2022, Chapter 17
Music 1, Analyzers 0
“Hey Dave, you know the new Modi Multibit 2 has an intermittent distortion problem, right?” I told Dave last March. I figured we’d have tons of time to get it taken care of while we waited for Analog Devices DSPs.
“What distortion?” Dave asked.
“Oh, you know, the intermittent thing, it’s like one channel gets garbled every once in a while.”
Dave shook his head, looking confused.
I pressed. “Dave, Cameron’s heard it. Evan’s heard it. It doesn’t happen all the time, but it happens enough that we should have a look.”
Dave shrugged and nodded and mumbled something, and I figured that was it.
And yep, in a couple of weeks, there was a rev to the PC board and a handful of new prototypes. One I kept, and it seemed to work fine.
One went to Evan, and he immediately kicked it back. “It does the same thing.”
Dave looked even more puzzled.
But Dave did as Dave does, and a new prototype came back in another couple of weeks, with another couple of boards. Those went to Corpus, where one went in my office but I kinda forgot about it, because by that time it was looking like November for the DSPs we needed to make the product, and one went to Evan, our resident breaker-in-chief, who said something vague about it being “better,” and in the ensuing craziness with shows, parts shortages, and new product intros, everyone kinda let it go.
And continued letting it go, as it became clear that Analog Devices would miss their November shipment date too.
And then Alex announced,
Hey we got enough DSPs to do a run of Modi Multibit 2 in December!
Aaaand that’s when Evan said, “Are we gonna do anything about the intermittent one-channel garble problem?”
That’s when Alex and me and Dave and a whole bunch of other people went, “What about what problem?” Because that was handled, wasn't it? We'd gotten new boards, hadn't we? Dave had fixed it, right? We were 100%, surely?
But Evan persisted. And Zach, our other lead Texas tech, said, “Yeah, it’s pretty easy to get it to do it on input or sample rate changes.”
I frowned. But I went home, got mine, plugged it in, and…
yep crap they are right.
This meant that Modi Multibit 2 went from a product that was just missing parts…to one that needed to be fixed before we could sell it. So I went back to Dave. “Hey Dave, you know, we really need to fix that intermittent problem on Modi Multibit 2.”
Dave, perfectly deadpan, said, “What intermittent problem?”
“The old one. The garble problem.”
Dave just looked more confused.
“You know, what you worked on at the beginning of the year.”
Still more confusion.
I sighed. “Come on, Dave, saying it doesn’t exist won’t make it go away.”
“It’s never happened on mine,” Dave said.
“It’s dead easy to make it happen,” I said. “Just run a playlist with different sample rates, or change the input a lot real fast.”
Dave shook his head. “I’ve tried that. It doesn’t do it.”
Argh. “Dave, just get with Cameron. He can show you. Then, please, can we kill this thing? Hopefully with firmware?”
Dave nodded. “It’s probably a DSP thing, I already did the board tweaks.”
So Dave went away…
…and came back, shortly, saying, “I can’t get it to do it.”
Yeah. This is my life.
“Dave. Everyone is having this problem. We told you how to do it. Did you go see Cameron?”
“No, but—”
“Go see Cameron and Asa. They can make it happen. See what they’re doing.”
And Dave went away again. This time for a while. I hoped the “while” meant something good was coming. I was in Texas, with two techs who were very eager for a solution, and Dave was in California, and even with our weekly regroups, sometimes it’s better to be in-person.
Finally, Dave came back. This time, with just a link to new Unison USB and DSP firmware.
“Well, that fixed it,” Evan said, still looking troubled.
“But what?”
“But now it pops on sample rate changes.”
Arrrrrgggghhhhhhhhhh. We were getting close to the production deadline. I was starting to get nervous about that one. “I’ll go back to Dave.”
Evan nodded, still with an odd expression on his face.
“What’s up?” I asked.
“Did you ever wonder why the problem only happened for us, but not Dave?” he asked.
“Do I want to know?”
Evan nodded. “It’s actually pretty funny. You see, the problem doesn’t happen on sine waves. Only on music.”
I stopped dead. “Wait—”
Evan laughed, seeing my horrified expression. He waited for me to put it all together.
“—so you’re saying Dave was testing only with single tones. Not music.”
“Exactly,” Evan said.
Holy crap, I thought.
A problem you could only catch by listening to music, not by using the analyzer. A weirdo glitch that somehow only triggered with complex signals.
That was, like, the most bizarre thing, ever.
And, thinking about it, it was really mind-blowing. Here's a product issue that doesn't show up with sine waves, but does happen with music. Something you'd never catch if you didn't listen to it—and, specifically, listen to it in a modern audiophile way, with a playlist of mixed sample rates, playing music. If you did automated testing only you never would have caught it! Suddenly I was very happy we listen to everything before it's packed and shipped.
Long story short, Dave fixed the popping (muting is way easier than chasing down one of the weirdest issues we’ve ever found in a product) and Modi Multibit 2 made its way to production. And all the fixes ended up just being firmware. No hardware changes were necessary.
So why was the glitch happening? It was a timing problem with very weird, parallel-input multibit DAC we’re using. Why was it only happening on music? Weeeeeeellllll…that we still don’t know 100%. But after extensive testing, the bug appears to really have been squashed.
And, if you’re keeping score:
Test Tones: 0
Music: 1
Here’s to Modi Multibit 2! I hope you enjoy it as much as I do.
