Passive EQ between Tube Stages

[systemtruck]

Hey all,

So here’s a pipe dream.

I’ve got an RCA BA-2 built, schematic below, and will be building a handful of them as channels in a mini console. I deleted the EQ, so you can just ignore/jump R11+C8.

I want to insert two shelf EQ’s between V1 and V2. It would be a shelf reduction of highs on one pot, and a shelf reduction of lows on another pot. No need for gain/boost of either range.

I believe I would be putting these on rotary switches so that you can adjust the capacitor values to select different frequencies instead of changing the resistor values for each frequency. That way the resistor values can always be stable, which in my mind sounds like a better situation for consistency of volume/loads, but I could be totally wrong of course.

It seems pretty straight forward to insert a single RC low pass shelf and a single CR high pass shelf.. choose values that work best for loads and the least insertion loss overall, and that’s that. But i would like to make each filter have more poles so that the function is a more powerful db/octave. 4 poles each would be the goal, so 24dB/octave. So this introduces some complication obviously….

I worked up something that seems to deliver, without much gain loss. A couple values are insane, and the frequencies and cap values chosen are just examples, 80Hz and 1.5k. But with these example values there is only 3.67dB of insertion loss even with both filters engaged. You can see the frequency that’s being measured at the top of the slopes, which is 360Hz, having -3.67dB. Bypassing the filters there is 0.5 dB of insertion loss still.

I changed the pot value to 1M, and inserted the low pass in the first half and high pass is second half since that made for better load relationships overall. I also multiplied each consecutive stage by 10x so that each stage doesn’t load down its preceding stage too much and cause weird performance. The original pot alone is 100K, so that’s the load that’s originally handing on the V1 output. This setup, I believe, loads V1 down to around 91K instead of 100K.

Is this just completely going to perform badly? What are the issues?

Like what about that first set of components… 10R and 10uF seems heavy or weird but i don’t know.

Is this approach just sheer madness???

Anybody else ever design a good 4-pole passive shelf situation? Let alone a DUAL one?

I would consider adding a cathode follower type tube stage into this if that is what allows me to make this happen best, but it seems like this is possible passively.

 

[ruffrecords]

You are right, many of those component values are insane which probably because you are tryiing to make the poles coincident.

I don't believe your 3.67dB of insertion loss. What source impedance are you using for the generator at the top left of your sim diagram? Even wired as a triode the 6J7 plate resistance will be several tens of kilo ohms.

I would suggest you think about whether it is necessary for the poles to coincide.

The problem with simple RC filters is they only give 6dB/octave per stage. Many passive designs use LC stages which give 12dB/octave per stage. Look up and read about constant K filters.

Cheers

Ian
.

 

[systemtruck]

Yes I was missing the output impedance which completely destroyed the whole thing when I added some, haha.

Thanks for the tips. I dove around constant K articles and landed on this…

https://en.wikipedia.org/wiki/Bartl...twork is bisected,the filter remains the same.

In Falstad I assembled the example circuit from it of a pi type low pass, shown below, and sure enough it works. The loss in the example is severe though, due to the fact that it’s going from 600 R to 50 R. (Seems like a weird example to use). So I flipped the two halves, set the output to 50R and load to 600R and it works perfectly. There is virtually no loss, and a good looking 18dB/octave slope at 15kHz. 3-pole isn’t quite the goal but perhaps it sounds good enough. Or maybe these can be cascaded.

Back to my context… I’m going to try to figure out the output impedance of V1, and will assume a 100K pot as the load and see what i can come up with for a 3 pole filter. Maybe the pot value can be increased to 250K without affecting the V1 performance too much? That would decrease the transfer loss.

I’d need to figure out the simplest and least expensive way to make it variable though. Hopefully i can dedicate either the inductors or the capacitors to be fixed, whichever is more expensive due to values, and then have the others be switched values.

 

[emrr]

I’m stuck on trying to understand the need. This doesn’t resemble any EQ i’ve experienced or longed for, regarding steepness. The theoretical plot posted most resembles what a Langevin 251 at full cut looks like, pretty sure most users are into the boost qualities with that EQ.

 

[systemtruck]

I’m stuck on trying to understand the need. This doesn’t resemble any EQ i’ve experienced or longed for, regarding steepness. The theoretical plot posted most resembles what a Langevin 251 at full cut looks like, pretty sure most users are into the boost qualities with that EQ.

This is an engineering approach of mine, for musical layering / arrangement reasons. I always want to have on hand some steep cuts of lows and highs, both for creative reasons to make extremely dark or extremely lofted via high cut, but also for somewhat surgical clutter reduction. For the latter, I like to cut upwards from the bottom in most basic sources/tracking, and I like to kill the highs as much as I can on abrasive sources.
Anyways that’s just me and how I roll. I haven’t used EQ lift / boost in over a decade. This build is just going to be for my own studio, I used to do a lot of session engineering in mid sized studios with great consoles and outboard, so I am well aware how eccentric this design is hahaha.

 

[Matador]

A regular 12AX7 can be buffered easily with a MOSFET follower to reduce the output impedance to less than a few hundred ohms without coloring the sound in a measurable way.

https://sound-au.com/project167.htm

 

[systemtruck]

I may revert to keeping EQ post the output transformer on each channel but it would be ideal for signal flow to have it be between the tube stages so I’m going to keep giving it some effort.
The idea of adding a tube seems like another logical fallback plan and if that’s the case I assume I could just design a “more” active EQ around that tube rather than it just being a buffer.

But as for purely passive ideas.. I’m wondering about a basic concept that I haven’t found an example of…

For context.. I think I could design a pair of 3 or 4 pole LC butterworth low and high pass like the Pi ideas in post #3. The only real hangup is cost and size of inductors that are required for high impedance. To make the inductor value change with switches would be just nonsensically expensive …. But what about selecting just the total largest inductor size required for all frequencies and adjusting its connection to the circuit with a pot strapped around it? Much like many eq circuits using a pot strapped around a cap to adjust EQ, can we also do that with an inductor?

I’m having trouble finding any examples of anything like this which leads me to believe the answer is no.

But if I can have a pot (or rather to be more accurate have a rotary switch set of specific resistors) be what’s setting the inductor’s applied values to the circuit then I can just use one or two inductors per filter and let the resistors do the work.

Is that even a thing?

 

[ruffrecords]

LC based filters is an incredibly complex subject, Many console manufactures use active op amp based circuits for the function which also allows them to be made variable using regular pots. But if you wish to remain passive you are going to have great difficulty designing a passive EQ where the frequency can be made variable using pots. Instead I would suggest you look at the CLC HPF filter used in the Neve 1073 as an example. It gives 18dB/Octave using a single muti-tap inductor to select 4 different frequencies.

Cheers

Ian

 

[systemtruck]

LC based filters in an incredibly complex subject, Many console manufactures use active op amp based circuits for the function which also allows them to be made variable using regular pots. But if you wish to remain passive you are going to have great difficulty design a passive when the can be made variable using pots. Instead I would suggest you look at the CLC HPF filter used in the Neve 1073 as an example. It gives 16dB/Octave using a single muti-tap inductor to select 4 different frequencies.

Cheers

Ian

I suppose I could look into tacking a bunch of smaller inductors in series to create a “multi tap” inductor that works for the desired frequencies and add up accordingly. Would be a lot cheaper than buying a full inductor value per frequency.

 

[Cqwet Dbdfte]

A regular 12AX7 can be buffered easily with a MOSFET follower to reduce the output impedance to less than a few hundred ohms without coloring the sound in a measurable way.

https://sound-au.com/project167.htm

"...because the voltage will rise to the full B+ when power is applied, ..."

Yes, if either a ramped startup PSU or a delay switched HV is not used. Not difficult.

 

[systemtruck]

I am considering winding some custom multitapped inductors for just this purpose. But are they inherently noisy and going to pickup all sorts of stuff?

It seems that most inductors on the market that are high enough inductance value to be used for high impedance circuits are also meant for high power/current circuits. And they are expensive.

I am assuming two things from this… 1, that high value / low power inductors used to be more available on the market when tube circuits were still a major part of the customer base. And 2, that the consumer costs of the inductors are at least partially correlated to the power/current capability of the wire used.

So I’m wondering if maybe some inductors can be self wound with cheaper low power wire, tapped where needed, but with enough turns/size to create these high values. Maybe even ferrite/iron core, if i can find the right raw material. (I’m still earning about this).

If this is possible, it would be fun and pretty satisfying to have the ability to shape the exact HPF and LPF filters needed with some custom multi tap inductors. I have a nice sturdy precision drill press that can be used at very low speeds. Seems perfect for creating tight windings around a shaft of something mounted in its chuck.

On the other hand, Hammond still offers high value inductors for say 20-25 USD. I actually have a couple arriving tomorrow for something else, 150H 8mA. The 156C. I wish it were easy to create taps in an already built inductor because that would be a very acceptable price to pay for a multitapped inductor per filter. I imagine I’d have to somehow measure inductance along its windings, and then tack-solder to a spot that measures what is needed. Sounds like a mess and making from scratch would probably be a better result.

@ruffrecords

 

[heatwalk]

Mesa boogie sells the inductors for their graphic eqs in a few values. I have considered using them recently for some custom eq builds but haven’t gotten around to it yet

https://store.mesaboogie.com/categories/no-category/electronic-components/inductors.html

 

[ruffrecords]

I am considering winding some custom multitapped inductors for just this purpose. But are they inherently noisy and going to pickup all sorts of stuff?

It seems that most inductors on the market that are high enough inductance value to be used for high impedance circuits are also meant for high power/current circuits. And they are expensive.

I am assuming two things from this… 1, that high value / low power inductors used to be more available on the market when tube circuits were still a major part of the customer base. And 2, that the consumer costs of the inductors are at least partially correlated to the power/current capability of the wire used.

High value inductors were not required in the past because most passive EQs were designed to operate at 600 ohms.

High current inductors are certainly more expensive than ones designed for the minute currents involved in audio

Your problems stem from the fact you have chosen to place the EQ between two high impedance circuits.

Cheers

Ian

 

[systemtruck]

Mesa boogie sells the inductors for their graphic eqs in a few values. I have considered using them recently for some custom eq builds but haven’t gotten around to it yet

https://store.mesaboogie.com/categories/no-category/electronic-components/inductors.html

Excellent, thank you. This could prove to be solution.

High value inductors were not required in the past because most passive EQs were designed to operate at 600 ohms.

High current inductors are certainly more expensive than ones designed for the minute currents involved in audio

Your problems stem from the fact you have chosen to place the EQ between two high impedance circuits.

Cheers

Ian

Yep, fully understood. And the engineering feat is surely interesting and hell i might even go for it.

I’m curious what kind of difference between output and impedance of 6J7 and load placed upon it i can get away with. From some post by PRR i see that the output impedance of any given typical preamp is usually anywhere from 0.15X to 0.40X the plate resistor. In this case, the plate resistor is 120K. So that would be a range of 18K to 48K. (And for what it’s worth, the plate resistance of a 6J7 is 11K in triode mode.)

The designed pot is 100K.

When using a calculator such as this one https://markimicrowave.com/technical-resources/tools/lc-filter-design-tool/ it becomes apparent from some experimentation that increasing the difference between the two impedances is very effective at reducing the size of inductors required.

So, to see how feasible this really is, i need to confirm the output impedance of the first 6J7 stage in a BA-2, and i need to know how high of a value i can change the 100K pot to be. I randomly chose some numbers, and used 35K as the source impedance and 250K as a pot value since that is a pretty common pot value in these preamps. This combination brings inductor values down into the several Henries territory maximum, as opposed to like say 20H+.

So if this EQ was ignored, and I placed a 250K pot into the circuit instead of the 100K pot, how much does that mess with the sound when loading down V1 less?

 

[emrr]

It shouldn't change sound, it will increase gain slightly while extending the low end passed by the coupling cap over an octave, as per the usual scaling of C versus R for same result.

 

[systemtruck]

It shouldn't change sound, it will increase gain slightly while extending the low end passed by the coupling cap over an octave, as per the usual scaling of C versus R for same result.

Then what’s my reasonable limit?
If there is no grid leak resistor on V2 then maybe V2 is depending on the pot wiper never being above a certain distance from ground.
Can I go up to 1M? Or even higher?
The inductor size for EQ designs gets more and more attainable the lighter the load from the pot is.

 

[emrr]

You don’t really see gain pots larger than 250k in full range audio, mostly. You start having variable treble with position the higher you go. A grid resistor can go as high as whatever for more gain because it’s not playing voltage divider as a second job.

There’s no grid leak because you’re already cathode biased. Grid leak implies grounded cathode.

You should consider V1 feeding a cathode follower driving the EQ. 3 valves. V2/3 could maybe be a 6SN7 type.

 

[ruffrecords]

Excellent, thank you. This could prove to be solution.




Yep, fully understood. And the engineering feat is surely interesting and hell i might even go for it.

I’m curious what kind of difference between output and impedance of 6J7 and load placed upon it i can get away with. From some post by PRR i see that the output impedance of any given typical preamp is usually anywhere from 0.15X to 0.40X the plate resistor. In this case, the plate resistor is 120K. So that would be a range of 18K to 48K. (And for what it’s worth, the plate resistance of a 6J7 is 11K in triode mode.)

The plate resistance depends not only on the mode but the operating point (higher quescent currents tend to lead to low plate resistances) but 11K is not a bad estimate for a 6J7 in triiode mode. The small signal output impedance of the stage is roughly the plate resistance in parallel with the plate load so you might as well just call it 11K

The designed pot is 100K.

When using a calculator such as this one https://markimicrowave.com/technical-resources/tools/lc-filter-design-tool/ it becomes apparent from some experimentation that increasing the difference between the two impedances is very effective at reducing the size of inductors required.

So, to see how feasible this really is, i need to confirm the output impedance of the first 6J7 stage in a BA-2, and i need to know how high of a value i can change the 100K pot to be. I randomly chose some numbers, and used 35K as the source impedance and 250K as a pot value since that is a pretty common pot value in these preamps. This combination brings inductor values down into the several Henries territory maximum, as opposed to like say 20H+.

So if this EQ was ignored, and I placed a 250K pot into the circuit instead of the 100K pot, how much does that mess with the sound when loading down V1 less?

Not sure if this pot is going to be before of after the EQ. The problem with putting it before the EQ is that the source impedance seen by the EQ is determined more by the pot position than the tube and will vary over a wide range. Better to have the pot as the final load for the EQ and drive the EQ from the known relatively constant 11K or so output impedance of the 6J7.

In this postilion, a 250K pot will work fine.

As an aside, the very first amp I built when I was 11 used a 6J7 and a 6V6. Great tubes.

Cheers

Ian

 

[spicemix]

RCA did a version of the BA-2 (MI-12241) that had a simple tilt EQ included, EMRR describes it long ago in this thread with schematic:
https://groupdiy.com/threads/rca-mi...e-a-ba-2a-with-eq-addition.16210/#post-215423

 

[abbey road d enfer]

There’s no grid leak because you’re already cathode biased.

A cathode-biased stage needs galvanic connection to ground.
One may choose to call the resistor different than grid leak when the negative bias developped into it is much lower than that developped in the cahode resistor, but still it's destined to rout the grid-leak current to ground (or a reference point).