Lightspeed Attenuator a new passive preamp

[jkeny]

Hi Rellum,
How much more work would it take to include an auto calibration feature in your set-up? i.e something that applied a series of currents to each LDR & recorded the resistance for each current. I read through your posts quickly but didn't really understand where the table lookup for each LDR was being stored & how it was created & used? I'll have another read

[jimbo51]

Originally posted by jkeny

were you lucky? What's the average hit rate? You can't get suppliers to agree to pick parts from the same batch, they will all refuse to do this.

Per the data sheet for part numbers with the "S" suffix:

"Packaged in ranges. Printed with part number, R2
followed by a letter. Individual ranges not available
separately. Range distribution is not guaranteed."


R2A 100 110 IF = 1 mA (guaranteed +/- 1 range)
R2B 110 122
R2C 122 135
R2D 135 149
R2E 149 164
R2F 164 181
R2G 181 200

Assume it's still not tight enough but a start?

[jkeny]

yes but electronics suppliers won't guarantee that they will pick all from the same range such as "G" for example - so then it's pot luck

[udailey]

Originally posted by jimbo51



Per the data sheet for part numbers with the "S" suffix:

"Packaged in ranges. Printed with part number, R2
followed by a letter. Individual ranges not available
separately. Range distribution is not guaranteed."


R2A 100 110 IF = 1 mA (guaranteed +/- 1 range)
R2B 110 122
R2C 122 135
R2D 135 149
R2E 149 164
R2F 164 181
R2G 181 200

Assume it's still not tight enough but a start?

This is where I got kind of ticked off at Silonex and Allied. They wanted 3.79 each for these. I talked to Silonex technical support about how they end up grading these. They apply a certain milliamps for a few seconds and make one single measurement then they feel like they can apply a grade to it. You can have 4 come in at grade "e" lets say and then end up with 4 that have wildly different measurements at other amounts of current. You have to do more than one measurement and 3 is a minimum. You also have to test them for a long time under the applied current so that the entire encapsulation comes up to a temperature that becomes steady in ambient. This is not just fiddly techno/audiophile freakiness. It really is necessary and for them to grade them based on one single test for only a few seconds is really ridiculous. If you bought all graded ones you would have a slightly better chance at getting matches. I buy ungraded and grade them myself as their grading standards are poor at best.
If youre not going to buy them already matched then at least buy a large handful of ungraded and try to buy enough that you get a price discount. Maybe someone will buy your extras so they dont have to buy so many to get more matches.
Its true that with many of the near matches the resistance at the same milliamps will pass each other back and forth on a curve. This will result in the image wavering left to right and is annoying.
Uriah

ps some type of autocalibration would be really really nice and would be a great help so that more people can make these attenuators. Of course it would cost so maybe it would be a draw as to whether to match or autocalibrate but I think it would be really neat to have a microprocessor do it.

[Tolu]

Auto calibration isn't possible I think!

The optocouplers are very temperature sensitive. You have the evironment temperature that has an influence and the part temperature that differs very much.

I did very much attempts with a 16bit DAC and the SR2S. Manual matching is the way to go. You can get 0.1 dB channel matching over the whole range! But it is the hell!

Try thermal coupling with a constant heat source of about 40° C.

[jkeny]

Udailey,
When I spoke to them they wouldn't even pick from the same batch run either so their "matched" premium prices are a joke - just another way to gouge the public

Tolu,
Why is auto calibration not feasible? Surely if the can be hand matched they can be machine matched? I'm talking about juts putting in any two LDRs into their working environment & letting a PIC processor build a lookup table for each which would give the correct current needed to produce a certain resistance and thereby a certain dB.

[udailey]

It would be cool if you could drop any two into one channel and then use your volume pot to change volume on that one channel. Then put two others in for the other channel and have a pic microcontroller measure resistance on all 4 then alter the current to the 2nd channel LEDs so that the resistance of both channel's Series resistors match and the Shunt match as well.
Uriah

[jimbo51]

Uriah,
Thanks for the reality check, more learnings for a newbie.

[udailey]

No problem. If you go back and order another dozen of the cheap ones you will surely find a suitable stereo setup when combined with the ones you have.
It might take a week or so if you measure one by one and dont try real hard Just use a 3V battery and a pot and never move the pot til you measure each one then another setting etc. Or use a few different resistors in series with the + from the battery. Give them time to warm up and then measure and put it in a spreadsheet. Then when all done graph it out and "look" for the best match.
Uriah

[rellum]

Hi JKeny

About the actual code, the dac words are manualy stored in a long list of RETLW xxxx instructions. Storing them in the program memory is the only method provided by this chip to create an easy acces to a lookup table. I've got the values by connecting a DMM across each LDR in ohmmeter mode and activating a part of code called "setup mode". With my handy remote control, I can select whatever 16 bit word wich is then instantly transmitted to the dac by the pic. When the correct resistor values are found and not wandering too much, I read both 16 bit values on the preamp display (in hexa mode) and put them on paper. It took hours to complete for 57 steps.


Auto calibration would require an extra AD converter and a DC reference voltage. Vref would be applied at the audio inputs and the ADC at attenuator's mid points. Relays should be used to keep isolation between analog and digital side in play mode. That way, knowing Vref value and the desired attenuation, we could measure the appropriate voltage at each mid point. An iterative programming method would vary led's currents to reach the perfect dac word and store it in eeprom for each step.
The PIC I'm using (16F628), has no inbuilt ADC and too less memory. Some bigger ones of the same family have ADC and Vref capabilities, but with only 10 bits of resolution. This is short if high levels of attenuation must be reached. Vref should be of low value and only applied during measuring to minimize temperature effect, wich is really a problem with those LDRs.
This is for an R-LDR attenuator. For an LDR-LDR one with constant input resistor value, Vref should be replaced by a precise constant current source and the voltage measured at the mid point AND input point.

For me, that would imply an entire redesign of my pic card and added relays on the analogue board.
What Tolu tried to tell you is that even if feasible, the end result, precision wise wouldn't give better results than a good simple hand matched process, probably even worse at low attenuation settings. What you think about is nice but I find it highly overkill.

When it comes to ageing, I always turn my hifi off when not in use and don't feel concerned.

And nowadays, such projects take too much of my time.
I still didn't find some to compare my pre with an ptivol/SKPre combo that a friend of mine seems to like a lot.


Cheers
Francis