I tend to agree that using LDRs almost seems like it would be more trouble than it's worth.
The main problem they are supposed to solve is simply the same one that all stereo attenuation schemes are supposed to solve: A stereo attenuator is needed. A couple of other possible problems that they might be able to mitigate are mentioned below.
One thing that LDRs should be able to do better than other passive attenuators is automatically adjust (or automatically be calibrated) for nearly-perfect matching of the two channels, at every attenuation level. But I don't know if that is a big-enough problem with other passive attenuators to warrant using LDRs, with the added complexity of auto-cal or auto-adjust.
And the same thing (automatic matching or auto-cal'd matching) could probably be accomplished a little more-easily with a dual pot and one LDR. Or, with more parts but less design effort, a stepped attenuator could be carefully matched.
One thing that LDR-based attenuators have, that none of the other types of passive attenuators have, is a complete lack of mechanical contacts in the signal path. But I don't know how important that is to sound quality, or how much better it might be in terms of long-term reliability and performance.
The main problem they are supposed to solve is simply the same one that all stereo attenuation schemes are supposed to solve: A stereo attenuator is needed. A couple of other possible problems that they might be able to mitigate are mentioned below.
One thing that LDRs should be able to do better than other passive attenuators is automatically adjust (or automatically be calibrated) for nearly-perfect matching of the two channels, at every attenuation level. But I don't know if that is a big-enough problem with other passive attenuators to warrant using LDRs, with the added complexity of auto-cal or auto-adjust.
And the same thing (automatic matching or auto-cal'd matching) could probably be accomplished a little more-easily with a dual pot and one LDR. Or, with more parts but less design effort, a stepped attenuator could be carefully matched.
One thing that LDR-based attenuators have, that none of the other types of passive attenuators have, is a complete lack of mechanical contacts in the signal path. But I don't know how important that is to sound quality, or how much better it might be in terms of long-term reliability and performance.
I think that's not the issue. When I first started reading about using LDRs in attenuators, I asked the following of the Lightspeed thread:
"May I ask for you opinion about the "sound" of LDRs -- how it affected your system when you switched to LDR input switching? Did you hear a difference? If so, what was that difference?"
The one-word answer I got was "transparent" and no one disagreed with that assessment.
So, it seems (I've never heard a set so I don't know), it is the sound quality that is the potential, and we are willing to sacrifice time and energy to solve the peripheral issues that the LDR presents. The important piece that complex drive circuitry can (potentially) provide that the Lightspeed cannot is close tracking between channels which obviates constant fiddling with the balance control. This lack seems to be the main problem with the Lightspeed. That, and very loose impedance control which can also affect sound quality.
So in my project I am assuming that the LDR's sound is worthy. My goal is to have the sound quality but also add to it
1) excellent tracking between channels
2) relatively tight impedance control
3) low current through the LDRs and the pot at any volume level to assure pot and LDR longevity
4) convenient remote control capability with readily available motorized control pots and circuits.
I think that the question of sound quality has been asked and answered. That's not the issue any more.
I hear what you're saying about sound quality, wapo. But I still have lingering questions and doubts. For example, I wonder about the magnitude of the improvement in SQ, which would help determine whether or not it justifies the effort.
Also, I generally like to discount listening test results such as "sounds better". I want the most-faithful reproduction, even if it "sounds worse".
And I haven't done enough research but keep seeing claims that the LDRs add distortion. But, if there is added distortion, there's not enough good data to know if they're still better overall.
Despite all of that, I do still strongly tend to "believe" that they could be made to give more accurate reproduction than other passive attenuator types, and could have fewer long-term problems or changes in accuracy.
I do have a big pile of VTL5C2 devices, here. Maybe someday I'll make the time to hack something together and do some measurements.
Also, I generally like to discount listening test results such as "sounds better". I want the most-faithful reproduction, even if it "sounds worse".
And I haven't done enough research but keep seeing claims that the LDRs add distortion. But, if there is added distortion, there's not enough good data to know if they're still better overall.
Despite all of that, I do still strongly tend to "believe" that they could be made to give more accurate reproduction than other passive attenuator types, and could have fewer long-term problems or changes in accuracy.
I do have a big pile of VTL5C2 devices, here. Maybe someday I'll make the time to hack something together and do some measurements.
The JFET option mentioned a page or so back offers that too. Depending how strict your definition of "passive", there's also the attenuator chips that use electronically switched resistor networks.
Moving away from passive solutions, I'm almost tempted to suggest analog multipliers (e.g. Gilbert cells), but I suspect noise would be a problem.
I'm not so sure about that. Things like the "light history" effect tend to put a spanner in the works. It means that identical LED currents will give different attenuation depending on whether, for example you just switched the unit on a few minutes ago or it's been left on overnight. Similarly, Turning the volume up to a certain level will give a different result to turning the volume down to the same level, all of which makes calibration a bit awkward.
In terms of things like history effects and response times, JFETs are far better than LDRs as variable resistance elements for attenuators. Temperature dependence is likely to go the other way though.
Changing the subject entirely (again)....
It seems the two main problems with channel balance in LDR attenuators are:
A) Matching and aging of the LEDs.
B) Matching and aging of the photocells.
The first of those can be neatly sidestepped by using one LED to illuminate two photocells. While that severely limits your options for trimming, it allows the interesting possibility of having a purely mechanical balance control, with no added electronic components at all (think moving shadow
).
It seems the two main problems with channel balance in LDR attenuators are:
A) Matching and aging of the LEDs.
B) Matching and aging of the photocells.
The first of those can be neatly sidestepped by using one LED to illuminate two photocells. While that severely limits your options for trimming, it allows the interesting possibility of having a purely mechanical balance control, with no added electronic components at all (think moving shadow
).If you really want to get around (or not worry about) all of the device-physics-related calibration problems then I think that the only certain way would be with LDR resistance measurement and feedback control system, which could probably be all analog if desired.
That would require out-of-band pilot tones, or DC, either continuously or possibly only when the knob position was changed. If done only when the knob position was changed, then some sort of memory (at least four bytes, or four digital pots) would probably be required.
That would require out-of-band pilot tones, or DC, either continuously or possibly only when the knob position was changed. If done only when the knob position was changed, then some sort of memory (at least four bytes, or four digital pots) would probably be required.
I like the pilot tone idea, but it would probably require a filter in the audio path at the output. It may as well be continuous, though. The idea of "on demand" recalibration when the the knob is moved is tempting, but I don't see a real advantage unless the filter is switched in and out of circuit during and after the adjustment.
Just off the top of my head (but hopefully close-enough to some potential reality):
I don't think that it would be very difficult to design and implement such a measurement and feedback control system. But of course there could be some gotchas that I haven't thought about.
After the details were evaluated, decided, and worked out for "continuous" or "only-at-change" interfacing to the LDRs, and what impedance range(s)/scheme to use, the actual measurement-and-control implementation should be fairly straightforward, possibly with an instrumentation amp across each LDR, a filter to discard everything except the measurement signal, an RMS-to-DC converter chip like the LTC1968 or a discrete equivalent circuit (if using out-of-band tones instead of DC), and an opamp as a differential integrator (possibly with a low-pass post-filter) to servo the LED current based on comparison of the setpoint to the measured resistance. The series and shunt setpoints could be derived from the output of a lightspeed-style pot, or, in some other way.
For better matching accuracy, it might be better to have only ONE measurement-and-feedback-control circuit, driving a digital pot or other memory scheme for each LED current, and time-multiplex (switch) it to connect to, and measure and set, each of the LED/LDR devices, in turn.
That way there wouldn't need to be any worries about the matching of the measurement and control circuits themselves. And as a bonus the parts count and cost should be significantly lower. But response speed and user perception might become a real concern.
There would probably be design issues to solve (or minimize) involving dynamic range and noise (and stability) at each stage in the measurement and feedback control chain. But I can't see any obvious show-stoppers, yet.
There might also be other implementation schemes that could be better.
I don't think that it would be very difficult to design and implement such a measurement and feedback control system. But of course there could be some gotchas that I haven't thought about.
After the details were evaluated, decided, and worked out for "continuous" or "only-at-change" interfacing to the LDRs, and what impedance range(s)/scheme to use, the actual measurement-and-control implementation should be fairly straightforward, possibly with an instrumentation amp across each LDR, a filter to discard everything except the measurement signal, an RMS-to-DC converter chip like the LTC1968 or a discrete equivalent circuit (if using out-of-band tones instead of DC), and an opamp as a differential integrator (possibly with a low-pass post-filter) to servo the LED current based on comparison of the setpoint to the measured resistance. The series and shunt setpoints could be derived from the output of a lightspeed-style pot, or, in some other way.
For better matching accuracy, it might be better to have only ONE measurement-and-feedback-control circuit, driving a digital pot or other memory scheme for each LED current, and time-multiplex (switch) it to connect to, and measure and set, each of the LED/LDR devices, in turn.
That way there wouldn't need to be any worries about the matching of the measurement and control circuits themselves. And as a bonus the parts count and cost should be significantly lower. But response speed and user perception might become a real concern.
There would probably be design issues to solve (or minimize) involving dynamic range and noise (and stability) at each stage in the measurement and feedback control chain. But I can't see any obvious show-stoppers, yet.
There might also be other implementation schemes that could be better.
Last edited:
Please keep feeding control strategies, no matter how weird they initially seem.
We can sift through them and find those that have potential to implement simply and the few genii among us can contemplate the complex solutions.
We never know which might be "best", until each strategy has been implemented and developed to it's limiting condition.
There may turn out to be 23 "best"
We can sift through them and find those that have potential to implement simply and the few genii among us can contemplate the complex solutions.
We never know which might be "best", until each strategy has been implemented and developed to it's limiting condition.
There may turn out to be 23 "best"
I was thinking about stepped attenuators yesterday. What about a stepped attenuator as the control for driving the LED/LDRs. The high level idea is you use a 12 step switch, with say 4dB steps for 48dB control. And then run the balance calibration on it, and add pots, or series resistors, as needed to get perfect volume balance at all steps. This would be more efficient than just making it with pots on every step.
Then have a secondary fine control, for say 6 dB total, to give control between steps.
Doing this with a board for resistors/pots would make it easier than trying to do it point to point on the switch.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.