Coupling cap value with grid chokes: any new experiences or thoughts?

[Deke609]

I couldn't take the 0.22uF - it just sounded too unbalanced. With more hours of listening, I realized that the amp wasn't so much more mid-focused, but rather tilted towards the low end with a corresponding loss at the high end. And it sounded slightly "slower". Put the 0.1uF caps back in and everything is nicely balanced again and back to the "usual speed".

I was trying to to get a sense of the "speed" issue and wondered whether it had something to do with time constants and transient recovery that PJ has posted about in the past, but I couldn't find a simple (or indeed any) LC equivalent to the R*C formula. And ended up going down a rabbit hole about critical damping, bandwidth and Q-factor which really hurt my head. So I still have no clue about what's going on.  But the amp sounds good 

cheers, Derek

[Paul Joppa]

There is a resonance (IIRC, you've posted it earlier) - 6Hz for 7000 H/0.1uF. That resonance is damped by the grid resistor if you have one in parallel with the grid choke. The height of that resonance relative  to the midband is the Q. The impedance for unity damping is (IIRC) equal to the reactance of either the choke or the cap; in this case about 250K.

You would learn some things (and so would we all!) if you compared with and without the parallel grid resistor, and compared a 0.047uF or 0.033uF capacitor.

[Deke609]

Many thanks PJ.

There is a resonance (IIRC, you've posted it earlier) - 6Hz for 7000 H/0.1uF. That resonance is damped by the grid resistor if you have one in parallel with the grid choke. The height of that resonance relative  to the midband is the Q. The impedance for unity damping is (IIRC) equal to the reactance of either the choke or the cap; in this case about 250K.

OK. Great. I was sort of on the right track then.  I did a calculation for critical damping last night -- twice, in fact -- and both times came up with slightly greater than 500K. But that was late at night -- I will redo that. Maybe I was using the wrong formula.

I will dig back into this. Your pointing out the magnitude:midband Q relationship is helpful and gives me some orientation points for further study.

  You would learn some things (and so would we all!) if you compared with and without the parallel grid resistor, and compared a 0.047uF or 0.033uF capacitor.

Exciting! I will do this. I'll need to get some new caps and break them in to do a good test.  I think I'll go with 0.47uF [edit - forgot a decimal place] 0.047uF. Adding them to my existing 0.22uF and baseline of 0.1uF, this will let me test 2X and 1/2X baseline. I don't know if there's any particular empirical value in this approach, but the symmetry appeals to me.   

I'll figure out a test setup for switching caps and toggling the parallel damping resistor in/out.  Mostly b/c I don't want to repeatedly lift the amp -- too heavy! 

cheers and thanks, Derek

[Deke609]

@PJ - I finally have time to experiment with this, but I could really use some guidance about what to experiment with and measure.

I have the 0,047 and 0,1 and 0,22 uF Vcaps (the 0,047 still need to be broken in - I will get that started either tonight or tomorrow).

I also picked up a handheld LCR meter that is good up to 10,000H, 100mF, and 100Mohms. And it will measure/calculate dissipation factor, Q, phase and ESR. Stated accuracy of 0,2%. Output signal is 0,6V. Test frequencies are 100, 120, 1K, 10K and 100K Hz (no sweep function unfortunately - meters with that functionality were well outside my price range).

I also have a scope and signal generator, and of course a DMM. I only mention all this so that you know what equipment I have at my disposal in case any of it influences your recommended tests.

The quoted specs for the grid chokes are 7,000H @ 12Hz and 1,300 ohms DCR. But I plan to measure both with the LCR to confirm.

There is an interesting thread about grid chokes on Audio Asylum: http://www.audioasylum.com/cgi/vt.mpl?f=tubediy&m=198414
Of particular interest (at least to me) is the first post by "Naz" about 7 posts down where he suggests tuning the circuit by varying both the cap value and damper resistor value. He also includes a Spice screenshot showing just that.  I am tempted to try doing a similar LTSpice sim for my three cap values -- perhaps even one that simulates a frequency sweep. At the same time, I'm wary of doing the sim for fear that it will lead to confirmation listening bias. At present, I have absolutely no idea or expectation about what would sound best, so maybe it's better that I go into this without any expectations whatsoever about what is going on under different test conditions.

I still need to order the damping resistors. Do you think it worthwhile to try different values as suggested by Naz? If so, what values do you suggest I try?

All of the above boils down to following question: what tests do you think would be interesting/informative to do?

many thanks, Derek

[Paul Joppa]

I agree that it would be less biased if you did the listening test first.

For a practical application, you'd like to know the frequency response, especially in the bass, so you could most easily just install the choke an the amp and drive the amp at various frequencies and levels - thus incorporating the driver source resistance and output tube Miller capacitance.

For Science, you'd want to know the parameters and how they vary. Here are some thoughts:

* The impedance magnitude is just the voltage across the choke divided by the current through it, so can easily be measured with the choke in place in the amp.

* The impedance of the choke will be inductive (rising with frequency) below resonance, and capacitive (falling with frequency) above resonance.

* Resonance will likely be in the 1kHz-4kHz range. It is easily identified by finding the minimum current as you vary the frequency with a constant voltage. This will also tell you the Q of the resonance.

* Above resonance, the impedance falls with frequency at 6dB/octave and is expected to be independent of signal level. Normally, the outer end of the winding is grounded; the innermost end has the full AC voltage, and has capacitance to the core as well as coil self-capacitance - so be sure to connect the core to the grounded end when testing. Here I am assuming it's a single-chamber bobbin, otherwise it gets complicated  :^)

* Below resonance, the coil is damped by the series DCR (independent of frequency or level), the eddy-current losses in the core (probably dependent on frequency but not level), and the magnetic hysteresis losses (most likely dependent on magnetic flux, which is proportional to voltage divided by frequency).

* The inductance is also a function of the flux level

* To sort these out, you'll need to do some modelling. The standard model would have a series resistance (the DCR) and two parallel resistances (eddy current and hysteresis).

* Without phase information, it's hard to separate the effects. One way is to assume the impedance has a fixed phase angle, meaning the impedance magnitude (after correction for the DCR) is proportional to frequency to some power between 0 and one; the exponent (and hence the phase angle) can be determined from the slope of impedance vs. frequency which will be between 0 and 6dB/octave. This gets tricky because you can't correct for DCR unless you know the exponent, which you can't determine until you've removed the DCR effect. Best bet that I've found is to use an iterative approach to find what fits best. A decent first approximation is to use the slope  from the region where the impedance is greater than the DCR but frequency is below resonance.

* If you do the above at a few different flux levels, you can get a sense of whether you need to separate out the hysteresis terms or not.

[Deke609]

Many thanks PJ. I may have caused you to write and explain more than I intended!

I was just following up on your post from about a month ago where you wrote:

You would learn some things (and so would we all!) if you compared with and without the parallel grid resistor, and compared a 0.047uF or 0.033uF capacitor.

... and at a very basic level was wondering whether you thought different parallel resistor values might be worth exploring ... or other basic things.

That said, your very thorough "For Science" response is quite helpful in walking me through some of the relationships. I was able to keep up until you got to parallel eddy current and hysteresis losses ... I have no idea how to model those and only the most rudimentary sense of what those terms refer to (a notion of core magnetic "resistance"and images of BH curves come to mind without much understanding).  So "Science" will have to wait for a long and possibly infinite time.   

many thanks, Derek