the power supply should be quiet enough. i tested it out by grounding the input of a single old amp channel board, powering it with the supply, and listening with my 32 ohm headphones, and there was almost zero audible hiss. no hum. frankly, i had to shut off my computer, hard drive and monitor (yeah, switching noise. can you believe it?) and close my door, and i still had to strain to hear the hiss.
so that's extremely good news, for the nerd inside me.
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
well, i swapped out the 820uF caps in the headphone amp power supply for two 2700uF caps, and i swapped the 47uF adjust shunt capacitor for a 470uF cap. all replacements were panasonic fc electrolytics rated at 35v... so my pre-regulator 120hz ripple is down from 1.2Vpp to 530mVpp at 300mA draw (close to the 500mA that the amp will draw) and the small peaks on the output dropped from 2.5mVpp to 1.5mVpp! that's good news. i still get quite a bit of high-frequency ripple, but since it's outside of the audible range, i'm hoping the supply will be silent enough for use. (otherwise i'd have to redesign... YIKES!)
but it should work. :)
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
 |
|
P(re).S.: actually the amp sounds pretty awesome, even with these flaws. i tested again by "mono-izing" my headphone cable and playing through both speakers in the headphones. lovely sound. a lesson for the kids out there: NEVER IGNORE THE HEAD-RELATED TRANSFER FUNCTIONS! :)
just to make sure i wasn't loco, i went ahead and set up a single channel of my old szekeres design with my 32 ohm cheapo portable headphones, and gave it the audio from my computer's headphone out. not surprisingly, it warmed up and i was listening to the first 30 seconds of owl city! still not surprisingly, the combination of woeful LF response in the old design and a lower output impedance than the amp was originally designed for meant almost zero bass. the highs were weak too, attributable to the (safety first!) low volume i was using.
"this sounds like bad news," you say.
sure, it does. but it's not. all the above issues are why i redesigned this thing. one obvious positive note here is that, using my benchtop dc supply, i had no audible hum! :) that's good, since the figures for hum are quite close between my bench supply and my custom supply. let's hope my ears weren't playing tricks on me! lol.
anyways, the supply and amp channels will probably be done soon. probably.
'til next time.
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
my power supply works! it's currently putting out 14.7Vdc, and i'm perfectly happy with that, since i wasn't intelligent enough to add a trimpot in series with the shunt resistor... good regulation and 2.5mV peak ripple; i'll be interested to see whether that ripple is audible in the amplifier. if it is, it may be back to the drawing board... :/ i hope it works out.
really i just need to finish the wood side panels of the case, and i can button it up and call the power supply finished! :D
more next time,
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
as you know i work from time to time on the gnome-gmail-notifier, a project that aims to provide new mail notifications for gmail users from the gnome desktop. i've been receiving an increasing number of stupid questions and requests lately, but this tops it...
earlier today i received an email, and was notified promptly by my notifier, that someone had posted an issue on my googlecode page. i decided to read it. the verbatim content of the issue follows:
I'm using gnome-gmail-notifier on Ubuntu 9.10 (Karmic). Unfortunately, the
alerts provided by this software only shows the total number of unread
e-mail I have.
If a new e-mail arrives, it only shows that the amount of unread e-mail
increased. I cannot understand how a notification system could be really
useful by only giving a number. I'd like to be alerted on WHAT I've just
received by e-mail and not HOW MANY pending e-mails I have.
What I expect from a e-mail notifier is an desktop alert showing the new
e-mail subject and the sender address. I'd like to request it as a feature
for gnome-gmail-notifier.
alongside this quite demanding "issue" was an attachment:
my response, not posted on the issue page itself:
you've made your request quite plain, leading me to my first grievance with this project. though i originally posted the gnome-gmail-notifier in a casual attempt to share my work on a gmail notifier for the gnome desktop, apparently, through its binary packaging into mainstream "desktop" distributions of linux, it has become almost a source of tedium for myself. i feel that i am spoonfeeding an application to linux users who may not even know the shell exists, much less how to effectively use it. if use of microsoft windows can be equated to de jure segregation, i'm sure that my situation can be compared to de facto segregation. in other words, is this not still vendor lock-in? instead of proposed patches against my source, the attachments to issues are screenshots of what users "don't like" in the application. pathetic.
the fact that you have posted your request in this sort of language on an open-source project's issue tracking page leads me to my second concern. the intent of open source is free and open communication of ideas. one specific caveat, however, is that this communication is only mutually beneficial when both (or all) communicating parties are knowledgable. in other words, a user communicating to a developer cannot relate his concerns on the level of the code on which the developer works; therefor, open source is naturally maximally effective when all parties can read, understand, and modify the source code in question. issue tracking, the patch and diff utilities, and revision control were all designed with this in mind. if users begin to "expect" a certain behavior from an open-source application with a feeling of entitlement, they are wasting their time using open-source software.
finally, what idiot has this many unread messages in their inbox and expects an inbox notifier to understand that they are simply too lazy to read their mail? the gnome-gmail-notifier is not a revision control system like subversion, it is designed for those who want to STAY ON TOP OF THEIR INBOXES. (in its current state, that is.) this is in accordance with the KISS principle. perhaps a future version of the notifier may integrate the behavior of other non-free notifiers, but this is currently infeasible.
honestly, i can't stand the stupidity of most linux users out there today, who have evidently downloaded an ubuntu ISO and expect it to "just work" and be exactly what they expect for zero dollars. sure, free software developers try their best to respond to the needs of their users, but if you want to be fully accomodated in your expectations of the software you use, without any programming knowledge whatsoever, you should have never switched to linux. go back to vista. NOW.
5 Comments | Post A Comment | Add to Memories | Tell a Friend | Link
well, i had wanted to be able to add quick tooltips to pictures on my project pages, like the main page, and i finally got around to doing it this morning. again, the code i use is formatting-agnostic; i simply create a php array with key-value pairs ("filename"=>"tip") inside a file called tips.php in the project directory, and the skeleton code does the rest! you can see the tooltips in action at geekysuavo.org/lassie, which is the new home of my headphone amplifier progress page. (yeah, i named it lassie. i was drawing a blank. random, i know.)
i even hacked out a quick method of automagically generating section links inside the ajax tooltips on the main page, but sadly the tips fade out cruelly before you can click them, and i'm too tired to fix the javascript for that right now...
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
"well i heard the tub running and saw a red light, so i figured there was a good reason the lights were off..."
- joe, one of my roommates.
quite true, quite true. i bought a 25W red-color bulb to use as ambient lighting during my pcb fabrication process, and it worked like a champ! the traces came out just as crisp and dark, and i could actually see! :P that solves it, i will always use a separate red lamp while developing boards. it just makes life easier.
and i'm considering trying the electroless tinning compound as well... we'll see. i was going to try and mix some on my own in the chemistry building, but it turned out the raw materials weren't readily available in the lab i work in... :(
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
well, i knew my redesigned szekeres amplifier boasts improved measurements over the original design, but i hadn't quantified them until now. i have to say, i'm quite pleased that my small changes have produced such improved performance. :)
 
since you can count them on one hand, my (essentially obvious) modifications are:
- increased input coupling cap from 1.0uF to 3.3uF (widens lf response)
- increased output dc-blocking cap from 470uF to 4700uF (widens lf response)
- increased supply voltage from 12Vdc to 15Vdc
- changed bias network from 100k/220k to 68k/100k (decreases current noise)
needless to say, this encouraged me to go ahead and hack out a regulated 15V dc power supply for the new circuit, which is now sitting on a PNG file ready for primting. yes, the design is my own, sporting PP-film-bypassed MUR820 rectifier diodes, a CLC prefilter, and an LM350K regulator with protection diodes. of course, this circuit and its toroidal transformer would be awfully jealous if i enclosed them in a cheap ABS plastic case when the unit they power is so fabulous, so i've started sketching what will be another beautiful aluminum and oak enclosure. of course. ~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
when better than a rainy day to spend some relaxing time in the machine shop?
so, i decided to drill out and cut the circuit boards i etched earlier in the machine and fabrication shop, and since the boards were just sitting there on my project table, taunting me, i soldered in the components and took it for a test run:
 
i turned the amp on with a dummy load, gave it a 1Vpk-pk sinewave, and the CCS resistor turned red hot and started smoking! as it turned out, the tabs on the TO-220 packages weren't adequately isolated from the heatsink, but the issue was solved temporarily by removing the screws from the mounting holes. i'll still have to find a way to use the isolate the circuit from the screws for the final mounting...
the good news is that the circuit works!!! :D i get an accurately reproduced signal at slightly less than unity gain, and the onset and majority of clipping is second-harmonic, which is great. the active elements don't even get too hot!
since the current version is working so well, i redesigned around a higher supply voltage of +15Vdc, a more reasonable bias point, and a much-improved low-frequency response.
more details here and on geekysuavo.org/follower...
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
well i've done a few things since the last post i made... golem seems to work as a guitar amp should, amplifying and grossly distorting a small signal, though i haven't actually heard it due to a lack of speaker cab and electric guitar... at least it doesn't explode though. :P that's always nice. (i'm still posting updates and info on golem at geekysuavo.org/golem.)
i've started work on a class-a mosfet headphone buffer/amplifier based on the szekeres design, which has been around forever. the case is a milled aluminum "channel" with hardwood side panels, and is done for the most part. i just need to get rolling on drilling and populating the pcbs i etched so i can do a test run of some signals through a dummy 50 ohm resistor. updates and information on the amplifier design and construction are available at geekysuavo.org/follower.
i still haven't gotten my netbook wifi to work on the TTUnet system now that they've changed over to WPA2... quite frustrating, to say the least. oh well, my room wifi still works great.
the gmail notifier is still in the process of a rewrite, but it shouldn't be long before i get a stable revision into the googlecode subversion... i mean, theoretically.
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
well, i've finished soldering the last of the connections in the golem amplifier today, and it'll be a mere matter of weeks until i can perform the first power-on and signals tests of my first tube amplifier! (heck, my first amplifier ever!) i've also posted pictures, schematics and basic explanations of the design and its history at geekysuavo.org/golem for everybody (everybody == those who are interested) to take a gander at! :) once i have results from the "hope this thing doesn't blow up" test, i'll post those as well, and perhaps sound recordings of the final tweaked product!
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
well, this is what happens when you've worked on quantum mechanics homework for the evening and try to make coffee for the next morning:
but it's all ok. i decided the coffee would taste better if i ground the beans first...
moving on, i've been looking at various tube amplifier circuits again, namely kt88 seul and e88cc phono mc preamps, so i decided to take a stab at a simple constant current sink (CCS) for the heck of it. providing a load-stable current sink for power output tubes brings their loadline into a more linear range for higher output efficiency, or at least that's my understanding. so... i hooked a the base of an mpsa42 npn i had laying around into a 10k/10k resistor divider (i know, not as good as zener, but i didn't have any), connected its emitter to ground with a 47 ohm resistor, and got 13mA (at 3.3V) through my lovely blue led. (Ice=(Vin*(R1/(R1+R2))-0.6V)/Re, where R1 and R2 are the base resistor divider and Re is the emitter-to-gnd resistor.)
cool, huh?
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
yeah, pwm from an lm324 running on a single supply with its triangle input coming from a function generator whose dc offset is manually tuned to give a 0V offset at the output isn't really my idea of hi-fi either, but it was a fun experiment.
soon i'll be able to hack a bit more effectively on class-d, using a few ad823 16mhz opamps and some lm393 comparators. it'll be interesting to see if a single ad823 can perform adequately as a triangle generator, or if a true implementation at 300kHz really does require a comparator... i'm excited! (first thing's first though: golem is almost done! :D)
well that post was useful, :/
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
so i finally tuned a pid controller for my nixie clock boost converters that use the atmega16. using gnu octave, i set up a simple simulation that crudely modeled a boost converter and built a simplistic pid loop around it which converged nicely, as any underdamped oscillator should do. fantastic, right? well, not quite. performing a 30-second simulation in octave using dt=2ms took way too long, so i wrote a C program. this also allowed me to constrain my variables to the types they would take on in the avr itself. hmm. strange behavior. the fact that my PWM duty cycle could only take on 8-bit values meant that the plant i was controlling had limited resolution. tweaking the gains allowed me to minimize oscillations at equilibrium to small peaks, which i then filtered out using a simple discrete-time low-pass filter. the result is an output voltage which is more behaved, more accurate, and better damped than the previous (if a < b then pwm++) method.
observe (pdf):
  
see? pretty cool.
~ brad.
p.s.: oh and the c-code runs ten times faster. i just simulated 30 seconds at dt=32us, which is the "real-time" speed, and it worked like a champ. :)
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
so i've been slowly revising my website with small tweaks and improvements, but this addition today has to be the best yet. until today, i had written separate php scripts providing "helper functions" for all my project pages, each requiring its own maintenance and having its differences to keep track of. each project also had its own css stylesheet and a header and footer. no longer! as of today, my led clock and nixie clock project pages are using the same global project function base and stylesheet. that means that the code i see for project pages is COMPLETELY markup-agnostic, xhtml 1.0 strict compliant, css 2.1 compliant, and easy to modify. my function base even keeps track of the html indenting and tidying! finally, on top of the benefits of a centralized codeset, the project pages automatically will contain floating section menus which link to each section of the (increasingly long) pages. what's more, the section menus are AUTOMATICALLY generated from my markup-agnostic functions!!! when i add a section called "introduction", the code keeps track of it for generation of the floating menu! AAAAH! IAMTHESUPERMASTERHACKER!@!@!!!!!
haha.
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
| 2009-08-11 08:54 |
| what's more... |
| Public |
| home |
 calm, now |
| and that right soon - shawshank redemption ost |
|
had to kill vino-server this morning. seems it was taking up the bulk of my processor for no reason and opening port 5900 even though i told it not to accept connections... so it's no longer running and my mind and cpu are at ease.
speaking to my last post, all of this work is to reach an end goal of a custom-designed crt clock. one other obvious necesity of a crt clock is an HT supply of +2kV. yep. so i've been looking at flyback converters online, but honestly the information out there is geared towards low-voltage applications. what's more, it seems i would have to build my own transformer for the design. (iron-powder toroidal core is probably going to be my first attempt.) sounds fun... egh.
well, it seems i've forgotten an important part of this whole thing: in order to work with the opamps i intend to place in the crt clock design, i need a dual-rail power supply. therefore, my next project. :) a dual-rail tracking 0-30V DC power supply based on the jung super regulator.
since my electronics are completely packed up at the moment, i have no way of doing the breadboarding required for a low-power prototype, so it will have to be later on that the project comes to life, perhaps next summer break...
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
ok, so i want to make a "radio-controlled" clock that determines the time based on a signal received from the WWVB radio station in fort collins, colorado. but there is a small problem. nobody out there has built one without either hacking a commerical "atomic" clock or buying the cheap $10 prebuilt receivers that do the dirty work for you.
what happened to do-it-yourself? i understand RF is tricky, but could winding your own loopstick antenna be that bad? i want to find out. so according to some back-of-the-napkin calculations, my custom antenna will have a ferrite with mu=800, l=4", d=0.37" (mu_rod=60). for a 5% (i know, high) ceramic capacitor at 6800pF, a resonant LC circuit at 60kHz would require a 1.035mH inductor. assuming 24awg wire wound a single time around the middle of the ferrite, my estimations say 99.45 turns should do the trick. a caveat to this is that using a cap that is actually 5% of its rated value will cause a 23dB drop in signal reception, so multiple caps need to be tested well to find the closest match.
because the signal from WWVB is really pulse-width modulated amplitude-shift keying (PWMASK, anyone?) and not audio AM, my thoughts are that a simpler receiver may be constructed... how about a buffer and a precision rectifier (or even just a diode detector), feeding into a comparator for digitization of the resultant signal? right now it's just a guess, since there is no antenna to test the idea with. but soon.....
more napkins wasted: if i used a time constant of 250us in the detector, the rise/fall time of the resultant signals would still be quite sharp enough for accurate input capture by the avr, which raises another question: why not use the analog comparator of the avr instead? good idea.
it sure would be nice if i could start hacking on this right now, but i'm stuck to wait until the next break, probably winter break or spring break. in the meantime, it's just an idea.
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
well, i've posted two videos onto youtube: a quick shot of my finished nixie clock and a late introduction to my all-digital thermometer. i used mencoder -oac mp3lame -ovc lavc -vf scale=640:480 -lavcopts vcodec=mpeg4 -o output.mpg input.avi, which seems to work well. soon i'll post a video of my finished led clock!
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
how's this for an avr programmer target? it has an LM7805 for +5V DC, headers for connecting the AVRISPmk2 and wires to a breadboard, and a 40-pin DIP ZIF socket for all DIP-package avr micros.
well, that's it. :)
~ brad.
Post A Comment | Add to Memories | Tell a Friend | Link
|
|
|
|
|
anxious
irritated
exhausted
okay
contemplative