SIAM 295b

[fragletrollet]

Fun project! Following

 

[GreyboxAudio]

Update: I've renamed the repository https://github.com/greyboxaudio/SIAM-295b and the schematic has also been updated with some minor changes to the gyrators. I'm also happy to announce that I plan to design a PCB this year

So about that PCB... While I now have enough experience with KiCAD & JLC PCB to order PCBs, I am still struggling to come up with a satisfactory schematic.
- The opamp in the shelving circuit creates a lot of high frequency noise and performs much worse than the discrete class a topology of the original. I'm refusing to accept "discrete class a magic" as the solution and want to ideally fix this issue at the source
- the mid band is weird, because the Q not only varies for each gain setting but is also slightly asymmetrical between boost/cut. This makes it tough for me to decide the right balance between simplicity and authenticity

But I've picked the project back up and will hopefully make some progress over the holidays.

 

[GreyboxAudio]

I rewired one of my PSUs from +/-12V to +24V and built the shelving filters with the discrete amp. I used all BC549C transistors and chose the nearest values for electrolytics (22uF instead of 25uF and 47uF instead of 56uF). It works but I need to put it in a metal case because the lack of shielding is messing a lot with the low end. I've also ordered a used Siemens V2284 card to gut for input/output transformers.

 

[zzzzz]

Looking good! Thanks for taking this on and your hard work. Excited for the end result.

 

[GreyboxAudio]

Happy new year! I now have some Haufe input/output transformers and have also built the input amplifier. So far everything works really well. Last but not least I've ordered a custom RM8 inductor for the mid band from don-audio.com.

 

[GreyboxAudio]

Glad to see they finally arrived! The main addition in Rev5 has been the Sallen-Key filter between the EQ stages. You could easily build that on a small piece of stripboard and add it to your existing circuit with some wires between R16/R18. The SKF reduces the high frequency noise and acts like the Dr1 inductor in the original W295b schematics, which I had ignored until that point. You could also try to increase the feedback capacitor of the Baxandall-OpAmp (C15 in your schematics) to 470p or even higher, which will mitigate the high frequency noise as well.
I also changed the rotary switch for the mid band frequencies to switch at the gyrators' inputs instead of switching the gyrators' ground connections, but it will work either way.
The additional 4M7 resistors can be soldered directly onto the switches and the output capacitor (C34 in my latest revision) is just for good measure to make sure there's no DC offset.

Speaking of the Dr1 choke it is most likely around 5.6mH. I ran a couple simulations to see what different values for Dr1 do to the circuit in combination with the original R101 & C35. With 6.8mH the frequency response will be flat but the high frequency roll off starts already below 20kHz. With 4.7mH the roll off is higher but there will be a small high frequency bump. 5.6mH seems to be a nice middle ground that also stays within the official spec (+/-0.5dB from 40Hz to 15kHz). A comparable Sallen Key low pass filter would be F=22kHz & Q=0.8

 

[GreyboxAudio]

My custom inductor and the choke have arrived. Everything seems to work as expected and I will start working on a housing soon. Hopefully this will improve the SNR.

 

[GreyboxAudio]

My W295b EQ is taking shape. I replaced the output transistor pair with BD139 instead of BC549, because the poor BC549 couldn't handle the heat when driving into a transformer.

 

[GreyboxAudio]

I also had an epiphany regarding the mid band: the W295b isn't actually setting the gain but instead it is damping the Q. In the original circuit the RCL circuit gets connected to one end of the resistor pair R75/R76 to boost/cut or between those resistors for a flat frequency response. In case of a +8 dB boost R84 would get manually adjusted to give a +8 dB boost (see first image). Now watch what happens when we connect the RCL circuit to the mid point of R75/R76 but treat R75/R76 like a potentiometer that's turned all the way to one side: We get the exact same result (second image)! This means that we can replace R75/R76 with a simple potentiometer, use a single resistor in series with the wiper to set the maximum gain and then have completely variable gain over the whole range. And by reducing the series resistance the maximum gain can even be increased to +/-10dB or +/-12dB, although with a very narrow Q (image 3)