[linuxworks]

but is that by hand selection and some test or simply by virtue that things that come off the same manuf line tend to be similar in characteristics?

I do agree with getting 'same manuf run' parts. that's GOOD stuff for almost anything! but I don't consider this matching, maybe its just an argument of semantics.

let me explain further. I recently built some amps that needed true transistor matching. and I bought a LOT of parts (in lots of 25 each instead of just 6 that might have been needed) and I got out my DMM and turned to the HFE setting and measured the gains of each of the transistors. wrote down the single number and sorted them by that. then picked pairs that were close to each other as well as weeding out the too-high and too-low ones from the pack. THAT is what I call matching.

headphones don't at all seem to be like that. there isn't a single number you can read - its that 'mountain graph' spectrum read-out that I'm talking about. you can't 'match' those. no 2 mountains are alike

 

[linuxworks]

Quote:

Originally Posted by jinp6301 /img/forum/go_quote.gif The difference between the grado 125s and the 225s are driver matching, and people really like the 225s and dont really like the 125s. Something to keep in mind

is this anecdote or is there firm data to support this?

any transparency on how they 'match' ? is this documented or just hearsay?

 

[jinp6301]

Quote:

Originally Posted by linuxworks /img/forum/go_quote.gif is this anecdote or is there firm data to support this? any transparency on how they 'match' ? is this documented or just hearsay?

From the Grado Labs website on the 225is

Quote:

Features are the same as the SR125i, new driver, cable and housing designs, but this headphone has an increased rear air flow, utilizes Grado’s larger ear cushion and has closer matched drivers.

increased air flow = metal mesh
larger ear cushion = bowls

 

[geremy]

With sennheiser, couldn't it just be that ALL 650 drivers or whatever have to meet a certain spec. to within 0.5 dB? Not that they individually match driver pairs...

 

[aos]

Quote:

got out my DMM and turned to the HFE setting and measured the gains of each of the transistors. wrote down the single number and sorted them by that. then picked pairs that were close to each other as well as weeding out the too-high and too-low ones from the pack. THAT is what I call matching.

Unfortunately, you are mistaken. It's the same situation as with headphones - there IS a graph. hFE is not just a fixed number for a transistor, DMM shows you a value at a particular operating point. There is a hFE graph (usually several of them) and you can find them in datasheets. Fancier transistor matching uses curve tracing (or whatever they call it) which is not really different than trying to match two headphone drivers based on their measured response. Likewise you can select a particular frequency - say 1kHz - and pick drivers that produce the same sound level within some tolerance and call that "matched".

 

[wang0215]

i think hd600 are more useful for CLASSICAL music~

 

[linuxworks]

AOS: I hear what you're saying but I've never been instructed to use data sets when matching transistors. the single scalar number seems to be what you use, since the test conditions are the same (temperature and voltage under test).

who in the world matches transistors by CURVES? do you know for sure that anyone truly does that? for audio??

speakers are going to be WAY more non-linear than transistors and I just cannot see how anyone can make the judgement call about which drivers are more 'close'. its a compromise call, at best, and I am not convinced that its worth the effort you'd need (and the # of samples you'd have to have onhand to find proper 'matches').

 

[AzN1337c0d3r]

Quote:

Originally Posted by linuxworks /img/forum/go_quote.gif who in the world matches transistors by CURVES? do you know for sure that anyone truly does that? for audio??

No you match by frequency range, within a tolerance, as someone said earlier in this thread.

Transistors replaced tubes, and tubes could be bought in matched pairs.

 

[AzN1337c0d3r]

Oh btw, you don't even need to match one driver to the other, you could just match to a "target". Reject the ones that don't make the grade and either use it in lower grade products (HD580 anyone?) or tweak them till they do match the target.

This actually makes the most sense. I believe the HD600 are claimed to be matched to within +- 1dB. Perhaps anything that doesn't meet this grade but is within +- 2 dB gets sent to the HD580s, and then anything that doesn't meet that is thrown away. They've probably improved manufacturing process so much that they were putting matched drivers into HD580 anyways and could discontinue the line and not waste too much money throwing away drivers.

 

[aos]

Quote:

AOS: I hear what you're saying but I've never been instructed to use data sets when matching transistors. the single scalar number seems to be what you use, since the test conditions are the same (temperature and voltage under test).

That's because if you match two transistors hFE at a particular point, there's a good likelihood they will remain close to each other in other points as well. But it's not a guarantee. And curve tracing requires more complex equipment (oscilloscope), more time and more skilled operator.

People have a need to simplify things and wherever possible describe an otherwise complex process with a single number. Usually this is good enough, but it is only an approximation. Without knowing how likely are two drivers to differ at, say, 10kHz if they are from the same batch and match at 1kHz you can't dismiss this matching procedure any less than you can do with transistors.