A4 teardown (and infrared pics)

Toumal · 2013-09-28T12:23:44.000Z

Today, for your entertainment, the Elektron Analog Four will shamelessly expose itself. Or rather, I took it apart and snapped naughty, naughty photos, even some thermal pics for good measure.

So anyway, I recently had to take my Access Virus TI keyboard apart for troubleshooting, and in the process I took some tasty thermal images which you can find over at The Virus TI Forums.
The bottom line was that the Virus TI was mostly a clean build, with few bodge wires i.e. they had to manually fix some problems in the PCB layout. Not the cleanest thing, but it usually happens with the initial runs of a new product. Other than that I found that the Virus TI was very well made, even though it did have some minor thermal issues, i.e. some parts are clearly operating rather close to their specification limits and drain power even while the TI is “soft-off”.

When I took apart the Analog Four I was ready for anything: Huge coils, discrete electronics, frankly I was expecting some frankenschematic with huge analog components and only one tiny little chip driving the display and doing the sequencer and stuff. Boy was I wrong.

This is an amazingly well built machine. It’s easily one of the best examples of electronic product engineering I’ve seen. First of all, the upper half comes loose with just 6 screws, and lifts up easily. The connectors are sturdy and won’t come apart on their own, and it’s just two ribbon cables really. The upper half contains the display buttons etc and not much else, so I didn’t bother disassembling that. Also… you know, it’s expensive and I wasn’t really interested to see tons of buttons on a pcb. But one nifty detail…

That’s right, space invaders. Oh and also, extra airvents. Just in case something gets hot, which never happens because this machine is just too cool.
Right then. Off to the main course:

A Spartan 3A FPGA, costs about $9, contains 50000 gates and over 1500 logic cells. This is very likely controlling the analog circuitry, doing stuff like setting parameters and the likes.

There we go: A Coldfire MCF54415 32 bit MCU. 250MHz, 2.1MIPS, 64k internal SRAM, and a ton of other stuff: integrated USB host controller, dual 10/100 ethernet, internal hardware acceleration for various encryption and hashing algorithms, SD card controller, SmartCard controller, various DACs, lots of timers and a ton of programmable IO pins. I’d say it’s safe to say that this is the main processor. And it’s a beast. Too bad about the ethernet support though, would’ve been cool to have the A4 have some sort of internet connectivity… maybe a web based UI? Would be awesome!

Next to that we have a Micron 256 SDRam chip (2WG22) and a Toshiba flash memory chip. My guess is it’s a 2GB chip but I couldn’t find the datasheet to verify.
On the other side we have this…

A 16MBit NVRAM chip from Everspin, based on MRAM (Magnetoresistive random access memory) technology, with a guaranteed data validity for > 20 Years. This is about as good of a long-term storage as you can get, beats the hell out of the usual battery-backup solutions you find in classic synths. Oh and right next to that is ANOTHER flash memory, this time a 64MBit chip (M29W640FB) from STMicroelectronics.

This is the largest part on the board. A giant ferrite coil from Coiltronics. Most likely not used for synthesis but for internal power regulation, a theory that’s further solidified by the presence of a LT3579 DC/DC converter and a few diodes. Also these parts are on the opposite side of the board, as far away from the analog sound circuitry as possible. Which brings me to…

HC373 is a D-type latch, there’s some additional logic circuitry there, nothing to write home about. Hmm, where’s the analog stuff?

Aha, Coolaudio V2164M Voltage controlled amplifier (2 per voice), T084 opamps, L339 Comparator, couldn’t find anything about the HJ4051. Your guess is as good as mine. Also on the board is one single Cirrus Logic CS42436 6-in 6-out ADC/DAC.

So yeah, not a lot of exciting stuff. The four voices are clearly visible on the PCB, but it’s difficult to make out the oscillator/filter/LFO sections. There’s a ton of opamps however, as well as lots of little caps that may well be part of oscillator sections. The thing looks absolutely beautiful but there’s not much to see, sadly.

But what about the thermograph? Maybe we can see something interesting going on there? Well…

This machine is cool. Too cool for IR photos. In fact you can hardly make out that I was playing with it just moments before. This is in contrast to my Virus TI…

But anyway, open up the A4 again and…

Nope. Nothing that gets warm really. Except for the DC/DC converter coil…

…and the power supply:

Now this is the Virus TI in comparison:

So there you have it. A peculiar, beautiful device. The analog circuitry is mystifying. Where and how it interconnects with DAC/ADCs and the digital world seems to be quite a complicated matter. My guess is that there’s analog oscillator/filter stuff going on at first, controlled by digital glue logic. The audio signal is apparently then sent through an ADC, the effect processing being done digitally, after which the four channels are converted back to analog for output. At least, that’s my guess. Maybe someone can correct me there and shed some more light on it.

There’s lots of different memory chips in there, including an innovative long-term storage for your kits and patterns. In terms of built quality, Elektron ranks up there with the very best, worlds apart from mediocre-quality stuff like the Kronos.

I can only conclude that the guys at Elektron are not only competent, but they also really enjoy what they’re doing.

P.S.: Give us a bigger display next time. Maybe an OLED screen? Oh and… pads for the sequencer buttons would be nice. Perhaps even with force detection. And it should whiten our teeth while we sleep. Thanks

void · 2013-09-28T12:42:56.000Z

nice one, thanks for doing this.

about the signal path: it’s stated on elektron.se that the path is completely analog, so no AD/DA conversion. The DAC/ADC is likely just for the FX sends, which are then mixed back to the analog stereo signal.

I’ve had mine open once and probed around a little to see if it would be viable to mod individual outputs. I was able to tap the individual voice signals, but my knowledge of electronics is too limited and I didn’t continue exploring this.

emdezet · 2013-09-28T12:54:54.000Z

So do you make multicopters and brushless gimbals, too?

Toumal · 2013-09-28T13:48:43.000Z

""void"":

about the signal path: it’s stated on elektron.se that the path is completely analog, so no AD/DA conversion. The DAC/ADC is likely just for the FX sends, which are then mixed back to the analog stereo signal.

That’s peculiar. I always thought they have to go digital to add the reverb/delay effects. These effects are most likely done in the digital world, don’t you agree?

So do you make multicopters and brushless gimbals, too?

And shoot guns and build geiger counters and rocket launch systems, yes. You know, hobbies!

void · 2013-09-28T14:30:32.000Z

""Toumal"":

These effects are most likely done in the digital world, don’t you agree?

Well, of course they are. What I meant is that the main audio signal from the 4 voices is not converted to digital and then back to analog, interrupting the analog signal chain, as you wrote:

""Toumal"":

…after which the four channels are converted back to analog for output.

I think this is not the case.
Imo, the ADC/DAC is used to tap the individual voices just for the FX sends, and then to convert the stereo FX signal to analog so it can be mixed into the main stereo signal.

pinkbabyhead · 2013-09-28T14:46:04.000Z

love it. thanks for posting this.

Toumal · 2013-09-28T17:18:22.000Z

[quote=“” void""]

[quote=“Toumal”]These effects are most likely done in the digital world, don’t you agree?
[/quote]

Well, of course they are. What I meant is that the main audio signal from the 4 voices is not converted to digital and then back to analog, interrupting the analog signal chain, as you wrote:

""Toumal"":

…after which the four channels are converted back to analog for output.

I think this is not the case.
Imo, the ADC/DAC is used to tap the individual voices just for the FX sends, and then to convert the stereo FX signal to analog so it can be mixed into the main stereo signal.[/quote]
Hmm well according to the official site you are right!

[i]Both oscillators of each voice are analog, with all their elaborate waveshaping completely analog, and with their pitches controlled digitally. The noise generator is digital. The two filters, and the overdrive circuit between them, of each voice are fully analog. The amplifiers are analog. The envelopes and LFOs are digital. Their design allow them to go well into audio range.

The signal path from the oscillators to the audio outputs is completely analog. The signal path from the external inputs to the audio outputs is also fully analog. The digital send effects are parallel to the main audio path, and the wet signals from the digital send effects are mixed with the dry analog signal before the main output stage.[/i]

Somehow I had it in my mind that you had to go THROUGH the effects processing, rather than mixing in with the original audio. BUT, this also means that the more “wet” your effects mix is, the less of the “true” analog sound remains.

The envelopes and LFOs being digital kinda explains the peculiar amount of logic ICs on the board. I bet the FPGA does a lot of that stuff as well. I doubt it’s being used to drive the display or do any of the UI input stuff.

What fascinates me is how SMALL a single voice section is in that thing. I bet you could easily modularize the whole thing and make it expandable to more voices. Heck, you could even start a new, modular kind of synth that offers slot-in expansions with different synthesis types, kinda like a Buchla but with each expansion adding one “voice” instead of requiring you to patch stuff like a telephone operator only to get a few bleeps and squeaks.

Just think about it, you could mix and match voices, you could slide in two digital sampler voices from an octatrack and three analog voices from the A4, or do all kinds of combinations.

Mmmm… daydreams…

Anigbrowl · 2013-09-28T19:33:47.000Z

""Toumal"":

P.S.: Give us a bigger display next time. Maybe an OLED screen? Oh and… pads for the sequencer buttons would be nice. Perhaps even with force detection. And it should whiten our teeth while we sleep. Thanks

What an awesome analysis. As for your feature request, I invite you to perform a similar analysis on a DSI tempest…min eshould be here on Monday and I’m looking forward to connecting it to the OT and A4 (which will be sending CV to the Tempest to overcome its lack of free-running LFOs).

plusn · 2013-09-28T20:11:42.000Z

Wow, thank you for the “inside job”. Excellent findings!

MX · 2013-09-29T10:52:18.000Z

Wow! Amazing analysis

switchbox · 2013-09-30T10:35:32.000Z

great…what is the j1 25 pole connector all about?..just for programming the fpga?..or for upgrading something? …maybe like a +drive?

jgb · 2013-09-30T11:16:24.000Z

Thanks! Interesting read, and lovely pictures.

jaalmadakan · 2013-09-30T11:30:07.000Z

"what is the j1 25 pole connector all about?..just for programming the fpga?..or for upgrading something? "

FPGAs are not “programmed” per se. The chip reads its configuration from a non-volatile memory chip at startup and reconfigures itself according to that. Kind of like a CPU program that would be loaded in RAM at startup.

The jumper is most likely used to load the configuration binary file in the memory and/or for functional testing during production… nothing really interesting for modding (unless you are willing to re-design the FPGA program from scratch).

Cocker · 2013-11-13T14:51:19.000Z

Wow. Unreal. Not really sure what I’m looking at for most of this but is there anything there that hints at future upgradeability / expansion ?

avantronica · 2013-11-13T14:57:02.000Z

[quote=“” Cocker""]
Wow. Unreal. Not really sure what I’m looking at for most of this but is there anything there that hints at future upgradeability / expansion ?
[/quote]

…"Next to that we have a Micron 256 SDRam chip (2WG22) and a Toshiba flash memory chip. My guess is it’s a 2GB chip but I couldn’t find the datasheet to verify. "…

make of this what you will ! ? 2GB, potentially ? hmm, room for all sorts in there, we can but hope

void · 2013-11-13T15:28:34.000Z

yea wondering what this is for… strange!

udenjoe · 2015-04-29T13:30:23.000Z

My guess is the larger memory is for the +drive, the next smaller is for your saved content that isn’t in the +drive. Then the other memory is for storing your changes constantly, your sandbox, undos, parameter changes, etc.
Just my blind guess.

vos · 2015-04-29T15:26:57.000Z

Very interesting - even if this is allmost Chinese to me - Thanx.
I allways had this feeling with Elektron instruments - they’re developed with passion and love - I wish I understood electronics - must be so interesting. Well it’s never to late to lurn I guess.
Great info Guys ! I’m very happy about the quality of the long - term memoty ( Hey !! I understood somethig !! )

Greatings

TrabanT · 2015-04-29T15:46:50.000Z

awesome thread! thanks for the work OP !!!

henning · 2020-02-19T21:19:58.090Z

Even today cool and interesting! Too bad the virus pics are not working anymore, probably they updated their forum?

It’s an interesting observation how Elektron builds quality hardware, and hopefully this is still true for current devices. But I assume it’s worth to invest in the more expensive devices, as surely there’s not much benefit for a company to make even the cheapest model:samples (that I got as my first Elektron) for 300€ as robust that it will last for 10 or even 20 years or more.

(found this while researching a bit for this discussion: Firmware + overlay = new machine?!)