Audio Asylum Thread Printer
Get a view of an entire thread on one page
|For Sale Ads|
OK everybody which phase inverter design is the best, as in best balance and equal output.
1. Split-load, Cathodyne, Concertina Williamson type.
2. Long tailed, differential pair Mullard type.
3. Paraphase and floating paraphase.
4. Cross coupled Van Scoyoc type.
5. Interstage transformer type.
My problem with Bonavolta's write-up is that he doesn't really explain what's going on in every case. Also, his "Schmidt" is really the same as his "Long-Tail Pair", apart from the bias arrangement. He refers to the Schmidt as "differential", which it isn't - there's only one input. One grid wags up and down with the signal, the other is riveted to ground. The coupled cathodes wag up and down with half the amplitude of the input signal, which is why you lose half the gain (because you're splitting the input). Use of a tail resistor to ground results in some of the signal leaking away at the cathodes, which is why a CCS is preferable as the tail, with its near-infinite impedance at signal frequencies.
I suppose this falls under IT. But what about a center tapped grid choke. I'm using one and have not mearured it's performance, except with my ears. My circuit is much like the Seth.
Have Fun and Enjoy the Music
"Still working the problem"
I can't say it's best, but I have good luck with the SS CCS as a cathode load for the long tailed pair. I've use it with 6BX7, 6BL7, 12BH7 and a couple of others. The CCS version I've used is the cascode mosfet type. I think it sounds good, the low level audio detail is tremendous, and has been reliable. I've also used the CCS as a cathode load for PP output stages so I have a full differential amp. The LTP produces good AC balance and if you want to use a pot in the cathode, which gives a bit of current feedback as well, you can get DC balance. As long as I get fairly close DC balance, I don't use the pot.
on what you need from the spiltter. The spilt-load type is a great performer but offers no gain and needs equal impedance on both outputs at all times so needs to be mated to a class A1 output stage that doesn't need a lot of drive such as pentodes/tetrodes.
The long tailed pair is better if you need to swing a lot of voltage but can be a bit harder to make outputs equal at all times. I tend to favor a cascode long tail pair commonly called a hedge type with a cathode CCS and a pot across the cathodes to adjust for equal outputs. This can be a "one stage" circuit that provides splitting and voltage gain at the same time. A negative voltage rail is needed to get the best. With enough voltage to work with it can drive most tubes.
The paraphase is rather out of style these days. The few older amps I listened to with them indicate they can do a better than average job in the 3D soundstage area. But frankly I personally wouldn't use them.
A transformer's main drawback is associated with cost. A good one isn't cheap and a poor one isn't near as good as a tube. IMHO, they are best used when they are needed. Sometimes for voltage reasons, sometimes for impedance reasons, sometimes to allow fixed bias without coupling caps, etc.
One option worth mentioning is to put the transformer in front of all tube stages where it need not be subjected to DC voltages/currents and use a hedge circuit after it. With both grids driven (normally one grid is grounded and the cathode driven) balance should be easier to achieve.
So, as usual in audio, there is no single best design. Instead it's about matching one part to the needs of the other parts and, of course, to your budget and abilities.
Good points from Russ!
I would add that a long-tailed pair doesn't have to have a negative source to work well. A large value resistor as the tail can be used with one end grounded, or I've even had excellent results with a simple cathode-coupled phase inverter (splitter is really not the proper term, as the signal is not ever split - inverter is technically more accurate).
I do recommend an AC balance control which is what Russ described (the pot across the cathodes). You can also put the pot across the plates, a la the Citation II. In either case, AC balance adjustment allows you to adjust for minimum distortion OR the best distortion pattern from the stage (my preference).
Russ's reply is great and I wouldn't add to it except that he omitted to mention the Van Scoyoc cross-coupled splitter. I've never found a clear description/critique of this splitter, not even in the venerable Radio Designer's Handbook. Mr. Stan White, who used it in his 1951 POWRTRON design, seems to think that it's the best but doesn't explain why.
I'd like to use the cross-coupled splitter sometime. I prefer an all-differential amp. I'd like to get the full gain from both halves of the splitter. It would also be nice to avoid having to lose voltage in the tail, as happens with a resistor or a pentode CCS (yes, I know I could use a SS CCS but I've never had any luck with them there "three-legged fuses" in tube circuitry).
The cross-coupled splitter offers good balance, good gain, good PSRR and the whole of B+ to play with. It's also very easy to apply negative feedback to its input. Its big draw-back is that it can't tolerate a high-level signal, because of the way the drive is applied from the cathode follower in the first half to the grid and cathode of the second part (the part that has the gain). However, provided that we use a sensibly high voltage and low current for the tubes concerned, our bias should be high enough to cope with a typical line-level signal of up to 2v RMS. The consequence of this limitation is that a cross-coupled splitter must be used early in the amp circuit, namely, as the input stage, which dictates an all-differential design.
I had a pair of ITT amps that used the 6BM8 in a Van Scoyoc. That was the most disapointing sounding amp I've ever heard. Even worse that the Sansui TU111 large tube integrated. YUCK!!
Most typical phase splitters\inverters do a great job at ballance and drive. Williamson, Dynaco and Mullard can all sound very good. I like transformers at the driver stage too. I don't think theres a silver bullet.
The cross coupled phase splitter can handle the same voltage swing as the equivalent plate loaded stage. Assuming reasonable balance, half the difference between the input voltages appears at the common cathode and the output voltage is the gain of the second triode times the difference between its grid voltage and this cathode voltage. If you do the arithmetic you will see that this makes it capable of full output swing.
They can also be balanced to any degree of precision required by ensuring the first triode of each pair has more gain than the second one and then reducing the gain of the first by reducing the cathode impedance. The only limit on the balance obtained this way is capacitive bypass at HF.
Non est ars quae ad effectum casu venit - Seneca.
"That which achieves its effect by chance is not art"
The balance is NOT good unless mu is very high. With a mu of 100, balance will be no better than 1%. Balance adjustment schemes to overcome this carry their own set of problems, normally unbalancing the distortion and bandwidth of the two sides to achieve satisfactory midband balance.
Must be because I don't have much experience with that one.
IIRC the original didn't use cathode followers in front and so one often saw it driven with a transformer for impedance reasons. But it's great to know it's been a good performer for you.
When you said you haven't had luck with SS CCs's is it because they sounded bad or they blow up or you can't get them set for the right current or what??? Sometimes hearing about what didn't work out teaches me more than hearing about what did work out:)
This post is made possible by the generous support of people like you and our sponsors: