Oscilloplasm - complex VCO pair

Earthside communication channel for 6U format devices

Moderators: Kent, slow_riot

Post Reply
slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Sat Feb 06, 2016 12:03 pm

One of the last things big jobs was designing the chassis. After some careful analysis I am leaning towards a single angled frame with an additional bend that extends backwards, bolting onto the PCB and parallel bend to improve the horizontal tensile strength.

My PCB layout and circuit design attitude is *very* well thought out with regard to electromagnetic noise compatibility. Although of course I will need to prove this to myself to be sure I'm making the right call and give some special attention to high impedance nodes that are attracting hostile signals!

Image

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1287
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Wed Feb 10, 2016 5:11 am

slow_riot wrote:My PCB layout and circuit design attitude is *very* well thought out with regard to electromagnetic noise compatibility. Although of course I will need to prove this to myself to be sure I'm making the right call and give some special attention to high impedance nodes that are attracting hostile signals!
What sort of problems do you foresee and therefore taking action to prevent?
EMC is a broad field.

Neil
Modular stuff :: http://www.njohnson.co.uk

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Wed Feb 10, 2016 11:14 am

neil.johnson wrote: What sort of problems do you foresee and therefore taking action to prevent?
EMC is a broad field.

Neil
People near FM transmitters demodulating signals into the audio path. One user complained about this with a euro module recently, given such a specific fault I can only put this down to a high impedance, high gain node with a long trace.

I am aware that opamps run at +100dB or thereabouts of gain, relying on negative feedback to achieve unity. So once you have a stray mHz signal coupled into the system accidentally, they can fall outside of the opamps operating bandwidth and be amplified, intermodulating down into the lower frequencies.

In my circuit I am using instrumentation amplifiers for the 1v/oct inputs,and in 2 other places which are running at 100x gain. These is going to need filtering (although the inamp design is very good about rejecting common mode noise)

I can reduce the bandwidth of the signal path to only what is needed, and lower impedances. High impedance nodes that are not driven by a low impedance source are max 20 mm.

Are there resources you can recommend which can show how to simulate the stresses of a worst case scenario environment?

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1287
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Thu Feb 11, 2016 1:57 pm

slow_riot wrote:People near FM transmitters demodulating signals into the audio path. One user complained about this with a euro module recently, given such a specific fault I can only put this down to a high impedance, high gain node with a long trace.
Well, for starters the one common problem with EMC is it is very easy to jump to the wrong conclusion a fix a problem you don't have, or what often happens is you actually make the situation worse! Unless you know what the coupling mechanism was you have no idea what you can do to fix it.

The most common way radio broadcast signals get in (e.g., FM radio, taxi/CB, etc) is some unfiltered input to a semiconductor junction (diode, transistor, IC) that then demodulates the RF down to AF, and then that couples into the signal path. Mobile phones are another fine example, albeit at different frequency bands and significantly more power.
I am aware that opamps run at +100dB or thereabouts of gain,
Yes, at DC
relying on negative feedback to achieve unity. So once you have a stray mHz signal coupled into the system accidentally, they can fall outside of the opamps operating bandwidth and be amplified, intermodulating down into the lower frequencies.
Once you get above a few Hz op-amps are actually quite good integrators. Which means their gain reduces with frequency - the magic number being the Gain-Bandwidth Product. So for most op-amps in synth world they have very little gain up in the 100MHz region, and more likely to be attenuating.
In my circuit I am using instrumentation amplifiers for the 1v/oct inputs,and in 2 other places which are running at 100x gain. These is going to need filtering (although the inamp design is very good about rejecting common mode noise)
You certainly want to look at filtering on your inputs and outputs. And remember that to RFI an input can be an output and vice-versa. Likewise for ESD any connection to the outside world is an input.

Common mode rejection is a good thing to have, but generally degrades with frequency, and at high frequencies you need to be very careful about impedance matching otherwise your CMRR goes down the toilet.
I can reduce the bandwidth of the signal path to only what is needed, and lower impedances. High impedance nodes that are not driven by a low impedance source are max 20 mm.
I guess you need to specify what you mean by "high impedance".
Are there resources you can recommend which can show how to simulate the stresses of a worst case scenario environment?
It really depends on what your objectives are. Do you want your design to pass official CE EMC testing? Are you concerned with low-noise design? Is your emphasis on emissions or susceptability?

Personally, I would start simple:
* no big loops in PCB traces
* ground planes
* filter at the edge of your module, both inputs and outputs
* keep your circuit as simple as possible - the more complex you try to make it the harder the problems and the harder still the solutions!
* minimise fast edges if you have any logic or comparators
* keep strong signals away from sensitive nodes

There is so much more to it - I have merely skimmed the surface.

Neil
Modular stuff :: http://www.njohnson.co.uk

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Thu Feb 11, 2016 4:18 pm

Thankyou Neil. The objective is low susceptibility to interference, and low emissions regarding self modulation and of surrounding modules.

I've been reading up about EMC from the beginning as part of low noise design, and many techniques have already been implemented for other reasons that can also help in the MHz region. But a lot are new to me, so thanks for the objective questioning.

Good to know about opamp behaviour in the high frequencies, that sure makes life easier.

The high impedance nodes I mentioned are some variable transconductances preceeding opamp integrators and I-V converters, which I'm guessing can present a 100k and lower source impedance to Gohm input. I suppose it's lucky I didn't pursue the idea of doing an oscillator with differentiation as opposed to integration then!

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Thu Mar 10, 2016 7:16 pm

OK final straight now. Almost ready to go to first PCB prototype, and then I can get onto doing sound demos of the complete instrument.

The price to buy is going to be around £650 .

Image

The prototype panel is still due another layer of colour. Each one will be hand painted.

Image

Image

User avatar
Cat-A-Tonic
Lobsters love Muff
Posts: 3699
Joined: Mon Feb 04, 2008 12:03 am
Location: Yokohama, Japan

Post by Cat-A-Tonic » Thu Mar 10, 2016 9:24 pm

looks amazing :goo: :wiard: :goo:

User avatar
davidh
Ultra Wiggler
Posts: 776
Joined: Mon Dec 06, 2010 3:58 am
Location: Europe/France/Alsace

Post by davidh » Fri Mar 11, 2016 1:54 am

:sb:

User avatar
DomMorley
Super Deluxe Wiggler
Posts: 1153
Joined: Sat Mar 10, 2012 3:27 pm
Location: UK

Post by DomMorley » Fri Mar 11, 2016 1:27 pm

Can't wait to hear it!
How I fill my days: www.dommorley.com

"Your head will collapse, but there's nothing in it..."

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1287
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Fri Mar 11, 2016 3:00 pm

DomMorley wrote:Can't wait to hear it!
+1 :tu:
Modular stuff :: http://www.njohnson.co.uk

User avatar
jdkee
Wiggling with Experience
Posts: 297
Joined: Sat Nov 08, 2014 3:25 pm
Location: Chicago

Post by jdkee » Sat Mar 12, 2016 11:26 pm

:love:

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Tue Mar 15, 2016 10:41 am

More pics... I've been updating the manual / user information on the website, I hope the text based approach is clear and detailed without being overwhelming. Once I have time I will probably condense the signal flow into a visual image but that's low on my to do list for the moment.

Image

Image

Image

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Tue Mar 22, 2016 3:12 pm

OK here's a demo of the carrier/output oscillator. Varying pitch and EVOLVE, no FM (even though it sounds like it).

It's a really awful demo, sorry. T record I have to monitor thru crappy earbuds and there is some weird overloading that i have tried to get rid of (maybe these earbuds or in my digital chain). I have otherwise been monitoring through a huge PA system and I can promise that it sounds better than anything else I've heard before.

Just the 1 oscillator sounds like a 2 osc FM pair already, and I did think whether the extra Mod osc was overkill, but it's totally unlike other FM sources due to it's purity, and sounds more like adding resonance on a filter than FM.

http://soundcloud.com/diffuse-1/oscilloplasm-demo

And also, this is just a mono output. The Carrier/Primary osc has 2 outputs, similar to sine/cosine, but the way the module works the phase relationship is constantly modulated. Devastating when each one is hardpanned L & R!

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Wed Mar 23, 2016 8:20 am

Given that the main oscillator can generate very complex harmonic content on it's own, I removed the hardwired linear FM bus and will change the front panel to better focus on harmonic control of the master oscillator. (VCAs no more!)

The harmonic control tunes the self relationship of the oscillator, and so compared to wavetable oscillators, there are settings which are in tune and settings which are not. The harmonic control has a linear input with attenuator, and a calibrated exponential input that will enable changes in harmonic content that correspond to 1v/oct, neat!

Image
Last edited by slow_riot on Fri Mar 25, 2016 11:01 am, edited 7 times in total.

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Wed Mar 23, 2016 12:58 pm

*slightly* better demo of the variable harmonic section. Classic VCO triangle in support. Need to work on the saturation characteristics.

https://soundcloud.com/diffuse-1/oscilloplasm-demo-2

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Fri Apr 01, 2016 9:34 am

The main oscillator is a (much) deeper variation on the classic state variable filter. Typical implementations use limiting on just one side, which is what I was using, but that was not enough showing hard clipping distortion when going into digital. Various options were considered and the best results were from using limiting on both integrators. It still can be a bit unruly but there is no way to improve further.

I tried to map out the harmonics but it is beyond me at this point! The harmonic evolution creates multiple resonant peaks sometimes at the fundamental and at octave harmonics of it, sometimes the fundamental dissapears, splits into 2 or more at dissonant or consonant intervals. It will not be instantly satisfying like a wavetable sweep but with mastery there is a lot that it can do, especially with the 1v/oct control.

I pushed the range of the master osc to 10s of Hz to 4kHz (the upper range giving more headroom for FM). The harmonic evolution will go above and below from LFO to 10kHz.

The onboard sine is now a typical wideband osc from LFO to 10kHz.

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Thu Apr 14, 2016 4:37 pm

I'm a few clicks away from ordering the first printed board, meaning I am 2 months away from taking orders. I'm willing to admit to being somewhat terrified of things going wrong but everything seems to be going well.

It was mentioned about eurorack manufacturers creating a standards body, my approach has been one of give and take with regard to things that already exist and other things that do not. Although both oscillators are run at +/-10V internally for maximum signal to noise, the outputs are +/-5V. The choice to use balanced bantam only was not taken lightly, and has been designed for full backwards compatibility with unbalanced systems. In particular the inversion switches on the 1v/oct inputs are built around the ease of doing this with balanced (you just flip the hot and cold wires around). Normally this would be totally inaccurate when driven from imbalanced source impedances, but using instrumentation amplifiers (with 1250 gigohm input impedance and without the Johnson noise that comes with raising the input impedance using resistors) should mean it works equally well regardless.

I'm using a novel technique of temperature compensation with thermal sensor ICs, in theory they work perfectly but in real life there is inertia, and the sensor output is set by a resistor which has it's own temperature dependence. Worst case scenario stress testing (armed with a can of freeze spray and lowering the exponential converter to -50°C) showed the limits of the technique, but at operating temperatures that are not a threat to human life I think the performance will be satisfactory.

Another point worth mentioning is that with this method of analogue synthesis starting with sine waves building up to a more complex waveform, at the board level, sines are much less able to interfere with adjacent circuitry compared to harmonic rich content typical in subtractive techniques, so crosstalk is uniquely low.

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Mon May 02, 2016 11:06 am

Assembling first printed board!

Image

Image

Image

User avatar
davidh
Ultra Wiggler
Posts: 776
Joined: Mon Dec 06, 2010 3:58 am
Location: Europe/France/Alsace

Post by davidh » Mon May 02, 2016 12:26 pm

nice :love:

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1287
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Tue May 03, 2016 6:31 pm

Nice photos. How many more board respins do you plan before the production version?

The quantity of AD534s is truly mind-boggling!!
Modular stuff :: http://www.njohnson.co.uk

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Tue May 03, 2016 7:04 pm

neil.johnson wrote:Nice photos. How many more board respins do you plan before the production version?

The quantity of AD534s is truly mind-boggling!!
Thanks Neil. 1 more respin for a couple of mistakes and edits for things whose visibility changed when moving from computer screen to real world and then to production.

1 edit was changing one AD534 to a 633 where I could get away with it. With this level of complexity of circuit the choice was either AD534/734 or AD633s plus 15 trimmers and a huge PITA in assembly. Even with the AD534s the accumulated offset builds up to 1 volt and so there is a DC servo in the main state variable loop to remove it.

Much thanks for your insights and honest appraisals.

User avatar
BugBrand
Knowledge of Bugs
Posts: 7110
Joined: Tue Jul 15, 2008 11:59 am

Post by BugBrand » Wed May 04, 2016 5:06 am

neil.johnson wrote:The quantity of AD534s is truly mind-boggling!!
I agree! I guess you're scouring the net for NOS parts rather than purchasing new (£30ish each from mouser!). Bloody hell!

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Wed May 04, 2016 8:36 am

BugBrand wrote:
I agree! I guess you're scouring the net for NOS parts rather than purchasing new (£30ish each from mouser!). Bloody hell!
Yeah it's the crux of the design for sure but there isn't another way around it (believe me I checked).

Grey market prices are good, if that dries up the pin compatible AD734 is about £14 each.

With the primary oscillator there are no waveshapers in the traditional sense, the 2 outputs are straight from the oscillator core, which is made from an extended state variable loop being controlled by another oscillator. So offsets/transfer errors/phase lag quickly accumulate.

Doing it in analogue does have some advantages, the 1v/oct control for harmonic was a big unexpected bonus. Even though it calculates the same equation as another wavetable oscillator on the market, that one just has the "nice" waveforms. This one has all of them. :twisted:

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Fri Jun 10, 2016 3:53 pm

Last frantic partial redesign... I hope! The behaviour on the first PCB was not the same as the hand wired prototype, with the oscillation tending to die out at higher frequencies, not good.

The problem with state variable oscillators is that when the integrators are "full" there is a tendency to push too much pressure into the feedback network. In this circuit the affect is that the primary harmonic tends to be accentuated, often to the point of clipping, whilst the supporting harmonics are not allowed to flourish. A tiny amount of negative feedback around one or both integrators helps to compensate for this effect. The amount of degenerative damping necessary varies with frequency, so the signal to be fed back needs to go through the voltage control element. The problem with through zero is that as the polarity changes our damping signal gets flipped around into an exciting impulse! The exact opposite of what we want. Putting the signal through it's own voltage control element adds another component but means we can maintain the phase where it needs to be.

As for the degeneration at high frequencies, we can diode clip the signal, and positively feed this back, meaning that there is a constant action amplifying low amplitude signals, the effect of which increases at higher frequencies. Further investigation necessary but I'm hoping to have cracked it.

Oscillators are absolutely resonant systems, and very delicate touches are needed to put control of the physics into the hands of the user.

The other delay has been doing a proper full metal chassis, which is now done.
Last edited by slow_riot on Thu Jun 23, 2016 10:02 am, edited 1 time in total.

slow_riot
Super Deluxe Wiggler
Posts: 4002
Joined: Tue Jun 30, 2009 8:58 pm
Location: uk

Post by slow_riot » Thu Jun 23, 2016 10:02 am

Here's the chassis, prototype being built now.

The use of the full vertical height of the faceplate is *huge* improvement regarding structural integrity.

Self clinching nuts and countersunk screws are necessary to keep a low mounting profile.

Image

Image

Power connector is 4 pin Molex 1490, -15V, 0V, 0V, +15V. The banana connector on the rear panel is for seperate shield/safety earth connection to the chassis and pin 1 (shield) connectors on the front panel. This makes it possible for the module to meet the standard proposed by AES48, where safety earth on a power supply can be used without fear of compromising signal integrity. The switch below it connects shield to 0V if you aren't using this connection.

Post Reply

Return to “Aether Machine”