I kind of suspect that, but I am green, that's why I post this thread to solicit feedback. The whole article doesn't smell right!!!! In particular Smoking-Amp shot down his hysteresis theory so utterly.
A very interesting thread, many thanks for all contributions! Alan0354 knows how to post the questions in the right way to extract most out of peoples knowledge
Hi Alan0354. You can buy Lundahl from K & K Audio - Lundahl Transformers, audio DIY kits and more in the USA. The prices are, I think, quite similar to the European prices.
Thanks so much for the link. Now I know how much I have to pay instead of have to worry about the import tax.
I am very interest in learning, when I see a link here, I do read it and think about it. Some I agree, but in this particular papers from Vaughn, I have problem swallowing it, so I post to seek other opinions. There are so many knowledgeable people here, I learn so much in the last one month.
Thanks, I am still thinking about your mu-follower. I did spend time reading and thinking to understand your mu-follower design from the last post. I really like the concept. It's just quite a bit of work to put it in as you need heat sink and all.
Now I have to think about the LTP differential output stage also.
"Now I have to think about the LTP differential output stage also."
You may want to look through these articles:
http://www.tubecad.com/2005/March/02/Which Tube Shall I Use.pdf
more detail:
http://www.clarisonus.com/Archives/...System Design Factors for Audio Ampifiers.pdf
The R tail fixes they mention only work at the low current end of the triode characteristic, where gm is curved upward.
However, some tubes do seem to fit the requirements over the full characteristic (gm, last page graph):
http://frank.pocnet.net/sheets/123/1/10JA5.pdf
http://frank.pocnet.net/sheets/123/6/6LU8.pdf
You may want to look through these articles:
http://www.tubecad.com/2005/March/02/Which Tube Shall I Use.pdf
more detail:
http://www.clarisonus.com/Archives/...System Design Factors for Audio Ampifiers.pdf
The R tail fixes they mention only work at the low current end of the triode characteristic, where gm is curved upward.
However, some tubes do seem to fit the requirements over the full characteristic (gm, last page graph):
http://frank.pocnet.net/sheets/123/1/10JA5.pdf
http://frank.pocnet.net/sheets/123/6/6LU8.pdf
Last edited:
So far, I find two configuration that look very interesting.
1) MrCurwin's mu-follower using MOSFET as source follower with the power tube to drive the OPT. This will a) give horizontal load line to the power tube. And b) low impedance of the MOSFET source follower driving the OPT that give very low output impedance to be more immure to load reactance.
2) The differential power tubes configuration in the Vacuum State amp where the total current between the two power tube remain constant. This involves CCS at the tail of the two power tubes.
BUT it is way too complicated to implement both as you have to have two MOSFET on the top and a BJT CCS at the bottom. ON top and more important, both chew up a lot of voltage headroom. Mu-follow literally eat up half of the voltage swing on the plate side. The differential power tube need CCS that chew the voltage on the cathode side. If I implement both, I am going to need a lot of voltage from the power supply. So I wonder which one is more important, that can make more of a difference. I looked at the picture of the chassis of the Vacuum State amp, it definitely does not have mu-follower as CCS on the top. This is supposed to be the best amp. This make me lean towards doing the differential power tubes with CCS tail. Then either strap the power tubes in triode or UL to lower the output impedance.
One think I don't want to do is to use cheap parts and try it out. That's the reason I do a lot of leg work in studying, reading and asking before I even start. I really don't want to build many amps, my plan is to build one tube and one SS only. So the first thing is to nail a good OPT that can accommodates all possible configuration. So far, I think Lundahl sounds very nice. I think I should get one with UL tap so I have option to do triode, UL, mu-follower and even SE ( they have OPT that can be used as PP or SE).
1) MrCurwin's mu-follower using MOSFET as source follower with the power tube to drive the OPT. This will a) give horizontal load line to the power tube. And b) low impedance of the MOSFET source follower driving the OPT that give very low output impedance to be more immure to load reactance.
2) The differential power tubes configuration in the Vacuum State amp where the total current between the two power tube remain constant. This involves CCS at the tail of the two power tubes.
BUT it is way too complicated to implement both as you have to have two MOSFET on the top and a BJT CCS at the bottom. ON top and more important, both chew up a lot of voltage headroom. Mu-follow literally eat up half of the voltage swing on the plate side. The differential power tube need CCS that chew the voltage on the cathode side. If I implement both, I am going to need a lot of voltage from the power supply. So I wonder which one is more important, that can make more of a difference. I looked at the picture of the chassis of the Vacuum State amp, it definitely does not have mu-follower as CCS on the top. This is supposed to be the best amp. This make me lean towards doing the differential power tubes with CCS tail. Then either strap the power tubes in triode or UL to lower the output impedance.
One think I don't want to do is to use cheap parts and try it out. That's the reason I do a lot of leg work in studying, reading and asking before I even start. I really don't want to build many amps, my plan is to build one tube and one SS only. So the first thing is to nail a good OPT that can accommodates all possible configuration. So far, I think Lundahl sounds very nice. I think I should get one with UL tap so I have option to do triode, UL, mu-follower and even SE ( they have OPT that can be used as PP or SE).
Last edited:
A very interesting thread, many thanks for all contributions! Alan0354 knows how to post the questions in the right way to extract most out of peoples knowledge
Hi Alan0354. You can buy Lundahl from K & K Audio - Lundahl Transformers, audio DIY kits and more in the USA. The prices are, I think, quite similar to the European prices.
Just looked at LL1679 4.5K : 4 ohm. 105W PP and 20W SE. It is $260US each. That seems like the closest fit for EL34 PP.
Do you know how to hook up for SE and what is the primary impedance. Do you just use half by still feed HT to the center tap and the primary impedance is 4.5K/4=1.125K? That sounds very low!!!!
Is the Ammorphous core makes a difference in sound? It is like double the price!!!
Last edited:
Thank you for this post. In wrapping my head around this are you saying that the same effect from the magnetising current applies regardless of the standing magnetisation.. or that it doesnt but simply that attention to the magnetising current leads to better compromises?
Magnetizing current (the "linear" portion) comes from just the primary inductance. Generally a P-P xfmr will have much higher primary inductance L because of the lack of an air gap, and so will have much lower magnetizing current. The core hysteresis shows up as a non-linear and power consuming distortion component on top of that. Both, or the complete magnetizing current, tends to distort the output voltage of the valves via their Rp. So a low Rp output, generally from some form of local feedback, will minimize that effect. The transformer itself faithfully transfers the input voltage to the output except for leakage L loss at HF and some copper loss. Its not like some unrecoverable error loss.
The hysteresis component in most OTs, SE and P-P is usually small enough not to be a big deal anyway at normal listening levels. But by fixing the larger general magnetizing current issue, one fixes the smaller hysteresis component effect as well. Transformers with too few turns for the voltage impressed, or operating at too LF, or with very large output V signals nearing core saturation, will develop large amounts of hysteresis. Hysteresis increases with voltage swing.
If one looks at a textbook magnetics curve, showing hysteresis at various DC operating points, the size of the hysteresis loops do indeed get smaller at higher DC level (for curves with magnetization, or H for the main axis). This is probably why the confusion arises. With H, or magnetization for the bottom axis, what is happening is the permeability goes down as the DC goes up, so a constant variation of magnetization H affects fewer remaining active domains. Fewer domains switching, less hysteresis.
But for a constant AC variation of --voltage-- the picture is different. Any given voltage variation demands the same amount of flux change (at a given frequency), so what happens with DC bias and consequent lower permeability, is the AC current goes up in order to swing sufficient domains to satisfy the AC voltage. When the DC gets the AC operating point up near magnetic saturation, the AC magnetizing current skyrockets to maintain the voltage, copper may even melt. Like plugging a 120 VAC xfmr into a 240 VAC outlet.
So on that earlier graph versus H, one would have to multiply the size of the local hysteresis loops shown by 1/permeability for an equal voltage excitation graph (instead of equal current excitation), bringing their size back to near constant. Magnetic materials are complex, and so it is difficult to say exactly what the outcome will be for hysteresis with DC without measuring an actual sample, but the 1st order expectation will be for similar hysteresis with DC bias versus none, since roughly the same number of domains have to flip for a given voltage. There are significant second order magnetic material effects of all kinds, so this is an obvious simplification, and I'm not claiming expertise at magnetics.
Hysteresis is a fairly minor effect to begin with in a well designed OT, so really much ado about not much. But if someone wants to spends $1000 on some exotic fix, I'm sure someone will take your money. Silver wire, amorphous core, nickel alloy, AC HF bias, gold plated, Litz wire, single crystal copper, ultra thin lamination, aerogel insulation......
The hysteresis component in most OTs, SE and P-P is usually small enough not to be a big deal anyway at normal listening levels. But by fixing the larger general magnetizing current issue, one fixes the smaller hysteresis component effect as well. Transformers with too few turns for the voltage impressed, or operating at too LF, or with very large output V signals nearing core saturation, will develop large amounts of hysteresis. Hysteresis increases with voltage swing.
If one looks at a textbook magnetics curve, showing hysteresis at various DC operating points, the size of the hysteresis loops do indeed get smaller at higher DC level (for curves with magnetization, or H for the main axis). This is probably why the confusion arises. With H, or magnetization for the bottom axis, what is happening is the permeability goes down as the DC goes up, so a constant variation of magnetization H affects fewer remaining active domains. Fewer domains switching, less hysteresis.
But for a constant AC variation of --voltage-- the picture is different. Any given voltage variation demands the same amount of flux change (at a given frequency), so what happens with DC bias and consequent lower permeability, is the AC current goes up in order to swing sufficient domains to satisfy the AC voltage. When the DC gets the AC operating point up near magnetic saturation, the AC magnetizing current skyrockets to maintain the voltage, copper may even melt. Like plugging a 120 VAC xfmr into a 240 VAC outlet.
So on that earlier graph versus H, one would have to multiply the size of the local hysteresis loops shown by 1/permeability for an equal voltage excitation graph (instead of equal current excitation), bringing their size back to near constant. Magnetic materials are complex, and so it is difficult to say exactly what the outcome will be for hysteresis with DC without measuring an actual sample, but the 1st order expectation will be for similar hysteresis with DC bias versus none, since roughly the same number of domains have to flip for a given voltage. There are significant second order magnetic material effects of all kinds, so this is an obvious simplification, and I'm not claiming expertise at magnetics.
Hysteresis is a fairly minor effect to begin with in a well designed OT, so really much ado about not much. But if someone wants to spends $1000 on some exotic fix, I'm sure someone will take your money. Silver wire, amorphous core, nickel alloy, AC HF bias, gold plated, Litz wire, single crystal copper, ultra thin lamination, aerogel insulation......
Last edited:
Having built a Williamson amp (with good results) and later slicing it in half I would say the excess of warmth I got at first was not welcome and should be dealt with. I still suspect the latter.. whether push pull or not.
Not much gets said about cojugating the load, but using passive components in parallel with the speaker to accept and damp the reactive elements certainly doesn't hurt. Makes the speaker look resistive.
Hi Alan,
I was confused at first as well with the Lundahl iron, thinking they could do "everything", but that is not the case. In case of the LL1679 it are in fact two different transformers, one will be called something like LL1679-PP, the other LL1679-70mA. And your USD 260 buys only one of them.
Well, rethinking, you can still take the LL1679-70mA transformer, connect all primaries in series, connect the B+ to the center tap, and one EL34 to each end, for PP operation. Now you will have an "enormous and superfluous" airgap, which will reduce inductance and power handling. As there will be much less standing DC current, you will get something more than the 40H and 20W (which might be sufficient for your needs), but you won't reach the 105H and 100W it could do without the gap.
Another point about the Lundahls. I am mostly a thinkerer, and really liked the idea of having one transformer being able to do multiple impedances. It appears, though, that the "jack of all trades but master of none" applies here as well. Read, for example, this post 814 SE Amplifier: Custom Output Transformers | Bartola Valves
Ale (member here) himself states that the impedance setting were not ideal, and I also think that putting primaries in series would result in better performance. So, yes, Lundahl are incredible value, I think, but maybe not as flexible as the datasheets make us believe.
Thinking a bit further. As you are in the USA you could contact Dave Slagle, from intact audio. His products are positively reviewed, and I once read he makes interstages that can be regapped by the user. Maybe he can do the same with output transformers? More expensive than Lundahl though.
good luck! Erik
Thanks for the warning, I guess I am going to go with the PP version. There is only a very slight outside chance I would want SE anyway. Particular reading the Vacuum State review. I am doing a first pass design on the differential power tubes with CCS tail already. If I do want to try SE, I can always do parafeed using mu-follower. then I don't have to worry about DC current.
Thanks
SE vs PP as to the topic:
I have designed and built 4 single-ended hi fi tube amps. All are different preamp and power tubes. They all sound great to me and I really enjoy listening to them all.
I have designed and built 4 push-pull hi fi tube amps. One transformer-coupled, no coupling caps. They sound great to me and I really enjoy listening to them all.
Push-pull may have the edge with regards to accurate/arcticulate bass, but that is not to say that the SE amps don't do as well. You just have to experience both with different types of music.
My belief is, as stated by others before, the difference is in the speaker system.
YMMV.
I have designed and built 4 single-ended hi fi tube amps. All are different preamp and power tubes. They all sound great to me and I really enjoy listening to them all.
I have designed and built 4 push-pull hi fi tube amps. One transformer-coupled, no coupling caps. They sound great to me and I really enjoy listening to them all.
Push-pull may have the edge with regards to accurate/arcticulate bass, but that is not to say that the SE amps don't do as well. You just have to experience both with different types of music.
My belief is, as stated by others before, the difference is in the speaker system.
YMMV.
Last edited:
As a means of controlling the level of even harmonics (reducing or contributing), a single ended early stage can be as useful as anything else.
The math on this one have so so many mistake it's not even funny. I derived all the equations along, it's just messed up. I gave up when the author said he assume Gamma = 0 to assume only 2nd harmonics!!! He is talking about pentodes here!!!! I can't even take it serious anymore!!! I don't even know why he pull out all the power series stuff for this.
Sorry about that. The paper got OCR scanned by Broskie and ended up with lots of errors in the math. I think the original paper has the math right, can probably find an original with Google.
Here is a link to Broskie explaining the messup:
http://www.tubecad.com/2005/March/10Mar2005.pdf
I'm still confused as to why Kiebert's paper says triodes and pentodes have a different 3rd order coefficient polarity. From looking at gm curves, it seems that either can have upward curvature, at least over part of their gm curve.
The point anyway was that just putting a CCS in the tail is not a magic cure all. Some finite impedance is more likely to give linearity. From looking at various gm curves like the Mosfet, when graphed versus gate V drive (Self, 4th ed. pg 502), the gm curve is a nice linear ramp. The derivative of square law V to I gives a linear ramp for gm. A power law above 2 gives an upward curving gm curve (on Vg base scale), and a 3/2 power law gives a downward trending curve. Since tubes are nominally 3/2 power law, but typically more like square law or above at low current due to grid/cathode wire proximity effects, their curves can vary in curvature. They get more toward a linear power law at high current, like Mosfets do. Cathode current saturation effects also round off the gm curve at the top.
The old low gm tubes can be closer to 3/2 power law. The later high gm tubes tend to start out above square law at low current and drop toward 3/2 power law or lower at high current. So the resistive tail can be useful in the lower current range. Unfortunately, most output tubes don't have gm graphs on their datasheet. But one can probably figure that most later TV sweep tubes are square law or above. One could try to measure that and calc off the plate curves. Some tubes are obvious though with clearly accelerating current gains as -Vg1 goes to zero.
As Broskie mentions above, the optimum tail R tends to be around 1/gm of the tube. I think a crude way of viewing this is that one is degenerating with a linear resistor, so one is effectively reducing the power law for the combo. If one gets it down to square law with linear gm ramps, it is ideal for class A combining. Since gm increases with current, it could be useful possibly to have a non-linear R that (dynamically, ie, AC R) drops with current. I have seen designs where the tube was degenerated with a TV damper (power diode) tube. Primarily they were using it to just get a bias voltage drop, but it has the additional effect of dropping the tube toward 3/2 power law. Unfortunately, all this degenerating stuff lowers the gm, and increases the Rp, so its only something for "touch-up" fixes.
Here is a link to Broskie explaining the messup:
http://www.tubecad.com/2005/March/10Mar2005.pdf
I'm still confused as to why Kiebert's paper says triodes and pentodes have a different 3rd order coefficient polarity. From looking at gm curves, it seems that either can have upward curvature, at least over part of their gm curve.
The point anyway was that just putting a CCS in the tail is not a magic cure all. Some finite impedance is more likely to give linearity. From looking at various gm curves like the Mosfet, when graphed versus gate V drive (Self, 4th ed. pg 502), the gm curve is a nice linear ramp. The derivative of square law V to I gives a linear ramp for gm. A power law above 2 gives an upward curving gm curve (on Vg base scale), and a 3/2 power law gives a downward trending curve. Since tubes are nominally 3/2 power law, but typically more like square law or above at low current due to grid/cathode wire proximity effects, their curves can vary in curvature. They get more toward a linear power law at high current, like Mosfets do. Cathode current saturation effects also round off the gm curve at the top.
The old low gm tubes can be closer to 3/2 power law. The later high gm tubes tend to start out above square law at low current and drop toward 3/2 power law or lower at high current. So the resistive tail can be useful in the lower current range. Unfortunately, most output tubes don't have gm graphs on their datasheet. But one can probably figure that most later TV sweep tubes are square law or above. One could try to measure that and calc off the plate curves. Some tubes are obvious though with clearly accelerating current gains as -Vg1 goes to zero.
As Broskie mentions above, the optimum tail R tends to be around 1/gm of the tube. I think a crude way of viewing this is that one is degenerating with a linear resistor, so one is effectively reducing the power law for the combo. If one gets it down to square law with linear gm ramps, it is ideal for class A combining. Since gm increases with current, it could be useful possibly to have a non-linear R that (dynamically, ie, AC R) drops with current. I have seen designs where the tube was degenerated with a TV damper (power diode) tube. Primarily they were using it to just get a bias voltage drop, but it has the additional effect of dropping the tube toward 3/2 power law. Unfortunately, all this degenerating stuff lowers the gm, and increases the Rp, so its only something for "touch-up" fixes.
Last edited:
This was the basis of a Western Electric patent ("Harmonic Equalizer"). There's a nice discussion of this in Valve Amplifiers, 4th ed, starting on page 184. A practical realization (including the three CCS diff amp that people who have never built it are adamant doesn't work) is shown later in the book in the Bulwer-Lytton amplifier.
Yes, that common mode signal appearing on a tail resistor, or in the B+ lead of a center tapped OT, should have interesting/useful distortion information in it. WE's "Harmonic Equalizer" puts it to work.
Here is a typical TV sweep datasheet, which I'm sure is at or above square law. But no gm curve, Da--it. It was a great performer when George (Tubelab) put it on Pete Milletts DCPP amplifier board. 6LR6 has a very similar curve set. Another great tube.
http://frank.pocnet.net/sheets/084/6/6HJ5.pdf
Here is a typical TV sweep datasheet, which I'm sure is at or above square law. But no gm curve, Da--it. It was a great performer when George (Tubelab) put it on Pete Milletts DCPP amplifier board. 6LR6 has a very similar curve set. Another great tube.
http://frank.pocnet.net/sheets/084/6/6HJ5.pdf
Last edited:
- Status
- Not open for further replies.