At last! The pretty good MIDI guitar

I have this thing for making a guitar sound like something different, like a piano, or a marimba. Those instruments produce short, “plucky” sounds like a guitar, so why not? One way to achieve this is to output MIDI instructions from a guitar, which has been attempted with rather limited success ever since MIDI was invented. Well, I’m happy to report that the wait might be over for the majority of us guitar (or ukulele) hacks. It’s a gizmo called “TriplePlay Connect,” and you can get it for under two hundred bucks. In this article, I make a pretty thorough review of it, including how to connect it for the quickest best results.

But first, a disclaimer. I did request a review unit from the manufacturer, Fishman Inc., but it seems my blog is too small-potatoes to even deserve a loaner. So I ordered one on my own cash (open-box, for $169 plus tax, shipped rather slowly from Tennessee), thinking that I might be returning it anyway. I won’t bug you with silly box opening videos  like most product videos on YouTube, but suffice it to say that it’s worth to buy open-box because you get a hefty $50 rebate just because the cardboard box has a ding in a corner.

The Connect is the younger sibling of the regular Fishman TriplePlay pickup, which adds wireless connection and a couple buttons, sells for 400 bucks, but otherwise is the same thing. It seems Fishman has finally realized that some inferior products are getting a lot of sales because of their cheaper price, and they could have that market. I’m talking about the likes of Jamstik, Artiphon, YouRock guitar, Jammy, and so forth. All of the products I just mentioned produce MIDI by sensing the position of your fingers on a (simulated) guitar neck, plus an additional sensor for the (simulated) string plucks. The TriplePlay differs in that it senses actual sounds made by actual strings, which then it converts into MIDI instructions.

And here’s what this allows it to do: pull-ups, hammers-on, slides, bends, in other words, everything one does on a guitar that would be pretty impossible on a keyboard. Sounds silly, but this is what allows the TriplePlay to pass the “suspension of disbelief” test that the other devices fail rather miserably. This is a rather subjective test where you start playing and get the feeling that the guitar strings have changed, because the sounds made by the synthesizer receiving the MIDI coming out of the guitar seem to come directly from the guitar. I’ll try to disassemble this into a more objective set of criteria, namely:

1. That things like pull-ups, hammers-on, slides, and bends are possible over MIDI, and reproduce the pitch changes that actual strings would have.
2. Low latency, defined as the time between the moment a string is plucked, and the moment the sound is heard.
3. Good dynamic fidelity, or ability to sound soft or loud depending on how hard you pluck or strum.
4. Tactile fidelity: the “strings” have to feel like strings under tension, or your connection with the guitar is endangered.
5. And of course, tracking, which means exact correspondence between the note you think you’re playing and what actually comes out.

There are other products out there that take a similar approach as the TriplePlay. The Sonuus G2M and its Chinese ripoffs take the raw output from a guitar jack and tries to figure out what it’s made of, with okay results if only one string is played, which turn rather disastrous if it’s a chord. The MIDI Guitar software does pretty much the same thing but works better because of the greater computational power of the processor involved in the detection; this one may be usable for chords unless they are fast (as they tend to be with a ukulele). The Roland GK-3 pickup is very similar to the TriplePlay in that each string has its own coil rather than mixing all in a single pickup. This in effect de-constructs chords so they aren’t any harder to detect than individual strings. But the GK-3 does pitch detection on a separate (and expensive) electronic box, and has a longer latency as reported by users. I explained in another article why this approach to tracking is challenging. In essence, a processor has to collect a whole period of oscillation of the string (better, a few periods), store it, and then analyze it in order to determine its frequency, which is then quickly converted into MIDI codes. All steps before the last take significant time and lead to latency.

To avoid latency, which would be worsened by their Bluetooth connections, other products detect finger position and pluck intensity directly, even before they cause a vibration. In fact, those products tend to not make any recognizable sound at all, loud or soft. Thus, the Jamstik and the Jammy have finger sensors on the neck. The YouRock guitar and the Artiphone replace the strings with raised rubber ridges with membrane switches below. The Yamaha EZ-AG and EZ-EG MIDI guitars have buttons instead of strings, which you press with your fingers. This is the approach taken by the Rock Band 3 controller (featured in this article). The Squier version of the Rock Band 3 controller has divided conductive frets so that the contact of a string with a given fret is like closing a switch. All of these devices sense plucks from the initial motion of the “strings” (real or simulated) under the right hand, which dies quickly rather than sustain for a while as in a real guitar. This leads to an important difference in the MIDI output: there’s no such thing as a meaningful “note-off” message. If you pluck an open string and the MIDI is sent to an instrument with constant sustain, like an organ or a wind instrument, the note will continue ringing until you press that string against a fret, which typically releases a “note-off” message with the same value as the previous “note-on” produced by that string. Not hard to adapt to, but it does harm that “suspension of disbelief” I mentioned earlier.

But where these fail the worst is on the right-hand dynamics. Simply put, it is hard to distinguish a soft muffled “chug” from a hard muffled “chug.” You do get some variation in volume, but it’s typically hard to control and requires unnatural right-hand technique. Forget about pianissimos. Forget about hammers-on, pull-ups, and slides; those will trigger the note-off message rather than change the pitch. Bends, when supported (it takes special force detection hardware), are sent as “pitch-bend” MIDI messages that have trouble sounding natural. But latency tends to be short because there’s very little processing involved: at most some determination of chug “volume,” plus back-and-forth wireless conversion (some Jamstik models, for instance, are reported to have serious problems on this account).

Having ripped its competition to shreds, I pass on to talking about the TriplePlay Connect. There are other articles that talk about what’s in the box, and how much difficulty is involved in getting it to run. I will only say that initial setup is rather critical: you must get the individual string coils at the right distance from the strings, and the shims in the box plus the couple screws you fiddle with don’t quite do it. In the end I had to adjust string height on the guitar bridge itself, but that did the trick. Why is this important? Because the TriplePlay has a delicate sensitivity to string volume, which goes out the window if some strings are “louder” than others because they are physically closer to the pickups. The included “TPConnect” software does include a utility to adjust string sensitivity in software, and it stays with the device even when not connected to this software. More on this later. Mounting the processor on the guitar is a snap, and the cable is, thank God, as long as your main guitar cable is likely to be. I never missed not having a wireless capability on the Connect (which the regular TriplePlay has).

The first thing you want to do is update the firmware, which is done through the special, and free, TPConnect software. There is a version for iPad, and also for Windows and MacOS. The iPad software contains a few sounds, which nonetheless will take 2.5GB of storage (beats me why they need so much room), but the free Windows and Mac versions don’t include any sounds. You can upgrade for 30 bucks, though, and get the sounds and other software. I had trouble updating from the iPad app, which doesn’t give you much feedback as to whether the update has been successfully concluded.

I found the TPConnect software to be nothing to write home about. Sure, you can split the fretboard and trigger loops, but what I want to do is play guitar. I might layer two sounds if they were interesting but, frankly, the selection was too limited to get me going (and 2.5 GB for just that!). Once you load the TPConnect app, the device is put into a mode where each string sends notes through its own MIDI channel, 1 through 6 (plus again 11 through 16 for instruments on “split” strings, which I didn’t use), and it stays this way even if you close out the app, until you reboot the Connect, which is done by unplugging it from the iPad and plugging it back. In this mode, bends can be sent as pitch-wheel MIDI messages, which the receiving synth will use properly if it’s set to a pitch bend range of 12 semitones (not all synths can do this). Apparently there is a way to change the pitch bend range of the device, but it’s undocumented and I couldn’t get it to work. Similarly, it seems there should be a button combination to switch between the default “trig” mode, which sends bent notes as new notes rather than a pitch bend, and the more natural “auto” mode (this is possible in the wireless TriplePlay pickup), but it’s also undocumented and I couldn’t find it. If you don’t load the TPConnect app, the device stays in “basic” mode and sends all MIDI messages through channel 1 (plus channel 2 for pedal effects) and no pitch bend messages, unless you boot it in “hardware mode,” described below.

And it’s in basic mode using channel 1, without the TPConnect app, where I finally got the much-desired suspension of disbelief, running third-party softsynths. The Connect’s volume wheel was recognized as such by a majority of my synths, and those that did not recognize it at first did so after a short “MIDI learn” setup on each app. The volume wheel is especially handy when you are mixing the original guitar sound with the synthesized sound. The Connect’s Up and Down buttons changed presets on a number of synths right out. I presume they can be reassigned to other functions as well. iPad synths where the Connect worked particularly well: Roland SoundCanvas, bs-16i, Addictive (not Pro), Animoog, Caustic, NLog Pro, Nave, Sunrizer, GarageBand, ThumbJam, Thor, Magellan, Yamaha FM essentials, all AudioKit synths, and more (I know, I have too many synths). My favorite mix typically began with a synth, to which I added a bit of dry guitar output (possibly with its own effects and modeling provided by the amp), with the iPad sound coming into the amp via the Auxiliary input. My amp is a Peavey Vypyr VIP 3, but I figure most amps will do just fine.

I did measure the latency by feeding both the dry guitar output and the MIDI into two separate tracks of Cubasis, and recording both while playing single open strings. The MIDI track drove Cubasis’s own sampler engine, for reduced latency, but I don’t think this mattered because what was recorded was the MIDI events, not the synthesized sound. Some strings appeared to make the conversion slightly faster than others, but overall the delay between the beginning of the dry sound and the corresponding MIDI event was a barely noticeable 13 milliseconds at worst, sometimes as low as 7 milliseconds. It did get noticed sometimes, though, but you had to try rather hard. Select a synth with a fast, plucky attack, and mix it in equal parts with dry guitar sound. When you do this you’ll hear something like a “slap” pedal effect on a guitar. The latency was unnoticeable in terms on playing, though fast strumming often resulted in dropped MIDI notes, which you might want to mask with some dry guitar, as described above.

The Connect keeps sensing vibration amplitude after the initial pluck, and it sends a note-off MIDI message when the volume becomes barely noticeable. It won’t update note volume as string vibration decays, however, so the sound stays the same until vibration stops altogether. Bear this in mind when playing winds or organs (which don’t decay naturally, anyway).

And then, you can boot it in “hardware mode” by holding the “up” button while you plug it into the iPad (or the PC or Mac, if you are using those). In hardware mode, the up and down buttons cycle through “hardware patches” which you can set up so MIDI instructions are sent over separate channels for each string, and pitch bend is supported. All “patches” are saved in internal memory in order to make this possible. Unfortunately, the iPad app does not allow you to edit the “hardware patches,” only the PC/Mac app does. Since the first 127 default hardware patches are of the “poly” type (Fishman’s lingo for using channel 1 for all strings), it’s pretty hard to make this feature usable for realistic string bends. You’ll have to download the PC/Mac app from this page, load the first hardware patch, double click it, and edit it so it is “mono” (one MIDI channel per string) rather than “poly,” then save it. Like the string calibration, hardware patches are saved in flash memory within the Connect, so they are usable as soon as you plug it in while holding the up button. Be careful here: if you hold the down button instead, you’ll reset the calibration and who knows what else. Other information on hardware mode can be found in this post from Shawn Bulen including file 1, file 2, and file 3 (actually, for the wireless device, but most things work with the Connect as well).

Documentation is nonexistent, whether in the app, the box that the hardware comes in, or the Fishman website, so users are forced to try and repeat until they figure out how this whole thing is supposed to work. This is a curious omission for a piece of hardware that is otherwise so polished and capable. Add to this that a majority of the promised software instruments are supposed to be purchased via IAP from the TPConnect app, but they they don’t exist when you click the Buy button. I am left with the impression that the instruments are already there, taking a big chunk of the humongous storage footprint of the TCPConnect app (if not, why is it so large?), but I’m not able to access them.

Bottom line: hardware score is an A, iPad software is a C. You really want to stay away from the TPConnect app after you use it once to calibrate the pickup. The good news is that third-party software synths accept the MIDi instructions send by the hardware with nary a hitch, so you can ignore the Fishman software and be a happy camper. The Windows/Mac software is also a lot better, and I made a review of it in this other article, but then you have to carry a computer to your gig, when you could be using your iPhone. Or you can use it to edit the first hardware patch, so you can have realistic bends for each string by booting in hardware mode, and then use whatever is on your iPad (or iPhone, for that matter).

So now I have a “pretty good” MIDI guitar (it would have been awesome if the iPad software had been better), and the total budget was $170 for the Triple Play Connect, plus $45 for the guitar (a used a mini Strat I picked up on Craigslist). I already had the iPad and the software synths, so no extra cost for that. Compare that to the upcoming Jamstik Studio MIDI Guitar, which will sell for $999 ($749 for the initial launch), which is also based on TriplePlay technology and, arguably, will perform very much like my rig (I promise a full review if I get my paws on one). I like my “pretty good” guitar so much that I’m going to sell my RockBand 3 Squier guitar, which is not going to get much use in the foreseeable future.