op amp gain; vary Rg or Rf, and why?

[emrr]

I was working on some early '70's RCA preamps with transformer input (15K sec), which use a discrete op amp in non-inverting mode. I note RCA chose to have Rg be a set value, and varied Rf via a stepped control, along with tailored Cf's. In almost every other similar preamp I've encountered, Rf is fixed, and Rg is varied. I'm attempting to consider the reasons for choosing one over the other. I'm assuming it's partially to do with a strict desire for the Cf break freq to sit at the same place.

I find this bit in a TI pdf:

The Noninverting Op Amp

The actual resistor values are determined by the impedance levels that the designer wants to establish. If RF = 10K and RG = 10K the gain is two as shown in equation 2, and if RF = 100K and RG = 100K the gain is still two. The impedance levels of 10 K or 100 K determine the current drain, the effect stray capacitance will have, and a few other points. The impedance level does not set the gain; the ratio of RF/RG does.

That touches on my only observation, which was that feedback current appears to change more drastically with Rf being varied.

To add to the fun, at some point it appears RCA changed the value of Rg, going from a minimum gain of 6.9 dB (2k/2k43) to a minimum of 12.5 dB (750/2k43). Max Rf is 42K2, and I am about to have a listen to that with Rg 430, which would appear to move gain range up to 16.4 - 40 dB.

I imagine the crucial part of messing with Rg in this case would be the effect on the Cf's. Perhaps I'm considering messing with input Z more than I realize.

Gain (officially) steps in 10 dB increments here, and I note those steps to be most accurate with Rg 2K, and increasingly higher with reduced Rg value.

I thought about changing the gain control to an Rg rheostat, but in the end I wish to avoid trashing the original cosmetics (personal problem), and simply change some bits on the 5 position 4 pole switch, which has an unused section. Harder, for sure.

At this point, I've arrived at a theoretical plan to rework the existing stepped switch for a wider overall gain range, changing two bridging (line pad) input positions to additional mic gain points. The HI+8 point maintaining the same Rf as MIC-40 and switching in additional Rg resistance for the one position, essentially becoming MIC-30. HI-20 going back to MIC-40 gain setting, and inserting a 20 dB U pad in the mic path. Should end up with roughly 17 to 58 dB in 10 dB steps, if I'm thinking about it correctly, and there are no booby traps. The most obvious danger I see lies is putting the Rg path through a stepped switch, and having an accidental open condition. This would have both Rf and Rg in MBB switching paths.

To be tried on the bench ASAP, any thoughts or derision appreciated.

[emrr]

Yep, these use Grayhill MBB switches, and I have 5 channels that have survived to this late date; 35+ years.

[JR.]

I was working on some early '70's RCA preamps with transformer input (15K sec), which use a discrete op amp in non-inverting mode. I note RCA chose to have Rg be a set value, and varied Rf via a stepped control, along with tailored Cf's. In almost every other similar preamp I've encountered, Rf is fixed, and Rg is varied. I'm attempting to consider the reasons for choosing one over the other. I'm assuming it's partially to do with a strict desire for the Cf break freq to sit at the same place.

When attempting to get best s/n at high gain you want low total impedance in feedback network (Rf and Rg in parallel). The pot varying the leg to ground give you lowest impedance at highest gain.

That touches on my only observation, which was that feedback current appears to change more drastically with Rf being varied.

In fact the current through the feedback resistor (R from output to - input), is established by resistor from - input to ground (and input signal). Varying the feedback resistor value translates this same current to more or less output voltage as you vary the resistance.

To add to the fun, at some point it appears RCA changed the value of Rg, going from a minimum gain of 6.9 dB (2k/2k43) to a minimum of 12.5 dB (750/2k43). Max Rf is 42K2, and I am about to have a listen to that with Rg 430, which would appear to move gain range up to 16.4 - 40 dB.

I imagine the crucial part of messing with Rg in this case would be the effect on the Cf's. Perhaps I'm considering messing with input Z more than I realize.

If I understand your general topology (non-inverting opamp). The input impedance will be mostly unaffected by feedback network resistors, and dominated by resistive termination.

It is common to put cap in series with gain resistor to ground to keep DC gain low so switch doesn't cause DC steps, or pot noise.

Gain (officially) steps in 10 dB increments here, and I note those steps to be most accurate with Rg 2K, and increasingly higher with reduced Rg value.

I thought about changing the gain control to an Rg rheostat, but in the end I wish to avoid trashing the original cosmetics (personal problem), and simply change some bits on the 5 position 4 pole switch, which has an unused section. Harder, for sure.

At this point, I've arrived at a theoretical plan to rework the existing stepped switch for a wider overall gain range, changing two bridging (line pad) input positions to additional mic gain points. The HI+8 point maintaining the same Rf as MIC-40 and switching in additional Rg resistance for the one position, essentially becoming MIC-30. HI-20 going back to MIC-40 gain setting, and inserting a 20 dB U pad in the mic path. Should end up with roughly 17 to 58 dB in 10 dB steps, if I'm thinking about it correctly, and there are no booby traps. The most obvious danger I see lies is putting the Rg path through a stepped switch, and having an accidental open condition. This would have both Rf and Rg in MBB switching paths.

To be tried on the bench ASAP, any thoughts or derision appreciated.

When it comes to blank page resistor value setting, the comment you quotes applies,, the general impedance of the feedback network, must be high enough that the opamp can drive it, and low enough that it doesn't add noise or interact with stray capacitance.

I am kind of guessing about your circuits... it would help to see a rough schematic of what you propose.

JR

[emrr]

When attempting to get best s/n at high gain you want low total impedance in feedback network (Rf and Rg in parallel). The pot varying the leg to ground give you lowest impedance at highest gain.

That's finally in layman's terms.

Perhaps it makes more sense to fix a constant Rf value, and use S3 for Rg steps.

Here's the original on the left, and the proposed on the right. Total gain figures include the transformer, which is 18.2 dB. R903 is Rg, and C902 is a 100 mfd tantalum. All of the "E" connections are the same for both, excepting those to switch section 3, which is unused in the existing design. Cap values are what's written on them; I haven't bothered to interpret yet.

It currently has a 44k2 load resistor on the 15K secondary, so we're looking at a fixed 590 ohms on the 200 ohm input taps. If I dispense with that load resistor, and my new input pad on the primary is 1440 ohms, then open input condition reflected to secondary should be 108K ohms, and otherwise roughly 13K2 with a 200 ohm mic. My proposed input pad would still be more ideal than the original setup.

link in case image cut

[emrr]

Safari is cutting off the right side of the image here, but Firefox is showing everything.

[MikoKensington]

If you control click, and open in a new window, all appears.

[mediatechnology]

When attempting to get best s/n at high gain you want low total impedance in feedback network (Rf and Rg in parallel). The pot varying the leg to ground give you lowest impedance at highest gain.

The above hits the nail on the head.

Safari is cutting off the right side of the image here, but Firefox is showing everything.

IE7 is cropping it too without a scroll bar. I try to limit the images I post to 700 pixels. I think phpBB3 doesn't properly auto-scale image attachments.

[emrr]

When attempting to get best s/n at high gain you want low total impedance in feedback network (Rf and Rg in parallel). The pot varying the leg to ground give you lowest impedance at highest gain.

The above hits the nail on the head.

Amazing how I can dig through a stack of books, and browse the web for awhile, and not come up with that info in any clear manner.

So, is there any reason why RCA would have chosen the other path?

[JR.]

Lower impedance Rs mean larger Cs for same pole frequency. Also the gain stage has to drive the total Rf+Rg to ground in parallel with input impedance of following circuits, If tube is not very low output Z and very low input noise, there is diminishing returns to dropping resistance further.

JR

[emrr]

the gain stage has to drive the total Rf+Rg to ground in parallel with input impedance of following circuits

Obvious enough, I hadn't thought about that part. These are mostly 75 ohm output rated op amps, and the following circuits are 600 ohm at minimum, mostly 10K. So not a problem.

Back to the drawing board; thanks. I will have to compare, and see what the ears say too.