<?=$alt; ?>
<?=$alt; ?>

Moog™ – 914 Fixed Filter Bank

Buy the PCB set:

914 Fixed Filter Bank - GIC
914 Fixed Filter Bank - GIC

Available Qty: 22
Price: $70.00
914 Fixed Filter Bank - Inductor
914 Fixed Filter Bank - Inductor
Does NOT include inductors. Contact Carsten Toensmann for inductors, http://www.analog-monster.de/index_en.html. See below for more information.
Available Qty: 35
Price: $50.00
Panasonic 50k log for 914 FFB
Panasonic 50k log for 914 FFB
Panasonic EVU-E2JFK4D54 - 50k, side adjust, metal bushing, logarithmic taper.
Available Qty: 62
Price: $1.50
Canare GS-4 micro coax cable
Canare GS-4 micro coax cable
This is a price per foot of the cable.
Available Qty: 48
Price: $0.50


Document archive, schematics, BOMs, layouts

This archive is for both the inductor and GIC version of the fixed filter bank. There is a separate BOM for each PCB and I apologize for that inconvenience.


Grayscale is offering a group panel buy of his design which is quite stunning. See here: Grayscale Group Panel Buy. Here is a link to the Front Panel Express panel layout: Moog_914_FFB_Panel_03


  1. Added more information to the documents archive. Please download the newest version.
  2. Added some clarification on the polarized capacitors in the Power section.
  3. Added some links and a little more about connecting the “DRIVE” signal between PCBs.


First, I have to give credit where it’s due for the inspiration.

  • Dr. Moog and his engineers for the original design.
  • Yves Usson for his work, design, and idea to break out even and odd cells.
  • Jurgen Haible (R.I.P.) for his approach using simulated inductors (GIC)¹.

My design clearly derives from the original Moog™ design and is as faithful to that design as possible. I spent more time pouring over those simple filter circuits more than any other project I’ve undertaken. Hours of simulation in MultiSim trying to characterize the circuit. Staring at photos and trying to trace signals and identify components. Finally realizing a resistor was missing which made all the difference because the resistor directly helps determine the “Q” of the cells. Then the reward of simulations that produced the proper frequency response and prototypes that finally worked the way they should with very little noise or hum. Last, dealing with some form and fitment issues so that the module could reasonably be built. What a trip.

So, thank you Dr. Moog, Yves, Jurgen, and any teacher or author who taught me, and especially my dad who taught me to love electronics as soon as I could hold the solder for him while building Heathkits.

The project is not cheap. Plan on $300.00 or a little more for inductors, and depending on what you choose, as much as $130.00 for op amps (see below). The GIC version is pretty straight forward to build and probably more robust if you gig your synthesizer. The inductor version clearly will be more true to the original, but soldering the extremely fine varnish coated wires and getting a solid solder joint takes patience and care. The cores are a bit brittle and fragile and care must be used when mounting them. In the end though, you will have an extremely useful and unique module for sound shaping.

Don’t hesitate to ask questions or give me feedback. I don’t do this to make money. I just want to help you make the music that I don’t have the talent to make myself.


This module is not an equalizer but it does boost and attenuate in frequency bands, so it is an equalizer. Honestly I don’t know what to call it. Reading the information I could find, mostly from the MoogArchives.com website (the following is in part paraphrased from the Moog™ docs reproduced there), this filter is a non-voltage controlled “modifier” with 14 band pass filters with a control which adjusts the level of each band. Each band has a 12dB slope. This creates peaks and troughs similar to a formant filter. There is also a low pass band and a high pass band.Each band has a “Q” of about 3.7 (thank you Yves for that tidbit). I think the intent was specifically for it to be not voltage controlled so that the timbre of an input signal would change as it moved through the frequency regions. One use might be to help replicate the fixed resonances of a particular real instrument, or for creating other unique effects by emphasizing certain bands mixed with an unfiltered signal.


A modifier module, typically in near the end of the signal path before final mixing, but it can go anywhere you have an audio signal to process. It is ultimately AC coupled so it is not of much use for CV processing. You can use the filter as an “all pass” filter and send the even signal to one stereo channel and the odd signal to the other for an interesting pseudo stereo effect.


There are two inputs:

  1. Signal. The summing resistor on this input can be chosen to handle higher level signals than originally intended by Moog™ or just high signal levels in general.
  2. Signal. Same as for #1. There are two summing resistors on the PCB so you can choose these to suit your needs. One big, one small so you can handle and mix both high and lower level input signals. I picked a relatively large summing resistor because my audio signals are typically higher than in the original.

There are three outputs:

  1. Main Signal. This is the sum of all the filter cells amplified back to line levels.
  2. Even Signal. This is the “even” filter cells starting with the 125Hz cell and summing with every other cell up to and including the HP cell.
  3. Odd Signal. This is the sum of the “odd” filter cells starting with the LP cell, summing every other cell ending with the 5600 Hz cell.


This module has 14 knobs (pots) which control the attenuation of each cell.


There are no switches


Connect an audio signal you want to process to one of the inputs, or both if you want to mix them first.


See the Component Notes page for some general information. See below for specific information relating to this module.


Use good resistors like metal film, as that will help keep the noise down


Use high quality capacitors, like a polyester film type especially for the GIC version in the GIC cells. Harry Bissell has a great capacitor reference reproduced here if you want to learn a bit about the different types and uses for capacitors.

914 FFB decoupling GICThe decoupling capacitors can be mounted on the underside (GIC version only) or in the DIP socket, if you use a socket. If you want to put the in the socket like in the photo, make sure you get a physically small, MLCC type. If you mount them on the solder side, do them first. Otherwise, good luck soldering them with an IC in the way…

Each capacitor which helps set the center frequency for the filter cell along with the inductor, real or GIC, and the capacitors which help set the value of the simulated inductors (GICs) is made from two capacitors in parallel if needed and there is room for two capacitors with two different lead spacings ( 0.1 or 0.2 inch (2.5mm or 5mm)). Moog™, Yves Usson, and Jurgen Haible do it this way, so why buck the system? I printed two values on most capacitors in different directions. I’ll explain why and how to pick the right values.

It appears that Moog™ designed the center frequency (CF) of the first cell to be a bit lower than the target CF and the CF of the second cell to be a bit higher. I have no proof of this other than looking at the schematics for the 907 and Jurgen’s design. Yves uses the same CF for the first and second cell. To do this, the capacitor for the first cell should be a bit bigger than ideal, and the capacitor for the second cell a bit smaller. But, if you think it is better to have the same CF for both cells, you can do that, too. I calculated values for either way.

If you look at the build photos, you’ll notice that on the GIC PCBs, there are a couple of capacitors whose value is not visible. For reasons I don’t understand, the value on this 180nF capacitors was printed on the side. It was only on these capacitors. You will also note the actual filter capacitors are the same on the inductor version and the GIC version. The GIC version has extra capacitors to help set the inductance value.

Remember that capacitors in parallel simply add their capacitance together.

On the PCB, you will see markings adjacent to the capacitor pairs that look like this from the first cell in the 175 Hz filter

330.8n -> 374.4n

This means the ideal capacitor is 330.8n and can move towards a value of 374.4n to achieve an CF that is 6% lower than ideal. The second cell is marked

551.4n -> 490.8n

which means the ideal is 551.4n but you can reduce that to 490.8n to get a CF 6% higher than ideal. Your combination of capacitors should add up to a value somewhere between these two numbers.

The GIC is similar, but since the GIC is the inductor, you want the capacitor to create an inductor of the ideal value, or err on the side of pushing the CF down for the first cell (bigger inductor), or up for the second cell (smaller inductor). These capacitor pairs are marked like this from the first cell in the 175 Hz cell

SEL >= 516.5n

which means create a combination of capacitors to be greater than or equal to 516.5n

The second cell is marked

SEL <= 309.9n2014-12-07 16.17.44

which means to create a combination of capacitors to be less than or equal to 309.9n

All right, you say thanks David for doing the math, but what two capacitors do I choose? “Ahhhh,” I say. “Look at the capacitors values on the silkscreen and you will see your answer.” On each of the two capacitor pairs, there are two sets of values. You should be able to tell which ones go with the ideal value and the 6% offset value by adding them together. The values on the left side of each capacitor are the ideal, and the values on the right side of each capacitor are the 6% offset. For the GICs, there is only a suggested combination for the ideal which is either “greater than or equal to” for the first cell or “less than or equal to” for the second cell.

Sometimes, there is only one value printed on a capacitor. That’s because using only one capacitor was the closest solution.

The printed values are standard values for polyster capacitors you will find at the lin2014-12-07 16.18.20k I gave you above. There are two lead spacings you can use on the PCB, 0.1 or 0.2 in (2.5mm or 5mm). There is a little capacitor symbol between two of the pads. If you use the smaller spaced pads, make sure your capacitor spans the two holes that are on either side of the symbol. It looks like this:


On this example above, you would put the capacitor in the two pads on the right side. The link I gave takes you to the 5 mm spacing, so there is no problem. You are welcome to do your own calculations and pick your own values. I arrived at these by creating a table of the combinations of most of the standard values available and simply hand picked the ones which best met the criteria. I’m sure there’s a smarter way to do it…

Email me if this is too confusing.


If you build the inductor version, you need inductors. You can wind them yourself (it’s SDIY, right?) or you can get them from Carsten. He has this all figured out and he will gladly send you a set of hand wound inductors to work in this module. If you do the GIC version, the inductors are simulated with op amps.

Should you happen to break a core, Carsten will sell you new ones or I found these ones from Mouser work at least for the small inductor size Carsten sells Small Inductor Core and these for the big ones, Big Inductor Core.


First, there are two ICs on the I/O PCB. These can be any decent generic op amp. They are labeled and the BOM calls for a OPA2227 which is a very, very low noise, expensive op amp. These are really overkill.

The output amplifiers for the Even and Odd cells are not part of the original 914, so use what makes sense to you. For example, an NE5532 would be fine.

Second, the GIC cells. GICs are prone to noise. Picking a cheap op amp here is a bad idea. You will pay the price with noise. Jurgen Haible specified a 4558 op amp. I specify an OPA2227 (or here at twice the price) which is a very, very low noise op amp and I’m really happy with how quiet the GICs are with these. They are expensive however, and (including the I/O PCB) there are 29 op amps in the GIC version. At $4.50 each, that means $130.50 just for the op amps. The 4558 is about 1/10 that cost. It’s your module, you be the judge. Just remember, you have several noise sources in the signal path ahead of the output amps, and the output has a gain of about 270. The more you can reduce the noise in the filter cells, the better. I can’t do much for how the Moog™ output amps works because I don’t want to change it. The Even and Odd output amps use a two op amp cascaded design in an effort to reduce noise.


The original used 2N2926, 2N3391A, 2N2925 transistors. The 2N2925, 2N3391A transistors are still made and can be had at Mouser. The 2N2926 is harder to find, but not anywhere near impossible. Honestly, I can’t tell you why Moog™ used three different types of transistors in the circuit. The 2N2926 is in the input stage. I see no reason why you can use any reasonable low noise, small signal transistor here. The output stage should have high quality, very low noise transistors due to the high gain in the stage. The 2N2926 is available at Cricklewood Electronics and maybe some other places, too.


The PCBs are laid out for a Panasonic PCB pot available at DigiKey and other sources I’m sure. Panasonic # EVU-E2JFK4D54, DigiKey # P3U0503-ND. You will probably have to trim the shaft, I did anyway. I used a Dremel and a cutoff wheel.

You will also need a couple of small’ish panel mount pots for the low pass and high pass attenuators. The hole in the panel design is made for an Alpha pot. If you choose a different one, be sure to check the mounting hole diameter. The top PCB is notched to accommodate these two pots.


There are 14 trimmers. I specify a Bourns 3362X series. This is a single turn trimmer. The foot print will fit many different trimmers, just be sure the trimmer is set up so that when you turn it clockwise, it connects to the right hand pin as seen when looking at the PCB from the rear (see the PCB layout). Some Bourns trimmers will connect to the pin on the left when you look at the trimmer adjustment.


For the panel I laid out, a good 3.5mm or 1/8 inch jack will work. I use the Switchcraft 42A Tini-Jax true 1/8 inch jack. These are switched jacks and they work with 1/8 inch plugs and 3.5 mm plugs. There are a lot of good 3.5mm jacks which will work.


There are no switches.


I provided locations for standard 0.1 inch pitch MTA connectors or you can just solder the wires. I highly recommend you do not solder both end2014-11-08 00.01_04s. On the GIC version, the connectors all mount on the component side and you can use vertical connectors. On the inductor version, due to how I laid out the PCB, you need to mount some of the connectors which would otherwise block access to the trimmers on the solder side. Also, I recommend you use 90 degree connectors on the bottom filter cell PCB and mount them on the solder side as indicated. It’s a tight fit, and you can mount them on the component side, too like I did in this prototype.

What’s important is that you connect the Even to Even, Odd to Odd, and All to All between the filter cell PCBs and the I/O PCB. The order of the cells makes absolutely no difference, so don’t worry about the order of the pins, with one big exception on the GIC version…

Here is a Mouser link to the type of 90 degree connector I used. This is the three pin version. Other links are on the Component Notes page.

TAKE NOTE ON GIC VERSION: The output of the low pass filter cell on the GIC is substantially higher than for the other cells. It’s summing resistor has to be much higher, about 133K as opposed to 33K. You must make sure that the pin connected to the low pass signal connects to a 133k (adjust to your taste) summing resistor on the I/O PCB. You have to do this for the ODD cells and for the ALL cells connector. It’s marked on the I/O PCB, but I forgot to add it to the silkscreen on the middle cell PCB. On the ALL and ODD cell connectors, it’s the pin on the far right, opposite end to the square pad. If you get this wrong, the low pass cell will be extra loud and what ever cell you connected to the 133k will be very, very quiet.


Here is a link to the page with build photos for the inductor version. A link to the GIC version will follow. The pages are also a sub-page to this one in the menu to the left.


I assume you know the basics of soldering. I like to insert the low lying parts first, like resistors, diodes, etc. After these, I install the IC sockets. Next capacitors, transistors, connectors. Use a good solder, either an organic flux, which you should wash regularly, or a no-wash flux.

Take a break every so often, wash off the flux if you are using a flux which required cleaning. Double and triple check orientations, pins, and solder joints.

Power Supply Regulation/Filtering:

The module needs +12 and -6 volts. I use adjustable LM317 and LM337 regulators to convert either +/- 15 volt or +/- 12 volt supplies to the proper voltage levels. Per the data sheet, the regulators should have a 0.1uF ceramic on the input and a 1.0uF Tantalum on the output. I added a 1.0uF Tantalum on the input to help filter a noisy main supply. NOTE: These four filtering capacitors, C503, C504, C506, and C507 can also be 10uf electrolytic capacitors per the data sheets.

You can probably always omit C503 and C506 with no bad effect (the polarized capacitors on the input side).

You can use either +/-15 or +/-12 as a main supply. If you use 15 volts, you will need both regulators. If you use 12 volts, you just need the negative regulator (LM337). Obviously, if your main supply is +12 and -6 volts, you don’t need either regulator. If you omit a regulator do the following:

  1. Put a jumper between the “I” and “O” pads.
  2. Omit the trimmer and the associated fixed resistor. R501 and F502 for positive and R503 and R504 for negative.
  3. Omit one of the polarized capacitors, otherwise you’ll have two capacitors in parallel. This won’t hurt, but it’s not needed. Omit either C503 or C504 for the positive side or C506 or C507 for the negative side.

If you do the power supply first (good idea) remember, you have to have a sufficient load on the regulator to make it regulate, like 10mA. So, if you do the power supply section first, you will also have to connect a temporary resistor between the output and ground sufficient to draw about 10mA (R = E/I or 12/0.01 =  1.2k, or for the negative side 6/0.01 = 600 ohms). Again, just be careful. Don’t hook up the filter cell PCBs on the GIC version or install the output amplifier op amps (especially if you use the expensive ones) until you’ve checked the supply voltages.

TAKE NOTE ON THE INDUCTOR VERSION: There is no power needed on the filter cell PCBs so you don’t need to install a 2 pin power connector on the I/O PCB. However, you should connect the ground wire of the DRIVE connector between PCBs to make sure they all have a ground reference.

TAKE NOTE ON THE GIC VERSION: There are several two pin power connectors which need to be daisy chained between the PCBs. Make sure each PCB has power.



The PCB pots mount all but the I/O PCB to the front panel. The I/O PCB has different holes so you can choose to mount it with spacers to the panel or to the bottom filter cell PCB. You can look at the photos of my build to see what I did. Keep in mind different clearances with all the connectors, etc. as you mount and run wires.

You ask, why the two big holes at the back of the I/O PCB and why the big rubber feet? My rack has a mesh metal bottom and the I/O PCB was very close to the mesh. I didn’t want a short, so I added to two big holes to fit some BUD press in rubber feet just to make sure I didn’t short. You don’t have to use the feet, but a photo of the Mouser bag is included in case you do.

IMPORTANT: The signal level at the summing node for the filter cells is very low and the signal level in the “Drive” signal wires is high. The inductors are also susceptible to stray electromagnetic fields. Run the “Drive” signal wires as far from the summing nodes and inductors as feasible. I drilled holes at the front of the PCBs to help with this. You can also twist the wires which should help. You can also use a small flexible shielded cable (what I did) on the inductor model which will help, but is a bit of a pain.

Coax to PCBNote on using coax:

I used micro unbalanced coax from Redco Audio, Canare GS-4. I sell it (see above), but would suggest you go to Redco.

When I ran the shielded wire, I soldered one end to the PCB and the other end to an MTA connector so I could separate the PCBs if needed. On the end I soldered to the PCB, I soldered the center conductor directly to the PCB. I then soldered a pin pulled out of a male MTA connector to the PCB, made a hook on the pin and then soldered the shield to the hook. The shield will not fit into the hole on the PCB and this seems to make a reliable connection.

On the other end, you will have to solder the shield and center conductor to an appropriate female MTA connector.

The holes in the PCB to run the  “Drive” signal wires will just fit the cable with a connector on the end. Just be patient and careful.

Coax to MTACalibration:

Wow… where to start. In theory, the calibration is very simple. Turn the trimmers clockwise to increase the “Q”. A good starting point is about 9:00 to 10:00.

In practice, you need some way to measure the “Q” of each cell. You need a real-time frequency analyzer (many scopes can do this) and there are tools for a PC from free to very expensive ($1000.00 or more) to do it. Honestly, I did a visual comparison of the frequency response of my design and Yves’ design using an iPad app, SigScope Pro, feeding a white noise source into the filter. I didn’t “calibrate” the filter necessarily, but through some calculations, simulations, and playing with the real circuit, decided setting all the trimmers to about 9:00 to 10:00 worked great. See the photo to the left to see the 5 trimmers on the “middle” PCB, or LP, 175Hz, 500Hz, 1.4kHz, and HP cells and trimmers on the GIC PCB. You can do the same for the inductor version. Remember, CW increases the “Q”, CCW decreases the “Q”.

2015-03-22 20.51.47Quick story. A good friend bought a Ferrari 308. He also owns and hot rods Porsche 914s (hmmm, no relation to this module :) ) he told me about the difference in tune ups between the two. The German Porsche had very precise and clear metrics and instructions for tuning the engine. The Italian Ferrari manual, on the other hand, would say “turn this screw until is sounds right” or “adjust this control until it feels smooth”.

In the spirit of Ferrari, adjust the “Q” to your liking. The original 914 had a “Q” of about 3.7 (thanks to Yves’ research and analysis), so if you want to have the Porsche version, you can do it. If someone has a good way to measure the “Q” of each filter cell, please let me know and I’ll share it here with many thanks.


There is nothing too special. I suggest using connectors on the flying wires from the jacks and pots to the PCB.

The mounting holes and spacing are setup for Alpha 16 mm or 12 mm pots. The jack holes are 0.25 inch in diameter.


1. There are a lot of references for Generalized Impedance Converters and some configurations other than what I used here. If you want some, let me know and I can send you some references. Otherwise an internet search will pull up some interesting information.