Adding NFB to LTPI

[EFK]

If I want to use a typical AC50 LTPI (1M grids, 1.2K cathode resistor and 47K tail) but I'd like to possibly inject some NFB (it will be on a switch), is there a preferred ratio or formula to use to split the 47K tail (approximately)? 23 and 23/4? 39 and 8? 27 and 20? ?????? There must be some reasoning behind this. Appreciate any ideas! Thanks.

[Chuck H]

The tail resistor will effect the PI output and balance. This is limited by the tubes max cathode resistance (which is halved in this circuit, I think) You can use a 47k tail and say, a 1k FB shunt and adjust the FB ratio to the shunt resistor. If you need a 20:1 FB ratio you would use a 22k dropper with the 1k shunt. This way your total tail resistance remains basically the same. There is no reason you have to use the seemingly requisite 4.7k shunt resistor for the FB loop, the loop just needs to represent the right ratio. If you plan to use a presence control in the loop the cap value WILL be different for a 1k/22k FB loop than it would for the typical 4.7k/100k feedback loop. The cap value will need to be changed because the impedance of the circuit changes the knee frequency of the cap. With a 1k/22k FB loop you will need a .47uf (not .047) in the presence circuit to get the same effect as the standard .1 cap used in your average 4.7k/100k FB loop.

You can also split the ratio of the tail as you mentioned. If you want the 4.7k resistor in there to keep other design considerations easier then just subtract 4.7k from the 47k tail. So a 43k tail with a 4.7k shunt would allow you to use the other usual cap and resistor values.

But why use this particular PI circuit??? Is it somehow well suited to your project? Output swing voltage?

Chuck

[EFK]

Thanks for that explanation Chuck! Perfectly clear explanation. I'm not committed to that PI, but I've noticed a few newer cathode bias 6L6 amps seem to use it and I've been thinking about trying it (as what I've got going at the moment is a 'spare parts' cathode bias 6L6 thing). I'd like to try something different than typical Marshall-type PI for a change.

Is there any reason NOT to use it?

[Chuck H]

The only reason I can think of NOT to use it would be that it has lower output than other 6L6/EL34 amps use. I see the values you propose used mostly in EL84 amps which have the lowest drive requirements of the typical guitar amp power tubes. Typically on the order of 25% of what a 6L6 would need. But if this is intended to be a clean amp and biased in class A or some such, maybe.

Chuck

[EFK]

Is there something else you could suggest? Let me outline my goals: I already have Marshalls coming out of my ears - early 70s aluminum panel 50W lead, JTM50, two JTM45s (one w/ 7591As which are GREAT tubes), etc. - plus I have an old late 60s Selmer TNB50 and a new copy I made, but I wanted to branch out a little and do up a cathode bias 6L6 amp to use for 'edge of breakup' type tones (the Marshalls really cover everything beyond that). Something loud but not as dirty, more American maybe. I think I've got the Brit thing covered. I've been thinking about maybe trying a 12at7 in the Pi also, and looking at some of those silverface schematics. When it comes to LTPI plans, there really aren't too many variations. I'm using a bunch of 'spare' parts, the PT is something like a JTM45 type but only 300-0-300. The OT is pretty darn big and more like a Fender type (I think) but does have 4-8-16 taps. I was planning to run the preamp voltages pretty high to keep it clean, don't want so much MArshall preamp distortion this time.

[Chuck H]

Hmmm... It's hard to suggest something so subjective. What kind of tone do you want to add to your gear??? Big box jazz, surf, maybe a tweed circuit for old school country twang or maybe that big mid scooped Fender clean or a BF hot rod like a Dumbl-ish thing.

Chuck

[EFK]

I guess what I should really be asking as pertaining to LTPI design is how to the numbers obtained here: Long-Tailed-Pair Phase Inverter Gain Calculator (and I'm speaking of the calculated inverted/non inverted voltage gain) relate to the characteristics of the PI? It's easy to plug in values and try to obtain some balance, but how do these voltage gain numbers relate to PI clipping and how do they relate to the ability to drive - either sooner or later - the power tubes?

[Steve Conner]

They don't relate in any way. The maximum unclipped output level is a completely separate issue to the gain.

To be sure, changing any single component will affect both of them at once, but in theory you can change one without affecting the other, by changing several components at once in the proper ratio.

The calculator uses a small-signal approximation, so it can't tell you the clipping level.

The tail resistor in a LTPI is a compromise between balance and output level. (You could get more output for the same balance, or better balance for the same output, if you returned the tail to the bias supply instead of ground, but nobody ever does this.)

[R.G.]

The devil who sits on my shoulder makes me say:

The gain, balance, etc, etc. of a diffamp (that's what a long tailed pair is most often called these days) depends on the impedance in its tail being high, as Steve notes.

There's another way to get high impedance without going to big voltages and big resistances - use an active device which is acting like a current source. An active current source can use a smaller voltage on the cathodes, leaving more voltage drive for the plates, increases gain as well as other stuff. It doesn't obviate using a split cathode resistor for feedback, either.

[Wombaticus]

The devil who sits on my shoulder makes me say:

The gain, balance, etc, etc. of a diffamp (that's what a long tailed pair is most often called these days) depends on the impedance in its tail being high, as Steve notes.

There's another way to get high impedance without going to big voltages and big resistances - use an active device which is acting like a current source. An active current source can use a smaller voltage on the cathodes, leaving more voltage drive for the plates, increases gain as well as other stuff. It doesn't obviate using a split cathode resistor for feedback, either.

To follow up on what RG was saying, if you really need excellent balance, you can implement a Constant Current Sink (CCS) using a transistor (or a valve for the purists):

The Valve Wizard

Scroll down to the last example.

[EFK]

Well this promptly went over my head! Way, way over... I'm a great copyist and sometimes creative, but do not understand much of the 'whys.' Let's put it this way - how noticeable would the difference be expected to be between the Vox-type 1.2K/47K and the Marshall type 470/10+5? I notice that many of the guys doing the cathode bias thing - not just EL84s but big bottles too - seem to stick with the Vox PI. However, many are also using multiple gains stages prior to the PI. I'm just using 1/2 12ax7 ---> tone and volume ---> 1/2 12ax7 ---> PI. Looking for something like tweed about noon or so dirt, just loud.

[Chuck H]

how noticeable would the difference be expected to be between the Vox-type 1.2K/47K and the Marshall type 470/10+5? I notice that many of the guys doing the cathode bias thing - not just EL84s but big bottles too - seem to stick with the Vox PI.

The 1.2k/470 resistor in your two examples are the bias resistor. It will determine when the PI cutts off and/or saturates. You can usually find a good value by measuring voltage at the cathode. Tube charts will tell you a good bias voltage to aim for or if you have a scope you can just look at the screen and adjust this value. Higher values bias the tube closer to cut off. This should be a pretty straight foreward choice because the circumstances choose the value for you. Some designers do seem to use this resistor to reduce gain also by intentionally biasing the tube colder for less output. You can experiment with this value to see how it colors the tone as long as you don't exeed a safe operating point for the tube.

The 47k/10k resistors in your examples are the tail resistor. This will determine the output and balance. As noted, a higher value will give less output and better balance while a lower value will have more output and less balance. There are other ways to balance the PI than a large tail resistor and achieve higher output with a lower value tail resistor. But tubes are so variable that the large tail resistor is probably the most consistent way to achieve good balance. You've probably noticed the 82k/100k plate resistors typically used by Fender and Marshall? This is one way to help balance the PI without going to a very large tail resistor. That's why the Vox amps with the 47k tail use 100k for both PI plates. Balance is achieved via the large tail resistor and there is no need to adjust gain with the plate resistors.

The amount of drive voltage at the PI grid and how much drive voltage you need at the power tube grid, combined with the limitations and bias requirements of the tube should determine the PI design. In other words, you wouldn't just arbitarily stick a Vox modelX PI circuit into a Fender amp and expect it to sound more like a Vox modelX. Rather, you would want to determine how the Vox modelX PI meets the drive voltage needs of it's host amp, and the bias characteristics, and design a circuit that behaves in the same relative way in the Fender amp. The plate voltage of the PI and power tubes, bias point of the power tubes and the drive voltage into the PI will all effect the recreation of the same characteristics in the Fender amp as compared to the Vox, so in truth the Vox modelX PI may be no better, or in fact worse for creating a Vox modelX sound in the Fender amp than the original Fender PI.

One reason one might use the 1.2k/47k PI values (lower output and better balance) with big bottles is that the amp it's being used in is class A (bias of the output tubes). Class A amps require less power tube grid drive voltage than do AB amps. They also create fewer watts and sound different primarily because of their operating class, not the PI design.

Does that explain anything?

Chuck

[R.G.]

To follow up on what RG was saying, if you really need excellent balance, you can implement a Constant Current Sink (CCS) using a transistor (or a valve for the purists):

I would have used a power MOSFET for the additional ruggedness, and a slightly different form of the CCS, with a bipolar transistor controlling the MOSFET current by stealing gate drive, but that's just me always wanting cast-iron, armor plated ruggedness and high performance. The single transistor circuit works well enough if you are careful to examine what happens to it in the 'edge' conditions. And either one is far, far superior to a resistor in being a constant current source.