Vol.2 is next, and most likely will consist of three individual songs rather than a related suite. I’m currently working on the first of these, ensuring that the melody is pitched right for my voice, and that I’m not plagiarizing myself (or indeed, any other artist). At least, not too obviously. This track I’m working on definitely wants to use a chord sequence from “Paradigm Shift” and I’m not sure that I can let it do that so blatantly.
Use Velostat to replace the pressure-sensitive cells under each key.
Change R43 to a variable resistor of about 27k and set it mid-way
The longer, illustrated version
Despite the R43 fix, the touch response has never really been usable. The mechanism Roland used clearly didn’t remain viable over time. The risk of screwing up the key bed completely has always deterred me from experimenting in this area.
And then, there was this thread on Computer Music Guide forum, “Roland A-80 (Aftertouch Repair)”. Back in 2014, noisevoid had the great idea to experiment with alternative pressure-sensitive rubber to see if they could restore touch-sensitivity to the notoriously unresponsive A-80 key bed. Their research identified a product from Zoflex as a possible solution…. and then they were never heard from again on that thread.
But the idea intrigued me, and after Syntaur alerted me that the “upper” key felt strip was back in stock, I decided it was time to go deeper. I placed an order for a sheet of the Zoflex product… and also did more research on pressure-sensitive materials.
One finding in particular looked very interesting: In 2016 Liam Lacey had implemented a “toy keyboard synthesizer” with pressure sensitivity using a product called “Velostat”. Intrigued, I ordered a sheet as an alternative.
Eventually the samples showed up. The Zoflex rubber sheet was about 1mm thick and quite “spongy” and quite expensive for a 10cm square sample, in beige. The Velostat sheet was larger, more like an A4 sheet of paper, and much thinner. And cheaper.
Here’s the plan:
Replace R43 with a variable resistor to allow for sensitivity adjustment
Remove the keys and the upper felt strips
Lift the plastic pressure strips enough to insert samples of alternate product
re-assemble loosely to test response
Replace R43 with a 100k variable PCB mount resistor
There’s no getting around it. Main Board A has to come out for desoldering. I’ve covered this process before. Six screws and several connectors need to be removed.
The odds of an alternative variable-resistance pressure membrane “just working” with the current 56k value seems unlikely. I obtained a small PCB-mount 100k resister and snipped off one of the outer legs. This gives us a replacement for R43 that will let us alter the resistance. I wasn’t sure this step was necessary but it seems prudent.
With that done, I re-assembled Main Board A back into position.
I created this animated GIF showing the mechanical action of a key press:
As you depress a key, the first thing that happens is that the grey rubber dome switch is activated. (You can see it on the right.) It has two contacts in it that closes two circuits in quick succession, indicating Note Velocity.
On the left you can see two felt strips that run under the keys. The black keys rest on the right-hand strip. The white keys rest on both strips. Under the felt strips are circuit contacts for each key press (white on the left, black on the right). Each contact is covered by a small rectangle of material that changes resistance with pressure. So, as you push down on the key, the resistance change is detected and translated to note pressure or “aftertouch”.
Remove the keys
This is where a key removal tool becomes very handy. I’ve actually made a short YouTube vid of the process:
With the keys removed, we can start finding out what is under those felt strips.
Remove the felt
At this point, my recommended procedure diverges from what I actually did. I was cautious and also not sure of what my next step would be. I wanted to remove the felt strip but leave the pressure strip in place. Also, I wasn’t planning on re-using the original felt strip, at least, not all of it. I figured I had a replacement strip that could be used for either the white keys or the black keys. I thought I only needed to keep 50% of the existing felt strips in re-usable condition.
The adhesive between the felt strip and the underlying plastic strip is quite strong and it was an awkward process to remove the felt intact while not also pulling the plastic strip away from the PCB. I really wanted to leave as much undisturbed and original as possible.
With the felt removed, we can see the plastic strips under which the pressure-sensitive resistive pads are mounted:
In hindsight, knowing that I was going to remove the plastic pressure strip eventually anyway, I should have removed the felt+plastic strip as a whole, and then carefully separated the felt from the plastic on a flat surface with more room to move.
Testing Substitute materials
I carefully cut out little rectangles of the Zoflex and Velostat material:
I then pried up the end of one of the plastic pressure strip and inserted the sample rectangles under it so that they covered the two carbon contacts beneath. Keys B7 and C8 were to be the test cases. I’m sorry, I don’t have a picture of this configuration, you’ll have to imagine it.
I then tacked the felt strip back in place temporarily and re-installed the two keys; lifted the key bed chassis back into place and re-connected it (remembering the aftertouch subsystem ribbon cable; easy to forget that one).
As a test patch, I initialized a program on the Novation PEAK and set the mod matrix so that aftertouch controlled the pitch of Osc1, bending it down a full octave at “maximum pressure”.
The Zoflex key was unresponsive.
The Velostat key was EXTREMELY sensitive to pressure.
Discard the Zoflex and experiment more with the Velostat.
WARNING: Turn off power to the A-80 before attempting to change the resistance of R43. Trust me. The A-80 will go berserk if you don’t.
By testing the change in touch response after each adjustment the value of R43, I could tell that reducing the resistance produced a corresponding decrease in sensitivity (a good thing!) A value of ~10k for R43 appeared to be optimal. Of course this will be subjective, your mileage may vary… I would think a varistor of 27k would probably give finer control.
What you shouldn’t do, and don’t need to know (Fast Forward)
I used a crude method of tacking the Velostat rectangles in place
I replaced the plastic pressure strips over the top and tacked them in place
I used the brand new replacement felt strip for the White keys
I assembled all of the key bed, and installed in the chassis
I tested the unit.
The Black keys worked great
The White keys would not trigger reliably – the new felt + velostat was too thick, preventing the rubber domes from actuating.
Also the difference in feel between white and black keys was disturbing.
Removed key bed from chassis, removed all white keys
Removed felt strip, which took the plastic strip with it because of the super-tacky double-sided tape I used
Decided to use the remaining original felt strip for the white keys, and re-do the pressure strip “properly” while I was at it.
Converting the pressure strips
You should apply this process to both black and white keys.
Having peeled back and removed the plastic pressure strips, I carefully cleaned the carbon pads with 91% isopropyl alcohol.
In the most-used keys (the central area), the original pressure pads had left white residue on the carbon pads. Use a light pressure and don’t go overboard, because you can remove carbon from the PCB and you don’t want to do that. Just clean them as much as you can, to encourage uniformity in touch response.
I then cleaned both sides of the plastic pressure strip with the isoprop, to remove as much of the residual sticky as possible.
The next step was to apply the double-sided tape to the resistive side – the side that will be laid over the carbon contacts on the PCB.
Using tools to keep use of greasy fingers to a minimum, I peeled the backing from the tape and placed the rectangles of Velostat over the pressure pads:
Carefully place the pressure strip back on to the PCB, locating it very carefully so that the pressure pads are over the carbon contacts for each key.
Firmly burnish the strip in place to encourage the adhesive.
Place a second layer of tape over the top surface of the plastic pressure strip (in hindsight, I could have done this before placing the Velostat rectangles.)
Peel the backing, and place the original felt strip over the pressure strip.
Re-install the white keys
And that’s it. Wonderfully responsive, polyphonic aftertouch.
The LCD display on my A-80 has always been a bit dim, but it is still usable. I think the backlight had dimmed over time, although I could still hear the whine of the voltage transformer that supplied it with power.
This noise is a known issue, and inspired by The Midi Maniac’s YouTube video, I ordered a replacement display from Gregor Walbeck E-Pianos in Germany. The part arrived in timely fashion, with very good instructions. Rather than re-create the excellent video, I’ll just add a few supplemental steps.
The standard first step for getting into the A-80 is to remove the screws from the underside in order to release the top panel and allow it to lift up and open. I’ve talked about this process before.
Having removed the 15 screws from the underside, but before opening the lid, there is another step that we need to do.
In order to reach the display sub-board, you will need to remove four screws that retain the metal sub-board chassis plate on which Main-A and Main-B pcb’s are mounted.
If you’re like me, you will have opened up the unit and discovered this plate is solidly attached by screws from the back of the unit:
These screws can be removed from the back side of the lid:
There is still one screw internally holding the plate in place, but it can be easily removed when the lid is open. I don’t have a picture of that, but you can see the MIDI Maniac removing it in his video. He doesn’t mention the four screws at the back, at all. They certainly add some stability to the internal circuit boards, but I can also see how they might have been added in a later production run, maybe.
Anyway – if your unit has these screws in the back, you will need to remove them. In my unit, these were two machine screws (outermost) and two self-tappers (innermost). I don’t know why they are different.
Word of warning: I had great difficulty re-inserting these screws. Aligning the mounting plate, then closing the lid in order to screw them in place was a trial. The threaded holes are in the mounting plate itself, not the thin back plate of the lid. There was no way to partially tighten the screws, then align the mounting plate on the points of the screws. I think if I had rotated the unit so that the lid was upright over the lip of the desk, I might have been able to re-fasten the screws while the lid was open. However, I persevered and eventually got it all tightened up.
Then, disconnect CN3 and CN4 from Main-A board. I needed to disconnect them to create some slack. Now, after unscrewing the single retainer screw in the center of the board, you can carefully tilt the mounting plate towards you to gain access to the display sub-board, nestled into its recess.
There are two connectors to remove from the display pcb: Power, and the ribbon cable. The display unit is retained by four screws, one in each corner. The only problematic one is the lower right, which I had to approach from under the bottom of the main boards’ mounting plate. There was only just enough clearance for an awkward operation. Not ideal, but I didn’t want to disassemble more than I absolutely had to.
The replacement display comes with eight plastic washers. I assumed that, due to the slightly thicker profile of the new unit, that these should be placed on either side of the pcb, so that the board was very slightly raised further out from the plexiglass window, and also protected from the head of each screw.
Okay, MIDI Maniac shows this reasonably well in his video, but in words: I inserted the screws on the right side, with the washers in place, loosely so that I could slide the new display board in from the left (the mounting holes in the PCB are actually slots). Then, using double-sided tape to hold the plastic washers on either side of the PCB, I could rotate the display into place, and then insert the screws on the left side and tighten them all up. (Remember to remove the sticky protection label from the top of the new screen before doing this!)
Then, simply attach the ribbon cable to the board. As per instructions, it does not need the power connector. This can be un-threaded back to the PSU board and curled up and zip-tied out of the way.
Now for testing:
Worked first time. Yay. Now the instructions are very clear about how to remove the two components on the PSU board that are no longer required to deliver power to the display. I snipped them off with a pair of nipper pliers. No More High-pitched whine!
So, either I have to pivot to writing music without D above middle C, or I have to repair my Roland A-80 again. The D4 key stopped responding. After a miserable couple of days procrastinating, I opened her up and took a look.
I’ve written about disassembling the A-80 keybed before, and of course the first thing I did was open that post and refresh my memory. Quick recap for first-time readers:
The A-80 is HEAVY and most of that is in the keybed itself;
It uses a rubber dome with two contact switches to trigger a velocity-weighted note event from each key press.
One thing I was sure about: After last time, I wasn’t going to try and lift the entire keyboard off the table, and also I was going to need to have it in a state where I could monitor MIDI output whilst testing key presses.
I pulled the keyboard out and unplugged it, then tilted it onto its back edge (on a soft towel) and removed the necessary 15 screws from the underside.
Right-side up, carefully lifted the lid. At this point I deviated from the previous process – I needed this to be in a playable state whilst examining the inner workings of the keybed. I unscrewed the 12 screws holding the keybed in position in the chassis (shown as II and III in the diagram above – III from inside the chassis, and II from underside. Then I lifted the keybed away from the lip of the chassis and raised it up on two metal blocks (about an inch high) on either side. This was sufficient to be able to remove the white keys whilst leaving all the connectors in place.
The key removal tool made quick work of removing the keys around the problematic “D”:
Not the sharpest knife in the draw, am I. When I saw that “!” in black sharpie on the base plate, I remembered from last time: D4 was the key in which I accidently damaged the rubber dome. I left it unrepaired last time, because it sat correctly, the invisible slit in the side of the dome didn’t appear to affect the key response. (And, indeed, it worked fine for 6 months.) Yet, here we are.
This was a great relief, to know that there was an actual root cause to the failure of D4, and not just something random breaking.
So, what to do? First, I tried repairing the slit (which was substantially larger than I remember) with superglue.
THIS WAS A BAD IDEA.
A) it doesn’t seem to bond the rubber dome together, and B) very great danger of getting glue on the conductive contacts under the dome or on the PCB. NEVER TRY THIS.
At this point, if I had no other recourse, I was prepared to re-purpose the rubber dome from the lowest or highest note, sacrificing it and yielding a functional 87-note keybed.
I had one more thing to try, however: Replacement rubber domes from www.bustedgear.com. I haven’t found any genuine article for the Roland A-80 in stock anywhere, but these replacement parts KCS10 are apparently good for the A-50 and they sure looked near-as-dammit identical. I’d ordered a couple of sets several months ago, on a hunch that I might need them in the future. THE FUTURE IS NOW
Using an exacto blade, I carefully removed the damaged D4 dome. The domes are installed as a strip, held in place by double-sided tape (true!) and the white key domes are subtly different from the black keys. Upon close examination of the replacement, I could see they weren’t an exact match with either black or white key dome, but I think they would be close enough. Anyway, the bad dome had to come out, even if I was to replace it with a dome from elsewhere in the keybed. It was worth trying these KCS10 domes, just to see.
Trimming the replacement dome to fit was finicky but the exacto blade earned its name. The replacement dome is slightly lower (not easy to see from this angle):
I carefully installed the D4 making sure not to dislodge the new dome, just sitting in place with the slightly tacky and very dusty tape remaining on the PCB.
Connect mains power and MIDI out, open Cakewalk with an instance of Pianoteq responding to notes, and HOT DAMN if it didn’t JUST WORK. Perfect? Pretty close. It does seem to respond a little “hotter” in terms of velocity sensitivity, which makes sense if the new dome is slightly lower/closer to the PCB. But you know what? The velocity sensitivity across the whole keybed has some variation in it, and this is DEFINITELY playable. It’s not wrong; it’s just slightly different. Not really noticeable unless I listen carefully.
Let me be more clear: I can forsee a time when I a) experiment with replacing a Black key dome, and if that works okay, then b) replacing the whole 7+octaves. That would give the A-80 a usable and consistent velocity response.
Okay, um, let’s take the key out and try and attach that dome in place a little more firmly. I used two tiny pieces of fresh double-sided tape and I guess it made a slight difference. Probably not a great idea, if one of the strips comes loose, it could work its way up or down to cover the contacts under the dome, and then this key will stop responding again. At least I will know how to fix it.
All keys back in place, keybed restored to original position, all screwed together, and hey we’re good for another take.
There’s more than one way to skin a cat, and other folks have talked about this before. The more, the merrier. Here’s how I constructed a project in Cakewalk to re-master a commercial CD.
I have a commercial CD release from 1986 that sounds horrible for several reasons: it was the Age of Digital Reverb and there’s too much of it. Also: it was the Age of Digital Recording, and CDs were new, and when Treble wasn’t something you required analog psycho-trickery to extract from magnetic tape. It’s a very shrill mix. I am fond of the actual music, though.
It was “remastered” back in 2001 when the Loudness Wars were in full swing, and the results were… over-compressed and also shrill. What to do? I still have the original CD release from 1986, fortunately.
Because this is a “re-master” and not a “master”, I’m going to treat all tracks the same way, using a common FX stack. I also want to be able easily to switch back and forth between “effected” and “un-effected” versions, for comparison purposes.
When evaluating quality changes of audio, it is important to match effective loudness between the two. This is because our ears think “louder” = “better”. And that’s a problem, because one of the things I will be doing is raising the overall level of the music. The original CD from 1986 is mastered at a fairly low average level.
So, the plan is to temporarily reduce the volume of the mastering effect chain during the tweaking so that the measured loudness is the same. This will require obtaining a loudness measurement from both signal paths.
I open a new empty project, and import the audio tracks from the CD. Cakewalk will create a separate audio track for each of the CD tracks.
Next, I create four stereo busses. The first is named SPLITTER, and I need to set all of the imported audio tracks to direct their output to this bus:
Press Ctrl-A to select all tracks. Then hold down the Ctrl key, and set the Output of one of the tracks to SPLITTER. All highlight tracks will set their Output the same way. (The Ctrl- key creates a “quick group” documented here.)
The other three buses are named MASTER FX, MASTERED, and ORIGINAL.
On the SPLITTER bus, I create two Sends: One to MASTER FX, and the other to ORIGINAL. (Both sends’ levels are left at the default of 0 db.) I set the Output of SPLITTER to “None”.
I set the Output of MASTER FX to the MASTERED bus.
Now I have two duplicate signal paths. Both MASTERED and ORIGINAL continue to direct their Output to the default, a hardware output (called “12” on my system).
The only drawback to this configuration is that Cakewalk will warn us upon opening the project, that SPLITTER is identified as a “Silent Bus”. I prefer the symmetry of the two Sends, rather than using SPLITTER’s own Output, so I pay the price of dismissing this warning dialog. (Feel free to adapt as necessary.)
The next step is obvious: we add some FX plug-ins to the MASTER FX FX Bin. My tools of choice are from MeldaProduction:
You probably have your own favorites. I’ll use these to apply a tilt to the frequency spectrum, taming the shrillness and up-ing the bass; and also some tasteful multi-band compression.
Speaking of LUFS… My Loudness Meter plugin of choice is – you guessed it – also from MeldaProduction: MLoudnessAnalyzer.
I put an instance of the Loudness meter in each of MASTERED and ORIGINAL, configured identically.
More detail: I configure the MASTERED instance for a “target” of -12 LUFS. I’ll use this as a reference as I mess around with the EQ and Multiband limiting, aiming to raise the overall level and apply a smidgeon of peak reduction (aiming for a still-conservative -12 LUFS).  I set the ORIGINAL instance identically so that I can match loudness-levels between the two, using the loudness metering.
Before we start playback, we need to set Exclusive Solo on (optional, but handy) so as to easily listen to one track at a time, switching between them with the Solo button:
And, just as important: Mute the ORIGINAL bus. Don’t forget to do this!
For convenience, I like to use a control group to link the Mute buttons of the MASTERED and ORIGINAL busses, letting me switch between them with one click:
Right-click on one of the Mute buttons and select Group > New…
Select a color for Group 1 (say)
Right-click on the other Mute button and select Group > Group 1
The final configuration looks like this:
You may notice the little colored dot on the Mute buttons, indicating they work together as a group.
Now I can take some time to fine-tune the mastering effect chain as described above. When I’m more-or-less happy with the results, I open both Loudness meters and watch the playback. I can see – as expected – that the MASTERED path is adding about +3 db of raw volume boost. Which is fine – but I also want to know if I’ve actually made the sound worse than the original. For that, I need to use my ears, and to remove the “louder-equals-better” bias, I need to reduce the level of the Mastered signal, so that the Loudness meters show the same values.
We do this using the OutptVlm control on the MASTER FX bus, reducing the volume to match. (See above, it’s temporarily set to -2.8 for the comparison review). 
Now I can start playback, and switch easily between the two signal paths by clicking on either of the two buses Mute buttons (they’ll swap in one click).
Yup – my version sounds better than the original.
Now, in order to export each track with the mastered fx in play, we need to ensure that we:
Switch to the MASTERED output;
Put the MASTER FX Output Volume control back to 0db;
Solo each track
Select the clip (to get the right timeline range) and Export to WAV or FLAC or whatever.
 I’ve re-calibrated, I don’t think -12 is conservative. But it’s an okay choice in this case, I stand by that.
 Alternatively, we could instead increase the OutptVlm on the ORIGINAL bus to match the MASTERED level. This has some advantages in that we don’t need to make any changes before the final Export. On the other hand, it is usually a good rule of thumb to “cut, not boost” to avoid unexpected overloading, so I’ll let the instructions stand as-is.
It’s past time for an update on Studio activity for album#3, I think. I haven’t settled on a final name, yet, but the working title of the album is “Circles”.
As described in the last diary update, I had lyrics for five or so songs, and some melodic ideas. Lyric notebook in hand, I took a couple of long walks up the local hill to find some solitude away from the studio equipment and distractions, and tried singing the verses and choruses at various tempos and pitches, until I found something comfortable that “worked”. Then I recorded them into my iPhone voice memo recorder, for reference. In the process I re-wrote a lot of the lyrics to fit the “right” meter.
Back in the studio, I listened to the results and translated them into a simple piano arrangement with two or three tracks for chords, bass, and some melody. I find a Rhodes electric piano patch is best for this process.
(Okay, I realize as I review what I just wrote, that this is not particularly insightful. It’s not Rocket Science, and represents a typical songwriting process. I’m not special, I get that.)
For the last two months I’ve been working solely on one specific track, taking it from the linear “Idea Bucket” project and building it out to a full sixteen minute epic, with four or five movements. The piece is called “Annulus” and it is based on a 2017 trip we took with a friend to Oregon, to view the total solar eclipse.
At the risk of getting way too conceptual, the five movements map to the five phases of a total eclipse: First and Second contact; Totality; followed by Third and Fourth contact, as the Moon’s disc passes and overlaps the Sun.
The corresponding musical sections are: Departure I; Arrival; Aperture; Circles; and Departure II.
These sections were all sketched out and re-arranged in the form of pure MIDI tracks of bass, piano, and percussion. Using MIDI instead of trying to record audio allowed me to play around with tempo and arrangements, and even some transposition, as I prepared to record the vocals.
Happy with the arrangement, tempo, and pitch, I recorded all the vocals, lead and harmony. Done!
(This is so different to how I used to create. I would spend hours and hours on the music before ever getting near the microphone to record vocals, only to discover that much of what I’d done had to be removed, or replaced, or rearranged to make space for the vocal lines. I’m going to try and avoid that way of working, in the future.)
To be honest, prior to recording the vocals, I had spent a lot of time with various piano patches and a LOT of reverb, working on some of the instrumental sections.
With the vocals done, I practiced and recorded fretless bass and chapman stick. Then, having built up both a familiarity with the music and also some callouses on my fingers, I wiped what I’d recorded and re-recorded the bass parts, better.
Next, drums. I find that bass followed by drums is best, because I’m much more likely to develop an interesting rhythm part as the bass part is refined, and then if I’ve already invested time in the drum tracks, I have to go back and re-do. So, bass comes first.
And now we’re caught up, because I “finished” the drum tracks this weekend. I solo’d the Bass and Drum buses and it all sounds pretty locked up.
I have a tendency to overplay (really?!?!) and so I know there will still be some refinement required, mostly removing unnecessary rhythm parts. But it’s a good place to stop for now.
Just for fun I took the Cakewalk project and shrunk it down to show the full extent of the project in full-screen:
There’s no guitars, and not much in the way of synths or other fairy-dust adornments. Just Vocals, Bass, Drums, and half-finished Piano tracks.
I’m thinking seriously about breaking up that 16 minute monolith into separate projects, for safety and simplicity and speed. I’m not sure where the breaks go – there are short instrumental bridges between each of the major slabs of composition and so I’d have to allocate them appropriately. Some thought required.
I think I’ll move on to recording guitars. Some of music currently realized as piano notes will have to come out to make room. There are some parts I hear as “hammond organ” in my head, so that’s coming up at some point. And definitely some more synth-y fairy dust.
Melodic progressive rock songs and instrumental interludes, a touch of 70’s influence but a product of the dystopian Now.
“Very smooth, hi-tech sounding delivery…” – Chris Jemmett, alt.music.yes
“This guy is awesome.” – Dazed, on the Carvin Forum.
“..on a rare occasion you just have to conclude that the prog world should be feasting upon the birth of a new and promising act. That’s exactly the case with this [first] album.”
– Theo Verstrael, DPRP.net
“I find this new album attractive, [..] slightly less appealing than the 2014 debut. But as that is often the case with great artists, let it not distract you from trying this fine album. Especially those that are interested in bands that play varied, cleverly made, well played and sung [..], this might just be your cup of tea.”
– Theo Verstrael, DPRP.net