Phase and NFB confusion

[Alan0354]

The LTPI is not a perfect differential amplifier, but it's about 90% of the way there.



In the LTPI there is some feedback connected to the bottom of the Tail resistor. This keeps the voltage across the Tail resistor almost constant. A constant voltage across the Tail resistor means the current through it is constant. This helps the PI balance. If the current through the Tail can be made to remain constant, the balance of the PI will be improved. The two tubes act like a current steering mechanism. An increase of current in one side causes an equal decrease in current of the other side. If the plate resistors are the same resistance, the voltage increase on one side will equal the voltage decrease on the other side.

That was quite ingenious. I don't think the constant current source is well know at the time.

I wonder why the SF Fender abandon the 100K/82K in the early 70's and use balance 47K instead. Does that make a difference in the sound? Again, I don't have the best ears, I experimented both balance and imbalance by a switch that I can switch back and fore. I can't tell the difference.

[jazbo8]

That was quite ingenious. I don't think the constant current source is well know at the time.

Leo's not that smart... making the tail resistor larger to provide better balance is just a property of the differential pair amplifier.

I wonder why the SF Fender abandon the 100K/82K in the early 70's and use balance 47K instead. Does that make a difference in the sound? Again, I don't have the best ears, I experimented both balance and imbalance by a switch that I can switch back and fore. I can't tell the difference.

The change to 47k was to accomodate the need to drive a quad of 6L6s, instead of a pair of 6V6s. One of the reasons that you did not hear a difference, was perhaps that the output tubes were not well matched to begin with. so PI's balance/imbalance was moot. For guitar amps, it is not really that critical to get a perfect balance, after all, we usually want more distortion not less...

[Alan0354]

Leo's not that smart... making the tail resistor larger to provide better balance is just a property of the differential pair amplifier.


The change to 47k was to accomodate the need to drive a quad of 6L6s, instead of a pair of 6V6s. One of the reasons that you did not hear a difference, was perhaps that the output tubes were not well matched to begin with. so PI's balance/imbalance was moot. For guitar amps, it is not really that critical to get a perfect balance, after all, we usually want more distortion not less...

I am referring to the 47 ohm that is part of the long tail. The NFB drive almost the same voltage as the input voltage across the 47 ohm, this keep the voltage across the 22K resistor constant to create a constant current source.

[jazbo8]

I am referring to the 47 ohm that is part of the long tail. The NFB drive almost the same voltage as the input voltage across the 47 ohm, this keep the voltage across the 22K resistor constant to create a constant current source.

I see now, should have read your post in context...

[J M Fahey]

Sorry, left post incomplete, weekends kill me .

Don´t know why crazy musicians get so emotional and upset about playing on Friday and Saturday nights.

What´s wrong with playing, say, Thursday 10 AM or something?



OK, for clarity I´ve redrawn the LTP PI in the simplest possible way, with everything referred to ground (so I added a negative supply), rounded values and assuming an ideal tube.

It works exactly the same as, say, an AB763 PI but this way is easier to understand.

The difference is that Fender had to simplify manufacture, so he used a single supply and to boot reference and bias voltages are floating tens of volts above ground, for the same reason.



The ideal dual triode has the following characteristics (close to some 12A*7 ) :

a) each triode passes 1mA with +100V on plate and -2V on cathode
b) transconductance (current variation after some grid voltage variation): 1 mA/V

1) main voltage supply is +200V DC

2) I added a negative supply: -100V DC

3) grids are grounded so : 0V DC

4) cathodes are at +2V

5) total cathode current : 100V/50K=2 mA

Here we have 2 important points:

a) Since 100V is MUCH larger than 2V cathode (bias) voltage, current is practically constant.

Which also means that:
b) if current through one triode rises by a certain amount, current through the other has to lower the exact same amount so total stays the same.

c) so now V1 and V2 are coupled by the cathodes, like it or not.
and
d) since current up in one causes current down in the other, they are coupled out of phase

Now to put some numbers into it:

Total cathode current: 100V/50K=2mA

Current per tube : 1 mA

Current across each plate resistor: 1mA

Voltage drop across each plate resistor: 1mA*100K=100V

Plate voltage: 200V(supply) - 100V(drop)=100V

No signal injected: everything is balanced, both plates are at the same voltage.

Now we inject some signal through the +IN, so V1 grid goes 500mV up.

Since we have 1mA/V transconductance, current through V1 rises to 1.5mA so drop through R1 rises to 1.5mA*100K=150V so V1 plate voltage lowers to 200-150=50V
So we have a negative going (negative phase) 50V variation (signal) on V1 plate.
And V1 gain is 50V/0.5V=100X

What happens on V2?

Since total current is constant and V1 rises 0.5mA, current through V2 has to go down 0.5mA .
So voltage drop through R2 is now: 0.5mA * 100K=50V and voltage at V2 plate is 200V-50=150V .
So now we have 50V higher, and gain is 50/0.5=100X in phase with input signal.

OK, now we know how/why this circuit gives us 2 identical out of phase signals and has quite gain.

Now to NFB and what happens with V2 grid.

Well, it behaves exactly like V1 grid, so if we rise V2 by +500mV now V2 plate lowers and V1 rises.

So now we confirm that a positive voltage on V1 grid makes negative signal on V1 plate and positive on V2 plate while a positive voltage on V2 grid makes positive signal on V1 plate and negarive on V2 plate .

So positive signals on V1 grid and on V2 grid are functionally out of phase because plate signals end up being of phase.

This is what Randall (and many more) found difficult to grasp.
It really sounds wrong ... until you get it

Go back to the beginning and reread it and at a certain point it "clicks" in your mind.

[Audiotexan]

Enzo, Loudthud, Juan, et al:

For me, it's posts like these that help SO much in making things clearer. Thank you ALL for being so gracious and willing to share the knowledge and experience with us here.

It's worth more than gold and platinum combined.
Regards,
Rob

[Leo_Gnardo]

Don´t know why crazy musicians get so emotional and upset about playing on Friday and Saturday nights. What´s wrong with playing, say, Thursday 10 AM or something?

You know Juan, Fri - Sat nights, it's money making time. Plus drinking and picking up girls. Strong motivation. And they don't want to look like a fool with a lame amp, so they continue to get those club gigs, or even bigger ones.

- - - - - -

Very thorough explanation Professor Fahey, that should be in the textbook, or at least a sticky somewhere. Good stuff!

[Alan0354]

To supplement what Juan wrote above, there are also a lot of explanations of differential pairs.

http://www.psut.edu.jo/sites/alasali...Text/P-ECL.pdf

Transistor Differential Amplifier Circuit Description, using a single input. Dictionary of Electronic and Engineering Terms

The theory of differential pair(PI) for transistor is exactly the same as tubes.

[uneumann]

[SIZE=2
In the LTPI there is some feedback connected to the bottom of the Tail resistor. This keeps the voltage across the Tail resistor almost constant. A constant voltage across the Tail resistor means the current through it is constant. This helps the PI balance.
[/SIZE]

LT thank you for this comment - this is the first good explanation I've seen in print for why NFB connects to the tail. Another mystery of the universe explained - and makes perfect sense once I see it.

It makes me wonder about creating a separate NFB signal that exactly matches the input (gain of 1) so that the tail is essentially perfectly bootstrapped for balance. The NFB to the inverting grid can be independent and set to any desired ratio.

[g1]

Thank you Loudthud, JM Fahey, & Enzo for the explanations. Current theft is a concept I can cope with and I can now get the gist of the circuit. However, Enzo's explanation about the second grid being akin to grounded and the cathode doing the wiggling for the second side is what my mind wants to lean toward.
I guess I was thinking from the perspective of the cathode, and that if the 2 cathodes are connected, then for the plates to be out of phase, the grids must be out of phase. In fact I have an SF Twin at the moment and just threw it on the isolation transformer to check. Indeed, with my ground clip at the cathode, the 2 grids are out of phase. With ground clip to ground, the 2 signals are in phase as Randall saw. Perspective is everything they say .

[JoeM]

Some of you may have seen an article in Glass Audio about using a CCS instead of the tail resistor. (LM334 I think).

[Dave H]

...Indeed, with my ground clip at the cathode, the 2 grids are out of phase. With ground clip to ground, the 2 signals are in phase as Randall saw. Perspective is everything they say .

If you look carefully at the in phase grid signals you’ll see that they are not the same amplitude so there is a smaller out of phase differential signal between the grids. The larger in phase component is common mode and is rejected. Measuring with respect to the cathodes eliminates the in phase component. It’s the differential signal that generates the out of phase outputs at the plates.

[loudthud]

It makes me wonder about creating a separate NFB signal that exactly matches the input (gain of 1) so that the tail is essentially perfectly bootstrapped for balance. The NFB to the inverting grid can be independent and set to any desired ratio.

I've never tried that, but I have tried the active current source to a negative rail. The one advantage worth having is the Diff Pair inputs at ground much like Juan's example above. The side effect of that is what happends when the B+ is in standby mode. Current wants to flow out the grids. So then you have to make the -100V rail also shut off in standby mode. Things get more complicated for no real benefit. The standard circuit is the standard because it works and it's simple. If there was a PI that sounded better (Hiwatt?), it would be the standard.

[Dave H]

I've never tried that, but I have tried the active current source to a negative rail. The one advantage worth having is the Diff Pair inputs at ground much like Juan's example above. The side effect of that is what happends when the B+ is in standby mode. Current wants to flow out the grids. So then you have to make the -100V rail also shut off in standby mode. Things get more complicated for no real benefit. The standard circuit is the standard because it works and it's simple. If there was a PI that sounded better (Hiwatt?), it would be the standard.

The active current source could be a transistor to ground then the -100V rail isn't required.

[Alan0354]

Some of you may have seen an article in Glass Audio about using a CCS instead of the tail resistor. (LM334 I think).

It's all marketing. It's so easy to do CCS. A NPN with resistor divider for the base to set the current is the best and easiest way. No power supply needed to power opamp. BUT then you cannot claim it's all tubes.

Tubes have a lot of short comings that you can overcome by mixing in transistors etc. But then you cheapen the amp!!! Musicman drive the power stage with cascode configuration that pretty much constant current source that you don't need to worry about matching tubes.

But there is no rhyme or reason, people just want all tubes amps. People here are a lot more technical oriented, they have more of an open mind. But if you go to a players' type of forum, they talk about all tubes all the time.