While my transformer based amp awaits some more soldering work, I nabbed a tin lunchbox and built an enclosure for a previous build, build around the Acupulco Gold/Tufnel/Sunny T dual LM386 design.
Mine includes a Big-Muff style tone control in between the two power amps, which has some mixed results. All the way to the low side, it produces some nice tones, but to make the high side listenable, I may have to solder a low-pass filter on the output. Small, full-range speakers do get shrill compared to guitar amp speakers.
There have been a few random posts with pictures of wires, some random schematics and some screenshots, so it might be time to give things some narrative.
For a few weeks, I've been trying to build a tiny 5 watt amp out of ICs with one particular feature: after hitting a small power-amp (an LM386), the signal gets routed through a transformer. This serves a couple of purposes, the primary being the emulation of the power amp transformer saturation that happens with an overdriven tube amp. After the transformer, the signal gets routed to two LM380 IC's, which bump the signal's power back up to 5 Watts or so.
Now, with the kind help of the folks at www.diystompboxes.com, I've managed to put together a working prototype on breadboard with some sensible component values. And it works.
Quick sketch. One day, I will learn Eagle.
Now, it produces a usable tone, but recording a demo has been hampered by having no heat sinks for the LM380's. I can play for a few seconds, and then the chips get too hot to be healthy, and I have to power down. This will be remedied in the future, and then I will record some tone demos.
But I can give some details on what's happening with the most critical component, the transformer. The unit I have is the Midcom 671-8001. Now, this was pulled out of an old Digital TV box someone was throwing out, so it's not really a high quality component. Its primary purpose was to figure out how such a component would fit into the signal chain.
The configuration I've settled on is a combination of two things. The transformer, the DC coupling capacitor for the 386's output and a small resistor form an RLC high pass filter. The coupling cap is necessary to remove DC from the amp's output, and the resistor is needed to ensure a stable filter (I think. Don't quote me on that.). Secondly, there's a cap on the output forming a low-pass filter to take some of the really nasty fizz from the LM380's input stage. After trying out several cap values, the transformer seems to behave like a 240Ohm resistor, which is close enough to the sum of its primary and secondary DC resistance values (108 + 120 = 138 Ohms). I'm probably missing some math that would explain it with more certainty, but it does seem like a significant coincidence.
Running a white noise signal from my phone to measure the frequency response while testing caps.
Expedient. Not scientific.
There are other things which remain unknown. The primary one is that I have no idea if the transformer is saturating. The LM386 puts out a lot of distortion, and the filtering effect of the transformer and its neighboring components is further altering the signal. For all I know, all I have is a needlessly complex filtering setup for my signal.
I've been posting somewhat haphazardly with regards to my amplifier goings on, and am trying to wrap my head around the concept of audio transformers. I promise this will settle down once I stop barfing my random technical thoughts online.
Anyway, I think I made a breakthrough today by running some white noise through the 1:1 transformer I pried out of an old digital TV box.
I know this is not the way to do stuff, but I don't have an oscilloscope and Waveosaur is free.
What you see there is the frequency response for white noise, and a telltale peak at 1242 Hz. The transformer was being used in an RLC high-pass configuration, with 240Ohms and a 100nF cap, which would indicate my transformer is roughly 0.16 Henries. Different cap and resistor values seemed to be consistent with this pattern, so at least I have that one part of the amp somewhat figured out.
Now to drive it from the 386 without blowing another power cap...
Just a little something I'm working on. I got a transformer hooked up to a small LM386 guitar amp I built, and so far the sounds its producing aren't unpleasant. More to follow (including some Python code, because I got tired of manually calculating the 386's gain and bass boost frequencies).
Did you remember to take the second left at RefSeq Genomic Accession ID?
I complain ad nauseaum about the various competing formats used to identify various biological molecules. Here's a bug#&*% insane helpful map from the people at bioDBnet.
I've been on a little bit of a communication kick lately, having against all reason actually enjoyed presenting at GLBIO. In truth, representing data visually is just a subset of the task of communication. Through visualization, we transfer data from a medium to our brains via the sense of sight. The peculiar custom of transferring data through standing in front of a crowd of strangers and talking for an hour or two is curious, but similar in many ways.
A few months ago of my co-workers was kind enough to direct me to the work of Edward R. Tufte, a critic and theorist who specializes in visual data representation. I am ashamed to say that until she literally put a book of his in front of my nose, I hadn't seen any of his work. I have since remedied this.
Among his writings, there exists a well-distributed, crabby critique called 'The Cognitive Style of PowerPoint' that I've recently taken to heart. By taking it to heart, I mean that I've used its principles to utterly dismantle the PowerPoint presentation I put on at GLBIO 2012. I redid that presentation in-lab today with the following set of basic rules:
Only use slides for things that slides are good at.
If you ask Tufte what slides are good at, he would tell you 'very little'. He spends a great deal of time discussing the many ways in which PowerPoint slides, and particularly their templates, just plain suck at communication. Data is parceled out in 10 to 20 line chunks, forcing the presenter to unnaturally partition their narrative. PP graphs are low resolution, illustrating only the most blunt of points, and used in places where a simple 'and then X happened' would have sufficed from the presenter. For the most part, a standard template for a PP presentation serves less as a means of communication than as an assistance device to organize the presenter, albeit within the bizarre constraints of 'one slide per topic' regardless of the scope or complexity of that topic.
With all of this in mind, I looked at the slides I had, and I started to pick them apart. Immediately gone were organizational slides. Unless a slide could communicate a concept more effectively than speaking alone, it was scrapped too. Out of 27 original slides, I kept 8, and those were severely cut down and held almost nothing but graphics.
8 slides to represent an hour long portion of my talk.
The results of this cull?
Twofold.
The ugly side of this was that I had underestimated the usefulness of PowerPoint as an organizational device. Stripped of a hierarchy of bullet points, I realized at midnight before I was to present, I had lost my narrative. The mess of bullet points that I transferred to five pages of printout were a poor substitute for a rehearsed talk. I stumbled over what should have been a flowing exploration of graph visualization practices.
Though this resulted in a few uncomfortable moments, it drove home a point Tufte made often in his critique: giving presentations is hard. Having PowerPoint, at best, turns a poor presentation into a boring one. Instead of spending hours of my time culling my slides, I should have invested that time into practicing my narrative and ensuring that the more efficient form of data transfer, the 150 word-per-minute speech, was as well-oiled as it could get.
The second result of my re-formatting was that I saw how effective a good graphic could be. If anything saved the presentation, it was this slide:
This horrible, horrible slide.
This slide was an illustration of bad visualization practices. It comes from this article in Nature, which is, ironically, about improving graph visualization. I have seen many different versions of the same concept repeated over and over again in graph visualizations: people misunderstanding the purpose of graph visualization entirely. Graphs excel at communicating to the viewer information about relationships between objects. In this example, none of the graph edges are remotely traceable, the relationships between the implied complexes are indistinct and there is no evidence to suggest WHY any of the complexes should exist. There is NOTHING achieved by this graph that could not have been done more elegantly with tables of protein names. It is the Michael Bay of graphs.
"...and then the protein complex transforms into a DEATH JET
that shoots FIRE while Megan Fox SWEATS PROVOCATIVELY!"
The effect of showing this graph was immediate: the entire room groaned. They understood, very quickly, the concepts I had been stumbling across in my unpracticed narrative. Including the slide was to my advantage. It saved my ass.
So, two lessons learned.
In closing, I'd like to leave you with something that kept me sane while I experimented with dangerous presentation techniques. The following is the first part of a lecture given by Louie Simmons, a trainer and competitive power-lifter. I listened to this lecture in between editing sessions to remind me that a good presentation doesn't need ANY slides. It needs content and a presenter capable of communicating it.
