Author Topic: EEVblog #450 - Ebay Unboxing  (Read 16009 times)

0 Members and 1 Guest are viewing this topic.

Offline dexters_lab

  • Supporter
  • ****
  • Posts: 1827
  • Country: gb
    • DextersLab2013
Re: EEVblog #450 - Ebay Unboxing
« Reply #25 on: April 06, 2013, 11:16:37 pm »
hi dave !

you may want a get this from ebay for a next teardown : http://www.ebay.com/itm/NASA-ARTIFACT-VPI-Vehicle-Power-Interface-Rack-Console-Hubble-Space-Telescope-/261090432601

i’d buy it myself to convert it into a fancy coffe machine but i don’t have enough room in my kitchen!

Nice, but i shudder to think what that cost the american taxpayer!
"A common mistake people make when trying to design something completely foolproof is to underestimate the ingenuity of complete fools." - Douglas Adams
https://www.youtube.com/user/DextersLab2013
http://dexterslab2013.blogspot.co.uk/
 

Offline Ferroto

  • Frequent Contributor
  • **
  • Posts: 289
  • Country: ca
Re: EEVblog #450 - Ebay Unboxing
« Reply #26 on: April 06, 2013, 11:34:16 pm »
just in case the fundamental constants change LOL!


 

Offline G7PSK

  • Super Contributor
  • ***
  • Posts: 3675
  • Country: gb
  • It is hot until proved not.
Re: EEVblog #450 - Ebay Unboxing
« Reply #27 on: April 07, 2013, 09:20:56 am »
Remember the photographs that came in an envelope printed "Photographs Do not bend" and they would come through your letter box doubled over and the postman wrote in pencil on them " Oh yes they do"
 

Offline BobC

  • Supporter
  • ****
  • Posts: 115
Re: EEVblog #450 - Ebay Unboxing
« Reply #28 on: April 07, 2013, 06:58:50 pm »
In the context of an electronics lab I think it would be useful to measure clock jitter. Apart from that, no idea.

With an MCS at hand, it becomes possible to perform measurements that would otherwise be impractical, generally due to needing an entire bench full of equipment made of unobtanium.  You soon develop the mindset of trying to convert almost any signal parameter into an analog pulse train by adding simple circuitry to convert the parameter(s) of interest into pulse characteristics.

Let's take your clock-jitter example.  Pulse analyzers are very good with short-time events, but less good with longer term events (their clocks are accurate only over relatively short windows).  Since jitter is, by definition, the irregular timing of individual edges, we would then want to convert each edge, both positive-going and negative-going, into a pulse.  There are many ways to do this, most of them needing less than $5 in parts.

The neat thing about working in the pulse domain is that you can get away with some really primitive pulse conversion circuitry, not even caring much about the component values.  This is true simply because we tend to care about differences between pulses, the relative measurements, rather than the absolutely correct values.

One of the hardest things to do in many analog circuits is to correctly analyze noise in multiple domains (voltage, current, amplitude, phase, harmonics, short-term vs. long-term, etc.).  Moving things into the pulse domain can simplify such characterization.  I once had to characterize extremely low frequency noise on a 2KV DC power supply:  The equipment being fed by the supply was exhibiting long-term semi-periodic irregularities, and every part of the system had to be examined to find the cause.  My job was the power supply.

The first obvious thing to do was to scale the output voltage down and feed it to a 6-digit DMM.  But the DMM itself wasn't stable enough over the long times we were interested in (hours to days).  Even the 12-digit meter in our calibration lab wasn't stable enough for long-duration measurements.

What I did have access to was a simple voltage-to-frequency converter (VCO), a well-controlled oven, and high-precision voltage reference that WAS stable for days and even years at a time (even inexpensive references are generally more stable and far cheaper than the expensive instruments needed to measure them).  I put the VCO into the oven, let it stabilize at a temperature a bit above ambient, fed in my precision voltage reverence, routed the VCO output to a simple pulse converter (also in the oven), and let the MCS run for a few hours.  I got a very tall and skinny Gaussian distribution of pulse times that represented the stability of the pulse conversion circuit. 

I then fed in the supply and took another few hours of data, and got another Gaussian distribution (I used the narrowest timing bins the MCS could provide).  The neat thing about Gaussian distributions is that they provide a mathematically simple way to extract meaningful characteristics from huge amounts of data.  I first normalized the distributions, then subtracted the reference data from the supply data, and got a distribution that reflected the instability of the supply. 

For the few hours data I took, the result was barely a wiggly line barely distinguishable from zero.  Were a variation present, I would expect the DoG (Difference of Gaussians) to have one or more distinct humps of its own.  I repeated the experiment for both the reference and the supply for a full day each, and got nearly the same low-level result:  The power supply was not the source of the data variations!

The only possibility that I was wrong would require the time-domain instability of the pulse conversion circuitry, the MCS, and even the voltage reference to PRECISELY match that of the supply under test: Not likely at all, unless you include an infinite number of parallel universes, in which case it is certain to occur in one of them.  But not in our universe.

In today's world of cheap, fast, wide ADCs with almost unlimited sample depth, tricks such as the above are needed less often, and the MCS market has dwindled away accordingly.  One unfortunate result is that technicians and engineers have become less familiar with domain conversions in general, especially quick & dirty ones in the lab, as well as the statistical analysis chops needed to make sense of the results.

Not that I'm complaining! I did the above while I was a technician at a company that made control systems and instrumentation for commercial and military nuclear power plants.  When I went to college, I majored in Computer Engineering (CS + digital parts of EE), and today I make a nice living applying some of those old tricks in the software domain, to make nasty sensors and touchy control systems play nice.

A primary key for me has been knowing how to transform between domains:  From voltage to frequency, from edges to pulses, from whatever you can't easily measure or analyze, to things you can.  To know when you've done the conversion correctly (by creating and performing tests that yield unambiguous results), and to detect when it is no longer correct.

That's why I'm an engineer:  I get to apply a world of tools to solve nasty puzzles, and I get to do it every day.  Learning a new tool, and using it correctly, keeps me jumping out of bed every morning eager to attack my next problem.  My favorite problems are the ones that kick my butt and make me feel like an idiot, for I know I will be less of an idiot when I finish!
 

Offline Smokey

  • Super Contributor
  • ***
  • Posts: 1604
  • Country: us
Re: EEVblog #450 - Ebay Unboxing
« Reply #29 on: April 07, 2013, 09:14:03 pm »
^^^^^  BobC wins post of the day.
 

Offline BobC

  • Supporter
  • ****
  • Posts: 115
Re: EEVblog #450 - Ebay Unboxing
« Reply #30 on: April 08, 2013, 04:29:40 am »
^^^^^  BobC wins post of the day.

I should probably share what caused the instability:  After exonerating the HVPS, I was asked to compare the problem data behavior to the variations in various environmental parameters: Line voltage, line frequency, humidity, temperature, EMI (electro-magnetic interference), and so on.

And wouldn't you know it, the observed variations approximately tracked the daily temperature changes in our test bay.  Once I shared this with the engineers, a heat gun and a can of cold spray soon identified the culprit:  Our instrument contained a logarithmic amplifier created using a matched Darlington pair in the feedback path of an op-amp.  The transistors were precisely matched electrically, but they exhibited thermal effects the initial design had failed to account for.  The transistor and op-amp parts were carefully selected to have matching thermal characteristics, but in reality they didn't: Specs lie!  Adding a suitable thermistor nicely eliminated the effect, with the thermistor, Darlington pair and op-amp all mounted to a common heat spreader.

That logarithmic amp was a true work of art: It was dead-accurate over 6 decades of operation.  Then they "improved" it to cover 9 decades, and I was asked to validate the implementation and create the calibration procedure.  While the amp was well shielded in normal operation, I had to remove the shields for my tests.

When you are operating down at the bottom of 9 decades, you need to measure to 10 or 11 decades, which means measuring quantities that start with "femto" and "atto".  I had to enclose my lab bench in a Faraday cage (copper window screen material), and had to avoid using any instruments that emitted significant noise (which included most digital instruments in the early '80's).  I had to feed the AC power for my instruments through a 60 Hz resonant isolation transformer to eliminate line noise, and also to get ground isolation (as Dave so brilliantly demonstrated in one of his videos).  I did need an extremely low impedance earth reference, so we drilled though the slab near my bench and drove an 11 foot copper-jacketed steel rod into the ground, to which the Faraday cage was connected.

This setup was so sensitive that even small arm motions would affect the readings.  Including the arm motions needed to write things down.  So I would set up a measurement, go to the other end of my lab bench, wait for the setup to stabilize, then use binoculars to read the instruments.  Took nearly two weeks just to take the data needed, and a month and a half to get the job done.

We didn't quite reach the 9 decade goal, but we did get a rock-solid 8 and a half decades.  Left our competitors in the dust: They needed two very expensive systems to measure the range we covered with one.  It took them most of a decade to catch up to us.  (A "decade", get it?  Nudge-nudge, wink-wink.)
« Last Edit: April 08, 2013, 04:32:19 am by BobC »
 

Offline Fezder

  • Regular Contributor
  • *
  • Posts: 103
  • Country: fi
  • Lean on, or was it learn on?
Re: EEVblog #450 - Ebay Unboxing
« Reply #31 on: April 08, 2013, 02:42:48 pm »
woa, nice length for a post! hey, not in negative way.....and all was in sense too :).
Both analog/digital hobbyist, reparing stuff from time to time
 

Offline stratogazer

  • Contributor
  • Posts: 15
  • Country: us
  • Rank Amateur
Re: EEVblog #450 - Ebay Unboxing
« Reply #32 on: April 08, 2013, 06:10:29 pm »
Dave,

One of the uses for a photon counter is in radiation dosimetry. Certain crystalline compounds absorb radiation. When these crystals are heated they give off photons proportional to the amount of radiation absorbed. The size of the crystal being known, one can calculate the amount of radiation absorbed by the crystal.

How ya gonna test it? :-DD
Learnin' little by little .  .  .
 

Offline M0BSW

  • Frequent Contributor
  • **
  • Posts: 674
  • Country: 00
  • Left this site 2013, they will not delete it ????
Re: EEVblog #450 - Ebay Unboxing
« Reply #33 on: April 11, 2013, 04:09:02 pm »
Send the seller this example


I actually caught Yodel driver throwing,one of my parcels over the gate, then the very following day City Link did the very same thing, despite seeing me come to the front door to take the very expensive ham radio transceiver off him, however he still threw it over the gate, "TWATS"
no one would or will tell me how to delete this account
 

Offline majki

  • Contributor
  • Posts: 21
Re: EEVblog #450 - Ebay Unboxing
« Reply #34 on: April 19, 2013, 12:06:19 pm »
Something related:

Hidden camera captures a package's journey from shipment to delivery:
http://www.dpreview.com/news/2013/04/18/hidden-camera-captures-packages-journey-from-shipment-to-delivery/print

The camera was controlled by an arduino to record 3 second video every minute and make longer videos while the box was moving.

 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf