Passive EQ between Tube Stages

[emrr]

Sure, and I’ve never in that case heard anyone call it a grid leak resistor.

 

[systemtruck]

11K output to 250K load is quite effective at reducing the inductor values necessary for 3 pole low pass filters of the constant K type in discussion, however, plugging some more numbers in it doesn’t help very much in the high pass arena. Still need some serious Henries for that. Also, once you attempt to make 4 pole inductor/cap constant K filters, which is the ideal for my personal stuff but perhaps too excessive, the 11K to 250K still then requires huge inductors for both LPF and HPF.

I admit i am tempted to use either a cathode follower as an intermediary stage or to put this constant K filter after the output transformer, both of which makes this thread off topic hahah. I could be wrong but I think the cheapest solution, and perhaps simplest, would be the latter… putting the constant-K EQ after the transformer and then load it down with the load that was otherwise going to load the transformer, such as a slug resistor which was already the plan.

Since I’m working in the context on a console, i will mention that I’m going to then being going straight into a network of resistors for mixing. My concern of course is the mixing network varying the load so much that the constant-K EQ’s preceding the mixing network at each channel would be thrown all over the place depending on the level/pan settings of everything around it. But I think i somewhat stabilized the loads from going all over the place by choosing to go straight into tube grids at the summing amps rather than using input transformers there which have a much lower input impedance. That way i have high impedance inputs that allow the mixing network some leeway to not load down the channel outputs as much. I haven’t pinned down mixing values yet, but maybe I can get away with not having to buffer anything, and just have the output transformer pass through the constant-K EQ and then have its slug to make it stay around 600R and the lighter load mix network not vary that load too much and the EQ’s will be fairly consistent.

One nice thing about multiple pole EQ’s, in my mind anyways, is that the repeated pole implementation helps to stabilize the frequency of the EQ and it can handle more variation in load than a single pole could. This could be something I’m making up though.

I know that a cathode follower buffer will be the only way to solidify things in either of those two approaches (in the post-output version I’m referring to a cathode follower being put at each channel’s entry to the mix network) and keep the EQ exactly the same at all times. So maybe that’s destiny. And if that’s the case, it’s either going to be inside the tube stages or it’s going to be after the output transformer. But if it’s a V2, and let’s pretend the cathode follower puts out a perfect 600 ohms output impedance… if i put the 600ohm constant-K filter there can i load it down with a 600R resistor at the grid input of V3? I’v never seen that kind of value entering a tube so it seems inappropriate.

 

[Cqwet Dbdfte]

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.

Every filter calculation requires defined source and load impedances.
100k source resistors makes it hard to design a reasonable filter, better to use a cathode follower or MOSFET source follower as buffers.
Gain tubes can be mu-followers for optimal linearity using constant current sources.
https://markimicrowave.com/technical-resources/tools/lc-filter-design-tool/
Works also for Hz.
Plan for a reasonable load and source impedance, not 600 or 250k.
Maybe around 10k. 600 requires a step down transformer.

 

[abbey road d enfer]

Sure, and I’ve never in that case heard anyone call it a grid leak resistor.

So, when does a grid resistor begins to be a grid-leak resistor?

 

[emrr]

the missing word is 'bias'. I only ever hear 'grid leak bias', and 'grid resistor'. You say grid leak, it implies the first. Context was:

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.

I haven’t looked for examples, but recall grid leak bias resistors to be very high in value.

 

[abbey road d enfer]

the missing word is 'bias'. I only ever hear 'grid leak bias', and 'grid resistor'. You say grid leak, it implies the first. Context was:

It does not answer my (rhetorical) question "when does a grid resistor begins to be a grid-leak resistor?"
As soon as there is resistance between grid and ground, a voltage develops across it, modifying bias. It may be negligible, but it exists.
Doesn't in any case this grid resistor leak electrons to ground?
IMO, the distinction between grid resistor and grid bias resistor is artificial.
Who am I to say this, not being a native english speaker?

 

[emrr]

Yes, it is about the importance of it in the balance.

 

[systemtruck]

I’m exploring the post-transformer option for filters since it may prove to be the most convenient.
But since it is a transformer which depends on a load to have any impedance, i am unsure how to implement a 600:600 constant K filter just after it.
I worked up an example 4 pole filter, then drew 3 versions of impedance placement.
For a tube output transformer, which one is the way? Or neither?

 

[abbey road d enfer]

since it is a transformer which depends on a load to have any impedance

The xfmr is driven by the triode's internal resistance, of about 10kohms. Using the formula Zs= (Ns/Np)².Zp, you can determine the reflected impedance at the secondary. I would not be surprised if it was close to 600 ohms.

 

[ruffrecords]

So, when does a grid resistor begins to be a grid-leak resistor?

When the cathode resistance is zero at which point you have grid leak bias.

Cheers

Ian

 

[ruffrecords]

Back on topic,
Another possibility is to have a 10K:600 transformer after the first stage followed by a 600 ohm constant K HPF/LPF followed by a 600:10K step up transformer into the second stage.

There is a schematic around of a 600 ohm 100% passive HPF/ LPF with switched frequencies. I am not sure who made it - maybe Altec?

Edit: It is the Altec9067B. Only three stage in each filter but nonetheless it should be instructive,

Edit++: check out this thread: https://groupdiy.com/threads/altec-9067b-passive-filter.51217/

Cheers

Ian

 

[systemtruck]

Back on topic,
Another possibility is to have a 10K:600 transformer after the first stage followed by a 600 ohm constant K HPF/LPF followed by a 600:10K step up transformer into the second stage.

There is a schematic around of a 600 ohm 100% passive HPF/ LPF with switched frequencies. I am not sure who made it - maybe Altec?

Edit: It is the Altec9067B. Only three stage in each filter but nonetheless it should be instructive,

Edit++: check out this thread: https://groupdiy.com/threads/altec-9067b-passive-filter.51217/

Cheers

Ian

Well now, that Altec 9067 is the closest thing I’ve seen to what I want to build. I’ll take 16 of those please !

The 10K:600 then 600:10K is nice, cost aside. The fact that the 6J7 is 11K out is lucky convenience for this common transformer designs. Cost would sure add up quite prohibitively if using good transformers though, doing a number of these channels. But again it’s really nice because it keeps the EQ inside the preamp and opens up the world of smaller inductors for EQ and doesn’t require any power changes/needs. My only tech concern would be noise being picked up from the drop down then gain-up. The BA-2 i built one of is so quiet that I’d hate to wreck that.

i couldn’t find any actual schematics of the Altec but did see that example you posted with the HPF and LPF. Yes this is more or less the way forward. Being down at 600R lets everything be realistic.

This reminds me… when using calculators like this one https://markimicrowave.com/technical-resources/tools/lc-filter-design-tool/ ..or other similar ones, some have the option to select filter style/type. I’ve been defaulting to Butterworth. Is that what most calculators are using?

I think I’m going to go 4 pole at 600R -> 600R. It’s just a matter of when in the staging. Being after the output transformer is the most cost effective way to do it since it’s already staged for it. I finally remembered that Collins 212B mixer which is a set of channels that have constant-K attenuators following the preamp output transformers and they dont have any shunt or resistor hanging on the transformer. So it guess it’s pretty straight forward. The only crappy thing about having the EQ here, other than the sensitivities of designing the mixing network that follows is, is the fact that if i want direct outputs after the EQ it’ll be unbalanced outputs unless i toss a 600:600 transformer after the EQ.

 

[emrr]

Well now, that Altec 9067 is the closest thing I’ve seen to what I want to build. I’ll take 16 of those please !

https://www.audiomaintenance.com/downloads/ezhlf3-500_colourbook.pdf

 

[systemtruck]

https://www.audiomaintenance.com/downloads/ezhlf3-500_colourbook.pdf

Nice. Another good model.
I checked out the interfacing and signal flow on page 17 schematic.
There is clearly a hot and cold and a chassis earth separate from this that some noise filtering is going down to. The cold is kept separate and “treated” as just as much of a signal as the hot. Looking at this has me wondering..

Can these constant-K filter designs all be strapped across a hot/cold pair from a transformer and the signal remains a balanced signal after? Like can the one I just posted go from hot to cold and earth can be separate?
Or if wanting a fully balanced filter, do we need to create a hot set of components and also a cold set of components? So double the design size basically?

 

[abbey road d enfer]

When the cathode resistance is zero at which point you have grid leak bias.

Cheers

Ian

It may be a language barrier issue, but it doesn't make sense. A 1-ohm resistance would turn a grid-leak bias into a cathode-bias?
I don't think so.
It's a semantic issue. There is no frontier, grid current always exists, and produces a negative voltage across the grid resistor. Saying grid leak bias happens only when there is no cathode resistor is a gross oversimplification. The actual situation is that the g-k voltage is the combination of voltage across the cathode resistor and the voltage across teh grid ressistor. It's a ontinuous process.
It looks like usage has sanctified a distinction that doesn't exist.

 

[ruffrecords]

It may be a language barrier issue, but it doesn't make sense. A 1-ohm resistance would turn a grid-leak bias into a cathode-bias?
I don't think so.
It's a semantic issue. There is no frontier, grid current always exists, and produces a negative voltage across the grid resistor. Saying grid leak bias happens only when there is no cathode resistor is a gross oversimplification. The actual situation is that the g-k voltage is the combination of voltage across the cathode resistor and the voltage across teh grid ressistor. It's a ontinuous process.
It looks like usage has sanctified a distinction that doesn't exist.

I think most people would agree that grid leak bias is when Rk=0 such that the grid leak resistor is the only source of bias. In general you either use grid leak bias or cathode bias by which you mean the dominant source of bias. With Rk = 1 ohm I am not sure which would be dominant.

Cheers

Ian

 

[ruffrecords]

Nice. Another good model.
I checked out the interfacing and signal flow on page 17 schematic.tly this
There is clearly a hot and cold and a chassis earth separate from this that some noise filtering is going down to. The cold is kept separate and “treated” as just as much of a signal as the hot. Looking at this has me wondering..

Can these constant-K filter designs all be strapped across a hot/cold pair from a transformer and the signal remains a balanced signal after? Like can the one I just posted go from hot to cold and earth can be separate?
Or if wanting a fully balanced filter, do we need to create a hot set of components and also a cold set of components? So double the design size basically?

A balanced signal exists only between the hot and cold. The chassis/screen is just that, a protective screen. It does not carry signal current.

It is possible to make any EQ operate on a balanced signal simply by connecting it between hot and cold. Many old passive filters did exactly this.

Cheers

Ian

 

[systemtruck]

Certainly very much of the complication with trying to make an equalizer of this style for a console is the gigantic efforts and costs if a lot of unique part values are involved. I’ve wanted a 4 pole filter and, as you can see from any calculator, making a constant-K 4 pole involves two inductor values and two capacitor values. Never mind the caps, the inductor situation gets out of hand when you’re trying to implement and purchase so many unique inductor builds.

I was messing around with values and I decided to see what would happen if instead of using the proper 4 pole design i went ahead and just put two 2-pole filters one after the other. So that is two identical caps and two identical inductors. I thought it was going to be a trainwreck in contrast to the true 4 unique value setup. Turns out, as opposed to the proper 4 pole it basically just adds a 0.5dB knee bump and then a similarly mild droop after, but then after that the curve and slope down is more or less the same nice 24dB/octave all the way. I tried this with both high pass filters and low pass filters.

So yeah, electronics engineers are not going to ever suggest this approach because it’s not clinical for any purpose and that bump can’t be removed. But it sure reminds me of resonance in any synth filter! And the fact that it’s only about 0.5dB is basically not ever going to be heard anyways. Done and done.

This helps turn things into realistic goals. I’m going to see if I can either A, make, or B, buy custom multitapped inductors that are all identical. Just one multitap design, and use these for both the LPF and HPF…. A typical classic EQ inductor having 7 taps actually offers 7+6+5+4+3+2+1 values if you use them in series in all lengths possible, or something close to 28 unique values. The complication would be switching, but that can be worked out. 28 is a lot to work with. I was thinking having a 20 step rotary switch, but even 24 could be used.

The classic pre-existing multitapped offerings may or may not work for this approach, but besides, being able to get down and dirty with the math and create a versatile set of taps all under 800mH-ish seems like a very effective move. The EQ ones from Cinemag are quite physically small, just around an inch in diameter. I’m curious about cost of a custom selection of values in that same package so I’ve sent a message.

But then there’s this https://www.mouser.com/ProductDetail/871-B65701W0000Y038
Which is something someone on this forum pointed to, and which has a massive “AL”, and i understand has a proportionately massive tolerance swing. But, maybe these can be wound by testing each value as i go and that does away with the tolerance issue? The huge AL allows a very low turns count, so each one of these would be a total of under 200 turns.

All of this is assuming that this EQ would come after the output transformer. I would love for it to be in the middle stage and use 10K:600 and 600:10K transformers around it, but that cost simply can’t be justified.

 

[abbey road d enfer]

A typical classic EQ inductor having 7 taps actually offers 7+6+5+4+3+2+1 values if you use them in series in all lengths possible, or something close to 28 unique values.

No. In order to realize many of the values, you need to disconnect windings, so taps are not an option. Separate windings is necessary, and the switching may be very complex.
I would think binary, with a set of values in 1, 2, 4, ...32 maybe. that would allow with a rather simple switch, to create values from 1 to 63.
Anyway, I wouldn't even start really thinking about it. The premises are not good IMO.

 

[systemtruck]

No. In order to realize many of the values, you need to disconnect windings, so taps are not an option. Separate windings is necessary, and the switching may be very complex.
I would think binary, with a set of values in 1, 2, 4, ...32 maybe. that would allow with a rather simple switch, to create values from 1 to 63.
Anyway, I wouldn't even start really thinking about it. The premises are not good IMO.

Switching complications aside.. how can that be?
Looking at any of the vintage EQ schematics, can’t we just move up the “bottom” ground/earth as well as the “top” of the selection and create a new value? Doesn’t everything floating underneath the new ground point just get nulled?