Partial Response, Maximum Likelihood

In the early 1990’s a very cleaver little formula become popular in computing and telecommunication fields. This little formula was called partial response, maximum likelihood (PRML), and it was able to identify the digital data stored in a very weak analog signal.

That might sound rather boring, but PRML was one of the turning points for stored and transmitted data. All of a sudden the amount of data that could be stored on a hard disk grew exponentially. Before PRML a big hard disk would have been 80MB, and it was limited because the space needed to store a bit of data was large due to the strength of the magnietic field needed to write a strong enough data bit. but as soon as PRML hit the streets a weaker magnetic field could be used and the space needed to store each bit was reduced dramatically. This allowed the capacity to jump up to 120MB, 250MB then 500MB and then 1GB, now drives are hundreds of gigabytes and often multiple terabytes per 3.5 inch disk (with multiple platters in each disk drive).

And as the capacity was growing, but the size of the disks stayed the same, data transfer rates also grew dramatically. This allowed Microsoft and others to create much larger bloatware. If it wasn’t for PRML maybe we would never have seen the paperclip in word, or be able to store just such an immense amount of porn and pirate movies on the internet.

And of course PRML was responsible for much more than just hard disks becoming massive. Another amazing PRML child was the digital cellphone. Before PRML cell phones were analog which meant that the quality of sound you heard in your ear was full of pops and whistles and people saying “hello can you hear me, this fu&*ing piece of Sh$t phone gaaaaaahhh”. And the phones themselves were much larger to deal with the much higher transmission powers.

But since PRML the sound you hear in your ear is crisp and clear, as it’s digital you either get a sound or you don’t. And with PRML checking every bit to work out if it’s a one or a zero amongst all that static and interference on the radio spectrum what you hear is clean. Of course cell phones still drop calls (yep, a lot less than decades ago, but still enough to drive us nuts). But when a call quality is so low that PRML can’t work out the signal, what you hear is silence, rather than the person at the other end screaming expletives into the mouthpiece.

As I’ve basically grown up with PRML, I’ve discovered another even more valuable use for the technology.

I use my own version of PRML on conversations people have with me. It’s so much easier than listening to every word. Now I just listen to the odd bit of a sentence and my PRMLing brain fills in the missing bits.

I find this process to be perfect, and I’ve convinced myself that I never miss anything important.

I suspect my wife does not totally agree…..

(133)

(1185)

Be Sociable, Share!

3 Replies to “Partial Response, Maximum Likelihood”

  1. That, presumably, would mean your wife is actually Mrs. Trellis, presumably of North Wales.

    I claim my five pounds, and would observe that discovery has clarified something that’s been bugging me for ages.

    Do you play the Swanee Whistle, by any chance?

  2. Trellis coded modulation is something entirely different. Obviously it helps encode more data on an analog signal. But it does not deal with separating the good data from the crap white noise and interference on every bit of the spectrum

    1. Indeed so, but Trellis coding enables the error checking.correction that makes PRML workable in the real world.

      I don’t know if cellphones use Trellis coding (ought to Google it really) but the approach was developed for the “super-fast” modems of our past lives (28.8kbits/sec — Whoopee!). and has been a core element of all high-packing-density transmission/storage systems since.

      Trellis coding, as applied to domestic spousal conversation, has clearly evolved naturally as part of human intelligence. I’ll admit it is an under-researched area, but the evidence is there, begging to be exploited by an eager PhD student in Hull or Poughkeepsie State (or wherever).

      Anyone with more than six months of marriage under the belt is well aware of predictable conversation patterns: “she said that, so she’s now going to say this, or perhaps that.” When this relates to stuff I’ve left out on the kitchen table — knives and forks, parts of old motorbikes, etc. — it becomes very efficient indeed.

      In fact, only five words are necessary to decode the entire conversation: “WHEN are you going to…?” It’s a perfect example of “natural” Trellis coding in action (and usually time to run).

      The brain is a truly amazing organ.

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.