The VAXMIDI by Infinite Response

In 2015 a Kickstarter was created for a project called “VaxMIDI” – a MIDI keyboard controller with attractive features, including polyphonic note pressure detection or “aftertouch”, a feature that traditionally has been difficult and expensive to produce, and/or protected by patents.

The second really interesting thing about this project is that it would be shipped in kitset form, to keep costs to a minimum.

The folks behind this project were reputable (read more about Infinite Response and the VAX77 keyboard). Strictly speaking, I did not need another keyboard controller in my studio, but as an engineer/musician, the project intrigued me.

I backed the project at the top tier, in for a 8-octave unit.

After the project reached its funding goal, we had to wait for the kit to be developed and manufactured before eventually receiving our kits. I was in the second wave of beta testers, and early 2017 I came home to a large box on the front doorstep.

I took over the dining room table for several months, assembling and testing each of the four 2-octave units.

image  image

Each base is an extruded aluminum section with a complicated cross section, supplying rails for the keys to swivel and rest on, and a clever mechanism of hammers that are flipped up by each key, and pass a slotted blade through an optical detector.

The aftertouch sensor is a long slot that widens as the blade is pulled through the sensor as the key is depressed more firmly.

There is quite a few places where this can go wrong: The sensor strip (of 22 detectors) can be misaligned; the distance between the hammer rail and the sensor base can be too high or too low; the felt on which the key steel rests can be too high… all these things have to be just right in order to position the slots correctly for the note and pressure detection to work correctly.

Additionally, there are challenges in assembling each base into the chassis and connecting the system circuit boards so that it stays together in a robust and stable configuration. Also, there are bugs in the firmware such that spurious patch changes were being emitted during testing.

Bottom line: My assembled unit is back in its big brown box, in my garage, until such time as we get some updates on how to address these issues. It’s looking more and more as though they will not be forthcoming. I think the company has long burnt through their Kickstarter funding, and combined with family health issues, I don’t foresee this project coming to a good conclusion.

That’s sad, because it shows a lot of promise. Technically, I think it can work, but it is really sensitive to manufacturing tolerances, and the skill and ability of the assembly technician (i.e. me) to overcome them only goes so far.

Re-releases and Resets

You may notice that things have changed color since your last visit. I got tired of the blue and decided a change would be a good idea for the next phase of existence.

Steel Tree 2.0

Okay, the biggest news is that I have created a 2.0 version of the Fruit of the Steel Tree album, as a digital release on BandCamp. The songs are also available to stream on this site - I’ve updated the Production Notes and you can read about the changes there.

Album #2

Album #2 is taking longer than expected, due to reasons that aren’t worth discussing. I have one track 100% completed, and I’m working on a couple of others that have been around for years but have never been finished. I have another long piece in demo form, 70% of it anyway. And finally, another large piece that is in the process of being written and I’m feeling very challenged by it both compositionally and thematically and it’s been hard to make progress on it.

Bottom line: I’m not giving up. There’s too much here that I value to throw in the bin. In order to get Steel Tree finished, I had to get serious and “reset” my activity and treated it like a project, requiring management and discipline.

I need another reset, to not despair of how long things are taking and instead satisfy myself with continual, gradual progress. And also, I’ll blog about it. The studio diary lives again (hopefully).

Pardon the dust

…while we remodel. Actually, it’s exactly the same decoration but placed over a new framework – BlogEngine 3.3, an upgrade from 2.5. Some of the theme styles aren’t working as expected; and some of the plugins aren’t working yet. There may be minor changes.

It has been over a year since my last confession… what’s new, you may ask?

Gone quiet? Not really

OK, I'll admit there has been something of a hiatus but this site has been extra quiet because mostly I'm posting updates on Facebook instead.

Check it out:

That said, I'm expecting to create a buttload of updates here in the near future.

A new arrival : Agile 53437 Bass

So here we have a lousy picture of an Agile Defiant 53437:


 It's not a trick of perspective; those frets really do fan out along the fingerboard. The bass B string has a scale length of 37", while the top G string is 34".



A work-around for Sonar's timeline editing behavior: Slippy Clips

Update Dec 2017:

In a recent update to SONAR, Cakewalk added a feature called "Ripple Editing". It doesn't actually help with the problem described below, which still exists in the 2017.10 release of SONAR.

Original Post:

So I have this project - "Working the Paradigm Shift" - that is mixed 7/8 and 4/4 time signatures. Cakewalk SONAR is largely a great product but I feel that it does let us down in the area of time line editing. Here's an example, and a work-around:

In this section of the project, I have a series of 7/8 bars followed by a 4/4 bar. Whilst practicing the keyboard solo, I realize that one of those 7/8 bars also needs to be 4/4 as well.

However, if I change Bar 24 from 7/8 to 4/4, all the subsequent clips change their musical position:

Notice the clips in bar 25,26,27 etc are now all offset. Clearly this is not what I want. I'm showing the click track here to make it very obvious what the problem is.

I don't know why this should happen. After all, the clips use the default time base of "Musical"; they are not locked to absolute time or anything tricky like that.

OK, so let's UNDO the time signature change, and try locking the clips first:

Nope, that still didn't work. In fact, even though the clips are "locked" to "Musical" time base, they still slip into the previous bar.

I don't think the clips should need to be locked at all; but even if they do, the SONAR documentation seems to indicate that the clips should not be changing their musical M:B:T position, if they are locked:

If the clip is set to the Musical time base, the clip's M:B:T position stays constant, and its Absolute position shifts. If the clip is set to the Absolute time base, the position does not move, but its M:B:T position shifts

So... this could be a bug, or perhaps I am missing something. I'll raise that with Cakewalk Support, but for now, here's the work-around:


1. Ensure that Bar 25 is set to 7/8, over-riding the default/previous bar time signature.

2. At Bar 24, insert 1 measure in the time line: 

Inserting a measure like this does not corrupt the subsequent clips' M:B:T position. This is expected and known behavior.

3. Change the time signature of the new Bar 24 to 4/4:


  • The original bar 24 (now bar 25) is inheriting the 4/4 time signature from new bar 24;
  • The subsequent clips have changed their M:B:T position, as we would expect, given what we observed earlier.

4. Change the time signature of Bar 25 to 6/8:

This should cause the later clips to re-align with the bars, because 7/8 + 7/8  =  4/4 + 6/8:


The "7/8" clip in bar 25 is wrong, but a) we don't care about it, and b) all the other subsequent clips are correctly aligned.

Now we can remove the "extra" 7/8 bar 25 from the time-line. This actually involves a known work-around for another issue, which is that you can't delete a bar from the time-line unless there is even data in the section you want to delete. Fortunately for us, we have a clip we don't need:

5. Re-size the "7/8" clip so that it is nicely situation within the bar boundary.

6. Ensure Snap-To-Beat (whole note is advisable) is enabled

7. Press Ctrl-A to select ALL TRACKS

8. Select the entirety of Bar 25, and select Edit > Delete Special from the menu.

9. Select all options, especially "Delete Hole" and "Shift by Whole Measures":

10. Click OK.

Bar 25 should now be removed, and now we have a Bar 24 in 4/4 and the rest of the project continues as it did before we started editing.

IN MY OPINION, THIS IS TOO DAMN COMPLICATED and should be addressed in a future version of SONAR. (Updated - see above)

Modifying the Roland A-80 to make aftertouch more responsive

The Roland A-80 keyboard controller is one of the few keyboards that supports polyphonic aftertouch. This makes it quite desirable for those who care about such things, and indeed, is one of the reasons I tracked down a unit on ebay.

Out of the box it is a fine 88-key 'board with a respected weighted action (it's not "hammer action" but it has a very nice response all the same), with sophisticated zone and tweakable response curves for things like velocity and the afore-mentioned key pressure.

On the down-side, it weighs a ton, and the aftertouch requires the keyboard technique of the Incredible Hulk. Seriously. In order to register any aftertouch messages emitting from the MIDI out socket, I needed to use force unattainable from mere human fingers.

What Good Luck: Fortunately, this is a known problem. See... there's this resistor on one of the circuit boards called "R43". Originally, it came from the factory with a value of 5.6k ohms, but when customers started complaining to Roland, the tech folks recommended changing it to 10k ohms. Later releases of the A-80 came from the factory with a 10k ohm resistor installed.

What Bad Luck: It seems that 10k is not enough. Other enterprising keyboard techs have Done The Research and found that higher-values give better results. Like, say, 51k ohms.

I knew about this issue when I sought for and purchased the A-80, fully intending to apply some DIY-love to the unit and make the mod. Two years later, I finally got around to it.

The first issue is figuring out how to open it up. I knew from some pictures on the web that the sides and panel opened up on a hinge (I could see the two hinges just by looking at the back panel of the keyboard) but knowing which screws to take out is a challenge. Future generations can save time by knowing that the screws you need to remove are all on the bottom of the keyboard. From the A-80 service notes:

The screw locations (I) are shown above high-lighted in yellow. Once you've removed those 15 screws (5 short, 10 long), the top panel will be loose, so carefully re-position the unit and lift the panel. It will hinge up on the back edge:

On my unit, there was a ground wire preventing the case from opening completely, but be warned! There are other connections to the circuit boards that may take up the tension if you unscrew the ground wire and haven't braced the top panel to prevent it from swinging open. (This is unlikely to happen because on a flat surface, the back panel rests on the table top.)

I had the unit mounted on supports that meant that the table top was actually lower than the underside of the unit - so the panel could free-swing further back from the unit. You're actually going to need to do something similar in order to get enough room to work on the unit.

There's a wide copper sheet linking the circuit board to the main chassis that will take up tension as the panel opens. The way I handled this was to unscrew the three screws attaching it to the chassis, then folding it back over the top panel and taping it down out of the way.

We can now see the circuit board that interests us:

Our resistor, R43, is on the left side. I've highlighted it with a red ellipse. 

There's really no escaping it: in order to replace the resistor, we need to detach this circuit board from the chassis. There are three connections you'll need to remove:

1. The ribbon cable on the top left (see the grey and black connector). This requires some care. It is basically just a slot that clamps down on the end of the ribbon cable. I actually detached it by accident and had a moment's panic when i thought I'd made a mistake. The picture above shows the cable already disconnected (sorry) but it is a simple operation to "open" the clamp by pulling the grey part of the connector away from the black part. MAKE SURE YOU NOTE WHICH WAY THE RIBBON CABLE IS ORIENTED. Then the ribbon cable just falls out. (It's trickier to reconnect it but if you're careful, there isn't much risk of getting it wrong. Just make sure the cable is fully into the slot on both sides before re-clamping.)

2. The other two connectors are highlighted by another red ellipse in the picture above. Much safer to remove but require more force than the ribbon cable. The plugs are grooved in a way that makes it impossible to reconnect them wrong.

There are six screws around the perimeter of the board that attach it to the chassis via separator posts. Two of the screws are holding cable retainers - make sure you remember where they are located before removing them.

Now the board folds back easily, allowing us access to the components from the reverse side:

There's a red arrow in the picture above (somewhat hard to see) that points out the location of R43. Make sure you know which it is! It is not labeled on the back side.

I don't have pictures of the solder removal process or the replacement resistor. If you don't know how to handle that part of the process comfortably, you should not be doing this operation yourself. I'm not going to write a tutorial on circuit repair techniques here.

The actually resistor replacement operation took about 15 minutes, after which I re-assembled the unit and screwed the panel back in place.

Bottom line: The mod works. Replacing R43 with a 51k resistor makes the aftertouch feature work as expected. Mission accomplished!

Special thanks to Plink Floyd ( for the excellent pics and entertaining rant. 

Telling V-Vocal what default parameter values to use

V-Vocal is a monophonic audio editing plugin developed using Roland technology, integrated into the SONAR product. It was packaged with Cakewalk SONAR from about version 5 or 6. The last version that included it was SONAR X2.

V-Vocal has been compared unfavorably to other "vocal editor" software such as the better known Melodyne or Antares Autotune, but it is actually a very capable tool when used judiciously.

V-Vocal still works in SONAR X3, although due to a change in company ownership, it is no longer part of the SONAR installation. However, if you have previous versions of SONAR installed, V-Vocal will be available. In fact, in my opinion, ironically V-Vocal works better in X3 than in any previous version of SONAR.

However, some quirks remain, and one of them is the default values used for several of the editing parameters. I'm always having to remember to adjust the FORMANT parameter in particular, every time the V-Vocal window is launched.

These default values can be changed, though, through some careful editing of an arcane .INI file. All credit for this hack goes to Tom Parker for this post on the SONAR user forum in December 2006.

I'm going to copy the instructions here for quick access, and in case the post is removed from the forum or lost in the cloud:

1. First, make a backup copy of the file C:\Program Files\Cakewalk\Shared Dxi\Vvocal\Skin\VVocal.INI

2. Then edit the original VVocal.INI file

3. Starting at line 130 you'll see a section "Display 8 ( Formant Control )".
   There are 3 values (on 2 lines) that need to be changed.
   Recommended: Change them all from 100 to 0:

   a) Ctrl181 3rd value (Value if you double click in the "PITCH FOLLOW" graphic dial center)
   b) Ctrl181 17th value (Initial value for the graphic dial)
   c) Ctrl183 17th value (The critical Initial value; this will be displayed in the box below the dial)

4. Save the changes, and re-start SONAR.