For PP load lines, look here (assuming that you haven't seen this site yet-look at the load matching sections):
education+diy
and post #11 here:
http://www.diyaudio.com/forums/tubes-valves/110963-ppp-opt-clarification-2.html#post1340175
I've been recently trying to figure load lines out as well, and I haven't found much coverage for output tube load lines in MJ 3rd ed, but the above links pretty much cover it.
Just in case anyone else, besides me, is following this thread and aren't familiar with the origins of this amplifier design, I stumbled on a website that provides the original article text written by Mullard, as well as the original figures, including assembly and wiring diagrams, and even a photo of the original prototype.
Mullard 5-20 Audio Hi-Fi Amplifier
Too bad that website owner used such a difficult-to-read colour and texture combination on his website.
..Todd
Mullard 5-20 Audio Hi-Fi Amplifier
Too bad that website owner used such a difficult-to-read colour and texture combination on his website.
..Todd
Too bad that website owner used such a difficult-to-read colour and texture combination on his website.
Mmmm ...
Why do you think they do that ...
Thanks for encouragement. At this juncture I would advise a pause to think and determine an optimal choice of main components, to the extent that such optimality exists. My feelings now:
1. As phase inverter at a lower voltage the ECC88 comes up rather strongly. No problem with the 6H6P1 and 5687, but ECC88 have somewhat higher mu. At 400V+, the 12AT7 can still be accomodated, as the required G1 drive for KT88 will now diminish; only about 60Vpp required.
2. BUT - going down to 30 - 40W: 6L6GC p.p. can just as well achieve that (I get 110W out of 4 x 6L6GC p.p.p.) with some saving in power supply requirements - particularly heaters, and very possibly cost. (Here in RSA a 6L6GC or equivalent is just about 50% of the cost of a KT88, thus for same $$ outlay one then has 60W plate dissipation per side instead of 35W for KT88.)
I would feel that there is somewhat of a waste to - er - waste a tube's abilities. I would consider KT88s falling into the 60 - 70W category at some 500V h.t. - alternately, as postulated above. I agree with the small audible difference between 40W - 60W, but the above considerations now come up. My personal feeling would be that once having decided on an output pair, get out of them what they are designed for.
3. I think we must now rather decide from the available/optimal tubes/transformers side, cost included. Cost will vary country/country. One would unlikely land with an untenable situation regarding the phase inverter.
[For the record, pentode (1st stage) anode voltage can be taken as low as 60V to accomodate a direct coupled LTP phase inverter. Also there is nothing that specifies that the LTP be direct-coupled to the 1st stage. It is simply a little more convenient. The famous Leak amplifiers of the same topology never did that. That would enable a lower tail voltage for the LTP.]
The references given by Boywonder are extensive - they appear kosher, although I have just glanced through the 'Education +" one; will need some time there! [Again just for the record, I use RDH ch 5:3-6 (pp 206 - 251)
for my transformer designs and have never had a flop. Here in RSA own design and local winding is quite less expensive and more accomodating than brand imports.]
I am confused (as usual). Is there a concern about the choice of the 12AT7 and KT88 combination based on their operating voltages? I don't yet understand the implication of the numbers on their datasheets.
It sounds to me, based on comments here, that optimizing one tube's comfort zone throws the other into a less comfortable zone. Is this correct? I'd greatly appreciate an explanation in lay terms.
Another question I have is about something Eli mentioned regarding the grid resistors on the KT88s needing verification. Can someone elaborate? We should fix it if something needs fixing. (sorry, I don't have the exact quote.)
..Todd
It sounds to me, based on comments here, that optimizing one tube's comfort zone throws the other into a less comfortable zone. Is this correct? I'd greatly appreciate an explanation in lay terms.
Another question I have is about something Eli mentioned regarding the grid resistors on the KT88s needing verification. Can someone elaborate? We should fix it if something needs fixing. (sorry, I don't have the exact quote.)
..Todd
Last edited:
Thats alot to consider, but i'm not sure how many more redesigns Tay can take! We must not lose our technical writer!
Here is the criteria that I feel must not be violated:
1. Monoblocks
2. KT88 in PP Ultralinear
3. 50 to 60 watts
4 Mullard style LTP topology
I'm open to any other needed changes.
Here is the criteria that I feel must not be violated:
1. Monoblocks
2. KT88 in PP Ultralinear
3. 50 to 60 watts
4 Mullard style LTP topology
I'm open to any other needed changes.
Last edited:
Johan,
More than convenience is involved with DC coupling. An additional coupling cap. in the signal path adds a high pass pole inside the GNFB loop. That exposes a builder to the same sort of instability observed with Williamson style circuitry. The very last thing wanted is to satisfy Barkhausen's criterion for oscillation. 🙁
I dearly want the 12AT7 triode somewhere in the small signal complement, for its highly beneficial effect on the net harmonic distortion spectrum. If the 'T7 is "in over its head", as the LTP, use it as the voltage amplifier and the ECC99 as the LTP. Of the "super triodes", the ECC99 has the highest μ and the smallest heater current draw.
A voltage gain of 30X+ from a resistively loaded 'T7 section is easy enough, especially if it's DC coupled to the LTP. Assuming a conservative (low RP) 6X voltage gain in an ECC99 LTP, yields 180X+ as the small signal open loop gain. IMO, that's quite sufficient. Never forget today's "standard" of 2 VRMS from the signal source.
Not a concern. Let's get it right and produce a great amp! (I enjoy the writing and drawing almost as much as the subject matter.)
No problem with me. And I like Eli's devotion to the 12AT7 too.
..Todd
I'm very open to this thought of Eli:
" If the 'T7 is "in over its head", as the LTP, use it as the voltage amplifier and the ECC99 as the LTP. Of the "super triodes"
Which leads to my original thought at the start of this. Just what it the perfect driver combination for the definative Mullard LTP KT 88 topology. I guess this is now the question at hand. I've never heard of Super triodes, but I like the ring of it!
" If the 'T7 is "in over its head", as the LTP, use it as the voltage amplifier and the ECC99 as the LTP. Of the "super triodes"
Which leads to my original thought at the start of this. Just what it the perfect driver combination for the definative Mullard LTP KT 88 topology. I guess this is now the question at hand. I've never heard of Super triodes, but I like the ring of it!
Last edited:
First and foremost is getting the gain structure right. Gain sufficient to drive both the "finals" and the GNFB loop is essential. However, too much gain is to be avoided, as it leads to "hair trigger" volume control trouble.
As I previously stated, 1 way or another, get the 12AT7 triode into the mix. The non-linearity of the 'T7 meshes extraordinarily well with PP "finals". Remember, PP O/P devices cancel internally generated even order distortion products. Combining the 12AT7's skewing towards 2nd order and PP "finals" leads to a NET harmonic distortion "waterfall" of 2nd > 3rd > 4th. The ear loves that spectral distribution.
OK, let me try to steer this back on track.
I think Tay has dropped his lower B+ idea, and i'm staying with a B+ of 465V for now.
Thats said.
It looks like the 12AT7 is ok in the Phase splitter role at that voltage,(is that right Johan?) so the current question is now this.
Running a B+ of 465V, do we leave things as they are, or would the best option still be to:
Move the 12AT7 to voltage amp duties, and sub in one of the following into Phase splitter duties: 5687, 6Н30П, 6H6Pi, ECC88, and ECC99
Let the debate begin!
I think Tay has dropped his lower B+ idea, and i'm staying with a B+ of 465V for now.
Thats said.
It looks like the 12AT7 is ok in the Phase splitter role at that voltage,(is that right Johan?) so the current question is now this.
Running a B+ of 465V, do we leave things as they are, or would the best option still be to:
Move the 12AT7 to voltage amp duties, and sub in one of the following into Phase splitter duties: 5687, 6Н30П, 6H6Pi, ECC88, and ECC99
Let the debate begin!
Brethren,
Wish you would not post so energetically in my night! 😴
Regarding the choice of tube for driver: It has to deliver the highest signal voltage output of the stages preceeding the power stage. Occasionally it has to deliver a little power or alternately have a low rp: This is when for fixed-bias output stages it must, apart from its own load resistor(s), also cope with a next-stage grid resistor of maximum 100K (and should you go push-pull parallel, very likely 50K). And for cleanest results, the 'loading' of said 100K through a capacitor should follow an even lower anode load resistor. (To best show these load lines one needs an Ia - Va graph - perhaps later.) Then, sometimes the ability to reach a low Va at Vg1 = 0. (This is where the 12AX7 mostly fails.) Then, as Eli showed, distortion pattern is of consequence; again because it operates at relatively high output signal.
Regarding direct coupling of the LTP:
Mmm .... Eli, a capacitive coupling to LTP input need not necessarily upset the stability 'Applecart". One can have a high next-stage grid resistor (several meg). I am thinking of where one can hardly avoid it because of different optimal operating points for the respective stages. (You will be shocked to learn that in my hi-end 100W design I cope with 5 l.f. poles - OPT included. And that is not counting the power supply - semi-regulated though. The Ø-band-limited pole in the G2 supply of my pentode input stage helps a lot to save the day, also I had to build a special 0,1Hz - 30Hz sine wave oscillator to test down there! But off-topic.]
Getting further technical; does one know that other triodes are worse than the 12AT7 in LTP? Will one not find that other similar triodes exhibit the same, if not identical traits? ....although that is probably what you are saying - they do not.
Tube-Mack,
Using the 12AT7 as input will not have the same 'pattern' as in phase inverter, as its distortion then goes through 'as is', unaffected by the p.p. topology.
Eli said:
Ah, imagined that basic rule of amplifiers would come up sometime:
If you want an amplifier, build an oscillator. If you want an oscillator, build an amplifier. (Guchelsnack's basic rule 1 of audio. G. is the son of Murphy, only worse.)
😀😀😀😀😀
Wish you would not post so energetically in my night! 😴
Regarding the choice of tube for driver: It has to deliver the highest signal voltage output of the stages preceeding the power stage. Occasionally it has to deliver a little power or alternately have a low rp: This is when for fixed-bias output stages it must, apart from its own load resistor(s), also cope with a next-stage grid resistor of maximum 100K (and should you go push-pull parallel, very likely 50K). And for cleanest results, the 'loading' of said 100K through a capacitor should follow an even lower anode load resistor. (To best show these load lines one needs an Ia - Va graph - perhaps later.) Then, sometimes the ability to reach a low Va at Vg1 = 0. (This is where the 12AX7 mostly fails.) Then, as Eli showed, distortion pattern is of consequence; again because it operates at relatively high output signal.
Regarding direct coupling of the LTP:
Mmm .... Eli, a capacitive coupling to LTP input need not necessarily upset the stability 'Applecart". One can have a high next-stage grid resistor (several meg). I am thinking of where one can hardly avoid it because of different optimal operating points for the respective stages. (You will be shocked to learn that in my hi-end 100W design I cope with 5 l.f. poles - OPT included. And that is not counting the power supply - semi-regulated though. The Ø-band-limited pole in the G2 supply of my pentode input stage helps a lot to save the day, also I had to build a special 0,1Hz - 30Hz sine wave oscillator to test down there! But off-topic.]
Getting further technical; does one know that other triodes are worse than the 12AT7 in LTP? Will one not find that other similar triodes exhibit the same, if not identical traits? ....although that is probably what you are saying - they do not.
Tube-Mack,
Using the 12AT7 as input will not have the same 'pattern' as in phase inverter, as its distortion then goes through 'as is', unaffected by the p.p. topology.
Eli said:
Ah, imagined that basic rule of amplifiers would come up sometime:
If you want an amplifier, build an oscillator. If you want an oscillator, build an amplifier. (Guchelsnack's basic rule 1 of audio. G. is the son of Murphy, only worse.)
😀😀😀😀😀
Possible option?
Really great sound, should mate with just about any output tube you choose.
Plenty of headroom from a low Rp driver.
Really great sound, should mate with just about any output tube you choose.
Plenty of headroom from a low Rp driver.
TubeMack,
No, as said (we are posting so rapidly), for 400V one can re-examine and bring the EF86 anode voltage down to give extra headroom for the 12AT7. Also, naturally the KT88 set would have a lower G1-signal swing at that h.t. So it should be quite feasible - that can be looked at. If it is decided to keep the 460V (or possibly even raise it) to use the KT88s to full value then there is no problem. Also, one needs to recalculate the load for the KT88s with 400Vht; it could still be made to give high power (lower Vac, higher Iac signal). But then we come to a new (lower) OPT primary impedance, and so on. That can also be calculated. As I previously said, we have to do with several different sets of acceptable conditions; it might depend on what is available.
It would appear that an OPT was available. What is the position regarding a power transformer? You guys will have to decide; I cannot comment from down here, it is not my market convenience.
No, as said (we are posting so rapidly), for 400V one can re-examine and bring the EF86 anode voltage down to give extra headroom for the 12AT7. Also, naturally the KT88 set would have a lower G1-signal swing at that h.t. So it should be quite feasible - that can be looked at. If it is decided to keep the 460V (or possibly even raise it) to use the KT88s to full value then there is no problem. Also, one needs to recalculate the load for the KT88s with 400Vht; it could still be made to give high power (lower Vac, higher Iac signal). But then we come to a new (lower) OPT primary impedance, and so on. That can also be calculated. As I previously said, we have to do with several different sets of acceptable conditions; it might depend on what is available.
It would appear that an OPT was available. What is the position regarding a power transformer? You guys will have to decide; I cannot comment from down here, it is not my market convenience.
"Sand" can be of great help in keeping LTP gain up, while allowing the use of comparatively small KT88 g1 to ground resistors. Keeping g1 to ground resistance reasonably low avoids destabilizing contact bias effects.
ZVN0545A source followers DC coupled to the LTP plates will accomplish the task. 😀 As small a value as 27 KOhms for the net grid to ground resistance is easily handled by the little FET. Just remember to size the coupling cap. to get a F3 <= 5 Hz.
Johan is correct about very careful design and execution allowing multiple high pass poles in the signal path to work well. Being a "belt and suspenders" kind of guy, I prefer to avoid the issue, from the outset.
ZVN0545A source followers DC coupled to the LTP plates will accomplish the task. 😀 As small a value as 27 KOhms for the net grid to ground resistance is easily handled by the little FET. Just remember to size the coupling cap. to get a F3 <= 5 Hz.
Johan is correct about very careful design and execution allowing multiple high pass poles in the signal path to work well. Being a "belt and suspenders" kind of guy, I prefer to avoid the issue, from the outset.
I'm also unclear over the constraints the 12AT7 LTP puts on the B+........seems like 460V or so is the lower limit based on the discussion but I have no idea why.
Taj: I can answer your question about the max value of the grid leak resistors. There are conflicting requirements for them, on one hand we want them large to minimize loss of gain since they form a potential divider with the output R of the preceding stage, and large value R's also allow smaller coupling caps to be used for a given lower cutoff freq. However, gas currents from residual gas in a tube need an easy path to ground to get rid of + charged ions when the residual gas molecules are slammed by the traveling electrons. That's why they are called "grid leak" resistors. When the resistance is very large (including the source resistance all the way back to the bias winding), the grid cannot stay -ve enough when lots of grid current is flowing, and this "less -ve" grid voltage causes more anode current to flow, blasting more +ve ions free, etc causing the tube to run away and self-destruct.
Typically the fixed bias max value for grid leak R is about 1/10th that of cathode bias, since cathode bias has some self stabilization capabilities that fixed bias does not. Interestingly, for the KT88, it lists different values for fixed bias above 35W and below 35W. So the 100K value is the most stringent of all operating conditions, which is listed for fixed bias above 35W.
With all of that being said, I'm still to much of a tube knucklehead to be able to calculate the source R all the way back to the bias winding, but it won't take much to exceed 100K ohms.
Last edited:
One more link for loadlines.....Steve Bench's site
Steve's Tube Pages
MJ 3rd edition = "Valve Amplifiers" 3rd edition by Morgan Jones. There is an accompanying book by MJ called "building valve amplifiers" as well.