sry not unity gain but unity coupled, but anyway it is unity. I don't know what Frank McIntosh said and when, but big respect to him for designing such a great circuit which is very unique indeed.
Back to the Plitron UC transformer; I am guessing that the two Cathode Windings have the same number of windings as the plate primary. There is a problem with that transformer; or they typo'd the specifications. It says 4K P-P; which means that if you hook up the Cathodes; then the primary impedance goes up to 16K; because when you series two equal windings then the impedance ratio goes up by 2 squared; or 4. So assuming that they meant 4K WITH the unity coupling hooked up then the Cathode windings are each 250 Ohms. Now the other problem is that the MC75 is actually trifilar; they did that to reduce the stress on the cathode follower drive stage; but if you run a 6LU8 driver (vertical sweep) into a pair of horizontal sweep tubes and use the 3rd winding to run the SG's on the HOT tube then you are really in for some great sound.
nuvistordave they used trifillar in older designs to be able to feed cathode followers with lower B+ voltage to survive it. In newer design they simply used 200V zener diodes to make this achievable with bifilar.
Bifilar with zener diodes in followers is smaller "evil" than trifilar, because trifilar takes huge amount of size in bobbin, and from practical point of view it is very hard to fit even bifilar (with double isolated wire which is quite big for its diameter) into big bobbin together with many secondary sections. With trifilar this is very hard if not impossible with reasonable low frequency characteristic, or very high quality core have to be used. F.e. on C cores you can determine bobbin size to fit it, but EI laminations are restricted in size and it is nearly impossible to fit it in there. On toroid it would be easier probably, but I would be scared of wires to damage isolation in the middle where they are quite fastened and tough.
Bifilar with zener diodes in followers is smaller "evil" than trifilar, because trifilar takes huge amount of size in bobbin, and from practical point of view it is very hard to fit even bifilar (with double isolated wire which is quite big for its diameter) into big bobbin together with many secondary sections. With trifilar this is very hard if not impossible with reasonable low frequency characteristic, or very high quality core have to be used. F.e. on C cores you can determine bobbin size to fit it, but EI laminations are restricted in size and it is nearly impossible to fit it in there. On toroid it would be easier probably, but I would be scared of wires to damage isolation in the middle where they are quite fastened and tough.
It is 4K P-P, but the coils are 1K each. You draw a 4K load line when you are doing your design (if you are connecting 5 Ohm speakers). Really 6.4k is the real value you would use if you are designing for an 8 Ohm speaker.
They lack the trifilar primary and tapped secondary so would not be a good drop-in replacement for a Mac amp. They are fantastic in a new design, though. They have ridiculously low DCRs on the windings. I achieved a 1 Ohm Zout without any global feedback. I haven't ever seen another open-loop amp with a Zout that low.
Agreed. I think we'll have to agree to disagree on our earlier points. To me, the fact that a 10V increase on the grid of the output device causes a roughly 20V decrease in Va-k of the output device makes it blindingly obvious that there is gain of roughly two. That is the definition of gain in a circuit to me, input signal change over output signal change. I don't think it makes a difference in the analysis that the two coils that this output signal change is being applied to are tightly coupled for this analysis of gain.
I can kind of see your point of view, though. You are taking a transformer-centric view of the analysis and I am taking an output-device centric view. I still think mine is the correct one for predicting Zout based on transconductance of output devices, etc.
Last edited:
good point
Hey, and thanks for participating in the discussion. I love learning stuff. And I'm not an expert on transformer design by any means.
I just don't see a way to draw load lines and do correct analysis from the output device's point of view for output stage design with your approach. That's my big hang-up. My normal approach has been to treat both coils as the total number of turns for the plate-to-plate load (as the Plitron datasheet does). Then I can draw load lines from my 450V B+ on a 4k load (or 6.4k if I am going to use an 8 Ohm speaker) just like I normally would with a conventional push-pull output stage.
To look at effective rp so that I can predict Zout of the amplifier, I would use the procedure in O.H.Schade's "Beam Power Tubes" to calculate effective curves with 50% feedback.
I wouldn't have a clue how to make sense of this treating the output stage as a cathode follower with 1/2 voltage swing in a load line analysis. It seems that it would create many unsolvable problems, so I question its usefulness.
How would you draw a load line for this output stage?
It is very tricky indeed, but probably somehow as you described, I would count impedances together to get one continuos load Raa, I did it by this way when designing transformers for example, but not for loadline. Anyway when you count it as one you will still get same numbers as counting it as parallel/half, because of ohms law. But load line would change angle...
My calculating approach when designing transformers was as follows: (Apart from frequency range, core size, magnetic flux and many other variables during counting...)
I had to determine correct primary wire size to determine current during B+ voltage to get it right. And I calculated it according whole load in transformer which are all primary coils in series (cathode Raa + anode Raa). Well the final stage is paralleled in different point of view but I will get back to it later.
When you calculate current over load with specific voltage you get loadline. But you can't use this calculation for example to get impedance ratio to Rz, because as mentioned before, it is not using whole voltage swing in output transformer, just half of it thanks to transformer windings orientation. Since it swings just half the voltage, you need twice the current, which is accomplished, but you make this twice current achievable by two wires = double windings = parallel operation = unity coupled design. And since vacuum tube is located in the middle of the load, and creating wave from this point, I think it sees just this half of the load at a time. I think we can act as there are "two different tubes" in place of one, in one half of the circuit. Because in positive (or negative) swing, there are always both vacuum tubes acting but from different symmetric orientation, but counted twice because there are two of them. I am not sure if I explained my thoughts correctly as I am not native english speaker, but I hope we understand each other.
So I think when transformer works as this "paralleled" operation, we could assume this for drawing the loadline as half the B+ voltage, half the load, but twice the current? What do you think?
I am not sure if I am right, it is just estimation.
My calculating approach when designing transformers was as follows: (Apart from frequency range, core size, magnetic flux and many other variables during counting...)
I had to determine correct primary wire size to determine current during B+ voltage to get it right. And I calculated it according whole load in transformer which are all primary coils in series (cathode Raa + anode Raa). Well the final stage is paralleled in different point of view but I will get back to it later.
When you calculate current over load with specific voltage you get loadline. But you can't use this calculation for example to get impedance ratio to Rz, because as mentioned before, it is not using whole voltage swing in output transformer, just half of it thanks to transformer windings orientation. Since it swings just half the voltage, you need twice the current, which is accomplished, but you make this twice current achievable by two wires = double windings = parallel operation = unity coupled design. And since vacuum tube is located in the middle of the load, and creating wave from this point, I think it sees just this half of the load at a time. I think we can act as there are "two different tubes" in place of one, in one half of the circuit. Because in positive (or negative) swing, there are always both vacuum tubes acting but from different symmetric orientation, but counted twice because there are two of them. I am not sure if I explained my thoughts correctly as I am not native english speaker, but I hope we understand each other.
So I think when transformer works as this "paralleled" operation, we could assume this for drawing the loadline as half the B+ voltage, half the load, but twice the current? What do you think?
I am not sure if I am right, it is just estimation.
Last edited:
Hi
The Amplifier i am building is the one with bifilar winding and 200V zeners (Mc275 Mk4). Almost all info says 4K / 4K5 Raa and not 3600R ?
Well if the KT88 has an Raa = 3600R and the primary has 2500turns in a standard setup. Then the secondary turns on the 8R output is calulated as:
2500: Square(3600Raa:8) = 117,85 turns
Raa after dividing the primary into bifilar 1250 + 1250turns:
((1250:117,85)^2) x 8 = 900Raa
Now the turn ratio is 10.6:1 at 8R output tap. Correct me if i am wrong.
Feedback ? Any one got an idea?
Best regards
Kim
The Amplifier i am building is the one with bifilar winding and 200V zeners (Mc275 Mk4). Almost all info says 4K / 4K5 Raa and not 3600R ?
Well if the KT88 has an Raa = 3600R and the primary has 2500turns in a standard setup. Then the secondary turns on the 8R output is calulated as:
2500: Square(3600Raa:8) = 117,85 turns
Raa after dividing the primary into bifilar 1250 + 1250turns:
((1250:117,85)^2) x 8 = 900Raa
Now the turn ratio is 10.6:1 at 8R output tap. Correct me if i am wrong.
Feedback ? Any one got an idea?
Best regards
Kim
Raa can be determined according to tube datasheet, value Raa=3600R is quite fine I think for 450V. I used KT120s so I even lowered it further.
You determined turns ratio according to impedance ratio which is fine. But from where you get those 2500 turns? By the way try to fit 2500 turns of double or tripple isolated wire that can handle at least 0,2 or more A, to standard bobbin with some reasonable layers of secondaries. No way you can fit it there trust me!
I will not share my winding info, but I can help you with some hints. To design really good transformer you have to calculate transformer nearly 10times from start to end and from end to beginning and all over again to fit things correctly not only mathematically and electronically, but also physically. You want to fit it there and divide logically into sections, sym. sections of primaries need to have same length of wire, you don't want secondaries sections to not create completely layer and such a things and so on and on...
You have to know what frequency LF and HF you want to achieve, what cores you can use, what magnetic flux they have, sizes of wire including isolations, bobbin size, and what primary impedance you want to have. Also how many secondaries you have, if they are tapped. According these data you will determine number of turns. HOWEVER, you can't calculate 3600Raa and just divide it to several primaries according to turns, because impedance is square root of turns ratio. You have to calculate it to already divided number, because you want to achieve 2x1800R or as I said before, thanks to huge feedback even lowered to half so 2x900Raa and calculate number of turns out of this number. According this you calculate wire diameter. Once you calculate this, you calculate if it fits to bobbin, you will find out it won't and you start making compromises with LF cutoff, or increasing core etc.
When you finally have primary turns you can calculate what electrical looses you will have how many % of them there are. When you have this you can finally calculate secondary number of turns according to required impedance(s) with specific wire diameter(s) including to compensate to these looses. Now you will find out it will not fit into bobbin as one continuous layer into 1 or 2 section(s) and you have to calculate again from end to start to make it fit and to make it better many many times. Because the better fitted layers and the more sections you have, the better high frequency response you get 🙂
But along these million variables you have to still think on achieving reasonable isolation to make it safe.
Apart from all I mentioned, it is really tricky to get right symmetric global negative feedback winding, which is not only voltage but I think it is also current feedback from some point of view.
You determined turns ratio according to impedance ratio which is fine. But from where you get those 2500 turns? By the way try to fit 2500 turns of double or tripple isolated wire that can handle at least 0,2 or more A, to standard bobbin with some reasonable layers of secondaries. No way you can fit it there trust me!
I will not share my winding info, but I can help you with some hints. To design really good transformer you have to calculate transformer nearly 10times from start to end and from end to beginning and all over again to fit things correctly not only mathematically and electronically, but also physically. You want to fit it there and divide logically into sections, sym. sections of primaries need to have same length of wire, you don't want secondaries sections to not create completely layer and such a things and so on and on...
You have to know what frequency LF and HF you want to achieve, what cores you can use, what magnetic flux they have, sizes of wire including isolations, bobbin size, and what primary impedance you want to have. Also how many secondaries you have, if they are tapped. According these data you will determine number of turns. HOWEVER, you can't calculate 3600Raa and just divide it to several primaries according to turns, because impedance is square root of turns ratio. You have to calculate it to already divided number, because you want to achieve 2x1800R or as I said before, thanks to huge feedback even lowered to half so 2x900Raa and calculate number of turns out of this number. According this you calculate wire diameter. Once you calculate this, you calculate if it fits to bobbin, you will find out it won't and you start making compromises with LF cutoff, or increasing core etc.
When you finally have primary turns you can calculate what electrical looses you will have how many % of them there are. When you have this you can finally calculate secondary number of turns according to required impedance(s) with specific wire diameter(s) including to compensate to these looses. Now you will find out it will not fit into bobbin as one continuous layer into 1 or 2 section(s) and you have to calculate again from end to start to make it fit and to make it better many many times. Because the better fitted layers and the more sections you have, the better high frequency response you get 🙂
But along these million variables you have to still think on achieving reasonable isolation to make it safe.
Apart from all I mentioned, it is really tricky to get right symmetric global negative feedback winding, which is not only voltage but I think it is also current feedback from some point of view.
The avalanche diodes (200V is NOT a Zener) are an interesting idea. I have 3 Mac OPT's that are QUADRIFILAR. The 2 p-K windings then there are 2 thinner windings. The primary impedance is only 200 Ohms to 8 ohms sec and sec is not tapped. Low ratio is how they got the windings done.
When the output tube is starting at 450V (plate to cathode) idle, how could we draw an operating point at anything but 450V? We can't just cut it in half. That doesn't reflect what is actually happening in the tube.
That's why I'm having trouble accepting the "Unity-gain" argument. The fact that the tube is working into two coils that have a unique winding arrangement doesn't change the fact that for every volt that swings at the grid of this circuit, there are two volts of change in Va-k in the tube, and that Va-k will swing ~400V with an input signal of 200V into the circuit.
I have a question for those who are making the unity-gain argument. Why is this a different amount of feedback than a standard cathode feedback push-pull transformer with a 50% CFB winding? Clearly that would be 50% feedback, right?
i do not see any problem here, the lockhart amp is also an implementation of the unity coupled design...
the challenge here is finding the original core material to make the OPT and the driver IT's....
smoking-amp, PRR has very good postings about this topic, smoking-amp still posts here and i hope he chimes in...
Well, I don't have any direct MAC OT winding info to offer, other than what I have heard from others before and read in articles. The scheme is clearly a classic 50% CFB setup, the grid voltage swing may look like a 100% swing possibly, but that is just the grid/cathode drive signal added to the 50% swing at the cathode.
The Lockhart transformer diagram seems lacking in # of interleaves, unless I missed some annotation there. But it is probably a good model to start from for a Mac clone. For any OT design, one wants to avoid any unnecessary windings on the core, they just increase leakage L and use up bobbin space. The zener/avalanche diode fix seems a good solution. If one doesn't need the paralleled mono-block feature, then I would just use the actual secondary for the global feedback too, it's more accurate.
Taking that minimalist tact even further, the Mac STILL has 2X more windings/turns on it than it needs. The optimum solution is called a Circlotron, it does the same thing as the Mac OT and is WAY easier to wind. (no bifilar/trifilar stuff needed, those bifilars are the SAME wire in the Circlotron, phenomenal coupling, half as many turns)
( I'm always mystified by all the attempts to clone the Mac OT every few years. More like an exercise in torture. One can do better now with much less effort.)
One can still use the bootstrapped loads on the driver stage with a Circlotron.
Going further, OT performance increases with lower primary Z. The Circlotron allows halving the Mac primary Z with no effect on the circuit or tubes. And then using high current tubes, like Mac did later with some models, allows going even lower on primary Z as well. Obviously, that puts us in TV Sweep tube territory (which Mac finally used ).
Once you are using TV Sweep tubes, old non-linear g1 drive is completely obsolete, why not use something even more linear than the best audio tubes, "Crazy Drive". Mickeystan posted a nice Crazy Drive amplifier recently, a good starting place:
http://www.diyaudio.com/forums/tube...nificent-television-tubes-67.html#post4675905
see curves of Crazy drive versus g1 drive below (last two pics, both same tube).
I would modify that version for a lower B+, with a lower Z primary OT ( and using a Circlotron OT setup instead, two B+ sources then). Put any front end drivers on it that can do 50% CFB swing + about 80V for the grid 2 drive.
Why not use some driver tubes that have some real punch (a'la Citation II video tubes) like 12HL7s. And high linearity too: (super trioded curves, see below, left) And while we are picking bits from the Citation II, might as well use some local N feedbacks too, instead of so much tricky global feedback. (why spends months trying to get all that global Fdbk stable?)
You might as well call it the SR-71 amplifier by now. We are finally leaving the BI-filar-Plane era.
The Lockhart transformer diagram seems lacking in # of interleaves, unless I missed some annotation there. But it is probably a good model to start from for a Mac clone. For any OT design, one wants to avoid any unnecessary windings on the core, they just increase leakage L and use up bobbin space. The zener/avalanche diode fix seems a good solution. If one doesn't need the paralleled mono-block feature, then I would just use the actual secondary for the global feedback too, it's more accurate.
Taking that minimalist tact even further, the Mac STILL has 2X more windings/turns on it than it needs. The optimum solution is called a Circlotron, it does the same thing as the Mac OT and is WAY easier to wind. (no bifilar/trifilar stuff needed, those bifilars are the SAME wire in the Circlotron, phenomenal coupling, half as many turns)
( I'm always mystified by all the attempts to clone the Mac OT every few years. More like an exercise in torture. One can do better now with much less effort.)
One can still use the bootstrapped loads on the driver stage with a Circlotron.
Going further, OT performance increases with lower primary Z. The Circlotron allows halving the Mac primary Z with no effect on the circuit or tubes. And then using high current tubes, like Mac did later with some models, allows going even lower on primary Z as well. Obviously, that puts us in TV Sweep tube territory (which Mac finally used ).
Once you are using TV Sweep tubes, old non-linear g1 drive is completely obsolete, why not use something even more linear than the best audio tubes, "Crazy Drive". Mickeystan posted a nice Crazy Drive amplifier recently, a good starting place:
http://www.diyaudio.com/forums/tube...nificent-television-tubes-67.html#post4675905
see curves of Crazy drive versus g1 drive below (last two pics, both same tube).
I would modify that version for a lower B+, with a lower Z primary OT ( and using a Circlotron OT setup instead, two B+ sources then). Put any front end drivers on it that can do 50% CFB swing + about 80V for the grid 2 drive.
Why not use some driver tubes that have some real punch (a'la Citation II video tubes) like 12HL7s. And high linearity too: (super trioded curves, see below, left) And while we are picking bits from the Citation II, might as well use some local N feedbacks too, instead of so much tricky global feedback. (why spends months trying to get all that global Fdbk stable?)
You might as well call it the SR-71 amplifier by now. We are finally leaving the BI-filar-Plane era.
Attachments
Last edited:
Geez, maybe you and that circlotron should get a room. 😀
But seriously, I have always wanted to build a circlotron. It think all of the power supplies would probably take more effort than the rest of the amp, but it does have some pretty good advantages.
How high is the plate voltage on those 12HL7 curves?
- Status
- Not open for further replies.