great! Using an HP 8903B I measured the distortion at 0.003% at full scale output at 10kHz.
 
By the way, a great piece of equipment for testing audio hardware is the HP 8903B. It isn't cheap, but you can find them used for well under $1000. It's great for measuring distortion and gain.
 
Maybe once my solid state amplifier becomes a product I can actually sell, I will dive in to the tube stuff.
 

 

[regal]

Quote:

 
Maybe once my solid state amplifier becomes a product I can actually sell, I will dive in to the tube stuff.
 

Not a good way to make a first impression on a DIY forum.

 

[dBs]

Quote:

dBs,
 
I've finished reading most of this thread - there is a lot of really good information here.
 
I plan on creating my own amplifier, although I am going the solid state route, with a +/- 15 volt supply (I currently am using the AMB sigma22 until I figure out a design of my own) which seems a whole lot easier than dealing with tubes. I've already purchased some decent opamps and some output buffers for prototyping - the LME49600 are about $10 a piece! However, they look great on paper, at 0.00003% THD+N with a 2000V/uS slew rate capable of delivering 250mW to each channel. I intend on driving HD650's in an unbalanced fashion, I will probably be using a Cirrus Logic DAC as a source with some kind of switch for bypassing the DAC/active filter and going right to the gain/buffer stage from an analog "line level" input.
 
I have already designed and tested an active lowpass filter with a gain of 1 and a cutoff frequency of 25kHz. It works great! Using an HP 8903B I measured the distortion at 0.003% at full scale output at 10kHz.
 
By the way, a great piece of equipment for testing audio hardware is the HP 8903B. It isn't cheap, but you can find them used for well under $1000. It's great for measuring distortion and gain.
 
Maybe once my solid state amplifier becomes a product I can actually sell, I will dive in to the tube stuff.

 
Personally, I think SS amps are more schematically complex and that's one of the reasons I'm going the tube route. I also REALLY like the tube sound from my very limited experiences with it. I've always been a fan of simplicity of design (though I like to be adventurous in the more complex realms on occasion), and that is definitely tube design.
 
Glad you are finding this thread informative. Beyond my degree, I know nothing of tubes or their amp design. I figured, I would be willing to sacrifice my dignity for a completely open air thought/design process from a beginner so that others can learn from my mistakes and find answers to questions that might be overlooked in other sources.
 
Quote:

Electra-Print is the go to DIY budget custom headphone amp OPT supplier.  His transformers are very good but not real expensive.  If you want expensive transformers no holds barr transformers use O-netics (contact Bud at DIYaudio).

 
Good call on Electra-Print. They seem very well priced and very flexible with their designs. I am trying to tailor my parafeed configuration to the well documented experimentation done at http://www.siteswithstyle.com/VoltSecond/Parafeed_fun/Parafeed_fun.html I requested info for the likely primary inductance on an output transformer with the following specs:
 
8k primary inductance
30mA
3W
flexible output impedance (they go from 4-32 and any of those is fine for full sized headphones)
 
In that article they specify that anything of >~40H per kOhm of reflected load.
 
I also asked if it would be possible, and how expensive, for them to whip up a couple 150H 30mA plate chokes.
 
It's nice that they are located in Las Vegas, only a few hours from me.
 
::UPDATE::
I got a reply back and it was pretty curt, lol. I wrote almost verbatim what was above and it seemed like he thought I was insulting his intelligence with a nerdy prank email or something. He recommended specifications of his own, so the question then arises do I trust the site with the quantified test results, or the gentleman from Electra-Print. I favor Electra-Print since I'm sure they've significantly more experience than a single test barrage.
 
The gentleman from Electra-Print indicated that the only specs that mattered for the OPT is the primary impedance, secondary impedance, and output power. He also recommended that the plate choke be around 50H.

 

[Leny]

Hello dBs, I've not read all of the VoltSecond information, but what I did read seems generally fine, so you may have possibly misinterpreted some stuff?
 
Maybe I misunderstand but it seems that in your request to Electra-Print you asked for a transformer capable of handling DC (i.e. 30 mA), however a parallel feed is always going to be capacitor coupled hence free of DC current in the primary. It may be that which confused the E - P guy. It is the DC or no DC situation which defines transformer design parameters in addition to the 'primary impedance, secondary impedance, and output power' as noted by E - P.
 
Having a transformer designed for no DC will result in proportionally higher inductance, all else being equal. So specify no DC, the turns ratio (primary impedance, secondary impedance) and the ouput power required, and then the inductance turns out to be whatever it turns out to be. Then choose your output coupling cap according to the inductance that you have ended up with. 
 
The (no DC) Sowter 8665 is noted as having a resulting primary inductance of 137 H (http://www.sowter.co.uk/headphone-transformer.php) and multiple taps (12, 6, 4 : 1). I doubt that you will find better. Expensive, but worthy of consideration because it does the right job right!