I read through both papers by Eddid Vaughn on SE vs PP.
http://www.x3mhc.no/dokumenter/SE-v-PP-Part1.pdf
http://www.x3mhc.no/dokumenter/SE-v-PP-Part2.pdf
I have no experience, I just read and try to understand before I build my amp. Mr. Vaughn is obviously biasd in favor of SE, but there are some valid points he gave also.
1) Can anyone give me link in the argument for PP over SE so I can hear the other side of the story? Anyone disagree with what he say and favor PP class A?
2) In page 8 of the first paper, Mr. Vaughn claimed NFB kills the space and air between the instruments, parts and voices. Is that true?
3) In page 4 and 5 of the second paper, Mr. Vaughn explain in detail the B-H hysteresis, that SE has a big advantage over PP because in SE, the OPT core never goes through ZERO FLUX. So you don't have to worry about the hysteresis distortion. This seems to be a very big advantage to get a good OPT and keep a net DC current large enough to keep the OPT core from ever goes to zero flux during the operation. Is there any argument against this for PP?
4) Follow up of question 3 above, how about using a CCS to always pull a current on one side of the PP OPT to keep the transformer from ever have to go to zero flux to avoid hysteresis. Yes, this means you have to have a bigger more expensive transformer that can take the DC current without getting into saturation. But its should be doable. Then you can have the high power of PP, but no disadvantage of distortion cause by hysteresis.
Thanks. I particularly like to hear the argument against his paper. I feel there got to be a good reason so many high end amps use PP instead of SE.
http://www.x3mhc.no/dokumenter/SE-v-PP-Part1.pdf
http://www.x3mhc.no/dokumenter/SE-v-PP-Part2.pdf
I have no experience, I just read and try to understand before I build my amp. Mr. Vaughn is obviously biasd in favor of SE, but there are some valid points he gave also.
1) Can anyone give me link in the argument for PP over SE so I can hear the other side of the story? Anyone disagree with what he say and favor PP class A?
2) In page 8 of the first paper, Mr. Vaughn claimed NFB kills the space and air between the instruments, parts and voices. Is that true?
3) In page 4 and 5 of the second paper, Mr. Vaughn explain in detail the B-H hysteresis, that SE has a big advantage over PP because in SE, the OPT core never goes through ZERO FLUX. So you don't have to worry about the hysteresis distortion. This seems to be a very big advantage to get a good OPT and keep a net DC current large enough to keep the OPT core from ever goes to zero flux during the operation. Is there any argument against this for PP?
4) Follow up of question 3 above, how about using a CCS to always pull a current on one side of the PP OPT to keep the transformer from ever have to go to zero flux to avoid hysteresis. Yes, this means you have to have a bigger more expensive transformer that can take the DC current without getting into saturation. But its should be doable. Then you can have the high power of PP, but no disadvantage of distortion cause by hysteresis.
Thanks. I particularly like to hear the argument against his paper. I feel there got to be a good reason so many high end amps use PP instead of SE.
1) I have no links, but I do favor balanced class A, provided there's no gNFB.
2) I find it to be true. Even if it's not true, best practise to design so you don't need NFB.
3) Hysteresis distortion is, in my experience, not a major source of distortion. It will not be the bottleneck of an amps performance, not in the top 5 even.
A lot of the PP nastiness (I'd say all) disappears when you
1) do balanced with inherently linear triodes, like 4P1L, 2A3, 45 etc., that don't produce a lot of higher harmonics to begin with
2) design so that you don't need gNFB.
Then it's just a super clean SE with a tad more power, basically.
2) I find it to be true. Even if it's not true, best practise to design so you don't need NFB.
3) Hysteresis distortion is, in my experience, not a major source of distortion. It will not be the bottleneck of an amps performance, not in the top 5 even.
A lot of the PP nastiness (I'd say all) disappears when you
1) do balanced with inherently linear triodes, like 4P1L, 2A3, 45 etc., that don't produce a lot of higher harmonics to begin with
2) design so that you don't need gNFB.
Then it's just a super clean SE with a tad more power, basically.
On this SE OT idea of less hysteresis distortion, I'm not buying that. From knowing some magnetics, when you put DC on a core, the best high permeability domains are the 1st ones to saturate, you are left with the low permeability domains which have higher anisotropy (stickiness). The hysteresis is just not so obvious on a core curve tracer for the gapped core, since the linear (air gap) magnetizing current is so much higher as to hide it. The DC is simply removing half the headroom fluxwise, leaving you with the worst tail on the performance curve.
But then even the linear magnetizing current is 90 degrees out of phase with the signal, so technically it's still distortion. A grain oriented toroid clearly is the way to go for low OT distortion with its high permeability. Local feedback around the outputs to lower Zout is another way to handle any hysteresis current, and should work fine for an E-I core. Many like the E-I sound, its classic as well. Less sensitive to DC imbalance.
There is a technique to remove most hysteresis from cores. Either the core is heat annealed with an orthogonal field (90 degrees to the way it will be used) or an applied 90 degree DC field is used during operation. Either of these close up the hysteresis curve, they also collapse the permeability significantly. Only used for very special applications. One might get a similar result by using grain oriented steel with the orientation the wrong way everywhere. You get a little of that at the back of an E lamination.
Another approach is to use a HF AC bias field, like used by tape recorders, to remove magnetic hysteresis.
Its been tried for audio OTs by inserting a low level HF oscillator into the amp input. It does heat up the OT some (consuming some Amp output power too).
A low Z, local feedback, output stage should come close to the same result.
There is (or was) a technique to make E-I laminations that had the grain orientation the right way everywhere. They used a doubly thick back section to the E and then doubly wide I lams that stacked against that in the OT. The flux could then cross from the E to the double I instead of going thru the back of an E. Result was an E-I core with permeability similar to a toroid setup.
Alternately, the I lams could be left out altogether, and the E lams (extra long ones) got pushed all the way together. That way there is no butt gap to cross in the lams.
But then even the linear magnetizing current is 90 degrees out of phase with the signal, so technically it's still distortion. A grain oriented toroid clearly is the way to go for low OT distortion with its high permeability. Local feedback around the outputs to lower Zout is another way to handle any hysteresis current, and should work fine for an E-I core. Many like the E-I sound, its classic as well. Less sensitive to DC imbalance.
There is a technique to remove most hysteresis from cores. Either the core is heat annealed with an orthogonal field (90 degrees to the way it will be used) or an applied 90 degree DC field is used during operation. Either of these close up the hysteresis curve, they also collapse the permeability significantly. Only used for very special applications. One might get a similar result by using grain oriented steel with the orientation the wrong way everywhere. You get a little of that at the back of an E lamination.
Another approach is to use a HF AC bias field, like used by tape recorders, to remove magnetic hysteresis.
Its been tried for audio OTs by inserting a low level HF oscillator into the amp input. It does heat up the OT some (consuming some Amp output power too).
A low Z, local feedback, output stage should come close to the same result.
There is (or was) a technique to make E-I laminations that had the grain orientation the right way everywhere. They used a doubly thick back section to the E and then doubly wide I lams that stacked against that in the OT. The flux could then cross from the E to the double I instead of going thru the back of an E. Result was an E-I core with permeability similar to a toroid setup.
Alternately, the I lams could be left out altogether, and the E lams (extra long ones) got pushed all the way together. That way there is no butt gap to cross in the lams.
Last edited:
The very best that PP has to offer ? Probably
http://www.vacuumstate.com/fileupload/dpa300B_brochure_lo_rez.pdf
Shoog
http://www.vacuumstate.com/fileupload/dpa300B_brochure_lo_rez.pdf
Shoog
Allen Wright used a CCS tail on the class A output stage.
Looks like Lundahl cut core OTs.
That may well sound terrific, but I see some issues with a CCS tail power stage.
Lets use the E55L tube datasheet, since it has the gm curves shown versus both Vg1 and Ip, and is a near ideal tube as well.
http://frank.pocnet.net/sheets/009/e/E55L.pdf
Top of page 9, the S curve is gm versus cathode current. With a CCS in class A this is the curve you want, since the CCS constrains the total P-P output current to a constant. Any current one tube takes, the other loses. So you flip the S curve around and slide it over the other S curve for class A. Sum the two to get operational gm in class A P-P. You get a curve with a big hump in the middle. This is not so good. This is why you need differential current feedback.
Now take a look at page 7 (either curve set, top is pentode, bottom is triode). These S curves are versus Vg1, which is what is relevant for P-P with grounded cathodes and a phase splitter drive. Again, flip one S around and sum the two. You can change the amount of overlap here (determined by tube bias). Full overlap would be class A. As you can see, it is possible to get quite close to constant operational gm in class A this way. This is good news. Most P-P class A amps are working correctly without coloration. Some differential current feedback could still improve the gm flatness a little, but it doesn't have much work to do here.
Not fully over-lapping the S curves, gives class AB. Differential current feedback gets a real workout to earn its living here. Most class AB amps are just operating in the partial overlapped region (so flattened gm) for normal listening levels, so they sound fine. And they give Kick-A-s performance when the sound gets loud due to the increasing gm (and odd harmonics at high level).
Looks like Lundahl cut core OTs.
That may well sound terrific, but I see some issues with a CCS tail power stage.
Lets use the E55L tube datasheet, since it has the gm curves shown versus both Vg1 and Ip, and is a near ideal tube as well.
http://frank.pocnet.net/sheets/009/e/E55L.pdf
Top of page 9, the S curve is gm versus cathode current. With a CCS in class A this is the curve you want, since the CCS constrains the total P-P output current to a constant. Any current one tube takes, the other loses. So you flip the S curve around and slide it over the other S curve for class A. Sum the two to get operational gm in class A P-P. You get a curve with a big hump in the middle. This is not so good. This is why you need differential current feedback.
Now take a look at page 7 (either curve set, top is pentode, bottom is triode). These S curves are versus Vg1, which is what is relevant for P-P with grounded cathodes and a phase splitter drive. Again, flip one S around and sum the two. You can change the amount of overlap here (determined by tube bias). Full overlap would be class A. As you can see, it is possible to get quite close to constant operational gm in class A this way. This is good news. Most P-P class A amps are working correctly without coloration. Some differential current feedback could still improve the gm flatness a little, but it doesn't have much work to do here.
Not fully over-lapping the S curves, gives class AB. Differential current feedback gets a real workout to earn its living here. Most class AB amps are just operating in the partial overlapped region (so flattened gm) for normal listening levels, so they sound fine. And they give Kick-A-s performance when the sound gets loud due to the increasing gm (and odd harmonics at high level).
Last edited:
Thanks for you reply, What do you mean by this? Do you mean design with low gain and wide band so you don't need GNFB?
If that is the case, then design the power amp with only LTP driving the power tube. Gain of this is not high to need GNFB. To get wider band, lower the output impedance of the LTP by using cathode followers to buffer the two outputs before driving the power tubes. OR using mu-follower like you suggested to me before to get the horizontal load line for the LTP and provide low output impedance.
Mr. Vaughn advice against gap core. I thought I read somewhere a gaped core make it more linear. I don't know what to think. Do you have any link I can read?
How do you do local feedback to lower the Zout?
You think the hysteresis actually gives part of the characteristics of the tube amp sound?
That sounds like a very good idea. How come I never seen people doing this? Is it true the hysteresis gives the sound characteristics of a tube amp?
Thanks for the detail reply. I don't know much about transformer. Do you know of any high quality OPT made in US. Lundahl is from Europe and I have to pay import tax on top of the cost.
Is anamophic core really better?
Mr. Vaughn claimed C-core is better because of the grain oriented in the right way. If it is true, any company in US I can buy from?
I need to spend some time on your second post before I response back.
Thanks
Last edited:
I use both Se and PP amps. Both can be excellent!!
Most important thing with SE is choice of speakers. If you prefer no feedback; buy well damped sensitive speakers..
Nothing worse than a SE amp driving speakers destined for high damping factor Transistor amps and the boomy sound that results.
Phil.
Most important thing with SE is choice of speakers. If you prefer no feedback; buy well damped sensitive speakers..
Nothing worse than a SE amp driving speakers destined for high damping factor Transistor amps and the boomy sound that results.
Phil.
From page 4 second paper:
"The offset DC creates a static magnetic field that holds
the transformer in a highly linear region of it's magnetization
curve, even at very low signal levels. It avoids
ever seeing a zero flux condition, whereas PP does not. "
This is a common mis-perception from looking at the AC magnetization curve which has a loop/kink in the middle. The assumption is that by applying DC bias one moves up the curve away from the kink. Unfortunately, its not true. No matter where one operates on the magnetization curve, for a specified AC voltage around the core, the same amount of magnetic flux must change, or the same number of domains flip essentially. What happens is (with DC bias) the AC sets up a new hysteresis curve around the new operating point. The domains still require energy to flip (causing hysteresis). Most any good magnetics book will show this. An old standard is the MIT "Magnetic Circuits and Transformers" book.
DC bias also makes the core permeability go down drastically (even before putting a gap in), since it eliminates half (the best half) of the useable magnetic domains (SE bias case).
The gapped core will make the effective core permeability more constant (air is constant), but by dilution. Ie, the permeability just goes down with more gap. Higher permeability reduces magnetization current absolutely, so is the best way to get rid of hysteresis distortion current. (by the way, it is only unwanted primary current, it does not directly distort the output V of the OT)
Hysteresis might contribute some to a badly designed or cheap E-I core OT sound. Its not a big factor typically, and local feedback can overcome that.
Local feedback can be several techniques. UL (ultra-linear Fdbk to the screens), CFB (cathode feedback winding), resistive feedbacks to the driver stage plates (Schade Fdbk), grids (crossed Fdbks for phase), or cathodes. The point is to lower the output Z of the tubes, so that unwanted magnetizing currents have little effect on their voltage.
C core and toroids typically will use grain oriented material with the easy grain along the magnetic path. Toroids will be very sensitive to DC imbalance though.
There are a few OT winders in the US. Someone will have to chime in on any ones that use C cores (not real common). I only know of one "local" toroid OT seller and they were using imported stuff wound with cheapo power supply type random winds (and had very poor BW performance). Toroid OTs need to be wound with special winding technique (like progressive wind) and tend to be expensive. Plitron makes high quality toroid OTs.
Amorphous material typically has lower losses, so less hysteresis. You would want to make sure it still had good permeability at low and high signal level. Another technique is to add 80% nickel alloy pin-striping lams to E-I OTs to give good performance at low signal levels. Then there is nano or micro crystalline material. These specialty alloys are typically quite expensive. Hi-B material is a Japanese (I think) grain oriented material one hears good things about and is economical.
"The offset DC creates a static magnetic field that holds
the transformer in a highly linear region of it's magnetization
curve, even at very low signal levels. It avoids
ever seeing a zero flux condition, whereas PP does not. "
This is a common mis-perception from looking at the AC magnetization curve which has a loop/kink in the middle. The assumption is that by applying DC bias one moves up the curve away from the kink. Unfortunately, its not true. No matter where one operates on the magnetization curve, for a specified AC voltage around the core, the same amount of magnetic flux must change, or the same number of domains flip essentially. What happens is (with DC bias) the AC sets up a new hysteresis curve around the new operating point. The domains still require energy to flip (causing hysteresis). Most any good magnetics book will show this. An old standard is the MIT "Magnetic Circuits and Transformers" book.
DC bias also makes the core permeability go down drastically (even before putting a gap in), since it eliminates half (the best half) of the useable magnetic domains (SE bias case).
The gapped core will make the effective core permeability more constant (air is constant), but by dilution. Ie, the permeability just goes down with more gap. Higher permeability reduces magnetization current absolutely, so is the best way to get rid of hysteresis distortion current. (by the way, it is only unwanted primary current, it does not directly distort the output V of the OT)
Hysteresis might contribute some to a badly designed or cheap E-I core OT sound. Its not a big factor typically, and local feedback can overcome that.
Local feedback can be several techniques. UL (ultra-linear Fdbk to the screens), CFB (cathode feedback winding), resistive feedbacks to the driver stage plates (Schade Fdbk), grids (crossed Fdbks for phase), or cathodes. The point is to lower the output Z of the tubes, so that unwanted magnetizing currents have little effect on their voltage.
C core and toroids typically will use grain oriented material with the easy grain along the magnetic path. Toroids will be very sensitive to DC imbalance though.
There are a few OT winders in the US. Someone will have to chime in on any ones that use C cores (not real common). I only know of one "local" toroid OT seller and they were using imported stuff wound with cheapo power supply type random winds (and had very poor BW performance). Toroid OTs need to be wound with special winding technique (like progressive wind) and tend to be expensive. Plitron makes high quality toroid OTs.
Amorphous material typically has lower losses, so less hysteresis. You would want to make sure it still had good permeability at low and high signal level. Another technique is to add 80% nickel alloy pin-striping lams to E-I OTs to give good performance at low signal levels. Then there is nano or micro crystalline material. These specialty alloys are typically quite expensive. Hi-B material is a Japanese (I think) grain oriented material one hears good things about and is economical.
Last edited:
Thanks a million!!! This makes a whole world of sense. It is not about the absolute sum neutral of all the micro dipoles. Each dipole has to be flip individually and each have it's own hysteresis. So no matter the total sum is biased one way or the other, each dipole has hysteresis that has to be overcome.
I don't quite follow. Do you mean higher permeability reduces hysteresis? So I should not get one with air gap if I can help it?
I am planning on UL. I think I can get more power using Class AB with the first 5W or so in pure class A. From what I read, because I can get more wattage, I actually get lower distortion at the real life level because distortion goes down with backing off the volume. At the same level of a triode pp, UL actually has lower distortion.
I am just an end user, so I can only buy OPT, and have to take whatever is on the market. So I really need some US brand to look at.
Thanks for your detail explanation, I really learn a lot from you.
Last edited:
I have a pair of JM Lab 913 speaker, it's a 3 way speaker with separate input for woofer and mid/tweeter. It is spec 4 ohm. I don't know anything else. I read that speakers with 3 way crossover is not suited for SE, is that true?
1)
It's nonsense so true is a word that simply does not apply.
Enough nonsense to stop reading anything else claimed in that "paper".
2) there's an excellent solution to transformer hysteresis, saturations, and other worries and that's plain going SS and forgetting about output transformers.
The rest of the World must agree somehow, because 99.999% of all amps in use on the Planet Earth are SS .
It's nonsense so true is a word that simply does not apply.
Enough nonsense to stop reading anything else claimed in that "paper".
2) there's an excellent solution to transformer hysteresis, saturations, and other worries and that's plain going SS and forgetting about output transformers.
The rest of the World must agree somehow, because 99.999% of all amps in use on the Planet Earth are SS .
Yes, I like that design a lot. The differential constant current source part of it is very interesting and makes a lot of sense.
I built Lynns OP stage w 2A3's biased well into Class A and loaded 3k3 ea via 6k6 p-p, OPT were good enough and it was great.
Built again with Gpimm CCS from the filament CT to ground via paralleled pots, wipers back to the respective secondaries for some DC balance (IT coupled) it was miles.. say way way better. Just try it, and Gpimm was all I had on hand, try A.Wrights IC by all means.
PSE for the same output power was more narrow, but deeper, more forgiving and more immediate. If I want to listen to rock, or AV .. DVD etc.. differential throws a wider image.. very impressive.. but for my taste PSE is better
Built again with Gpimm CCS from the filament CT to ground via paralleled pots, wipers back to the respective secondaries for some DC balance (IT coupled) it was miles.. say way way better. Just try it, and Gpimm was all I had on hand, try A.Wrights IC by all means.
PSE for the same output power was more narrow, but deeper, more forgiving and more immediate. If I want to listen to rock, or AV .. DVD etc.. differential throws a wider image.. very impressive.. but for my taste PSE is better
Last edited:
- Status
- Not open for further replies.