Yes, I agree that's a factor, and the cheaper or older the pot, the more problematic it becomes.
I built a high-stability variable power supply using a PIC for that very reason -- you can use any pot with any value for control, and voltage output is regulated by math computed internal to the PIC which is based upon a read of the potentiometer input. The important part is that you can use software to smooth or average to any degree that you want, trading off rate-of-change for steady-state stability, and the tradeoff does not need to be a problem with the right software solution.
With this method, for a regulated power supply there is no worry about voltage changing or spiking due to a wiper coming off a worn track, or noise in the output caused by a less-than-perfect wiper-to-track connection, and likewise for an LDR attenuator, you can remove the pot anomalies from the equation (which is one reason why LDRs are of interest in the first place.)
Hmmmm. How did you establish that the cost (or age) of the pot was a critical parameter? Folk with years of experience with this technology say different. Who to believe?
And why?
Hmmmm. When I first heard the original perspective from Marlowe, it did sound reasonable to me (aside from the absolute correlation between cost and sound), but I'm no expert. Also, his comments do tend to support what I myself have experienced with pots as well as published data I have read regarding the relative longevity of various materials used in the tracks -- carbon, cermet, plastic, etc.
Perhaps you could at least summarize the wisdom you have gleaned from the folks with years of experience and also reference some sources so one could go and read up on your perspective?
Who to believe? And why?
* As you know, there is no electrical link between the pot and the LDRs in this class of circuit. As the role of the pot is unusual, I don't off-hand see how prior experience with more conventional applications is relevant.
* By definition, longevity does not matter until the device starts to age. How long do you expect that to take?
And so on. What I asked was "How did you establish that the cost (or age) of the pot was a critical parameter?" No answer as yet.
My hunch is that it doesn't matter much but, if I'm wrong, it'd be good to know.
My "wisdom" in this instance relies entirely on remarks posted on this thread by the various designers of successful implementations of the design. These are as accessible to you as they are to me. As it's you, not me, who's challenging conventional wisdom here, I'd respectfully suggest it's up to you to prove your case not the other way round.
All I did was ask what the basis was (other than assertion) for the claims made. Seems a fair question.
I am nonplussed that you would make such a remark as "there is no electrical link." Simply because the connection is via photons rather than electrons emphatically does not mean that there is no link between the pot and the LDR. Even at the common sense level, you do realize that the the pot controls the LDR, right? Of course there is a link.
I think that you have not experimented with these devices nor analysed the relationship of the pot to the devices. While I'm not an electronics guru, I have worked with these devices over the past month or so, trying to measure and control their behaviour down to the .000005A level, because you have to be able to do that in order to control resistance in the 500K~1M ohm area which I want to do. With some trials and a lot of errors, I have been able to both measure and control in the range, right at the edge of the ability of my equipment to measure. I can assure you with full confidence that at that degree of sensitivity, ANY disturbance of the current applied to the LED will have an affect on the LDR. No question about it, it's definite. And the same holds true at higher current levels, though certainly less obviously.
With regard to different qualities of pots and longevity: a cermet trim pot can begin to deteriorate after maybe as little as ten or fifteen wipes and be in real trouble after a hundred or so turns, and with other materials it's much longer but not an infinite period. Look it up for yourself.
Normally, there is a correlation between price and quality, but there are exceptions and that is why I qualified my earlier support of Marlowe's statements. But I can say categorically that there are indeed distrinctions in longevity between materials, and pots do deteriorate over time and usage, albeit at different rates.
Marlowe claims that he can hear the difference between pots, and that better quality pots sound better. That may or may not be the conclusion that I would reach, but since the use of LDRs is driven by the desire to eliminate pots from the signal path, maybe there is at least a grain of truth to his claims. Given what I know about pots and the sensitivity of LDR controls, I, for one, would be very leery about challenging his statement outright.
While I certainly enjoy reading this list and respect everyone's right to post and do not put myself in the same category as several of the truly bright and well-trained people who contribute, I also am fully aware and accept that there is a very broad range of experience levels and technical expertise here, and it doesn't take a rocket scientist to see that. Some of what is published on diyaudio is brilliant, and some of it is pure hearsay and out-and-out technical nonsense. If you are going to challenge someone and do it in a less-than-kind manner, you'd better be really, really, sure of your facts. And you are comfortable using the general posts on this list as your "expert" witness?? Please tell me it ain't so.
The cost is somewhat misleading, but a better pot is supposed more expensive in general.
Just do believe yourself.
Try a different pot and you will know it.
Last edited:
Yes, Lightspeed is a good volume control scheme and a better pot makes it even better.
Please calm down. If my tone seemed aggressive, I apologise for that - it wasn't meant to be. You're right - I've built a Lightspeed and "tweaked" it a bit but I have not "experimented" with it. And, yes, I'm out of my depth designing with pico-amp circuits so you have the advantage on me there as well.
I read Marlowe's comment that "I have tested a number of potentiometers in my Lightspeed attenuator . . . the sound varies greatly from one pot to another . . . the more expensive the pot, the better the sound".
This didn't seem entirely unreasonable but it wasn't intuitively obvious either and would cost money to try. So I asked:
* How did you establish that the cost (or age) of the pot was a critical parameter?
The replies didn't help much. Yours told me to "Look it up for yourself" and Marlowe's to "Try a different pot and you will know it".
It isn't so. I took even Marlowe's claim with a pinch of salt.
OTOH, George Stantscheff and Uriah Daily, both familiar with the circuit, are on record as saying that the quality of the pots doesn't matter much. That doesn't mean they're right but their experience lends weight to their posts. When someone says different, it seems reasonable to ask "What led you to conclude that?"
One of the beauties of the Lightspeed circuit for me is that, though it sounds superb, it is simple and very cheap to build. That usually means that a lot of thought has gone into the design. A high-end pot would/could double the cost. What for?
What an interesting idea. In order to achieve an ideal 5K fixed Zo volume control for example, the series LDR requires a potentiometer range that is entirely different from the shunt LDR. You could achieve that with accuracy by using fixed resistors for each attenuation point. That means you'd need four decks for a stereo control, but easily doable with relays.
On the other hand, you would have to choose between either selecting each resistor individually to match the LDR, or go back to using matched LDR sets. For me, not happy alternatives.
you're months if not a year too late.
A switched attenuator has been suggested for controlling the LED current.
There is then no problem with current limitations and degradation of the wiper.
Doubling the number of switch poles allows both sides to be precisely matched for stereo balance at every step.
The range in attenuation available with a switched control is very much wider than can be got from a dual 100k log pot.
In these three areas, a switched control offers significant improvement in the LED/LDR usability.
The only remaining criticism that is not addressed is the high output impedance of the LDRs and it's resultant effect on how it should be integrated with the receiver.
A switched attenuator has been suggested for controlling the LED current.
There is then no problem with current limitations and degradation of the wiper.
Doubling the number of switch poles allows both sides to be precisely matched for stereo balance at every step.
The range in attenuation available with a switched control is very much wider than can be got from a dual 100k log pot.
In these three areas, a switched control offers significant improvement in the LED/LDR usability.
The only remaining criticism that is not addressed is the high output impedance of the LDRs and it's resultant effect on how it should be integrated with the receiver.
Well, the idea was new to me, anyway.
Further, I would disagree on some of your points:
1) If you look at the LDR resistance values required for certain attenuations while maintaining a constant Z and the amount of current required to achieve that resistance, you will find that your attenuation is limited on the one end by the minimum resistance of 40 ohms or so at 20ma, and on the other end by the vanishingly small amount of current required to achieve high values (say, 200~300K ohms) of resistance and the very great difficulty in achieving accurate control at those levels. The end result of all of that is that the comfortably controllable attenuation range is about 40~45dB, and you'd have to go to T attenuators and paralleled devices to do better. I've looked at this pretty hard, and unless my math is wrong, this is the situation.
I concede that if your objective is something that acts like a real pot -- a fixed total resistance of some value -- say 5K -- between limits and a wiper that moves between those two limits, it's not hard at all if you are willing to tolerate the minimum 40 ohms at each end. It is the constant Z requirement that drives the requirement for the wide range of resistances, but a constant Z is highly desirable if you are going to use these devices without buffers.
2) The fixed resistor solution actually does not address at all the problem caused by the wide tolerances found in these devices. For a pre-computed resistor value network to work properly, the LDR must present a predictable response to a given resistor value, and it does not do that. So, how to get around that issue? Not with fixed resistors with values based on theoretical calculations because that does not take into account the individuality of the LDRs. My PIC based solution involves determining the actual current required to achieve a given level of resistance for each individual LDR, and then using that predetermined amount of current to achieve a predictable resistance. I can't see any other realistic way to achieve that goal of consistent response across multiple channels at every level of attenuation.
300k has already reduced the current to the LEDs by two thirds. 1M0 would reduce by a factor of ten. Yes, precision is needed, that's what the switched offers.
Why does it follow that constant output impedance better suits an unbuffered LDR?
a switched attenuator controlling the current of each LED can be set up precisely to match stereo channels.
I'll say again,
the switched attenuator to control the LED currents overcomes three quarters of the alleged problems of the LED/LDR volume control.
The control circuit I'm developing can do this with a little bit of variation -this is due to voltage/current rapidly changing on the LED but I find it settles down. The variation of course is minimized if the control pot is rotated slowly and 200k-300k is achievable even up to 35Meg. The feel of a nice pot being rotated to vary volume still appeals to me rather than an up/down button especially on a hi-fi system
Well, we're still not on the same sheet of music.
First, if you consider that max current for the LDR is 20 ma and the approximate current (for about 200K) is .01ma, we are not talking fractions -- we are talking several orders of magnitude of difference. That is a very big deal because the rest of electronics is linear and that is a very, very, wide spread. Audio numbers (dB) are logarithmic to match our hearing, and that can make the numbers look trivial until you really look into measuring and controlling, and then the problem is definitely non-trivial.
Second, let's look at the 1K trimmer pot as used in the Lightspeed as it is used to "match" channels, and to do that we need to deal with math.
Please understand, I hate math. If God had intended me to be an engineer, he would have made me capable of doing math. He didn't, and I have to rely on my long-suffering friends to help me with the calculations. But here is the deal:
The logarithmic scale which matches human hearing is, when graphed on a logarithmic chart, a straight line. Let us assume for the moment that a Silonex device is a linear device (which it's not). But if it were, the challenge would be pretty simple: determine the difference between the logarithmic straight line and the LDR straight line, and compute the difference. This difference is defined in math as the difference in slope and the difference in offset from zero between the two lines -- the famous straight-line formula.
You would then devise a formula to calculate the difference in slopes and the difference in offsets to determine the correction required to rotate the slope of the LDR to match the slope of the logarithmic audio, and then find the correction in offset so that the LDR line and the audio line are fully aligned, and then you would have a fully repeatable alignment.
The problem is, the LDR is NOT a linear device, nor is it perfectly logarithmic. You can see that in the charts that jackinnj published on my thread at posts 19 & 20. So, what that means is that you cannot define the LDR using a single straight line formula. You have to 'approximate' the curve using multiple straight line formulas that match the various points on the curve. Plenty complicated. Then, you have to match those segmented straight line formulas to the appropriate segments of the straight line logarithmic line that represents human hearing. Then -- and here is the kicker -- you have to remember that the LDR tolerances are so loose that there are very large differences between devices so the straight line formulas that work for one device do not work for others! You CANNOT deal with all of that by just calculating some resistor values and putting them into a relay bank. You have not solved three-quarters of the problem. You have solved the too-much-current-through-the-pot problem, and you have solved the need to provide different slopes to the series and shunt LDR problem. You have not solved the main problem of loose-tolerance LDRs, which is blocking all development past the simple stereo pair solution.
Now, let's talk about that 1K trimmer pot. What it does is very simple -- by adjusting it, it matches a single point on the hearing line between two channels with a single point on the curving LDR line, and ignores everything else. It doesn't adjust the multiple slopes the length of the attenuation range, and it doesn't correct the offset either. It doesn't match channels at any point on the long slope except at that one point. You will have a perfect match at one single point on the entire graph, and increasingly worse matches the further you get away from that point. And that match is only between channels -- it does nothing to match the LDR response to human hearing.
My PIC-based solution in effect does two things simultaneously and solves all of the problems -- it creates a 1/2dB fixed resistor attenuator that fully controls two separate slopes for the series and shunt LDRs and puts all LDR current through a robust mosfet variable resistance for each LDR instead of a fragile potentiometer track and, through software adjustments, it compensates for LDR variations at every point in the full slope so that LDR resistance matches the desired resistance regardless of the the tolerance problems discussed above. Because current can be set in software at each step to whatever is required, there is ample room to change resistance values to deliver a near-perfect flat Z device.
Last point -- why does constant Z matter? Because you need to know the system impedance in order to properly match your LDR volume control to your source device and your load amplifier. A simple potentiometer has a very wide ranging Z which is not desirable; my PIC design is intended to deliver a very stable Z both to the input and the output, thus obviating the need for buffers at either end.
Whew! If you got this far, thanks for listening, I hope I've made my position clear without being obnoxious. I don't mean to force anyone to my perspective, but I feel that I have the facts on my side. This post contains the kernel of what I know, I have nothing further to add, so it's my last word on the subject.
Well, I can't agree with that. I'm talking about the technology of using the LDR, and Lightspeed is certainly the starting point for all of that but not the whole enchilada by any means. I did not get the sense that Andrew was limiting his discussion to the Lightspeed design, either, but I won't put words in his mouth.
Best wishes.