JLH said:
Doanworrybouddit. For the cathodyne, the main imbalance is a capacitance problem. For any reasonable design, you have the first turn-over at ~250KHz, with a second at ~1.8MHz -- too far away from the audio band to make any significant difference.
One o'scope design actually used a pentode-based cathodyne in its horizontal scan circuit that was good to 8.0MHz.
That A Number One weakness with the cathodyne is the lack of voltage gain. Output swing limitations, and the intolerance for driven stage grid current can be worked around. This explains the Williamson design wherein the cathodyne was moved back in the signal chain, and the finals being driven with a differential.
Hi Miles,
Do you know of any single tube PIs with gain? The Concertina actually delivers 1:1+1. Shouldn´t that be considered 6dB?
Just joking, just want to say that the Concertina is a magnificient "invention"!
Anyone tried a Iron Concertina? Could be done with a LL1667 or 68. Looks great on paper! A 6C45 at 10mA with 220V B+ will give you a 200V ptp swing at each output with less than 50ohm Zout!
A LTP takes a doubletriode.
By adding a second triode to the concertina you actually get 6dB higher gain than the LTP
. The gain of a Seesaw or Paraphase equals the 2-tube Concertina.
On the downside, as I see it, could eventually be that you need higher B+ to get the same output swing as a LTP with negative tail or a Paraphase.
They all sound different so you will anyway have to judge them by your ears.
Do you know of any single tube PIs with gain? The Concertina actually delivers 1:1+1. Shouldn´t that be considered 6dB?
Just joking, just want to say that the Concertina is a magnificient "invention"!
Anyone tried a Iron Concertina? Could be done with a LL1667 or 68. Looks great on paper! A 6C45 at 10mA with 220V B+ will give you a 200V ptp swing at each output with less than 50ohm Zout!
A LTP takes a doubletriode.
By adding a second triode to the concertina you actually get 6dB higher gain than the LTP
. The gain of a Seesaw or Paraphase equals the 2-tube Concertina. On the downside, as I see it, could eventually be that you need higher B+ to get the same output swing as a LTP with negative tail or a Paraphase.
They all sound different so you will anyway have to judge them by your ears.
revintage said:
The 6db(v) gain applies if you measure phase-to-phase, but not if you measure phase-to-AC ground, which is what you do most of the time anyway. As for single tube PI's with gain, no not really. I've seen some pretty esoteric PI designs, but there are none that work better than either a cathodyne or the LTP/Schmidt/differential.
It's still one hole in the chassis.
By adding a second triode to the concertina you actually get 6dB higher gain than the LTP
The A Number One problem with this and related topologies is an imbalance in harmonic distortion between phases. The "corrected" tube always has lower distortion than the driven side, whereas you don't get that problem with the cathodyne or LTP, especially if you use an active load in the tail.
So far, all I've ever used is the LTP, and this one sounds just great.
See, in your world it´s LTP, which I respect. It is true for you but why should it be for all off the rest of us ?
In my world that is the least preferable. No matter what technical explanation one might lean against.
They each and everyone has their imperfections. And the LTP is in no way symmetrical either.
So my PI-choice is transformer albeit the best are terribly expensive.
? In my world that is the least preferable. No matter what technical explanation one might lean against.
They each and everyone has their imperfections. And the LTP is in no way symmetrical either.
So my PI-choice is transformer albeit the best are terribly expensive.
With an ECC40 a simple tube followed by a concertina works very well as long as it's not asked to drive heavy loads, which in this case it isn't. In my headphone amp, this is what I use as the frontend and I've yet to build an output stage which realizes the full potential of the input/PI stage. I haven't tried an LTP for the input stage yet, then again I don't have a reason to do so until I get the output stage up to scratch.
It can be, if you use a good CCS in its tail and tightly matched plate load resistors. That will force any change in voltage at one plate to be exactly equal and opposite to the change in voltage at the other plate = perfect AC balance. The only possible lack of symmetry is in DC balance, if the two triodes are not well-matched. Adding an unbypassed cathode resistor in each half, above the CCS, will usually iron out mismathing provided that it's not too severe. The LTP, properly designed and executed, is IMHO the best all-round tube-based splitter. It gives decent voltage gain, generous headroom, low distortion, superb balance and good PSRR. All the other designs using tubes are worse in some respect.
I think second best is the cathodyne but, of course, it lacks headroom and can't tolerate heavy or unequal loading, which makes it unsuitable for driving OP stages directly (as in the Dynaco designs). The questions raised about its balance were adequately answered many years ago by Albert Preisman (see http://www.aikenamps.com/cathodyne.pdf).
I don't think any other type of tube-based splitter is worth bothering with, and that includes the Van Scoyoc and the paraphase.
Transformer splitting can be excellent, if the transformer in question is of high enough quality and you don't try to include it in a negative FB loop, but it ain't cheap and IMHO is not worth the cost.
The beauty of the concertina is it's guaranteed performance when replacing tubes. When used with a low u Williamson diff driver this will sort out any upper frequency misbalance defects. By the time the high freq capacitances really upset the stage, the output tranny will be a useless performer.
A point to bear in mind, the Williamson diff driver can provide an o/p Z around 2-5K, enough and low enough for most output stages.
For HiFi, the concertina with a triode diff driver is the obvious choice but do the stage gain calculations. The cherry on the icing is to use a triode front end with concertina (instead of true pentode stage), and use a step-up input tranny.
For MI, the LTP is ok but anode resistors have to be tweaked for balance and all the amps I've examined with this configuration had varying amounts of misbalanced output volts resulting in double figure thd without the owner noticing it!
So does it matter ?
richj
A point to bear in mind, the Williamson diff driver can provide an o/p Z around 2-5K, enough and low enough for most output stages.
For HiFi, the concertina with a triode diff driver is the obvious choice but do the stage gain calculations. The cherry on the icing is to use a triode front end with concertina (instead of true pentode stage), and use a step-up input tranny.
For MI, the LTP is ok but anode resistors have to be tweaked for balance and all the amps I've examined with this configuration had varying amounts of misbalanced output volts resulting in double figure thd without the owner noticing it!
So does it matter ?
richj
Not tweaked (assuming a good CCS in the tail), just matched. Same with the concertina. Each has its advantages and disadvantages, so it's not like there's a universal "better." They both can have essentially perfect balance with the same amount of matching (two resistors). Concertina has lower output impedance. LTP has gain. Both are subject to error if the load is not symmetrical.
Revintage, if I'm understanding your meaning correctly, it reminds me of a problem I encountered in my all-differential approach, wherein I wanted to use a 6SL7 LTP splitter, DC-coupled to a 6SN7 differential stage driving EL34s in triode PP - a bit like a Williamson with a different splitter, I suppose.
I was discouraged from using simple DC coupling because of a mismatch between the triodes in the 6SL7, causing a DC imbalance that was then amplified by the 6SN7, ending up with a serious DC offset at the 6SN7 plates. One day I hope to replace the 6SL7 with 6SU7 for better matching, but the simple DC coupling also made it difficult to attain good operating points for the two stages, because the grids of the 6SN7 were at the plate voltage of the 6SL7.
I found the answer for me was to use step-networks between the diff stages, so that only a fraction of the 6SL7 plate voltage was on the 6SN7 grid. This made the operating points much easier to set up, reduced the effects of the 6SL7 mismatch and still gave good LF stability for NFB.
hey-Hey!!!,
I'd avoid the williamson-style amplifier-concertina in favour of an LTP. A 6SN7 would do very nicely here, one grid grounded. It is also easy to convert to XLR/balanced input if you bring the P-I duties to the linestage at some point. I'd also bypass the 6CB6 screens to their cathode node.
OT, I'd go for a different pentode too, 6AC7, 12HL7, EF184 come to mind, and will deliver more gain with lower plate loads. It depends on exactly how much gain you wind up needing based on your P-I stage.
cheers,
Douglas
I'd avoid the williamson-style amplifier-concertina in favour of an LTP. A 6SN7 would do very nicely here, one grid grounded. It is also easy to convert to XLR/balanced input if you bring the P-I duties to the linestage at some point. I'd also bypass the 6CB6 screens to their cathode node.
OT, I'd go for a different pentode too, 6AC7, 12HL7, EF184 come to mind, and will deliver more gain with lower plate loads. It depends on exactly how much gain you wind up needing based on your P-I stage.
cheers,
Douglas
ray_moth said:
--don't agree. That's a bad excuse on performance grounds for abandoning the concertina. I'm against DC'ing tube amps stages as DC drifts can be large especially those using AC heater voltages. Caps in the system aren't necessarily a bad thing and I use caps for both interstage and output stage coupling. If I can get 36dB global nfb with interstage caps, then there is nothing wrong with the design.
In every case I've found the high output Z of a LTP strangles and slews HF performance. Driving 2x 6550 with 50K grid leak is an impossible task without massive thd. The LTP EF184 phasesplitter used in the Radford 100W amps is a design that despite many attempts I never get a repeatable performance from. One has to use massive amounts of global nfb to get the linearity & thd down to respectable limits.
One forgets the beauty of the concertina is virtually nil thd resulting in far lower global nfb i.e 14-20dB.
As I use multiple output power pairs, noway can an LTP deliver the goods, when compared to a balanced class A Williamson diff driver. This is the best consistent solution for driving output stages.
richj
Your experience is interesting. That is a lot of NFB. The Citation II achieved a similar level of NFB, but only a small proportionl of that was global. It, too, used AC coupling throughout (although the splitter was an LTP not a concertina).
The difficulty I find myself facing is that, because the outputs of a concertina can't be DC coupled and this type of splitter needs to be followed by a gain stage for most applications, the topology and coupling arrangements used in the Williamson are almost inescapable. I find this a difficult topology in which to apply a lot of GNFB and still retain stability.
One way of improving the stability could be to stagger the poles at the coupling caps. Another way could be to lessen the amount of GNFB required, by using an additional inner NFB loop from the OP tube plates to the driver cathodes (which then can't be differential any more). This inner loop can be either capacitively coupled, as in Normanhurst's example, or directly coupled, as in Fred Nachbouer's example (which requires some juggling with the drivers' bias).
I am still experimenting with modeling these things in LTspice. I have not abandoned the concertina, which has some very strong points in its favor, notably its simplicity and its excellent balance if executed in the right way.
ray_moth said:
Or use the design in the first post (6CB6a => EL34, with plate to plate local feedback) and put a concertina before that ? Perhaps you can do without global NFB in this way (?)
I think about 1 half ECC40 voltageamplifier followed by ECC40 concertina, followed by design from first post (4 holes in the chassis).
Yes, why not? Try it and see if it gives you what you want.
Problems with pentode operation are high output impedance and production of higher odd-order harmonics - i.e. poor damping and unpleasant distortion. Some speakers behave well without all that much electrical damping from the amp but, unfortunately, not many.
NFB is usually required to combat the shortcomongs of pentode amps. Global NFB is sometimes the only satisfactory solution, because in many cases it's the only way to get sufficient gain inside the loop for NFB to be effective. Local or partial feedback can do the job only if the gain within that loop is sufficient for your needs; otherwise, you may be forced to go for global NFB or give up on pentode mode altogether and go, instead, for ultralinear or triode mode.
Problems with pentode operation are high output impedance and production of higher odd-order harmonics - i.e. poor damping and unpleasant distortion. Some speakers behave well without all that much electrical damping from the amp but, unfortunately, not many.
NFB is usually required to combat the shortcomongs of pentode amps. Global NFB is sometimes the only satisfactory solution, because in many cases it's the only way to get sufficient gain inside the loop for NFB to be effective. Local or partial feedback can do the job only if the gain within that loop is sufficient for your needs; otherwise, you may be forced to go for global NFB or give up on pentode mode altogether and go, instead, for ultralinear or triode mode.
Nobody has mentioned implementing the Concertina with Sand.
For 1:1 here with no voltage gains, what would be the harm?
Many would implement a LTP splitter with a sandy CCS, and
not think anything wrong of it.... But Concertina always gets
overlooked for the same evil sandification ritual, I dunno why.
Concertina is a lot more straightforward, when drive swings
required aren't too rediculous. May not be ideal for screen
driving type voltage swings, but fine for driving a negatively
biased grid or two. Needing less than 1/3 B+ swing, and
perhaps a bit closer to half could be swung here by sand.
And I have to say, the impedances top and bottom are VERY
different, but it only takes one extra follower to make them
equal. And still one less active device than the LTP with CCS.
Yet even though the drive impedances are different. As long
as the loads are the same, and symetrically opposite in slew
loading, the voltage drops top and bottom will be equal and
opposite. Doesn't matter the drive impedances are unequal.
If the load can't tell they are unequal, the difference is not a
"problem" and doesn't need fixing.
True the top end does not respond to "Feedback" from its
own load (except a very small amount in case of a Triode).
But the feedback enforced from below should be equal to
what the top "would have seen", if it had been sensitive
in that same negativey feedbacky sort of way... Do you
see now why both loads have to be equal? So that "fake
NFB" at the top also responds equally.
For 1:1 here with no voltage gains, what would be the harm?
Many would implement a LTP splitter with a sandy CCS, and
not think anything wrong of it.... But Concertina always gets
overlooked for the same evil sandification ritual, I dunno why.
Concertina is a lot more straightforward, when drive swings
required aren't too rediculous. May not be ideal for screen
driving type voltage swings, but fine for driving a negatively
biased grid or two. Needing less than 1/3 B+ swing, and
perhaps a bit closer to half could be swung here by sand.
And I have to say, the impedances top and bottom are VERY
different, but it only takes one extra follower to make them
equal. And still one less active device than the LTP with CCS.
Yet even though the drive impedances are different. As long
as the loads are the same, and symetrically opposite in slew
loading, the voltage drops top and bottom will be equal and
opposite. Doesn't matter the drive impedances are unequal.
If the load can't tell they are unequal, the difference is not a
"problem" and doesn't need fixing.
True the top end does not respond to "Feedback" from its
own load (except a very small amount in case of a Triode).
But the feedback enforced from below should be equal to
what the top "would have seen", if it had been sensitive
in that same negativey feedbacky sort of way... Do you
see now why both loads have to be equal? So that "fake
NFB" at the top also responds equally.
Shoog said:
Agreed 100% !
Lemme push this idea a bit further:
Consider a pentode LTP PI with a common CCS in the cathodes and another one in each plate
And ... no, you don't have two CCS in serie just because the feedback resistors divert current from the LTP plates to power stage ones !
But you can set the feedback resistors at any value you want without being bothered by the available current for the driver wich is always provided by the current sources in the anodes.
Just a dream ... for now ...
Yves.
P.S.
In his paper about partial feedback, J.Broskie opened the way suggesting a choke to feed the driver's anodes, not so different
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.