using obnly one half of a 12AX7?

[Wombaticus]

If you want to use that other 12AX7 half then look at Fender schematics from the Rivera era.
He routinely used parallel 12AX7 triodes as the input stage. Benefit? twice the gain. 0.7 (root 2) of the noise. Ideal for that 1st stage of any git amp.
schematics for Rivera era Fender amps
Cheers,
Ian

Unfortunately, the Johnson noise associated with grid stoppers of that size will dominate the overall stage noise even when tubes are not paralleled. And because of the pecularities of the RMS addition required for uncorrelated noise sources, it works out that only the dominant noise source really matters, as the contribution to the total of lesser sources gets swamped. So paralleling does not, in fact, help with noise unless notably small grid stoppers are used. Note also that this problem is exacerbated by any increase in stage gain, as Johnson noise at the grid will be multiplied by the gain of the stage. Finally, the way these stages are typically set up, gain is typically not doubled. Remember the Gm is doubled / ra is halved. But even if you keep a constant plate resistor, the "gain multiplier" when paralleling will be given by:


[(mu * Ra)/(Ra + 0.5*ra)] / [(mu * Ra)/(Ra + ra)]

simplifies to:

(Ra + ra) / (Ra + 0.5*ra)

Where ra = ra when unparalled.

You will find that plugging in some typical numbers shows that paralleling can result in a nice bump in voltage gain, but nowhere near the two fold you propose....which would require a non- existent plate resistor.

[markr14850]

Thinking out loud...

If we parallel the stage, input capacitance will double, yes? With this, we could cut the grid stopper in half for the same overall filtering effect, and reduce the resistor noise.

[Wombaticus]

Thinking out loud...

If we parallel the stage, input capacitance will double, yes? With this, we could cut the grid stopper in half for the same overall filtering effect, and reduce the resistor noise.

Absolutely. Depending on tube and gain, you are still going to need something on the order of 10k or less to start to get in the range where the noise levels become comparable. Which might not be enough HF rolloff to effectively block all RF. The best way to implement a low-noise input in general is with a smallish grid stop followed by an additional bit of added input capacitance, perhaps on the order of a couple hundred pF.

[Gingertube]

Wombaticus,
That "only the dominant noise source" is signifcant is a good point.
Have you done calcs to conform that the thermal noise of the grid stop dominates the tube noise or ios that base on practical experience? I had thouight about this, particularly in connection with the Rg1 resistor which of-course is much higher value and will have much higher noise until I "clicked" that Rg1 is shunted by the source impedance of the guitar pickup.
I want to understand this as I'm embarking on a high gain preamp design for a local fellow and my design brief is that it must be low noise, everything we can do in that 1st stage to drop noise gives big benefits.
Cheers,
ian

[Wombaticus]

Wombaticus,
That "only the dominant noise source" is signifcant is a good point.
Have you done calcs to conform that the thermal noise of the grid stop dominates the tube noise or ios that base on practical experience? I had thouight about this, particularly in connection with the Rg1 resistor which of-course is much higher value and will have much higher noise until I "clicked" that Rg1 is shunted by the source impedance of the guitar pickup.
I want to understand this as I'm embarking on a high gain preamp design for a local fellow and my design brief is that it must be low noise, everything we can do in that 1st stage to drop noise gives big benefits.
Cheers,
ian

The result is not mine - my knowledge of this is based on some as-yet unpublished material from a tube authority familiar to this group. But so far as I'm aware, his calculations and the necessary empirically derived constant for flicker noise are based on formulae and data from the wisdom literature. Anyway, without speaking too much out-of-turn, the basic result follows from making a pessimistic assumption about the empirical constant required for calculating flicker noise. Flicker noise is the dominant noise source intrinsic to the triode, and unfortunately, the one that most variably depends on the method of tube construction and metallurgy -- making it the least amenable to direct calculation.
To do the full calculation, one also applies the standard formulae for thermal and shot noise. The upshot is that for a very standard 12AX7 input stage (100K plate, 68K grid, 300V B+), the EIN will be something like 3.5uV with the grid stop, and on the order of 1.2uV without. The "without" number, incidentally, is not too far off from what one might expect from a FET configured for audio. And mind, this is assuming triode with mediocre to poorish noise characteristics, not some specially-selected low noise device. Similar calculations show that you need to get the grid stop down to about 10K to make the contribution of Rg on the same order as that of the tube (meaning that going smaller would be pointless). In that case though, the HF rolloff required to block RF begins to become inadequate, so you will probably want to add some grid-ground capacitance to achieve the -3dB point you desire (calculated in the usual way)... it's a good general principle to remember that capacitance is essentially free from the noise considerations that plague resistors, so whenever a filter can be reconfigured to favour capacitance over resistance, there's at least the possibility of a noise benefit.

[trobbins]

The benefit of ferrite inductors comes in to play as well for their low audio range impedance - to get a significant impedance (eg. 50k) in the low hundred kHz range would require a multiturn core of 'low frequency' ferrite like 3E5, so it won't win on cost - but given the application has no DC current then that certainly helps keep the impedance high.