I was wrong, it's only 1/3 of the story.Even for the correct current, the resistance value may not be the same from device to device. If you get the resistance values to match at one point, they may not match at other points.
This is EXACTLY why I explained that you MUST determine THE CORRECT current FOR EACH attenuation STEP and obviously for ALL LDRs !... These measures must be done either manually or automatically, then they are STORED and lastly USED to reach the correct resistor values, i. e. the required target attenuation factor !...
This is ALL the story. No more, no less.
Then you may have an impedance issue. But, this is mostly addressed by the use of a "good" buffer...
Last edited:
Let's get this right, and stop all the ** being told here, feeding the same varying currents to each led segment does not guarantee the LDR portions will all have the same resistantaces at those currents, so getting the currents the same is only a third of the problem, other third is the matching resistances at those currents, the last third, this is to be done over a 5 x spread of 1mA to 20mA to each led.
Cheers George
Cheers George
I don't have any pics since the lightspeed and the shunt are in plastic radio shack boxes with not very precise holes. It is on my todo list to make nicer boxes and clean up the wiring between the two boxes.
I did buy one of dvb-projekts lightspeed board but I overheated the ldr on one channel and I had a hard time with unsoldering the ldr so the board is sitting unused at the moment. I ended up making it with perfboard and using 2 pin terminal strips for the led/ldrs to avoid having to solder these parts. The terminal strips sound fine to me. I use them for the audio connection in my simple se. I have seen dvb-projekts shunt regulator but I thought that was for high voltages only. Quanqhao boards have one 1.5-5volt connection which is what I used.
In the production Lightspeed Attenuator I use 2 x 4 pin DIL opamp sockets side by side so all ldr's are close/touching each other, this alows me to insert the ldr's "NSL32SR2S's" without soldering or cutting their legs, then I calibrate with an audio signal generator and dual trace scope at all levels. Then once I am sure they are all close matched, I solder each leg with heat sinks quickly with a fine very hot tip "into the sockets". Then calibrate them again once cool. Then pot the whole lot together with hard potting wax, wait for all to cool again and then check the calibration one more time for good measure.
Cheers George
Cheers George
OK George since a couple of years ago you helped me, it's now my turn to help you... At least to clearly understand what I did to improve the idea of the LS MkII.
To be clear, I'll explain for one channel single-ended LS MkII. The attenuation Factor is noted Fi, then the series resistor of LDR is noted Rsi ohms obtained with Isi micro-amps of current and for the LDR shunt the values are respectively noted Rshi and Ishi. Of course, each measure is done with all the required attention (long time and temperature controlled, etc.). This will take some time to be done. As you would probably agree, for any given attenuation value Fi, you'll have a combination of the 2 currents for the LDRs. Right ? Then suppose by one mean or another you stored elsewhere these 2 values. Now, imagine you do the same for ALL the attenuation steps and obviously you store ALL these "doublets".
Afterwards, if you need a F attenuation value, you just have to feed the LDRs with the "adapted" currents, i. e. those you already stored elsewhere !
In fact, this is the theory, and obviously the values of the resistors in LDR will vary to some degree... This is why a new "calibration" is required from time to time (i. e. check the correct currents).
The final step of this improvement will obviously be to monitor the resistor value in real time and feed the LDRs with the required currents for a given attenuation value. Not an easy task indeed...
Last edited:
I would only add that I think using a liquid solder flux is a fantastic way to improve your chances of keeping the LDRs safe. I have found differences before and after soldering and I attribute it to overheating the LDRs which is why using some sort of tweezer/heatsink during soldering is a great idea. If you add liquid flux the soldering happens very very fast. Can be acheived in less than 3 seconds. Flux the pads.. heat the pad with the iron.. with the iron on the pad apply the solder to the iron... roll the iron over to the LDR lead... as SOON as the solder jumps to the lead remove the iron immediately. This is all you need to do to keep your hard matched LDRs matching once in the circuit.
If for some reason you need to remove the LDRs then use solder flux again. This is EVEN IF you have a rosin core solder. Also use a good copper mesh wick.
So the removal procedure would go like this: Visualize what you are going to do.. Pray.. Now apply flux to the joint as well as to the wick.. Lay the wick on the joint.. Heat the wick directly above the joint with a decent bit of downward pressure to the iron onto the joint.. when solder starts to flow you will keep the iron where it is while you drag the wick across the joint underneath the iron but remember to keep the iron in place. The wick is moving between joint and iron and it is sucking up more solder. This should take about 5 seconds.
One you have remove all or almost all of the solder you will probably still have a film of solder that is holding the LDR to the pad. You heat directly and pull the LDR out of its position on the board then clean up the board.
Good luck and really its not that tough you just have to prepare and have in your mind the procedure you will follow, then follow through exactly in this manner and you will keep your LDRs from burning.
Uriah
I should also mention that of great importance and to be remembered is what George said about a very hot tip. A relatively cool tip will take a long time to solder/unsolder. Again use flux but also use a very hot tip and things will happen very quickly which is what we want. Count on your heatsink and blowing on the LDR to remove the heat before it gets destructive. Cool tips will be damaging. Your 10w/20w iron is not hot enough. A 30W+ iron will do the trick. I keep mine around 350-400C.
Anyone that has tried to contact me in the last few days please excuse my lack of attention. I will get to everyone today and have been busy with holiday preparations.
Uriah
Anyone that has tried to contact me in the last few days please excuse my lack of attention. I will get to everyone today and have been busy with holiday preparations.
Uriah
Uriah....at last something funny !
Never mind this being the removal procedure this is what I did building the bl....y thing !!
Talk about pray, visualize and ...talk to myself during the whole nerve racking process and then blow like a loony on the LDR's !!
Apologies for this technical contribution
Andrew
I just want to add a thought on balanced operation and the Lightspeed.
Due to my current (other) project I've been forced to really take a look on balanced topology.
Although I'm a n0ob it seams to me there are two viable options for balanced operation.
1) Trasformercoupling and SE Lightspeed
2) Fully balanced build with servo
Just my $0.02, I don't know if it's worth anything?
Either way you still have to match the LDR to get L/R linearity.
Due to my current (other) project I've been forced to really take a look on balanced topology.
Although I'm a n0ob it seams to me there are two viable options for balanced operation.
1) Trasformercoupling and SE Lightspeed
2) Fully balanced build with servo
Just my $0.02, I don't know if it's worth anything?
Either way you still have to match the LDR to get L/R linearity.
As for transformer coupling, it is a step backwards from the purity of the Lightspeed, they ring and have limited bandwidth as well as other problems with inductance resistance and other drawbacks.
For servoing "sensing" the output level and auto correction for any level L/R change, this works but has a detrimental effect on the sound as well being attached to the output, a bit like dc auto servos some say and myself you can heard them when compared to a manual nulled out offset voltage in a direct coupled active stage.
The best is still to pain stakingly match the LDR's at 5 different levels over the volume span anything else is a compromise on sound quality.
Cheers George
I suppose you're right, it would be important to get a good trafo and it would still be affecting the sound. (The best one could hope for is negligible influence.)
As for the servo, I never said it was for L/R action. I was actually thinking it would be hard to implement that way. So, strictly servoing the actual balanced signal, not L/R.
(I suppose it could be done but it would be ugly and not optimal.)
Going for matched LDR's only will probably be the most transparent but matching will be important and you risk not having perfect balanced performance. In the end it's all about the price you're willing to pay.
It opens up for some interesting experiments though.
As for the servo, I never said it was for L/R action. I was actually thinking it would be hard to implement that way. So, strictly servoing the actual balanced signal, not L/R.
(I suppose it could be done but it would be ugly and not optimal.)
Going for matched LDR's only will probably be the most transparent but matching will be important and you risk not having perfect balanced performance. In the end it's all about the price you're willing to pay.
It opens up for some interesting experiments though.
Even matched for 5 points isn't enough to insure a correct attenuation range ! Unfortunately...
It must be determined, for each attenuation step, the required current for ALL the LDRs... This is mandatory otherwise a mismatch will happen. Obviously, for balanced operation (and I know what I'm speaking...), these correct values are definitely required.
Last edited:
Let's not be hasty and dismiss georgehifi that fast.
There are problems but they might be of a theoretical nature. Mathematically correct might not be necessary for good irl operation.
Bearing in mind general speaker behaviour, small deviations might be negligible as long as the sonic signature is pleasing. Theory and reality doesn't always add up.
There are problems but they might be of a theoretical nature. Mathematically correct might not be necessary for good irl operation.
Bearing in mind general speaker behaviour, small deviations might be negligible as long as the sonic signature is pleasing. Theory and reality doesn't always add up.
Any form of monitoring the resistance of the LDR's means it's going to be tacked on the signal path "not good" and this would be the only other way one could keep all LDR's tracking together, been there and done it. The only and best way is by matching them at the minimum of 5 points, this gives an overhaul deveation of around +- 1db, undetectable when using the NSL23SR2S which go down to 40ohms at 20mA low enough at a whisper for any system, save a pair of 110db horns or JBL's and such with high gain amps.
Cheers George
Cheers George
I think with any good idea, other ideas are going to branch from it. Whether they be better is unknown. Is a VCCS controlled one better? Is the buffered one better? There are so many things involved in the sound quality that it's really hard to be sure which is the most dominant. Just for example, the Lightspeeds; if the original concept was just to remove the wipers from the signal path, it's hard to believe that research has proven this to have very good results if it had not been tried in practice.
Last edited: