[patrick-turner]

On Saturday, 10 November 2012 01:23:43 UTC+11, John L Stewart wrote:
John Byrns;963727 Wrote:

snip

Here are the curves with Local NFB as you have referenced. +-------------------------------------------------------------------+ |Filename: 6L6 Local NFB Curves RDH4 p396.jpg | |Download: http://www.audiobanter.com/attachment.php?attachmentid=313| +-------------------------------------------------------------------+ -- John L Stewart

The curves show that at Ia at 70mA, Ea = 350V, Ra' = 2k0 approximately.
Ra without FB is variable, but approx 33k, and gm = 5mA/V approx so µ = 165
Ra' = Ra / ( 1 + [ µ x ß ]) = 33k / ( 1 + [165 x 0.1]) = 1.88k.
Obviously, just how you read the curves gives a variable answer, And Ra gets lower for higher Ia. But setting up a tube and measuring it is the best way to know.

Using a shunt FB network with two resistors requires more voltage drive than the use of series voltage FB with a CFB winding based on keeping the same ß for both applications of FB. In other words, for the same amount of applied FB and Ra', the shunt FB is less sensitive. Plus the shunt FB has low Rin.

But there is one advantage with the shunt FB. The FB is derived with resistors which have no leakage inductance, so that the HF response is very extended and flat without ringing. Say you have an EL34 with FB network R1 = 22k, and R2 = 100k. A DC blocking cap of say 0.47uf could be used between R2 and anode.
In this case ß = 22 / ( 100 + 22 ) = 0.1875. (Its like having 18.75% CFB.) If the EL34 had Va = 200V+ and RL = 4k5, in pentode A = 30, so Vg = 6.7V- so you have 206.7V across 100k and 45.47V- across R1, so Vin ahead of R1 = 45.47 + 6.7 = 52.174V-. The closed loop gain = 200V / 52..174 = 3.8, or about the same as a 300B, and you should find Ra to be slightly lower than a 300B while both tubes would make 9W. Distortion in the EL34 without any FB in pentoad at 9W just under clip would be about 12%, and with FB it should be about 2%, but there are more harmonics present than in the case of 300B, or say a KT88 in triode which works similarly to 300B. Using a UL tap up to say 50% with EL34 in addition to the shunt FB OR the same amount of CFB for the same % of UL, you get slightly higher Ra' and marginally less A' needing maybe 65Vin, and THD remains about the same but H spectra is much more like the triode. The triode 2H without FB is more than with the 18.75% shunt FB around the pentode.
Company bean counters hate paying to make the driver stage produce more voltage at higher current while staying linear to cope with the shunt FB. They prefer to apply global FB which allows OP tubes to be driven easily plus the GNFB reduces all the defects of driver gain and OP tube and OPT - equally, so life is easier with GNFB.

Where you have local FB, AND GNFB, you usually cannot apply much GNFB because despite the local working well, phase shift beyond the ends of the band is rapid for F change, and you end up being constrained so that total of local and global = same = between 12dB and 20dB, so if local FB = say 12dB, then maybe 8dB of global is difficult and still get unconditional stability.
There is not one single free sandwich to be eaten along the journey when travelling with NFB. However, I have repeatedly made amps with CFB and with very wide BW OPTs that many makers never make except perhaps McIntosh and I found it possible to apply up to 35dB GNFB with 12.5% CFB also in OP stage, before the amp became impossible to stabilise.
Patrick Turner.