Nordic Nerd

SimFish, the Aquatic Life Game

2012-01-20T15:46:00.000-08:00

I'm working on a game named SimFish, a simulation of aquatic life, where your role is basically to control the environment of the fish, and see how it affects them (sometimes dramatically). Here is a video:

R/C Receiver and new video

2011-11-02T00:22:00.000-07:00

Fully assembled radio receiver, with servo outputs and blue

LED on the left, PIC16F688 microcontroller in the middle,

and ARF7244A receiver on the right.

New video of the R/C rover in action. TV reception
is still poor, as we lack an RF power amplifier.

R/C Car Project

2011-10-25T22:48:00.000-07:00

My cousin bought an hobby grade R/C car for a project we're doing. Above is a video of me controlling the front wheel servo using a PlayStation 3 controller, via my computer and an MCU (microcontroller).

I have spent the last few days perfecting the MCU's software, implementing a serial protocol with error protection, and writing fail-safe subroutine to ensure that the car won't go haywire if it receives corrupted data. Today, I went and picked up a 500 mW transmitter/receiver pair, and interfaced it with the MCU, making it wireless. Finally, I installed this system in the R/C car, and my cousin carefully installed the transmitter to a piece of cardboard. Below are pictures of the results.

Transmitter and battery

Closeup of transmitter voltage regulators

Birdseye view of transmitter

Closeup of transmitter radio interface

MCU and receiver

Birdseye view of MCU and receiver

Active Speakers - Update

2011-10-25T22:33:00.000-07:00

The audio amplifier project has been put on hold until Sverre can procure some extra funds. The latest progress is that I have tested all the electrical connections on the power stage with a multimeter, and they seem to be correct. Attempted to buy what I thought was a dual/bipolar bench supply to test the power stage with, but the supply turned out to have floating terminals, so I had to return it.

Active Speakers - Power Stage PCB Soldering

2011-08-27T07:37:00.000-07:00

Active Speakers - Power Stage PCB Arrived

2011-08-24T10:22:00.000-07:00

It has arrived! The power stage I designed now has a real circuit board. Stay tuned for more photos as I begin to solder this baby together!

Active Speakers - Crossover PCB Design

2011-08-12T03:44:00.000-07:00

This is my design for the crossover filter, with a steepness of 24 dB/oct. GND, V+ and V- are power rails, LIN and RIN are the left and right inputs, and LHPO, LLPO, RHPO and RLPO are the left/right high/low-pass outputs. Component values are deliberately left out in order to make the design reusable. Filters with various crossover frequencies and Q values can be designed by changing the values of the various capacitors and resistors.

Circuit schematic from LTspice that formed the basis for the PCB layout above. I made a custom 4-in-1 symbol for the TL074 quad op-amp, together with a special 14-pin sub-circuit (using this SPICE model) with the pins connected like on the real chip, for easy layout in FreePCB later on.

Active Speakers - Power Stage PCB Design

2011-08-10T21:21:00.000-07:00

I have submitted this PCB design of the power stage to Silver Circuits for prototyping. There is a total of 6 amplifier chips, back to back in a 3 × 2 arrangement. ULW+, ULW-, URW+ and URW- power the woofers, while ULT and URT power the tweeters, providing ~300 watts of playback power on a 4.6 × 2.25 inch circuit board. The heat sinks on this thing will be ridiculously large...

The LTspice schematic for the above PCB layout. As you can see, it wouldn't be easy to recognize that the two drawings are the same circuit. Structure and clarity are the main concerns on the schematic, while space and trace length are paramount when laying out a PCB. The LM3886 symbol was custom made from a generic opamp symbol in LTspice, and the model was found in a forum and modified so pin numbers match with the actual chip.

Active Speakers - SPICE With Actual Audio

2011-08-07T23:11:00.000-07:00

Using the WAVE file feature in LTspice, I ran a simulation with an actual audio source, a 9.161 second clip from Blame It On The Boogie by Michael Jackson, chosen for its clear transients, and being one of my favorite recordings. These simulations run at about 1 second per 2 to 5 ms of processed audio, so patience is recommended.

A simulation with a Thiele/Small circuit is not complete. It does not take room acoustics or baffle diffraction into account. It is equivalent to placing the speaker in an anechoic chamber or a tall outdoor pole, and recording the on-axis response (minus baffle diffraction) with a measurement microphone. The resulting WAVE file is best played back on headphones or in-ear monitors with a flat frequency response, and while not a perfect simulation of the listening experience, it should reveal any obvious flaws in reproduction.

I determined from this that the slight drop in bass below 400 Hz needed to be corrected, so I added a low-shelf filter to the circuit and re-ran the simulation. If you listen carefully to the bass, you might hear a difference. I also created a new 3.696 second clip from On The Floor by Jennifer Lopez, chosen for deep bass and being contemporary music, and ran a simulation on that.

Judging from these two clips, some electronic bass extension is definitely warranted.

Active Speakers - Enclosure #2

2011-08-06T23:34:00.000-07:00

Many sites about baffle step and baffle diffraction draw their information from an AES paper by Harry F. Olsen, published in 1950, named Direct Radiator Loudspeaker Enclosures, where various enclosure geometries were tested for their frequency responses. If you neglect to read the actual paper and trust the websites out there, you'd be forgiven for believing that Olsen only tested spherical, cubical, rectangular parallelepiped and cylindrical enclosures, because that's everything those sites cover.

As it turns out, Olsen actually tested 12 different geometries! He found out that the most uneven frequency response was produced by a cylindrical enclosure with the driver mounted on one end of the cylinder. The second-worst contender was the cubic enclosure. The rectangular parallelepiped enclosure, very common because it's easy to build, was half decent, but not really a dramatic improvement over the cube. The second best enclosure was a combined parallelepiped and truncated pyramid enclosure. A little harder to build, but nothing a good carpenter couldn't handle. Finally, the spherical enclosure was the best, with a perfectly smooth frequency response. You see some high-end mass-produced speakers using this design.

Based on the measurements in his paper, I made some changes to the design I posted previously. The results can be seen in the drawing above. The front baffle edges have been softened, from 90 to 45 degrees, to reduce diffraction. I also took the opportunity to model the bass port to proper scale. Aesthetically, I actually feel that this is an improvement, and it's only slightly more complicated to build. A small price to pay for such a nice improvement in sound quality.

Active Speakers - Baffle Diffraction

2011-08-05T19:35:00.000-07:00

It would seem that centering the woofer might be a problem. See this article. I'm considering possible solutions to this. The issues pertaining to golden/acoustic ratio of the enclosure, plus uneven baffle response, might be solvable if we move the bass port to the rear. This might also mask some of the harmonic distortion induced by the bass port, but comes with a penalty of increased wall reflections. A spherical enclosure would sound awesome, but look funny and be difficult to build. Rounded corners can apparently help. Speaker design is, and always will be about compromises...

Active Speakers - Thiele/Small SPICE Simulation

2011-08-04T16:45:00.000-07:00

This is a plot of a SPICE simulation of a single amplifier channel (3 LM3886es) powering the bass and treble drivers. The treble input voltage is 0.15 times (-16dB) that of the bass voltage, because the tweeter is a lot more efficient. The responses are much flatter than they look, because the dB range of this plot is very narrow (for precision). The tweeter is capable of operating at 1.5kHz, so our crossover (not modeled or simulated yet) can be set much lower than previously feared. I'll post a schematic of the amplifier later on.

A plot of the summed outputs of the bass and tweeter drivers. The tweeter amp input was phase inverted to compensate for group delay from the woofer. The system would have severe distortion in real life, because both drivers are being fed a full-range signal. Crossovers will correct this, and flatten out the bump at 1.5 kHz.

Active Speakers - Basic Amp Parameters

2011-08-03T21:38:00.000-07:00

I have decided to base my amplifier on the LM3886 operational power amplifier from National Semiconductor, since I'm not in the business of reinventing the wheel. I have had success with it in the past, and it's a popular chip with self-builders. I'm going for an active bi-amped system, with the amplifier in a separate chassis and a 4-conductor cable going to each speaker enclosure.

This is a diagram of one amplifier channel. The amplifier will have 2 channels, powered by an unregulated ±25V dual supply. The low-pass and high-pass filters will be 4th order Linkwitz-Riley filters. I might add an all-pass filter to the treble chain to phase-align it with the bass chain. The bottom LM3886 in the bass chain will operate in inverted mode, with the chips providing ~100W into 8 Ω altogether. Depending on the difficulty, I might make the crossover frequency manually adjustable. I expect to set the crossover point to about 3kHz. This is unfortunately right in the middle of the presence band, but driver limitations will probably make it necessary. The treble driver will be powered by a single LM3886 amp, providing ~50W into 8 Ω.

Active Speakers - Enclosure #1

2011-08-03T12:40:00.000-07:00

I previously simulated and picked out a 7 cm long, 5 cm wide bass port, but the deviating characteristics had me somewhat worried, so I decided to do some research on bass ports. I found out that many consider short and wide bass ports ideal, if paired with the right sort of driver. This bodes well for my chosen design.

I found a second simulator that focused more on bass ports, which confirmed that my design would have the anticipated half-power frequency of 43.65 Hz. It also let me know that my speaker's acoustic power is 0.24, and that the air through the bass port would travel at 0.13 mach at the maximum rated power, and that this is below the whistle threshold of 0.16 mach. A tad close, but the bass driver should never operate at full power anyway.

The treble and bass drivers have overall diameters of a = 105 mm and b = 176 mm respectively, and the bass port is 81 mm, or c = 131 mm with 50 mm of needed clearance, so our enclosure will need to be at least a + b + c = 412 mm tall. For tolerance and aesthetics, we'll add p = 10 mm of padding, so p + a + p + b + p + c = 442 mm is our minimum height. Our minimum width is 176 mm, the diameter of the woofer, or p + 176 + p = 196 mm with padding.

I wish for the front baffle to have the golden ratio (φ), so let's set our width to 442 mm / φ ≈ 273.17 mm.

Using these dimensions, 2.73 dm × 4.42 dm ≈ 12.07 dm² is the area of our front baffle. To satisfy the goal of 30 liters, 30 dm³ / 12.07 dm² ≈ 2.49 dm must be depth of our enclosure. Solving forwards, we get 2.73 × 4.42 × 2.49 dm = 30.05 dm³.

The full-box golden ratio x : φx : x/φ, and the full-box acoustic ratio x : 2^1/3x : 2^2/3x were considered and rejected, because they were impractical. The golden ratio enclosure wasn't tall enough for the components, and the acoustical ratio enclosure was too tall for a small apartment speaker. A depth of ≥0.8 times the baffle width should produce good results in any case.

This is a drawing of the front baffle, as imagined by me. The bass driver looked best when geometrically centered, with the treble driver and bass port set 10 cm apart from it each, meaning that I shifted everything down a little compared to the calculations above. The opening of the bass port is still over 5 cm above the bottom of the enclosure, so no problem there.

Active Speakers - Drivers & Simulation

2011-08-03T11:07:00.000-07:00

My roommate Sverre Palmstrøm, who kindly humors my long technical monologues, became interested in commissioning a pair of active speakers from me, and gave me a budget of 835 US dollars for materials. I had previously built a subwoofer with a dedicated amplifier, and had moderate success with it, until it short circuited, because two wires were touching on the stripboard. No more stripboards for me, in other words...

I decided on a 2-way design, for simplicity, and also compactness, because Sverre's apartment is small. Many good near-field monitors are 2-way speakers, which bodes well for audio fidelity. After browsing the dynaBel webshop for days, I landed on a Seas U18RNX/P mid-bass driver from Norway, chosen for its popularity among self-builders, and a Sammi DT-25B40 from Korea, for the good quality-to-price ratio.

Next came some speaker enclosure modeling. I inputted the Thiele/Small parameters from the Seas'es datasheet into a web based simulator, and experimented with closed-box and vented-box enclosures. Here follows a few plots from the simulator...

The frequency plot tells us that a traditional closed-box design (blue) will give us the worst bass response. Figures show the half-power frequency at 79.55 Hz for a 10.56 liter enclosure. The lowest frequency produced by a bass guitar is about 41 Hz, and we want to at least be near that. The simulator further proposes a 6.45 liter vented-box design (black) with half-power at 51.98 Hz. I must reject this design because the enclosure would be impossibly tiny (18³ cm), with an impossibly long 24.28 cm vent, 5 cm wide, and is still far off from our design goal. Some manual tweaking of the enclosure and vent dimensions lead to my final design (red), a 30 liter enclosure with a 7 cm vent, 5 cm wide, with half-power at 43.65 Hz. As you can see, my design comes with the penalty of some -1.2 dB of barely audible bass reduction.

The step reponse of the closed-box design has shorter and fewer oscillations, as expected, but would come at the penalty of a poor bass extension, as shown further up. My vented-box design has more ringing than the simulator's proposal, but nothing major. I hope and believe that the better bass extension will be far more pleasing than a minor reduction in ringing.

No big surprises in the group delays of the simulator's proposals. A group delay of less than 0.030 seconds is said to be acceptable. The peak in my design is unlikely to be audible.

My design compares favorably to the simulator's vented-box proposal in this impedance simulation. Lower voice-coil impedance means higher power transfer.