G7 - high grid R vs lower R's in classic designs

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

volki

Well-known member
Joined
Jun 9, 2004
Messages
178
Location
Berlin
hi jakob (and others as well, of course),

i've been comparing your design and gus' reverse-engineered schem of the brauner to my library of classic schematics... one striking difference is the unusually high grid resistor (and actually the polarizing resistor, too). in the classic designs, you find a few hundered megohms at maximum. according to textbooks, this is supposed to reduce the problem of grid leakage current shifting the bias too much. also, a superimposed random current (shot noise) from the grid wouldn't create a noise voltage as high as with 1G...?

on the other hand, i hear that noise level gets down with increase of grid resistor. could anyone elaborate? Low pass behaviour with the tube input capacitance effectively in parallel, or something?

at last, as to variation of polarizing R's... is there a disadvantage to make it lower than 1G, other than possibly affecting frequency response? is resistor noise an issue here? is 1G possible with the now available metal film R's, while classic mic's had to use noisier carbon (-composite) ones?

cheers,
volker
 
I have seen an analysis of this. Somewhere. It may be in a book at work, and this is a weekend (yay!).

I think the first issue is: tube grids leak enough that a few-hundred-meg is as high as you can go and stay stable. OTOH, small FET leakage is so low (smaller device size, cooler operation, and lower voltages) that you can run gigs of grid resistor.

But it seems to me there was a more obscure reason why some amps use 100Meg and some amps use 1Gig.

> noise level gets down with increase of grid resistor. could anyone elaborate? Low pass behaviour with the tube input capacitance effectively in parallel, or something?

Yes. The thermal noise in the resistor is shorted-out by the capsule and grid capacitance. 30pFd at 1KHz is 5Meg reactance (???check my math). So the noise of a 100Meg resistance is attenuated 100/5= 20:1 lower, a 1Gig resistance 1,000/5= 200:1 lower. The 1Gig resistor has 3 times higher noise voltage than the 100Meg, but that still makes the 1Gig 3 times lower when loaded with the capacitance. (I think that is roughly right, neglecting a few 10%s and √2s.) Of course when you get gate resistance noise down below capsule acoustic-resistance noise or tube noise, increasing the grid resistance hardly affects total noise output. That may be another reason not to use Gigs on tubes: their self-noise will typically be higher than a good FET. Huge resistors on grids don't improve noise, cost more, and let the tube drift off of optimum bias. (At some critical value of grid resistance, a tube will "runaway", bias drifts far away from what you expect. Which-way depends a lot on stray gas, cathode temperature, grid-cathode dimension tolerance, etc.)
 
prr, thanks for the explanations!

[quote author="PRR"]I think the first issue is: tube grids leak enough that a few-hundred-meg is as high as you can go and stay stable. OTOH, small FET leakage is so low (smaller device size, cooler operation, and lower voltages) that you can run gigs of grid resistor. [/quote]

i'm aware of the difference with fet's. rather, i was trying to find out why G7 and brauner use 1 gig resistors obviously without any drawback, where classic designs (partly with the same tube, check e.g. u67) only use a few hundred meg. maybe it's in your book at work...? =o) i figure this also might have to do with some recent developements of some tiny aspects of design tweaks, maybe...?
 
I think this more to do with problems in making stable and reliable ultra high ohm resistors in the period when the "classics" were made.
Is there any measurable difference in performance between 300M and 1G?
Steve
 
PRR
Oliver has written that the bias of the u47 was picked at 1.1V as to be just over the work function of the cathode oxide at its uderheated voltage. If I understand correctly by just beening over the work function voltage the amount of electrons leaving the cathode can be somewhat controlled this should cut down th grid charge alot he also wrote that there are internall sheilds in the V14 if I understood what I read correctly.

Just use enought electrons and limit where they go for grid R 1gig or greater

It turns out the type of oxide used has alot to do with what tube can be used for a tube microphone.

I am guessing on this 6072a s don't like to be underheated So when they are run at 6.3V the grid resistor can not be increased to much. If you look at the AKG circuits using 6072a the grid R is often not a very high value 8meg 33meg not 200meg or greater. I am guessing this was designed for the grid charge and/or the small size of the transformer used.

The AC701 oxide was designed for 680c operation and to run at 4v no more no less If I understand Olivers post correctly.

When I build a tube microphone cardiod only I like not to use grid R at all(like the Royer circuit) the circuit seems to have less noise with no grid R?
 
ok-- to sum this up real quick:

- the higher the grid R, the less noise due to grid current
- when reaching the point where self (shot) noise of the tube dominates the above, there is no point in further increasing grid R from a noise perspective
- also, too large grid R's can limit biasing possibilities and stability

this is opposed to the following:

- to aviod possible low freq roll-off caused by the input Z in conjunction with capsule capacitance, very high input Z is needed
- capsule polarizing R, grid R and grid-cathode input resistance are effectively in parallel and need to be sufficiently high to provide an overall high Z (a possibly present blocking cap for dc polarizing can be neglected if sufficiently large)

for raising Z, a tube may be underheated to a certain percentage (depending on type). the benefits are
- grid-cathode input resistance increases
- grid current decreases, so higher grid R's can be used for selected bias point, which additionally increases Z while retaining noise floor

a side effect of the smaller space charge in the underheated tube is increased output resistance, which has to be considered for an appropriate o/p xformer etc. i suppose another side effect are different load line characteristics, possibly reducing linearity (someone jump in please)?

regarding unterheating of certain tube types:
- ac701 and other different miniature tubes (EZ1000, 6111...) and nuvistors will be damaged by underheating, due to special cathode design
- 6072 (akg) & EF86 (brauner, gyraf, U67) always seem to be heated at specified 6.3v in mic's, althoug oli archut states EF86 can be underheated down to 30%
- heating of EF94 = 6AU6 seems to vary from 6.3v (e.g. schoeps m201) to 6.0 +/- .2v (e.g. schoeps m221b)
- EC 92 is always heated at 5.8v in gefell mic's
- VF14 is underheated at 38% in U47

any comments / additions?
 
gus, you're right, didn't look hard enough onto your drawing. :oops:
that might explain the high grid R?

still, G7 is not underheated, same high grid R anyway. biasing...?
well, maybe i'm getting too much caught up in detail anyway.
 
I have a related question. Most tube data sheets I've seen give a maximum value for grid resistance that is usually WAY under what I've seen used in mics (~1M)... How come???

I'm very sorry if this is a stupid question or if it has been discussed elsewhere, but I can't find it or figure it out...

Peace,
Al.
 
> give a maximum value for grid resistance that is usually WAY under what I've seen used in mics (~1M)... How come???

If you run a tube up to its voltage and power ratings, and use an oversized grid resistor, then if grid leakage goes runaway the tube will melt.

Most triodes, kept down to 100V or using significant plate and cathode resistors, just won't melt no matter what.

So a 12AX7, rated I think 1 Meg max grid R, was often run with 10Meg grid, 470K plate, and 150V supply (and no cathode resistor) in low-level stages. The exact bias point will vary a LOT from tube to tube, but it will never be harmed, and if signal level is much below 100mV then all likely bias points will work.

Radiotron Fourth has some info on grid current causes and effects.
 
Good question, this one.

The initial reason for the 1G in the G7 was that this was the easiest-to-get approximation of the 400M resistor used in the only other EF86-based mic I had schematics of, the U67. (this proposes 5.7 to 6.3V heater voltage)

And the 1G seems to work very well for EF86 bias - I tried it in our "real" U67 as well, with no (noticable) ill side-effects. Using 100M seemed to be less ideal soundwise - allthoug I never got around to measure resulting noise from this.

But I also found out - by experimenting - that underheating the tube will work really well - particulary with the (now relatively inexpensive) PF86, the 300mA version of the EF86

Jakob E.
 
Hi Jakob,
Good timing on your last comment. I just purchased ten NOS Phillips Miniwatt PF86 tubes to use in a G7-type mic. I bought them on eBay for about $1.50 each (incredible, compared to the price of EF86 tubes).
What did you find the best voltage to be for the heaters?
Thanks in advance.
-Scott
 
Scott,

I think I ended up a bit below 4V for the heater (for the TelePF86's I tried). But these tubes reacts quite differently to underheating, so you'll probably need to experiment.

Jakob E.
 
Hi guys
On my C800 I have less noise without 1G grid to ground. Needs to put less value or it can works without. Also (on Gus sugestions for another mikes) I connect capsule lead direct to grid. I underheated 6AU6 to 5,7V.
Duka
 
haven't some guys put 2gig or even 3gigs for that resistor??
what is the optimum value??? (EF86)
thanks
Gil
 
The effect of underheating a PF86 will be more dramatic as the voltage is lower and the current higher. Based on a pure wattage calculation, an EF86 run at 5.7v would equate to a PF86 at 4.07v.
Whats nice is that you can get PF86 in top brands NOS for peanuts.
Steve
 
i was always wondering if the tube's characteristics (g.m, S.d, etc.) and thus the distortion behaviour changes with underheating?
 
It's hard to tell, as measuring on microphones is very complicted. It might very well be that some parameters are disturbed, but if it improves overall impression in carefull listening tests, I think it's safe to do.

Jakob E.
 
From what I can find out. You need to know the oxide and cathode metal alloy before if you want to underheat using science.

The best way to tell is to just try it. If you start to get problems try a different voltage. One thing I have noted is that most tubes I have tried will work at 5.7VDC for 6.3AC rated heaters. around 5V seems to be the lowest you can go with 6.3 V tubes.

How did I get 5.7V?

I have a bunch of 7805 regulaters, if you add a 1n400x diode in the ground leg they go to about 5.7 or so, so even with the voltage drop in the cable you are around 5.7 at the tube.

From posts on the web don't underheat
AC701
6111 In the m147 m149
 
I have a feeling that the PF86 (300mA series heated version) has different cathode material than the straight EF86. At least, the ones that I've come across react much better to underheating.

Jakob E.
 
Back
Top