Hi,
I have never tried to design an output transformer, but the problem seems to be trying to get 10octaves into the passband and minimise the compromises at both ends of the audio spectrum.
I suspect much of the compromise can be avoided by designing for only 5octaves. Similarly the bottom 5octaves can be more cheaply served by a dedicated bass/mid transformer.
The Audax HD3p tweeter has an air cored transformer to transfer all the drive power and is reputed to sound quite nice handling just the top 2octaves.
I have never tried to design an output transformer, but the problem seems to be trying to get 10octaves into the passband and minimise the compromises at both ends of the audio spectrum.
I suspect much of the compromise can be avoided by designing for only 5octaves. Similarly the bottom 5octaves can be more cheaply served by a dedicated bass/mid transformer.
The Audax HD3p tweeter has an air cored transformer to transfer all the drive power and is reputed to sound quite nice handling just the top 2octaves.
For what it is worth, here are my suggestions on transformers.
What range of frequencies applied to a transformer is not the only determinant. If you use a SE circuit, then it would be advantageous to divide the OPT's up into a minimum of three and a maximum of five separate transformer coupled circuits.
The bottom range, from 20 to 300 with an EI cored, minimum winding capacitance beast, just to keep the total gap under 10mils.
Upper Bass (300 to 800Hz) or upper bass and Mid transformer (300 to 3500Hz) could be an E/I with M3 core or an amorphous C core. The amorphous C core device will have to have very special attention paid to its dielectric circuit, so that as the E Field collapses to B Field, the ultra fast rise time amorphous material is not driven into an extreme half wave signal spike, with attendant ringing. An E/I M3 core will work as well or better here. This is keeping in mind that ultimate transient response is NOT the best plan at these frequencies.
Low mid to Hi mid (800 to 3.5KHz) An E/I M3. 48% nickle or amorphous core will all work well here, with the best sound leaning toward the M3 core.
Mid HI to Hi frequencies. For the Mid HI only, 3.5k to 8 k an E/I nickle core (48%) will be the best bet, micro gapped to flatten permeability induced rise in inductance/distortion. A small amorphous core transformer will also work well here, but again special attention to coil dielectrics will be needed.
For hi frequencies (8kHz to 40kHz), it depends upon the driver. For Raal ribbons, amorphous C core wound as a multi section and externally connected tapped auto former, an E/I core nonsaturable reactor, wound as a multisection isolation transformer and conected as a tapped auto former, or as an isolation transformer and using 48% nickle core, or an MPP torroidal core transformer such as Tribute supplies. However a transformer using core here must have very fast rise time, so multi sections for amorphous core or nickle and whatever Tribute does with their exceptional torroids is needed.
Air cores are not particularly linear devices, because of field clumping, and we are talking about ultimate performance devices, as I understand it. This does not mean they are poor choices, but a properly designed ferrous core device will always sound smoother, more transparent and with cleaner transient rise and decay.
For a push pull device, E/I in M3 or Nickel, with a properly designed core, that does not exhibit hysteresis induced reminence, will be a superior choice and can actually handle the entire frequency range with just one out put transformer. You can divide them up into catagories that depend upon the speaker drivers and lean towards nickel for the 3.5kZ and up realm, but this will be the sonically superior choice.
Bud
What range of frequencies applied to a transformer is not the only determinant. If you use a SE circuit, then it would be advantageous to divide the OPT's up into a minimum of three and a maximum of five separate transformer coupled circuits.
The bottom range, from 20 to 300 with an EI cored, minimum winding capacitance beast, just to keep the total gap under 10mils.
Upper Bass (300 to 800Hz) or upper bass and Mid transformer (300 to 3500Hz) could be an E/I with M3 core or an amorphous C core. The amorphous C core device will have to have very special attention paid to its dielectric circuit, so that as the E Field collapses to B Field, the ultra fast rise time amorphous material is not driven into an extreme half wave signal spike, with attendant ringing. An E/I M3 core will work as well or better here. This is keeping in mind that ultimate transient response is NOT the best plan at these frequencies.
Low mid to Hi mid (800 to 3.5KHz) An E/I M3. 48% nickle or amorphous core will all work well here, with the best sound leaning toward the M3 core.
Mid HI to Hi frequencies. For the Mid HI only, 3.5k to 8 k an E/I nickle core (48%) will be the best bet, micro gapped to flatten permeability induced rise in inductance/distortion. A small amorphous core transformer will also work well here, but again special attention to coil dielectrics will be needed.
For hi frequencies (8kHz to 40kHz), it depends upon the driver. For Raal ribbons, amorphous C core wound as a multi section and externally connected tapped auto former, an E/I core nonsaturable reactor, wound as a multisection isolation transformer and conected as a tapped auto former, or as an isolation transformer and using 48% nickle core, or an MPP torroidal core transformer such as Tribute supplies. However a transformer using core here must have very fast rise time, so multi sections for amorphous core or nickle and whatever Tribute does with their exceptional torroids is needed.
Air cores are not particularly linear devices, because of field clumping, and we are talking about ultimate performance devices, as I understand it. This does not mean they are poor choices, but a properly designed ferrous core device will always sound smoother, more transparent and with cleaner transient rise and decay.
For a push pull device, E/I in M3 or Nickel, with a properly designed core, that does not exhibit hysteresis induced reminence, will be a superior choice and can actually handle the entire frequency range with just one out put transformer. You can divide them up into catagories that depend upon the speaker drivers and lean towards nickel for the 3.5kZ and up realm, but this will be the sonically superior choice.
Bud
Hi
Agree 100% about caps but definitely not about the opams – once you fix the power supply issue ( and some more faults of th "B"
) these opamps are really lovely – transparent – musically – natural – no opamp sludge at all.
The power supply issue, by the way, is one of the really big hurdles also in all tube amps and even with voltage cascoding circuits I found it something better not to be overlooked.
The NAIM circuit can perform great as XO buffer – if driving and being driven properly, its a little bit prone to ringing.
Just out of curiosity - anybody who knows which AD / DA and opamps used in the Xilica ?
Using transformers I am always a little bit suspicious about harmonics like outlined some time back
http://www.kinotechnik.edis.at/pages/roehre/el34/kleinuebertrager/thd_kleinuebertr.html
scroll down to the bottom- there you find a 100w output transformer driven at 760mV rms at 30 Hz and 80 Hz at no load.
Harmonics are only down to around 60 dB which is kind of surprise at least to me.
BudP, thanks for this great insights - do you have some measurements you can share ?
Greetings
Michael
Lynn Olson said:
Agree 100% about caps but definitely not about the opams – once you fix the power supply issue ( and some more faults of th "B"
) these opamps are really lovely – transparent – musically – natural – no opamp sludge at all. The power supply issue, by the way, is one of the really big hurdles also in all tube amps and even with voltage cascoding circuits I found it something better not to be overlooked.
The NAIM circuit can perform great as XO buffer – if driving and being driven properly, its a little bit prone to ringing.
Just out of curiosity - anybody who knows which AD / DA and opamps used in the Xilica ?
Using transformers I am always a little bit suspicious about harmonics like outlined some time back
http://www.kinotechnik.edis.at/pages/roehre/el34/kleinuebertrager/thd_kleinuebertr.html
scroll down to the bottom- there you find a 100w output transformer driven at 760mV rms at 30 Hz and 80 Hz at no load.
Harmonics are only down to around 60 dB which is kind of surprise at least to me.
BudP, thanks for this great insights - do you have some measurements you can share ?
Greetings
Michael
Sorry, I do not. I lack proper test equipment, protocols and enough time. There is a possibility that Rick Chin, local AES head here in the NW USA, will provide all of these in the relatively near future, as we work up a presentation on "voiced" output transformers for guitar amps.
EC8010 did post a scope pic of one of my PP OPT's being driven by 2A3 tubes and a 10 kHz square wave, on another thread. I am not sure of the test parameters such as load etc, am sure the OPT's were not designed for 2A3's and so had a 26 kHz ringing I have not seen when driven by their design center tube.
As for harmonics, well, E/I core, that has not had it's zero crossing lag, due to B saturation hanging at full magnetic potential until H null and beyond relieves it, is going to have to show strong additional harmonic problems. Lynn actually has E/I transformers built to solve this problem, as do Gary Pimm, Gary Dahl and many others. Only Gary Pimm has any test data that I am aware of, but Lynn should still remember what these devices sound like, well enough to comment in a qualitative sense.
Bud
Hi
Will gonna be interesting for sure.
My two measurements shown were taken from a toroidal type transformer - the Plitron 2100-CFB – no DC was involved
http://www.plitron.com/audio_sr.asp
Whats the trick here ?
Greetings
Michael
BudP said:
Will gonna be interesting for sure.
BudP said:
My two measurements shown were taken from a toroidal type transformer - the Plitron 2100-CFB – no DC was involved
http://www.plitron.com/audio_sr.asp
BudP said:
Whats the trick here ?
Greetings
Michael
migeO,
B sat referred to is the return leg of an AC driven BH curve, where B hangs on and keeps core acting as if it were a permanent magnet, until H has reached null and proceeds on the opposite polarity swing, causing the permanent magnet stasis in the core to collapse. Not an application of DC.
Trick is just to cause this collapse as a passive result of loss of H, from the B/H sat moment to null in H. Eliminates all of that crashing nastiness, zero crossing error and huge volumes of antenna type mag field reacting to non linear phase, mag field transform between sec and pri. Method is a trade secret. Only because I have to make part of my "money to live indoors with" from it.
Bud
B sat referred to is the return leg of an AC driven BH curve, where B hangs on and keeps core acting as if it were a permanent magnet, until H has reached null and proceeds on the opposite polarity swing, causing the permanent magnet stasis in the core to collapse. Not an application of DC.
Trick is just to cause this collapse as a passive result of loss of H, from the B/H sat moment to null in H. Eliminates all of that crashing nastiness, zero crossing error and huge volumes of antenna type mag field reacting to non linear phase, mag field transform between sec and pri. Method is a trade secret. Only because I have to make part of my "money to live indoors with" from it.
Bud
PigletsDad said:
phase_accurate said:
OK, then here are my two cents worth. Two circuits UG/Buffer. They are very simple and have already been built in large numbers.
1st the tube/hybrid version:
This is only the starting point. Somebody with an eagle eye will recognise this from the Gainclone/Chip Amps forum. Lots of these were built as Buffer for Inverted Gainclones. Split supply is the key to its simplicity.
But a more complete version was designed for the JLTi Integrated amp. This added the extra key missing elements:
Only a little bit more complex. I have used this as a Unity Gain Preamp stage (CD Players with full 2VRMS don't need much gain, right?) and it has typically embarrassed a number of commercial 10K-buck Line Stages. I can testify that it sounds very good indeed!!!
Note the missing components values, but it is not hard to figure out what does what. So Pedja Rogic took the circuit typology and converted it to an all-FET version seen and described here:
http://www.pedjarogic.com/gc/index.html
I am sure he won't mind me posting it here (he borrows from me, so tit-for-tat):
What's going on here? I can't type T I T as in tit-for-tat? Auto censored? Must be American prudeness!
I wonder how many words are censored - the three letters used in the phrase I typed has nothing to do with female upper body attachments!!!
OK, back to subject.
Even the tube version does not need HT, based on 6922, no need to get terrified working with high voltages. Good way to get into tubes in a safe way.
Joe R.
Tom Danley said:
Hi Tom, nice to see you on board!!! Thanks for the tip about the Xilica XD4080. All of my amps are fully balanced transformer coupled, using Jensen or higher-grade transformers throughout. It's one way to get hum, noise, and RFI induction down to extremely low levels, and thoroughly break all ground loops.
The way this discussion is going, a pro-grade EQ/crossover for the 500 Hz and lower range looks promising, and some kind of basic Aikido-alike for the highpass part of the signal. Sorry to be so down on op-amps, but I went through a lot of those things back when I was looking for a linestage for the Amity power amp.
They all sounded more different than I expected, to my surprise, and the old ones certainly had the grungy, grainy, and muffled sound I remember so well from the early Eighties. The Nineties-era opamps sounded a lot better, but still not that close to a straight-wire bypass - my standard test for any linestage.
Although come to think of it, the tube linestages (the ones with coupling caps) weren't all that great either. My dissatisfaction with all of these things is what led to the Raven preamp, which turned out OK but not fabulous. Hmm - I really don't want to re-visit all of that hassle again.
Returning to the SRPP discussion, I have a simple-minded model of what tubes do - they're nothing more than voltage-controlled variable resistors. The voltage potential of interest, of course, is applied to the grid and cathode - the tube isn't interested in arbitrary references like "ground" or "B+".
Looking at Broskie's circuit, the B+ is fixed at 300V and the plate of the bottom tube is at 150V (quiescent, no signal). If the plate swings up to 200V, the momentary internal resistance of the lower tube has gone up. There is a complementary decrease in the upper tube (it only sees 100V across it), so net current flow through both tubes is more or less constant. If no current is drawn by the load, the complementarity is good.
This is similar to the see-saw current flow and complementary changes in internal resistance of a classic push-pull circuit in Class A with transformer coupling. The difference between the two is when a load is applied. With a SRPP, the load disrupts the balance, while a true push-pull circuit does not lose balance.
If the current drawn (by the load) is great enough to shut one of the tubes off, a true push-pull circuit merely reverts to Class AB, and can easily deliver several times the quiescent current to the load. By contrast, the SRPP is indeed a series circuit, and current-clips when excess current is demanded. That's a big difference. Just visualize how the circuits behave when the load is shorted - true push-pull turns one tube on as hard as it will go, while the other circuit simply clips.
John Broskie analyzed the current flow of the SRPP and recognized the sensitivity to the load, which unbalances the SRPP. Using a cathode-bypass cap also unbalances a SRPP, as he points out. Since the bottom tube is so heavily loaded (it's seeing a load equal to Rp, which is pretty low), the only hope for linearity is complementarity, which the Aikido delivers, along with very intelligent noise cancellation. The complementarity is mirrored in the output stage, which also has complementary (see-saw) changes in resistance between the upper and lower tube.
If I understand the Aikido correctly, if it is built with identical tube sections in all four parts, the audio-frequency changes in current flow drawn from the supply should be constant - this essentially eliminates power-supply coloration, which is a response to changes in current demand from the audio circuit. This is similar to a shunt regulator, which has a complementary current demand to the audio circuit it is powering, thus buffering current-demand surges from the earlier parts of the supply.
Joe, many thanks for illustrating the cascoded follower circuits. These are simple and very very good, and I daresay probably better than any op-amp you can buy. All Class A, plenty of current delivery if you're smart, pretty much unbreakable (YAY!), the only real limitation is the (inevitable) output-coupling cap.
There's a terrible temptation to use a much smaller cap at the input and a cute servo circuit to steer the output to zero volts (DC), but designing good servo circuits (that are reliable and never fail) is an art unto itself. Better to spend the moola on an exotic cap instead.
There's a terrible temptation to use a much smaller cap at the input and a cute servo circuit to steer the output to zero volts (DC), but designing good servo circuits (that are reliable and never fail) is an art unto itself. Better to spend the moola on an exotic cap instead.
Lynn Olson said:
Hi Lynn
We have done exactly just that, a variation of that even before Pedja did his all-FET version, and that is what a whole lot of owners of Vacuum State converted Sony SACD players have been listening to since 2002, and no output cap. Must be getting close to a thousand players, at least 500 plus!
Reliable? Ultra-reliable!!!
But ours mixed Fets and Bi-polar. Darlington pair Bi-Polar as the emitter follower is better sounding that Pedja's Fet. But there will now be some DC on the input anyway, so servo is a natural.
The key to good sounding servos is to make them S.....L....O.....W.... and then tell the user to be patient and don't have the power amps ON. Give it five minutes.
Joe R.
phase_accurate said:
So Q2, Q3, Q4, Q5, R1, R2, R3, and R4 are simply current sources, right? Lots of ways to build them, as many or as few parts as you want.
The actual audio circuit looks like it is Q1, Q6, Q7, D1, and D2. D1 and D2 can be replaced by matching transistors if you're really into precision temperature-tracking. Be interesting to see if the real thing has distortion residuals of -100 dB ... certainly would be nice if it did. No caps, I like that a lot.
Joe, thanks for the real-world experience on servos. Congratulations on the reliability - these things can be system-killers if anything goes wrong. Sounds like a real belt-and-suspenders approach would be a normally-closed shorting relay on the output of the circuit. Since it's a shorting relay, contact corrosion is a non-issue, and it ain't rocket science to tell it to open when the circuit is DC-stabilized. (I worry about weird things like a hot-start from one of those nasty 1-second powerline dropouts - that can send a gigantic transient through the whole system before you can run over and turn it off.)
Sometimes I really like my transformers - just turn it on and use it, no worries about caps discharging or servos stabilizing. Yes, I have hot-started my power amps - the power in Silverdale, Washington was truly 3rd-world quality, ugly 1~2 second dropouts followed by power surges. This happened several times a year, as often as not with no warning. (Poor Gary Dahl probably had several during the last windstorm.) I eventually had to power the 36" HDTV off a sinewave UPS to prevent it from being destroyed from one of those powerline transients. My power amps had to fend for themselves, thus the hot-start-without-system-transient requirement.
The power here in Northern Colorado is much more reliable, but then again we get 120-mph winds (microbursts), 1/2" hailstones, and whiteout-visibility snowstorms on a regular basis. If the Colorado infrastructure were built as flimsy as the Northwest the whole place would fall down around our ears.
Attachments
The worry about system transients from powerline-induced hot-starts is not an idle concern. These things happen, and with the kind of standing voltages inside tube preamps, really huge (20~50V) transients can be sent to the power amps. This can destroy or badly degrade the input transistors for nearly any transistor power amp (which is why using tube preamp can void a transistor-amp warranty).
I use transformer-coupled triode amplifiers, but the thought of a full-sized system transient going into a 103 dB/metre efficient loudspeaker is more than a little frightening. Since the speaker is going to be built from prosound components (even the RAAL is good for 200 watts clipping), it wouldn't be damaged, but the transient itself would sound like the crack of doom at those kind of efficiency levels.
I use transformer-coupled triode amplifiers, but the thought of a full-sized system transient going into a 103 dB/metre efficient loudspeaker is more than a little frightening. Since the speaker is going to be built from prosound components (even the RAAL is good for 200 watts clipping), it wouldn't be damaged, but the transient itself would sound like the crack of doom at those kind of efficiency levels.
Lynn's analysis of the circuit is correct. I assume he means to substitute D1 and D2 by transistors used as diodes, right ?
But even tough the output offset voltage of this circuit is minimised it will still not be zero. Maybe a symmetrical version would help here. I will think about such a variant.
Regards
Charles
But even tough the output offset voltage of this circuit is minimised it will still not be zero. Maybe a symmetrical version would help here. I will think about such a variant.
Regards
Charles
Hi,
Am I right in thinking that the voltage across R5 sets the output bias voltage?
6mA across 600r =3.6Vdc.
Will trimming the CCS currents, force Q1 collector to a lower voltage by reducing the current flow through R5?
How low can we run R5 current and still get correct loading?
If the lower connection of R5 were returned to a lower voltage, maybe -3.6V using a pair of LEDs, would the circuit still work or work badly?
Am I right in thinking that the voltage across R5 sets the output bias voltage?
6mA across 600r =3.6Vdc.
Will trimming the CCS currents, force Q1 collector to a lower voltage by reducing the current flow through R5?
How low can we run R5 current and still get correct loading?
If the lower connection of R5 were returned to a lower voltage, maybe -3.6V using a pair of LEDs, would the circuit still work or work badly?
Urk. I didn't realize DC is passing through the load. Not just any old DC, but DC that sets the bias current. This rules out the circuit for the input transformers on my power amps - transformers intended for studio line-level use will not tolerate any DC at all on the primary.
It won't hurt the transformers - no real problem there - but midrange and bass distortion will go up hundreds of times. If DC is kept out of the transformer, distortion is low enough that they qualify for the input of the Audio Precision test system - or making the classic recordings of the Fifties and Sixties.
It won't hurt the transformers - no real problem there - but midrange and bass distortion will go up hundreds of times. If DC is kept out of the transformer, distortion is low enough that they qualify for the input of the Audio Precision test system - or making the classic recordings of the Fifties and Sixties.
I have to say someone here should take the Marchand XM1 circuit and tell me why it's compromised. With the right sounding resistors, opamps. caps and power supply it sounds lovely = as in as good as good triode line stages, ect.. I am not a 'circuit dude' so please look at it and point me in the right direction. I already have done the best parts swaps! What am I missing?
I have used passive line level with r/c and with also with little military chokes. also used the audio research ec3 tube crossover with good caps- neither sound as good
XM1 Marchand
I have used passive line level with r/c and with also with little military chokes. also used the audio research ec3 tube crossover with good caps- neither sound as good
XM1 Marchand
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.