Cool! Circuit Lab dot com, free browser-based circuit design

[Dakota]

Am I late to this party? Early? Anyway, I'm thrilled to have recently bumped across this, and figure this is just the sort of thing Tape-Oppers are into, so now shared with you:

https://www.circuitlab.com/

It's along the lines of LTSpice type circuit design and test software, but not nearly as in-depth or time matured, as it's only been public for less than a year as far as I can tell. It will likely grow in depth as more people get involved.

It is, however:

* free!
* cross-platform easy, runs inside your web browser, no program to install
* learning curve as easy as any circuit design aide could be expected to be
* can be logged into from anywhere, your stuff in progress is saved on their server
* your stuff can be private, link only, or public - so you can share and collaborate with others really easily
* it makes nice clean circuit diagrams and test-result charts you can easily export
* it's *really* fun if you go for the geeky fun, which I do...

I've already made an account and have been a happy busy beaver making and simulation testing and fine-tuning audio circuits. Hooray.

This might really catch on - but it needs more audio/music DIY people involved there to really get it rolling -

Thoughts?

[Dakota]

Related news, there is also:

http://www.docircuits.com/

Which is very similar in many ways to Circuit Lab, but seems to have even less people involved at the moment, and less components that have been modeled yet. But it does have some fancier looking GUI aspects, and an amusing "burn" animation if your circuit cooks a component.

Also on poking around, it's said that MIT has something similar in-house for some of their course work. Are there common roots under these three? I don't know.

[Jim Williams]

The current issue of Electronic Design magazine has a summary of all the freeware spice programs.

TI provides their version, TINA-TI, their version of the DesignSoft's TINA Design Suite.

Analog Devices offers their freeware Spice from National Instruments, MultiSimComponent Evaluator. It's paid for version is very common in university labs, that's why schools can't use the freeware version. You can though, it's probably the one you see at MIT.

The most advanced and popular freeware Spice is Linear Tech's LTSpice, designed by staff engineer Mike Englehardt, who still works there and maintains/updates the program regularly. It has replaced the popular P-Spice from the 1990's.

However, these are simulators for analog designs. Although modern models have improved with parasitics now modeled, it's only a sim and any decent design EE needs to breadboard any spice design to confirm performance as real world parasitics can wreck havoc with well planned models.

Read up on what Bob Pease from National Semi said about Spice. He trusted them about as much as I trust a polititian.

[Dakota]

Thank you for the info, Jim! Very informative.

[Dakota]

However, these are simulators for analog designs. Although modern models have improved with parasitics now modeled, it's only a sim and any decent design EE needs to breadboard any spice design to confirm performance as real world parasitics can wreck havoc with well planned models.

Absolutely! It's really helpful to use the simulations to get into ballpark predicted ranges of the design goals, but in the real world, complex interactions can and will come up that the simulation didn't see coming at all.

[The Scum]

For the record, I use spice when starting designs - before I commit anything to breadboard, I usually simulate parts of it. I have an infinite number of virtual opamps, caps, etc to destroy in spice, before I start to commit to the real thing. But it's a stepping stone, often so i know what components I need to track down in meat-space.

So I just went and compared this to my usual, good ol' LTSpice.

Sure enough, click on it and start editing.

I built a simple Spice-stress-test circuit - non-inverting opamp with a load resistor. +/- 15V rails, 10V 1kHz sine input. Circuit Lab gave me a Tl082 by default...and it has models for a few of the other "usual suspects"...LM324, 741, etc. Same for BJTs, JFETs.

LTSpice contains a lot of LT parts...and some generic parts that can be tweaked to represent parts that aren't included with a little datasheet reading. Jim W likes the LT1013, right?

The CircuitLab editor is easy to use...it uses a clever "rubberband line" to do the wiring - friendlier than many other circuit applications. Cut/copy/paste are supported using ctrl+c/x/v. The editor in LTSpice is, um, unique...it feels like maybe it was written for X11/Motif, and then ported to Windows. It has it's own concepts of cut/copy/paste/move/undo, nothing like you'll have used anywhere else...

With all resistors at 10K (the easy, unity gain test), both behaved similarly.

The one thing to note here is speed of simulation. A 3 second, .000001 second step simulation took about a minute in LTSpice. It was going to take 3 hours in CircuitLab. Thankfully, in either of them, you can tell it to cancel, and you'll get the results calculated up to that point. And this is a trivial circuit, that shouldn't be too intensive to simulate. I can test the truthfulness of the models

I altered resistors for a gain factor of 10...and both simulators did about the right thing - they both clip the output near the power rails, as expected (a really bad spice opamp will swing outside the rails, a completely incorrect result.)

Then I dropped the output load resistor to .1 Ohm. Agian, both behave reasonably - they don't drive infinite current into a near-short (which, again, a bad simulation will allow). The CircuitLab model current limited at 300 mA...not accurate for a TL082 (datasheet says ~50 mA current into a short), but exactly what you'd expect with the stated output impedance of 50 Ohms.

So CircuitLab is usable, but dog-slow. I also don't see any way to use one of LTSpice's great audio features: the ability to use a WAV file for input, generating WAV output.

[dfuruta]

I've been using circuitlab for a while now. It's great for drawing up schematics, but:
the simulator is very slow and not totally trustworthy (I've run into problems when feedback gets more complicated, for example a BJT with collector into the inverting input of an opamp, feedback to the emitter)
the component selection is very limited
you can't export a netlist, so you can't take your schematic and put it into a pcb layout tool
the editor gets bogged down when working with lots of parts or, strangely, at zoom levels other than 100%.

I really like how easy it is to use, though.

[Dakota]

Fantastic analysis, The Scum and dfuruta! Thank you so much. Keep in mind I'm not an EE and highly value the perspective of those more skilled.

Easy and cheerful to use... slow simulation much under the regular expectations of the spice family. I've been choosing simulation times just long enough to see what I'm looking for and not longer than that.

Totally needs the .wav input and output option. I'd be stoked for that.

If you all feel like it, perhaps put in feature requests on their forum? This thing feels kind of like facebook before it caught on... facebook for DIY audio EE would be pretty cool... bouncing ideas, synergy, fun... needs critical mass to get rolling...

Hold it. Facebook isn't cool.

Anyway, a fuzz circuit in progress - CircuitLab time domain export:

[The Scum]

Neato - a "capital H" waveshaper if ever there was one.

Your facebook analogy raises a good point - since this thing is hosted online, what are the guarantees that your circuits aren't being swept out the back door of the program for some other use? (Simluate using a TL082, all of a sudden you get banner ads for TI) Since they require an account to save or print a circuit, I assume it gets stored on their server?

It seems like an interesting option for the open hardware world. As someone who designs proprietary circuits for hire, I have a responsibility to maintain the confidentiality of those circuits. I'm unsure of it's suitability in the regard.

Is there a reason you're not using LTSpice?

[dfuruta]

Is there a reason you're not using LTSpice?

Can't speak for Dakota, but I don't have a Windows box. I'm using Kicad in a virtual machine for layout, and I've used ngspice, but it's really user-unfriendly. Circuitlab is great for doing something quick, since it's a lot faster and easier to use than Kicad/Eagle/most more powerful layout programs.

[Jim Williams]

I'm more like Robert Pease, I found far more irritations with similators that they are worth for me.

To me they are like a entertainment puzzle, fun if you have the time, but mostly just an exersize. Once you get competent at designing analog and it's limitations, you get those feelers about those designs that may be questionable, especially if the math doesn't line up.

So I do most of my "simulations" on paper and in my head. Only a real world breadboard means anything to me on a complex design. Spice a complex feedback design and all sorts of strange "predictions" come up. Use any higher current stuff and ground traces and currents do all sorts of stuff the simulators never predict.

Breadboarding push-pin boards I often use because changes can be done about as fast as entering a new model on a PC but you get instant real world results, not a "best case model" not incorporating all the parasitics a real world circuit has.

They are a good learning tool but as Dirty Harry said, "A man's got to know his limitations" still holds true.

[Dakota]

Neato - a "capital H" waveshaper if ever there was one.

Your facebook analogy raises a good point - since this thing is hosted online, what are the guarantees that your circuits aren't being swept out the back door of the program for some other use? (Simluate using a TL082, all of a sudden you get banner ads for TI) Since they require an account to save or print a circuit, I assume it gets stored on their server?

It seems like an interesting option for the open hardware world. As someone who designs proprietary circuits for hire, I have a responsibility to maintain the confidentiality of those circuits. I'm unsure of it's suitability in the regard.

Is there a reason you're not using LTSpice?

Thanks, The Scum! Yes, it's a severe waveshaper, lots of harmonics - modeled there with two parallel comparators with hysteresis, final version would be four comparators.

And indeed, circuitlab shouldn't be considered completely private - I wouldn't put proprietary corporate secret designs in progress there... but the usual modern day common sense rules and expectations apply. I do notice I get more electronics parts ads on facebook than I used to, ha...

(Aside - I use the ghostery plugin in my browsers anyway to block as much ad targeting and tracking as possible anyway, y'all might want to think about getting that, it's great. http://www.ghostery.com/ )

But yes, great for the open source and DIY collaborating crowd.

I'm not using LTSpice yet because I'm just easing into this, and I'm liking the learning curve, I can deal with circuitlab for now. I'll be moving over to LTSpice as well as I grow. I'm self-taught in electronics (grew up with my Dad being a ham radio and general electronics hobbyist, learned by hands-on messing around), so my knowledge base is really uneven - this has been hugely helping round out my grasp on things.

And Jim - just having all this in your head already is awesome. I can certainly do that for composing music, but am nowhere near that in electronics...

[Jim Williams]

I'm working on some new mic pre designs using complex feedback loops and current feedback opamps. The problems that come up building this stuff are never predicted nor modeled by simulations, too many "real world" complex poles and parasitics to be completely modeled.

Bandwidth is both your friend and enemy. When you are attempting to amplify a feeble mic signal 60 db (1000 times) while operating at 30 or more mhz bandwidth on a tiny 2" square pcb, something's got to give and spice won't show it.

A scope does, instantly. Then you get to figure out how to fix it.

[The Scum]

I need to speak up in favor of simulations. They're another useful tool to have in the box, and like any tool, there is room for mis-application.

The sad fact is that the state of the art in electronics is now far beyond anything that gets used in audio. Cellphones, and the infrastructure behind them are driving everything at the moment. GHz transmitters and receivers, high capacity batteries, and ever increasing flash memory density are driving the technology.

I was on a project team a couple years ago, designing a piece of instrumentation for use in wet environments. It was to be battery powered, and needed to be able to monitor and disable portions of the system if water somehow got in. The power system in this thing was pretty complex, maybe 20 ICs. Long before we even considered ordering parts, we were building Spice models, and playing with the circuit parameters, tweaking the behaviors we wanted, and selecting an initial set of parts so that we could build a first prototype.

The parts in question were the sort that are almost impossible to prototype with...tiny SMD chips, with integrated heatsink/pads on the bottom of the package. The footprint for each one evaluated was unique, needing a custom PCB layout to support it. The simulations allowed us to iterate over several generations of design without the cost or time involved in building physical prototypes.

When we moved on to hardware, we were able to take some of those virtual test benches, and adapt them into physical ones...there were cases (like the overcurrent short detection) where the actual circuit had nearly identical behavior to the sim...far better than the senior EE overseeing the project expected(he was both a SPICE doubter and evangelist...he had a keen sense of when a simulation wasn't truthful, and had a number of tricks for forcing a sim into "real world" poor behavior).

It's probably telling that we designed in a bunch of TI and LT parts, for which high-quality SPICE models are available.

Before that I was at a company that did custom ICs for 1/2/4/10GHz data transmission for Fibre Channel networks. The team that wrote the simulations for the chips was actually larger than the team designing the chips. The simulations were invaluable for finding what would have become 6 or 7 figure mistakes if we had simply jumped straight to hardware. Those sims ran overnight, every night, for years worth of development.

The PCBs that hosted those ICs recieved similar scrutiny from the respective team. Altium has very good (and very expensive) tools for simulating high speed designs.

So it feels like analog audio is unique in the sense that we can work with parts that are large enough to prototype with, at bandwidths that aren't incredibly critical of the physical construction.