YADTMI - yet another dual triode mixing idea

[SoulFetish]

So basically a mixing Buffer as an inverse of the common plate load mixer. You might have guessed, it's for a wet/dry signal mixer. I was hoping for some feedback on the design, and if it's worth a damn. It's basically the output amplifier stage for the unit.



The great news is I get to start breadboarding the submini stand alone reverb circuit next week. I just ordered the spring reverb unit from CE, lemme' catch you up to speed:
Following through on some direction from nickb and trobbins, I was able come up with a workable design for both the 12VDC heater and +240V HT supplies to run off a single 9VAC input supply. The power supply has come along quite nicely I think. But because the performance is so heavily dependent on careful component selection, it took a bit of time and research to dial in the simulation. I just need to run some real world test to confirm what I've simulated. But I was able to find and source all the components I needed. Here is a screenshot of an 800mS simulation of the Heater supply along with mouser links to the parts:



SB540 Schottky Diodes
5600uF/25V
2200uF/25V
(The esr ratings on those caps are ridiculous )

[Helmholtz]

Looks like another "exotic" circuit

I think it would work but might miss some unwanted interaction. It's probably easiest/safest to built and test or simulate it.

Have you considered a more straight forward design like a grounded cathode stage with plate to grid voltage feedback, producing a low impedance "virtual zero" input a the grid (needs individual mixing resistors in series with the signal sources)? The second triode could be used as a cathode follower providing low output impedance.

[SoulFetish]

Looks like another "exotic" circuit

I can't seem to help myself. I'm an experimentalist at heart.

I think it would work but might miss some unwanted interaction. It's probably easiest/safest to built and test or simulate it.

Have you considered a more straight forward design like a grounded cathode stage with plate to grid voltage feedback, producing a low impedance "virtual zero" input a the grid (needs individual mixing resistors in series with the signal sources)? The second triode could be used as a cathode follower providing low output impedance.

Yeah, totally. This was one of the first ideas I had months back (and probably still the strongest candidate). After doing some initial reading on signal mixing, the advantage of channel isolation provided by a virtual earth was really appealing. Plus, the low output impedance is inherent without the need for another buffer stage. What I thought would be the best configuration given the parts I needed to use, was to make use of the very high open loop gain I could achieve with the extra pentode available. But I ran into two snags.
First, I'm not sure how to complete the design of a local inverting feedback stage using a pentode. The second, every time I started thinking about it, I felt myself giving into the temptation to turn it into an active tone control stage as well.
But here is the basic idea of what I was thinking for it. I'm would appreciate any info, suggested improvements, or corrections:

[Helmholtz]

I'm would appreciate any info, suggested improvements, or corrections

Now this circuit looks promising.

Remarks regarding the pentode circuit.:

- You need input coupling caps and some DC ground reference for the grid
- I would add a grid stopper for stability and a cathode bypass cap
- For component values you can use the table contained here: https://frank.pocnet.net/sheets/084/6/6943.pdf
- If you choose the feedback resistor as well as the mixing resistors to be 470k, gain will be around 1.

(this list not claiming to be exhaustive)

[Dave H]

- You need input coupling caps and some DC ground reference for the grid

There's a DC ground reference through the feedback resistor and output load resistor.

[Helmholtz]

There's a DC ground reference through the feedback resistor and output load resistor.

Correct, I missed this path.
It was just a general remark. As shown without coupling caps the grid will see (high) positive voltage.

[SoulFetish]

Correct, I missed this path.
It was just a general remark. As shown without coupling caps the grid will see (high) positive voltage.

No doubt, I just forgot to draw the coupling cap.

So I did a little more work. According to the info provided by the datasheet, the circuit as drawn below (with screen/cathode fully bypassed), the open loop gain should be around 77.4 with an input signal of around 700mV RMS.



With that information, how do I determine optimal values for the components highlighted in Blue?

[Helmholtz]

With that information, how do I determine optimal values for the components highlighted in Blue?

Depends on how much you can load your wet/dry signal sources and how much gain you need. In post 4 I gave an example. The total gain as seen from each source will be roughly given by feedback resistance divided by series mixing resistance. If you don't need amplification, all resistors could be 470k.

Cap value depends on desired lower corner frequency and total load resistance. The relevant resistance for the HP calculation is total output load resistance in parallel with the feedback resistor.

[Dave H]

Cap value depends on desired lower corner frequency and total load resistance. The relevant resistance for the HP calculation is total output load resistance in parallel with the feedback resistor.

That's what I thought at first but the capacitor is inside the feedback loop so won't the lower corner frequency be reduced by the loop gain?

[Helmholtz]

That's what I thought at first but the capacitor is inside the feedback loop so won't the lower corner frequency be reduced by the loop gain?

Good point!
In fact, the output coupling cap is part of the output impedance of the circuit and output impedance is reduced by the loop gain. So the capacitance should appear to be increased by the same factor.

(For someone who knows how to simulate tube circuits it should be easy to do a frequency sweep.)


Once again I am thinking I should delay my replies until you posted first .

[SoulFetish]

Good point!
In fact, the output coupling cap is part of the output impedance of the circuit and output impedance is reduced by the loop gain. So the capacitance should appear to be increased by the same factor.
.

Right. In fact, the bandwidth is extended entirely, so I'm wondering if bypassing the feedback resistor entirely with a small value capacitor might be a good idea. But according to some reading on local feedback stages, even though the capacitance appears to increase, the capacitor value should be chosen so as not to have any influencing reactance at audio frequencies.
Aiken suggested of 10x the value of the desired corner frequency (50Hz seemed reasonable, so I chose that as well). Also, I opted for a closed loop gain of about 1 ¹/₂.
[IMGhttps://images-wixmp-ed30a86b8c4ca887773594c2.wixmp.com/f/cb9dde6e-d447-45a3-ae50-40254df453b6/ddmc35h-430d77f1-5aed-4073-b745-ae552929cdef.jpg/v1/fill/w_1006,h_794,q_70,strp/pentode_virtual_earth_mixer_values_by_soulfetish_ddmc35h-pre.jpg?token=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJzd WIiOiJ1cm46YXBwOjdlMGQxODg5ODIyNjQzNzNhNWYwZDQxNWVhMGQyN mUwIiwiaXNzIjoidXJuOmFwcDo3ZTBkMTg4OTgyMjY0MzczYTVmMGQ0M TVlYTBkMjZlMCIsIm9iaiI6W1t7ImhlaWdodCI6Ijw9OTAwIiwicGF0a CI6IlwvZlwvY2I5ZGRlNmUtZDQ0Ny00NWEzLWFlNTAtNDAyNTRkZjQ1M 2I2XC9kZG1jMzVoLTQzMGQ3N2YxLTVhZWQtNDA3My1iNzQ1LWFlNTUyO TI5Y2RlZi5qcGciLCJ3aWR0aCI6Ijw9MTE0MCJ9XV0sImF1ZCI6WyJ1c m46c2VydmljZTppbWFnZS5vcGVyYXRpb25zIl19.SplrHz_9OhWVjTxQ ojxZYKvJJ9F4-fMhEhIPrxvUG7g[/IMG]

But the problem is that most design examples are assuming a perfect zero impedance driving circuit so as not to affect the Ri value (and altering the feedback ratio.)
I would run into level issues in this case because the dry signal is derived from the output of a Mu Follower, where as the Wet is taken off the plate of a cascode. This is a significant difference. So, I could just make the value adjustments in the series resistance, or I could buffer the recovery amp and essentially equalize the driving impedance. Like so...



But, as much as I like the impedance buffer, I hate the idea of adding another stage and loosing the opportunity for a bit more + gain in the recovery amp. What do you think?

[Helmholtz]

so I'm wondering if bypassing the feedback resistor entirely with a small value capacitor might be a good idea. But according to some reading on local feedback stages, even though the capacitance appears to increase, the capacitor value should be chosen so as not to have any influencing reactance at audio frequencies.
Aiken suggested of 10x the value of the desired corner frequency (50Hz seemed reasonable, so I chose that as well). Also, I opted for a closed loop gain of about 1 1/2.

Any capacitance between the plate and grid acts as Miller capacitance and will appear multiplied by the open loop gain between grid and ground, producing a low pass. E.g. 10pF would result in an UPPER corner frequency of around 300Hz!*

Determing the lower output corner frequency requires to know the total (including external) load impedance.

I don't understand why you use a huge 0.33µ output coupling cap. Especially as feedback increases the effective capacitance by the open loop gain. For a lower corner frequency of 50Hz you might want something like 200pF-500pF depending on load impedance.


Edit: * Sorry, I forgot to take into account that also the feedback resistor is Miller transformed to the grid divided by 77. so the upper corner frequency will be not be 300Hz but more like 35kHz.

[Helmholtz]

I would run into level issues in this case because the dry signal is derived from the output of a Mu Follower, where as the Wet is taken off the plate of a cascode.

As your plate resistor is 15k, the output impedance of the stage can't be higher than that. I think a source impedance of 15k is low enough and doesn't require an extra buffer. Even a difference in output impedance of say 100k could be easily balanced by the mixing resistors.
The output impedances just add to the mixing resistors.

[Dave H]

I don't understand why you use a huge 0.33µ output coupling cap. Especially as feedback increases the effective capacitance by the open loop gain. For a lower corner frequency of 50Hz you might want something like 200pF-500pF depending on load impedance.

I don't have a pentode circuit drawn up. Here's the frequency response of a 12AX7 unity gain stage with 500k feedback and load resistors and a 500p cap. It's -3dB at 20Hz.

[SoulFetish]

Any capacitance between the plate and grid acts as Miller capacitance and will appear multiplied by the open loop gain between grid and ground, producing a low pass. E.g. 10pF would result in an UPPER corner frequency of around 300Hz!

Determing the lower output corner frequency requires to know the total (including external) load impedance.

I don't understand why you use a huge 0.33µ output coupling cap. Especially as feedback increases the effective capacitance by the open loop gain. For a lower corner frequency of 50Hz you might want something like 200pF-500pF depending on load impedance.

That was a typo(should have been .033µF), but either way, it should be large enough for full bandpass. I think a good case can be made that it's much simpler and more predictable to set the Low frequency rolloff at the input once the input impedance is calculated. Because it is in the feedback loop, it should not be used to set the corner frequency.
Load impedance is unpredictable in that it could be loaded in any number of ways because it's the output stage of the unit.
But, in reality, I decided to use the large value because each source of information I could find to learn how to design it told me to.
From Aiken's "Designing Single-Stage Inverting Feedback Amplifiers"

"Since the output coupling capacitor in the above circuit is inside the feedback loop, it is only marginally important in the actual output frequency response, as any rolloff due to it's capacitive reactance will tend to be taken out by the feedback, as long as there is enough excess gain at the frequencies of interest. However, if the reactance is large compared to the output impedance, it will introduce a frequency-dependent increase in output impedance, which will affect the above gain equations. This increasing output impedance will both increase the gain and the output impedance of the feedback network, so too small a capacitor will either peak the response at low frequencies, or roll off the gain at low frequencies, depending upon the actual output impedance and the capacitor value in relation to the feedback resistor. In addition, if the feedback resistance is too low in comparison to the reactance of the capacitor, the frequency response will also be adversely affected.

For these reasons, it is best to choose the output coupling cap reactance based upon the feedback resistance value. It should be around ten times the value required to achieve the desired frequency response, and can be calculated as follows:


¹ Co = 10 * 1/(2*pi*f*Rf) = 10 * 1/(2*pi*50*1000000) = 0.032uF (choose 0.033uF as the nearest standard value)
¹ altered to use component values in the above circuit

Once again, this capacitor should not be made any larger than necessary, or the transient response of the stage will be poor."