• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Sizing Coupling Capacitors

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I read a good thread here on sizing coupling caps on this forum but I didn't completely follow it and I think I need a little guidance.

I have a newly aquired Magnavox 175 amp. It has a pair of 12AX7 phase inverters and four 6V6 in push pull stereo. I think the coupling caps have some room for improvement.

The input caps between the RCA jacks and the 12 AX7 grid are .007 uf ceramic disk in series with a 47K resistor. I have a handfull of .47uf vitamin Qs that could be used for this. Is that too much capacitance for this application?

Between the 12 AX7 plates and the 6V6 grids are .047uf caps. How about using the 0.47 vitamin Qs here too?

Link to Schematic

Thanks
 
In some designs it is important to have the coupling poles placed higher in frequency than the power-supply decoupling poles. Otherwise low frequency oscillations occur due to positive feedback through the plate power supply.

However, in some designs, the feedback on the plate supply line will be negative so the amp will remain stable if the coupling pole is placed lower than the decoupling.

If you up the interstage coupling cap values in that amp it will be a coin toss if there will be low frequency oscillations because of that odd-ball phase splitter(when you consider part and tube variation). If you do get oscillation, adjusting R10 and R21 should stabilize it.

If you want to upgrade the interstage caps in that guy without doing further tweaking, I'd probably play it safe and use ones of equal value but higher quality.

You can use .47uF for the input cap in any case though, since it is not in between stages. It will be larger than necessary, but it won't hurt anything.
 
When low frequency oscillation is not a threat, I generally go by this rule of thumb Keep R x C = .1

1uF for 100k

so when resistance doubles, capacitance will half
.5uF for 200k (.47uF close enough)

when resistance halves, capacitance will double
2uF for 50k (2.2uF close enough)

Some people would consider my choice of cap on the large side, but I consider keeping a close to flat phase response as important as flat amplitude. Phase will shift before amplitude begins to fall.

When using negative feedback you can go smaller on the capacitors that are in the loop and get the same results. Since feedback helps correct the phase difference.
 

This is to calculate the -3dB point. -3dB is half power. Of course, when you have an RC time constant in a feedback loop things change.



Bottom line: go ahead use your .47uF vitamin Q's for the 12ax7 input. For the rest of the coupling caps, keep them the same as schematic values if you want to play it safe. Upgrading quality is ok, as long as the values are the same.

ceramic sucks.
Very true
 
Thanks for your comments Jeb, they are very helpful. I'm a rookie so bear with me here.

As for the first caps (non-interstage), if a guy had let's say 2 a .047uf and a .47uf of identical construction and quality, wouldn't the larger be the choice? Are there cons? Is the gain effected by the cap value (RxC)?

And why would we want to calculate the -3 dB point? In the equation, what is it that equals (1/2 pi RC)? Is this the answer to my question above about gain?

As for the interstage caps, I have a hand full of .047 orange caps that I can change out for the interstage. I just don't trust the old oil/paper caps not to catch fire. I have also used Sonicaps in crossover networks with audible (to my ears) results. Now I know this is a hot topic, but if a guy wanted to experiment (compare the $1 orange drops to the $5 Sonicaps) , would this be an appropriate application for the Sonicaps?

Thanks
 
The coupling cap in conjunction with the next stages input impedance form a high-pass filter. This blocks DC and frequencies below a certain point, but will allow frequencies above a certain point to pass. Since the blocking is gradual with frequency and not absolute, they developed a point to use as reference. This point is the -3dB point and is referred to as the cutoff frequency, which is the frequency at which the amplitude is attenuated by 3dB.


http://en.wikipedia.org/wiki/High-pass_filter

Example frequency response of high-pass

An externally hosted image should be here but it was not working when we last tested it.


Cap size does not do anything to the gain above the point where it begins to roll off. The cap size will just change the point at which it begins to roll off.
 
As for the first caps (non-interstage), if a guy had let's say 2 a .047uf and a .47uf of identical construction and quality, wouldn't the larger be the choice?

Generally, Yes. But there comes a point where you have more than needed to do the job right. You can actually remove that input cap and replace it with a jumper, as long as all of your sources have low DC offset.


I have a hand full of .047 orange caps that I can change out for the interstage

If your referring to Orange drops, they do pretty good. Just make sure the rated voltage on them is high enough. (I should take my own advice, as I accidentally used one in a circuit last week that had too much voltage).
 
OK, I get it. But I do have another question or two.

I now understand that a consideration in monkeying around with cap values in this situation is that one will alter the cutoff frequency, much like a high pass filter in a crossover network. I ran the calculation on this application of the existing .007 uf cap (measured value) with the 470 K resistor in parallel and came up with 48 hz, a credible choice. However the drawing calls for a 1000mmf which I interpret to be 1000pf or 10(-6) resulting in a 338hz cutoff which would seem a poor choice, unless you have a problem with turntable rumble to deal with.

At any rate, it would seem that using the larger caps (.47uf) will drop the cuttoff frequency to less than 1 hz thus for all intents and purposes eliminating the filter entirely. Since I'm not running a cheesy turntable, perhaps that's no big deal. I could drop the resistance value to 10K to bring up the cutoff, but that would create an impedance problem for the source.

And there may be more to this than I gleen. You did say "in conjunction with the next stages input impedance" but the next stages impedance would seem to be to be in series not parallel. Is that not correct? What am I missing here? I do have to consider all of the parallel resistance, not just from that one resistor, correct?

And I was refering to Orange Drop caps. I type faster than I think. And I will scope the source and see if I can listen to it with the cap bypassed. That will be revealing.
 
That is generally true Andrew, but if you refer to his particular schematic it seems they are using m to represent micro rather than u. Otherwise the caps they are using are ridiculously over sized. I've seen seen some computer grade can capacitors marked 10,000 mfd, that made me think damn, a 10 farad cap! In reality it was 10,000uF
 
And there may be more to this than I gleen. You did say "in conjunction with the next stages input impedance" but the next stages impedance would seem to be to be in series not parallel. Is that not correct? What am I missing here? I do have to consider all of the parallel resistance, not just from that one resistor, correct?

Here is some examples.
Look at the channel 1 input. C6 forms a filter with R4. At low frequencies, the input tubes grid is considered to be open circuit. So that means R3 can be left out of the equation, since it is in series with the tubes grid (an open circuit). That would mean 1 / (2 x pi x .01uf x 470k) would give you the -3dB cutoff.

For the next stage, we assume the 2 tube grids in that circuit to be open. So, C8 forms a filter with R13+R10. That would be 1 / (2 x pi x .047uF x (22k + 470K)) or simplified, 1 / (2 x pi x .047uF x 492K).
 
AndrewT said:
1mmF is 1milli milli Farad =1uF
1uuF is 1 micro micro Farad= 1pF.
1000pF=10^-9F=1nF

I found that confusing too. This (mmfd) was magnavox's way of expressing a picofarad. I had to measure the caps value to confirm this. Elsewhere they use mfd as uf as was common then.

The actual value of the cap is .007 uf. which is just outside the tolerance of the .01uf (10,000pf) cap.
 
Thanks Jeb. Good examples, I got it.

And you are correct, that cap value is 10,000 pf not 1,000 as I had incorrectly posted in my calc to arrive at the 340 hz cut. Instead, that gives us a design low pass cut at 34 hz which seems quite reasonable. The actual cap has, however, drifted just to the point of attenuating the lower bass notes. I'll make a point of listening for that when I listen without the cap and after I swap with new caps.
 
Capt'n Dave:

Another consequence of excessively lowering the f3 cutoff with a higher value cap is that the excess low freq stuff can end up saturating the output transformer (and causing distortion) in a smaller transformer, so too high a cutoff and no bass, too low and possibly distortion. I believe that you typically want the cutoff freq somewhat below 20hz (like in the single digits), you are not going to hear much below 20hz anyway.

You can certainly use sonicaps as coupling caps as long as they can stand the voltage. To be safe, it's nice to use caps that have a voltage rating exceeding your B+ voltage so they don't pop on start up when it's possible for them to see the full B+ voltage. I'm too much of a noobie to know if the B+ voltage rating for coupling caps is also a good idea for tube rectifiers (it's a good idea for SS rectification).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.