Siemens W295b Confusion

[krabbencutter]

I've been trying to DIY a W295b and it's confusing the heck out of me. Like some folks already mentioned in the forum, the High/Low Shelf is basically an active Baxandall EQ. I've simulated it in LTspice and put the circuit on a breadboard and everything works so far. When comparing it to the Soundtoys Sie-Q the low shelf behaves completely identical, but the high shelf is way off. My breadboarded high shelf behaves like a standard baxandall high shelf but the High Shelf in Sie-Q works much more like a 17KHz Constant-Q Bell filter.

Picture 1 shows a 6dB boost of SieQ (pink) and my breadboarded circuit (blue) from 1kHz upwards
Picture 2 shows my simulated circuit
Picture 3 shows a 12dB boost of SieQ (purple) in comparison to a 17KHz Constant Q Bell curve (pink, slightly offset)

Since Sie-Q is the only available reference, I don't know if Soundtoys decided to design the high shelf differently or if I'm missing something on the schematic?

 

[dogears]

Looks like you're missing a cap. C33 in the original.

You can read about the effect of this cap in Peter Baxandall's original 1952 article, figure 8.
https://learnabout-electronics.org/Downloads/NegativeFeedbackTone.pdf

 

[Bo Deadly]

You can do it without a center tapped switch. But then I think you might need to double the pot value and adjacent resistors.

If I were doing this, I would first simulate the W295B exactly with discrete parts and study the noise and impedance of the inputs and output while sweeping over different boost / cut and generally study the overall behavior of it. Then I would probably use op amp in / out but add resistance to match impedances and just use 11 position switches with the center tap cap. I could see that center tap changing the behavior enough that if I wanted to call it a W295B, it would want it in there. But if, for whatever reason (cost), a pot was desired, I would simulate the non-center tap pot version and adjust the values until the two circuits simulated as closely as possible.

You might also try simulating the two capacitor version of the treble control as well to see if that gets you even closer to the original behavior (probably doesn't).

 

[krabbencutter]

Looks like you're missing a cap. C33 in the original.

You can read about the effect of this cap in Peter Baxandall's original 1952 article, figure 8.
https://learnabout-electronics.org/Downloads/NegativeFeedbackTone.pdf

You are absolutely correct. After reading PRR's comment about the center taps for the 30th time in two days, I finally realized my mistake Siemens W295b
He's talking about potentially omitting the bass-pot tap but my mind always skipped the "bass pot" part and I thought both taps were redundant with modern components.

 

[Bo Deadly]

He's talking about potentially omitting the bass-pot tap but my mind always skipped the "bass pot" part and I thought both taps were redundant with modern components.

You mean treble pot. The center tap is on the treble control.

 

[krabbencutter]

You can do it without a center tapped switch. But then I think you might need to double the pot value and adjacent resistors.

If I were doing this, I would first simulate the W295B exactly with discrete parts and study the noise and impedance of the inputs and output while sweeping over different boost / cut and generally study the overall behavior of it. Then I would probably use op amp in / out but add resistance to match impedances and just use 11 position switches with the center tap cap. I could see that center tap changing the behavior enough that if I wanted to call it a W295B, it would want it in there. But if, for whatever reason (cost), a pot was desired, I would simulate the non-center tap pot version and adjust the values until the two circuits simulated as closely as possible.

You might also try simulating the two capacitor version of the treble control as well to see if that gets you even closer to the original behavior (probably doesn't).

Hm ... the center tapped 150uF capacitor is indeed crucial to the design. I think I'll go with a dedicated Boost/cut switch as a compromise. That way I can keep a standard potentiometer for full control over the gain range.

I should mention that I'm attempting to do this in the "Poor Man's" spirit. So no magnetics, no transistors and as many off the shelf parts as possible

 

[Bo Deadly]

Hm ... the center tapped 150uF capacitor is indeed crucial to the design. I think I'll go with a dedicated Boost/cut switch as a compromise. That way I can keep a standard potentiometer for full control over the gain range.

I should mention that I'm attempting to do this in the "Poor Man's" spirit. So no magnetics, no transistors and as many off the shelf parts as possible

It's not crucial. It just helps achieve a good null at the center position but mostly because of resistor tolerances. A vaguely good modern potentiometer will yield a good null. With a pot at some point around the center it will be a perfect null. And if you use a pot with a center detent, you can trim one of the side resistors to make the center detent the perfect null.

Just try simulating with 10k and double the resistors. Does it shift up? It should same cap / larger resistor should shift everything up.

 

[dogears]

Hm ... the center tapped 150uF capacitor is indeed crucial to the design. I think I'll go with a dedicated Boost/cut switch as a compromise. That way I can keep a standard potentiometer for full control over the gain range.

I should mention that I'm attempting to do this in the "Poor Man's" spirit. So no magnetics, no transistors and as many off the shelf parts as possible

https://www.mouser.com/ProductDetail/Bourns/PTT111-4425A-B502?qs=h2IHEVivlqBmjzhMc6zJRQ==
Should work I think.

 

[krabbencutter]

It's not crucial. It just helps achieve a good null at the center position but mostly because of resistor tolerances. A vaguely good modern potentiometer will yield a good null. With a pot at some point around the center it will be a perfect null. And if you use a pot with a center detent, you can trim one of the side resistors to make the center detent the perfect null.

Just try simulating with 10k and double the resistors. Does it shift up? It should same cap / larger resistor should shift everything up.

As far as I'm aware C33 changes the response of the treble knob quite significantly. And it needs to be connected to a center tap.
You can see the comparison in the screenshot. It's the same exact circuit with the same gain settings, but the first one has a center tap with a 150uF cap connected to it.

 

[Bo Deadly]

Again, try changing the pot and resistor values like I said twice already.

 

[krabbencutter]

Again, try changing the pot and resistor values like I said twice already.

I've spent the past two days simulating, testing, comparing and swapping components. Nothing so far has worked and neither did your suggestion. Connecting C33 however did work. But feel free to poke around in my simulation.

 

[Bo Deadly]

Well you're right. More than you probably realize.

Here is how to do with a pot:


However, aside from the fact that lowering the limiting resistors probably loads the source amp too much, it is also nothing like the W295B with the center tap to ground which lowers to Q as more boost is applied with serves to effectively shift the frequency down:



I also noticed something important. In the center position there is a huge resonant peak (according to simulation) at ~120kHz. This will not happen with the original circuit (because of the 56p from the input transistor collector to base C9). You can use an op amp but you will need to add a 100p as shown above. I would argue that this makes a strong case for building the original discrete circuit but I wouldn't tear you down for using an op amp. This cap also serves to shift the frequency down as more boost is applied. So again, I think I would model the discrete circuit first to see how that behaves exactly because there's a lot going on here. Then you will have a lot more confidence about making an op amp version that behaves in the same way.

Basically if your circuit (or plugin) doesn't do this, you can't call it a "W295B". And the regular bax circuit simply does not behave in the same way. So you MUST use a switch with the center position grounded if you want to emulate the W295B precisely.

 

[krabbencutter]

I've been recently revisiting the W295b circuit and I still can't wrap my head around the mid band. There's some kind of feedback loop with 2x 510k resistors and the gain knob will switch in a resistor on either side of the feedback path, followed by an LC to ground. But the +/-6 and +/-8 positions have the same resistance of 2k. And as far as I can tell all the remaining connections will be left open. So how can it be, that switching in a 2k resistor gives 6dB of gain and another 2k resistor gives 8dB of gain? What am I missing here?

 

[dogears]

I've been recently revisiting the W295b circuit and I still can't wrap my head around the mid band. There's some kind of feedback loop with 2x 510k resistors and the gain knob will switch in a resistor on either side of the feedback path, followed by an LC to ground. But the +/-6 and +/-8 positions have the same resistance of 2k. And as far as I can tell all the remaining connections will be left open. So how can it be, that switching in a 2k resistor gives 6dB of gain and another 2k resistor gives 8dB of gain? What am I missing here?

They’re trimmers. They are supposed to be adjusted to the exact value required for the indicated boost / cut.

 

[krabbencutter]

They’re trimmers. They are supposed to be adjusted to the exact value required for the indicated boost / cut.

Oooooh now it makes sense. So the value just indicates the base resistance of the trimmer? My brain was hard stuck thinking the given resistance values indicate the actual calibration Thanks a lot!

 

[krabbencutter]

Got it working in LTspice now