>
basically what he did was take the 70 volt PA system output (600 ohm) of a standard push pull output xfmr of the day and use that..
No, and you know why. The "impedance of 70V" depends on the power. For a 70 Watt amp, the 70V tap needs a 70Ω load. For this ~10 Watt amp, if it were a 70V tap, it happens to be 500Ω.
But 600Ω windings were very common as Program Amplifiers. Given +8dBm telco line level, 16dB headroom to rated power, rated power about 10dB below clipping, 6dB pads on all lines, you need about a 8-10 watt amplifier. Rated low-THD at 1W/+30dBm, 200mW/+24dBm (+8dBm plus 16dB headroom) on the line, 6dB pad, 10dB below clipping, that's a 10 Watt amp. Couple 6V6.
>
70 volts sounds good for the grids. lol
Around that much is needed. Look at the
6386 data sheet, page 3, you need 40V or more to sink the Gm very low. It isn't that simple: the 6386s are pre-biased to keep them below the kink at ~500uMho, the rectifier is pre-biased to stay off for small signals. And if the rectifier amp saturates, the peak escapes to ruin the disk cut; better to wham those grids so hard the 6386 cuts-off and flats the peak. I don't recall what I figured but it was OTOO 50V-70V needed from the rectifier.
We get the required drive power/impedance from the attack/release resistors. I do not have the Fairchild plans on me, but the calc is something like:
We have eight leaky grids, so the DC path from grids to grid bias voltage must be about 1Meg/8= 125K; in fact it uses 220K. This is also our maximum release resistor. The largest ratio of attack to (short-peak) release is 12,500. So the output impedance of amp and rectifier "should be" 17 ohms. If we also needed 50V peak, we'd need 50^2/17= 147 Watts peak (75W RMS). In fact we never want max GR in a single cycle. We don't feed peaks 20dB over threshold and expect brick-wall flattening. The specs mention 10dB reduction, a reasonable maximum for reasonable mastering.
Actually, for more than small GR, the Narma limiter is current-limited on attack. Figure the maximum current a 6V6 will deliver at zero grid and ~200V G2. Multiply by transformer ratio. Your rectifier driver must approximate that max current level.
>
I should have the same amout of FB.
The
amount of feedback is not critical, isn't even any kind of constant:
If indeed the 12AX7 is biased to class B, then below threshold there is no gain, thus no feedback.
When the 12AX7 starts working, the amp forward gain depends on the load. The load is a rectifier-capacitor thingie. If the instant audio voltage is less than any stored charge on the cap, the load is infinite. The 6V6 has large output impedance, and for large load impedance its gain is large. When the instant audio voltage is greater than any stored charge on the cap, the load is low impedance, actually a large capacitor. 6V6 gain is low, and darned hard to calculate.
The net effect seems to be that the output impedance at the core is much much lower than 600Ω, as you would expect from 12BH7-6V6 cascode with about 1/4 feedback from a 150Ω or 600Ω winding. 15*10*0.3*0.25 is about 20dB of NFB, a reasonable value for something with iron in it. The amp output impedance at the iron is like 60Ω, the output winding adds ~30Ω, at high current the diodes may be 10Ω; we have about a 100Ω source when not current limited. 2uFd and 100Ω in position 1 gives 0.2 milliSeconds, just what the spec says. If you use almost any sort of tube amp with NFB and 600Ω transformer, you will be in that ballpark. If you use a transistor amp, you might need some dummy resistance. (But if you allow transistors, you may as well do all this rectifier stuff at 1/10th the voltage, save a LOT of power, and use a DC amplifier to boost the control voltage 10X to the 6386 grids. Narma did not have a DC amplifier any simpler than a 10-Watt AC amp.)
>
if I disconnect a feedback loop and the gain doubles, the amp had 6dB of feedback.
This is meaningful for linear amplifiers with linear loads. The rectifier is very nonlinear, and we suspect the amp input is sometimes nonlinear. And there is the added side-effect of control amp banging on the vari-gain stage and its own input.
I don't suggest you break this loop. Aside from possibly meaningless results, the control amp "should never" be in overload for more than milliseconds, since any large control output will whack-down the vari-gain stage. The control amp will swing VERY large voltages, and has no load (either you broke the rectifier or it quickly charges the time-caps and then runs no-load). It is possible to punch-through an output transformer that way. It is not inevitable, and I've run Fishers unloaded; but if a Fairchild punched-through its vintage transformer, tears would flow.
The rectifier amp should have output impedance like 100 ohms. Its peak current should be, I think, 400mA. The 6973 with 200V G2 saturates about 100mA, and I think you have 1:4 current ratio between one side of the plate winding and the 600Ω output winding. Peak output voltage should be about 90V-100V.
The output impedance is
not "600Ω"; the source impedance is less, the load impedance is usually more except when it is much less. "600Ω" is about the mid-point of the effective impedance when it is working hard.
