I've been thinking about the topic being discussed and the interest in this phenomenon... well... here are several thoughts on the subject...

first thought... we should separate out anode (plate) chokes from grid chokes. As they perform different functions and are used and spec'd differently. the plate choke must ordinarily be designed to accomadate unbal dc plate currents and thus generally requires an air gap. A "grid choke" typically is used with a dc blocking condensor and is designed as an "ac only" device. The magnitude of L that you want from each may also be quite different.

second thought....

the example you provided... quoted below...

:::300pF in parallel with 80pF input capacitance of 300B is 380pF. And with high impedance driver (cascode, pentode, high rp tubes common cathode...), say Rout~15kOhms, we have f-3= 1/(2*3,141*15000*380*10^-12) = 27,9 kHz , limited frequency response of the driver.:::

this is with a very high impedance tube... and generally when you want to drive a 300B grid your going to use a tube with a much, much lower internal plate resistance. Say anything btwn a low of 800 ohms to 5000 ohms or even 7000 ohms would be much more typical (I think)...

with a low impedance driver and a high shunt capcitance the -3db cutoff frequency would then still be much greater than 27.9khz.

as with all devices and circuits... the designer must optimize his/her circuit and use a topology that makes sense and produces good results.

For example... it is much tougher to build a plate choke for a tube with an r sub p of 15K... and generally these tubes are operated at much less plate current.... so they are not even generally spec'd or used to drive the grids of triodes which do have higher internal capacitances... so with triodes most designers aim at using a tube with much lower r sub p and perhaps greater bias currents.

if your going to use a tube with an r sub p of 15K I would *most* (not absolutely though) likely not recommend that you use a plate choke and instead might recommend a plate resistive load instead.

third thought.... the whole issue of shunt capacitance also applies to any transformer that carries an audio signal through it...

push pull output transformers will have shunt capactances across the primary...

single ended output transforemrs will have these same capacities...

so... why the focus on shunt capacitance (in seeming isolation) of a
plate choke or a grid choke?

In most real world apps... shunt capacitance is only one of dozens of parameters and quantities that must be taken into account in achieving a good design...

whenever the focus of any discussion on the merits or demerits of a design boils down to ONE parameter... it's bound to lead to mischief...

the trick in design is not to aim at or focus on optimizing any single parameter... whether it be minimizing shunt capacitance, maximizing inductance, minimizing DCR, or minimizing flux density level, or etc...

good designs are designs that pay attention to a whole range of factors or considerations that go into a design and optimizing for the "whole" of the device and not just any single paramter...

as an addendum to this post I will put up a link to a post written by voltsec which I think you may enjoy reading.

cheers,

msl

and that's also why I posted the link to voltsec's paper... cause he makes the same observation...

the problem with isolating one spec (any particular singular spec) is that it invites mischief in the design process...

almost any design (in a wide range of fields) is made easy if the customer only cares about one specification and will trade any and all to achieve it....

but the pumping up or pumping down of any singular spec (in isolation) in an audio transformer just leads to "specsmanship" not necessarily a good overall well balanced design that will work well and sound good.

if you look at your reasons as stated above and voltsec's as stated in his post (down near the bottom of his post).... that pretty much captures my take on specsmanship games.

We try to optimize a range of values and parameters and etc... that produce (hopefully) a good quality, balanced, great sounding transformer. Others can design to isolated arbitrary specs... and take their chances on acheiving a good overall design.

msl

I've been thinking about the topic being discussed and the interest in this phenomenon... well... here are several thoughts on the subject...

first thought... we should separate out anode (plate) chokes from grid chokes. As they perform different functions and are used and spec'd differently. the plate choke must ordinarily be designed to accomadate unbal dc plate currents and thus generally requires an air gap. A "grid choke" typically is used with a dc blocking condensor and is designed as an "ac only" device. The magnitude of L that you want from each may also be quite different.

second thought....

the example you provided... quoted below...

:::300pF in parallel with 80pF input capacitance of 300B is 380pF. And with high impedance driver (cascode, pentode, high rp tubes common cathode...), say Rout~15kOhms, we have f-3= 1/(2*3,141*15000*380*10^-12) = 27,9 kHz , limited frequency response of the driver.:::

this is with a very high impedance tube... and generally when you want to drive a 300B grid your going to use a tube with a much, much lower internal plate resistance. Say anything btwn a low of 800 ohms to 5000 ohms or even 7000 ohms would be much more typical (I think)...

with a low impedance driver and a high shunt capcitance the -3db cutoff frequency would then still be much greater than 27.9khz.

as with all devices and circuits... the designer must optimize his/her circuit and use a topology that makes sense and produces good results.

For example... it is much tougher to build a plate choke for a tube with an r sub p of 15K... and generally these tubes are operated at much less plate current.... so they are not even generally spec'd or used to drive the grids of triodes which do have higher internal capacitances... so with triodes most designers aim at using a tube with much lower r sub p and perhaps greater bias currents.

if your going to use a tube with an r sub p of 15K I would *most* (not absolutely though) likely not recommend that you use a plate choke and instead might recommend a plate resistive load instead.

third thought.... the whole issue of shunt capacitance also applies to any transformer that carries an audio signal through it...

push pull output transformers will have shunt capactances across the primary...

single ended output transforemrs will have these same capacities...

so... why the focus on shunt capacitance (in seeming isolation) of a
plate choke or a grid choke?

In most real world apps... shunt capacitance is only one of dozens of parameters and quantities that must be taken into account in achieving a good design...

whenever the focus of any discussion on the merits or demerits of a design boils down to ONE parameter... it's bound to lead to mischief...

the trick in design is not to aim at or focus on optimizing any single parameter... whether it be minimizing shunt capacitance, maximizing inductance, minimizing DCR, or minimizing flux density level, or etc...

good designs are designs that pay attention to a whole range of factors or considerations that go into a design and optimizing for the "whole" of the device and not just any single paramter...

as an addendum to this post I will put up a link to a post written by voltsec which I think you may enjoy reading.

cheers,

msl

Neat post. It invites so much discussion... that it could keep us busy for quite some time...

I'll try to keep some of my comments and observations and speculations a bit more brief...

first observation:

presumably the data derived is a product of the WHOLE circuit... not a subset of the circuit or an individual stage or etc...

which leads to (and this of necessity to an extent) that the resultant behaviours apply to the whole circuit and not to any specific part of the circuit...

for example... it appears that the output transformer's primary is treated as having simply a reflected impedance of 3,000 ohms... but it to could be modeled like the grid choke... by adding the other relevant parameters of an output transformer.

the power supply is not included in the analysis... it too may have an effect on the amplifiers performance in the real world. In fact... in a study that Paul Joppa did modeling resonances the worse case he came up with was a series fed se output stage which resonanted quite distinctly with the last cap (8 mfd) in the series circuit.

simplifying (which all models do by their very nature)... has it's limitations and advantages... knowing all of these and ferreting them out isn't always so easy.

but let's run with the spice analysis you provided...

::::Our driver has some good properties (low Miller capacitance, amplification ~35, good sound ), but unfortunately, has Rout~Ra, or 12kOhms in this example.::::

which means that if you choose this drivers it will present a set of different challenges and design requirements than if you had chosen an alternative driver tube. fair enough. but we should all be aware of this.

::::The simulated frequency response we can see in lower diagram - high frequency started to fall after 20kHz,::::

I put up a small post it note to try to decipher the X and Y values and I was getting somewhere btwn -1db and -2db (with my imperfect eyes) at around 50 khz.

If this is correct.... not nearly the end of the world or in-and-of-itself anywhere close to rotten in my book. I'd want to listen to the circuit and decide not decide it's merits on this spec.

:::and we have LF resonance (~8dB) on 10Hz. For later, if we want to avoid this and have a linear response down to 2Hz, we must use a much larger coupling cap Ci, about 4,7µF.::::

the above was with use of a .22uf cap. On Joel Tunnah's site he has a resonance calculator as well as a Q calculator...

in line with p-spices result... the resonance calculation also agrees that .22uf and 1700 henries will resonate at 8.23 hz.

if we change the cap value... make it larger... here are the calculated resonant frequencies (keeping 1700 henries constant)

.5uf will resonate at 5.46 hz
1uf will resonate at 3.86 hz
4uf will resonate at 1.93 hz

interestingly, as voltsec sez in his post... when you change one parameter invariably you will effect or change other parameters as well...

let's look at the Q.... again calculated on Joel Tunnah's website...

1700 henry choke with a dcr of 8,000 ohms... then by varying cap size not only do we change the resonanct frequency but we will change the Q ...

.22uf cap with 1700 henries and a dcr of 8,000 ohms will have a Q of 10.988

.5uf cap with 1700 henries and a dcr of 8,000 ohms will have a Q of 7.289

1uf cap with 1700 henries and a dcr of 8,000 ohms will have a Q of 5.154

4uf cap with 1700 henries and a dcr of 8,000 ohms will have a Q of 2.577

so that not only does the resonanct freq change with the cap size but also the Q... so the magnitude of the resonant peak and behaviour of the resonance will change...

all of which goes.... in my view... to the point of... this is where the designer and his\her skill comes into play... it's knowing how to opitimize the circuit... what values to use... what tubes to use... and how to optimize a whole range of values to acheive a specific performance goal.

and it could very well be possible... that given certain design requirements or certain choices of other components... that perhaps LC coupling is not optimal in the particular application being looked at...

but... juggling the parameters and knowing which parameters need attention or need to be optimized is part of the "skills package" that the designer brings to the table. Some designers bring more enhanced "skills packages" to the design table... others bring less developed or refined "skills packages" to the design table...

be interesting to run some of the pspice simulations with the differing cap values shown above and see what ya get...

on the surface.... my first hunch is that your circuit is close to being pretty darned good... maybee a bit more fine tuning will get it where you want it to bee...

cheers,

msl

And for HF - use a grid choke with smaller Cw, or another driver - with lower Rout. If we use, say, common cathode 6C45Pi, our Rout would be much smaller, about 1/10 then cascode, and now HF "falling" problem is gone, but 10Hz "hump" would be even larger - need larger Ci...

That's a nice one-sided way of characterizing everybody else save yoursleves( or what ever group whose opinion you choose to champion ).

Good engineering design is not defeated by publishing specifications when asked. It has absolutely no influence on the design process if the answer is an honest measurement of the device in question.

If the answer sucks, and makes one's product line look 2nd rate is another story entirely. Leaving the question un-answered does leave the impression...

It is best to assume that the customer designing a circuit knows enough to look at a given product and make up his own mind. The idea that magnetics are sooooo mystical that only a designer with Xx years experience could possibly know what will sound good in a new circuit makes me laugh every time....oh well, good thing it( design) isn't really complicated.

Getting a straight answer OTOH...
cheers,
Douglas

I've been thinking about the topic being discussed and the interest in this phenomenon... well... here are several thoughts on the subject...

first thought... we should separate out anode (plate) chokes from grid chokes. As they perform different functions and are used and spec'd differently. the plate choke must ordinarily be designed to accomadate unbal dc plate currents and thus generally requires an air gap. A "grid choke" typically is used with a dc blocking condensor and is designed as an "ac only" device. The magnitude of L that you want from each may also be quite different.

second thought....

the example you provided... quoted below...

:::300pF in parallel with 80pF input capacitance of 300B is 380pF. And with high impedance driver (cascode, pentode, high rp tubes common cathode...), say Rout~15kOhms, we have f-3= 1/(2*3,141*15000*380*10^-12) = 27,9 kHz , limited frequency response of the driver.:::

this is with a very high impedance tube... and generally when you want to drive a 300B grid your going to use a tube with a much, much lower internal plate resistance. Say anything btwn a low of 800 ohms to 5000 ohms or even 7000 ohms would be much more typical (I think)...

with a low impedance driver and a high shunt capcitance the -3db cutoff frequency would then still be much greater than 27.9khz.

as with all devices and circuits... the designer must optimize his/her circuit and use a topology that makes sense and produces good results.

For example... it is much tougher to build a plate choke for a tube with an r sub p of 15K... and generally these tubes are operated at much less plate current.... so they are not even generally spec'd or used to drive the grids of triodes which do have higher internal capacitances... so with triodes most designers aim at using a tube with much lower r sub p and perhaps greater bias currents.

if your going to use a tube with an r sub p of 15K I would *most* (not absolutely though) likely not recommend that you use a plate choke and instead might recommend a plate resistive load instead.

third thought.... the whole issue of shunt capacitance also applies to any transformer that carries an audio signal through it...

push pull output transformers will have shunt capactances across the primary...

single ended output transforemrs will have these same capacities...

so... why the focus on shunt capacitance (in seeming isolation) of a
plate choke or a grid choke?

In most real world apps... shunt capacitance is only one of dozens of parameters and quantities that must be taken into account in achieving a good design...

whenever the focus of any discussion on the merits or demerits of a design boils down to ONE parameter... it's bound to lead to mischief...

the trick in design is not to aim at or focus on optimizing any single parameter... whether it be minimizing shunt capacitance, maximizing inductance, minimizing DCR, or minimizing flux density level, or etc...

good designs are designs that pay attention to a whole range of factors or considerations that go into a design and optimizing for the "whole" of the device and not just any single paramter...

as an addendum to this post I will put up a link to a post written by voltsec which I think you may enjoy reading.

cheers,

msl

I enjoyed the dialogue. Keep us posted of how your circuit comes along and how it fares by ear!!!

msl