Taming The Treble

[romberg]

Thanks for pointing out the other/better way to use a cap like this. I'm making an educated guess that any stray high frequency noise from the power supply would be on the plate if the cap was not there at all. So, putting one end of the cap on B+ is no worse than it was before (as far as noise from that power supply node goes). Putting it to ground would flush the highs from both the plate/signal and the power supply. Is that the gist of it?

Thanks,

Mike

[gingertube]

Not quite:

For susceptibilty to power supply noise you look at from where the output is being taken - the anode

You can see that, when considering teh power supply noise, the anode is at the output of a voltage divider, top arm of the divider up to B+ is 100K and the bottom arm is the impedance back to 0V from the anode.

Ignoring any any load resistors, like the grid leak to the next stage, that (bottom arm) impedance is eqaul to the internal impedance of the tube "rp" plus any unbypassed cathode resistor multplied by the tubes mu+1.

In this particular example the amount that the B+ noise is divided by is negligable.
Why?
because the rp for a 12AX7 at very low current (which this will be due to the RK=10K) will be about 100K and the RK is unbypassed so looking down from the anode it looks like a 10K (Mu + 1) value. Mu for 12AX7 is about 100. So that bottom arm of the divider is effectively 100K + 10K x 101 = 1.1Meg.

So in this case 1.1 Meg / (1.1Meg + 100K) of any rail noise will appear at the anode (just over 90%). Adding the cap across the 100K means that 100% of the rail noise will apper at the anode (above the frequence at which the capacitive inpedance is low compared to the 100K anyway).

That shows, that in this example it makes pretty much a rodents hind quarters difference which way we wire the cap.

More usually thou' the 12AX7 stage will be run at larger current which means that the rp value will about 60K ansd the RK will have a bypass cap effectively taking it out of the circuit.

In THAT case the lower divider arm becomes simply 60K and 60K/(100K + 60K) or less than 40% of the B+ noise will apear at the anode. In that more usual case the connection of the cap from anode to 0V rather than directly across the 100K makes a big difference.

This is one of the reasons why the HiFi Boffins use current sources (VERY high impednace) in place of the anode load resitor and use low rp triodes. Any B+ noise is then divided to a tiny value. The other reason they do that is that the source of distortion in triodes is the variation of rp with signal swing. The ratio of the rp variation vs the load resistance determines the level of distrortion so that HIFi Boffin scheme also gives very low distortion, not what we want in a GIT Amp.

Bit of a rave - feel free to tell me to pick up my marbles (if I can find them) and go home.

Cheers,
Ian

[bluesky636]

Personally, I think this is all irrelevant. If the power supply is putting out enough noise that the location of this cap makes any difference, the power supply is either poorly designed or faulty. In any case, the power supply should be looked into and problems corrected because any noise would be present in each stage whether the cap is there or not. BTW, I have a BS in Electrical Engineering, earned when tubes were still part of the curriculum.

[Tillydog]

the amp is quieter with the cap across the plate resistor than it is without it. Hiss is significantly reduced with no real loss in highs ...

Out of curiosity, did you try a grid stopper on V2b? I found a small one made a big difference on my Express - similar to what you're noted for the cap. Both the grid stopper and bypassing the plate would serve to kill any HF oscillation in that stage, so I'm wondering if that could be the cause.

[bluesky636]

the amp is quieter with the cap across the plate resistor than it is without it. Hiss is significantly reduced with no real loss in highs ...

Out of curiosity, did you try a grid stopper on V2b? I found a small one made a big difference on my Express - similar to what you're noted for the cap. Both the grid stopper and bypassing the plate would serve to kill any HF oscillation in that stage, so I'm wondering if that could be the cause.

My final schematic is posted on page three of this thread.

[Tillydog]

My final schematic is posted on page three of this thread.

I know!

It doesn't show a grid stopper on V2b, and I was wondering whether you had tried one and rejected it, or not tried one at all.

No matter....

[bluesky636]

My final schematic is posted on page three of this thread.

I know!

It doesn't show a grid stopper on V2b, and I was wondering whether you had tried one and rejected it, or not tried one at all.

No matter....

Sorry. Misunderstood the question. I didn't feel the need for one.

[Tillydog]

No worries!

Thanks.

[romberg]

Tillydog what value of stopper did you use? I wanna try this too.

Mike

[Tillydog]

I used 100 ohms (just by chance).

The difference was immediately noticeable in terms of reduced hiss and harshness, and was the final piece in the jigsaw for this amp for me (and I had been tweaking it for over 2 years). It didn't sound *bad* before, just... unimpressive...

I would be interested in whether it makes any difference for you.

As an aside - I've just realised that the OP's schematic shows both halves of V2 used, but only one half of V1 (rather than vice versa). I never noticed that before. Maybe it's significant...?

[bluesky636]

Using the Amp Book Calculators, you can determine the frequency response of a gain stage with either a grid stopper or cathode bypass cap.

Using a 1000pF plate bypass cap like I did has the following results:

Here are the frequency response plots for the plate resistor bypass capacitor. The first one is with a 100pF cap (that was the smallest the calculator provides. No "0" pF.). Note that the high frequency -3dB point is above 10kHz.

[img763]http://farm4.staticflickr.com/3782/9387 ... 35b4_o.jpg[/img]

Here is the frequency response with the 1000pF cap. The high frequency -3dB point is now about 3kHz.

[img760]http://farm8.staticflickr.com/7429/9387 ... c1c2_o.jpg[/img]

Here is an explanation of the calculator which can be found at

http://ampbooks.com/home/amplifier-calc ... capacitor/

[img850]http://farm8.staticflickr.com/7439/9387 ... 01df_o.jpg[/img]

[bluesky636]

Here are the plots for the same stage using a grid stopper instead of a cathode bypass cap.

http://ampbooks.com/home/amplifier-calc ... d-stopper/

With a 68kOhm grid stopper:

[img813]http://farm3.staticflickr.com/2857/9410 ... 2351_o.jpg[/img]

With a 100Ohm grid stopper:

[img815]http://farm6.staticflickr.com/5497/9412 ... cfcc_o.jpg[/img]

The output impedance of the previous gain stage and the gain of this stage can be determined using this calculator:

http://ampbooks.com/home/amplifier-calc ... impedance/

[bluesky636]

As an aside - I've just realised that the OP's schematic shows both halves of V2 used, but only one half of V1 (rather than vice versa). I never noticed that before. Maybe it's significant...?

My amp is based on the Weber JAVA which has the option of using the second half of V1 as a cathode follower to drive the tone stack, which I left out:

https://tubeamparchive.com/viewtopic.ph ... highlight=

[Tillydog]

With a 100Ohm grid stopper:

[img815]http://farm6.staticflickr.com/5497/9412 ... cfcc_o.jpg[/img]

Yeah, I'm convinced the effect of the small grid stopper was to suppress some sort of high (radio) frequency oscillation or ringing which was interacting with the audio frequencies, rather than a straightforward rolling off at 'bat ear' frequencies.

I hunted for parasitics with the 'scope, but could never find any, so the above remains conjecture, but the change was easily audible.

[bluesky636]

Probably wouldn't be a good idea to combine a cap bypassed plate circuit with a large grid stopper in the same gain stage.