Corrective filter in 4xEL84 amp

[gingertube]

Output transformers are an impedance translator device. Within the frequency response range of the transformer it will not matter whether the R+C "Zobel" is placed:
1) across the secondary
2) across the full primary
3) 2 off across each half primary (each of the push pull halves).

You are trying to do 2 things with these RC networks.
1) linearize the impedance vs frequency load (compensate for the speaker looking inductive at higher frequencies).
2) Dump the higher frequency power into this network instead of into the speaker.

Choosing the R value depends upon what you are ultimately trying to do.
If the "zobel" is being used across a triode stage anode resistor (for stability in a circuit with feedback) then you want to adjust the amplitude response vs frequency whilst leaving the phase response largely unaffected. In that case you use a value of R equal to 1/10th of the resistance being shunted. The frequency response will be down to 1/2 amplitude (-6dB) when the zobel impedance is the same as the load impedance impedance (ignoring vector arithmatic for now to give the idea - that will be when Zc = 9/10RL and R = 1/10 RL) but notice that the 45 degree phase shift will be when Zc = R which will be at a frequency about 9 times higher than the -6dB amplitude frequency. Note also that the minimu amplitude will be 1/10th (-20dB) at frequencies when Zc is effectively zero.

If you are purely trying to compensate for the rise in load impedance due to the speaker appearing inductive at higher frequencies then you might use as higher value R. If you use r = RL then the maximumn amplitude reduction you can achieve is to 1/2 of the original (-6dB).

If you use a value of R > RL then the amplitude reduction will be smaller again.

Also note that the slope of the roll off is always the same , 6db per octave.

The response is what is called a shelving response, there will be a low frequency gain set by the basic circuit, a high frequency gain set by the value of R in the zobel and a 6dB / octave slope between these points with the 1/2 response point set by the capacitor value.

With 10K + 2n2 across a 10K Raa transformer primary then maximum high frequency roll off will be 6dB and teh half response point will be at 7234Hz.

Assuming an 8 Ohm speaker the exact same response could be achieve by using an 8 Ohm + 2.7 uF across the secondary.

So select R for the amount of roll off you want to achieve, then select the C to set the half response frequency.

If you want to get a bit "scientific" then may I recommend the following reading / experimental method(s).
http://www.siteswithstyle.com/voltsecon ... xfmrs.html


Cheers,
Ian

[Tillydog]

Tilly, keep in mind that in this exact (contextual) instance, the other impedances involved are also going to be doubled/halved accordingly, due to an extra set of power tubes in parallel.

Ahh, OK - I read the original comment as being that 10k/0.002 was directly equivalent to 5k/.... etc. My bad. So, yes in the above context, it would seem proportionate.

I'm with Ian, in that (to a first approximation, at least) I believe it makes little difference which side of the transformer you chose to add the components.

My experience has been mainly with things that perform perfectly into a dummy load, but develop bad habits driving a speaker, hence I believe that the problem I have experienced stems from the speaker impedance interracting with the OT, rather than anything 'coming from the amp'. Everyone's MMV!

[husky]

I totally understand all that, but again, the Carmen Ghia and Badger both use the 10k/.002uf values (which is the same as 5k/.0047uf). I think it's better the change the cap value rather than the resistor for power dissipation reasons.

Actually I use a 1n/10K
You really need to dial it in for taste since the OT Impedance and tubes will have a big effect on the results.

A resistive load isn't going to expose the potential of parasitics like a fully reactive load will.

[dave g]

Unrelated to conjunctive filters, but relevant to the EL84 fizzyness:

Most EL84 amps use a bias voltage around -10 volts. Contrast this with 6L6s, EL34s, etc. which require on the order of -30 or -40 volts (in the typical fender/marshall push-pull output circuit).

EL84s require far less signal from the PI to drive them into distortion (in the typical push-pull arrangement this happens when you drive the grids positive w.r.t. the cathode) than other power tubes. If you try to use a PI that has enough swing for EL34s or 6L6s, you are going to find yourself driving your EL84s way too hard. This is where all of the crossover distortion, blocking distortion, hard clipping, and bees-in-a-paper-bag fizziness problems associated with EL84s come from.

The AC30 PI that is a perennial favorite for EL84 amps runs at slightly lower voltages than the Fender or Marshall variants and uses a larger tail resistor, so it does have a bit less swing - but still too much for EL84s to handle.

There is nothing wrong with sticking in a post-phase inverter MV to prevent those grids from getting driven too hard. Heck, the easiest thing to do is just replace your 220k grid reference resistors with a voltage divider. You could start with 100k/100k and attenuate more if necessary.

[Gaz]

I totally understand all that, but again, the Carmen Ghia and Badger both use the 10k/.002uf values (which is the same as 5k/.0047uf). I think it's better the change the cap value rather than the resistor for power dissipation reasons.

Actually I use a 1n/10K
You really need to dial it in for taste since the OT Impedance and tubes will have a big effect on the results.

A resistive load isn't going to expose the potential of parasitics like a fully reactive load will.

Thanks, that's interesting because 10K/.002 was too much for me as well. All the samples of the Badger still have plenty of bite. In regards to my original question, how do you adjust the filter for the 4xEL84 version? Do you change the values to 5K/2.2n? Thanks.