do the sums.
100k pot and 100r source resistance gives output impedance of [100,000+100]/4=25k025
The Lightspeed if ~7K is roughly a quarter of that.
If you can't do that bit of mental arithmetic, then get your calculator out and learn to use it.
The pot varies the _current_ through the LED in the LDR, and LDRs are designed to vary their resistance according to the current driving them. And they vary their resistance non-linearly, by a LOT, over the range of allowed currents.
Some people here need to carefully re-read my post #12, in this thread, which states, among many other things:
"Lightspeed's simulated output impedance varies from about 37 Ohms to about 14.6 kOhms, as the attenuation level is varied from maximum to minimum." (Or maybe it's the other way around...)
And note that the variation of the output impedance in not only at the extremes of the attenuation range, as someone suggested. It varies continuously, over the whole range, and whenever the overall attenuation is changed.
I have not expressed a judgment about whether or not the variations in output and input impedance are bad or good (yet). I have just stated what I know about the behaviors of the circuit, and raised questions about whether or not the behaviors MIGHT have any negative effects.
Some people here need to carefully re-read my post #12, in this thread, which states, among many other things:
"Lightspeed's simulated output impedance varies from about 37 Ohms to about 14.6 kOhms, as the attenuation level is varied from maximum to minimum." (Or maybe it's the other way around...)
And note that the variation of the output impedance in not only at the extremes of the attenuation range, as someone suggested. It varies continuously, over the whole range, and whenever the overall attenuation is changed.
I have not expressed a judgment about whether or not the variations in output and input impedance are bad or good (yet). I have just stated what I know about the behaviors of the circuit, and raised questions about whether or not the behaviors MIGHT have any negative effects.
I think that George also uses his own definition of "measured output impedance", if I recall correctly from some posts where he tried to describe why someone else was measuring it wrong, and how to measure it like he did. Or maybe the 7k was the average.
Everyone is free to download my LTspice simulation files and see for themselves, or tell me that I have simulated it incorrectly.
Or, it would be simple enough to pick a couple of attenuation levels and substitute the resulting LDR resistances into the resistive dividers and calculate the output impedances manually, and see if they are different.
Tom
No. The R of the LDR behaves as just a pure resistance, and acts just like a resistor, except that the resistance can be varied by varying the current through the LED inside the LDR.
And of COURSE the shunt R (of the shunt LDR) also affects the input and output resistances/impedances. As far as the signal is concerned, it is just a two-port network containing a (variable) resistive divider.
One problem is that the LDRs' Rs' resistances do not change linearly, versus the LED currents. It is more like logarithmic. So, when the LED currents are varied "inversely" (to each other), the sum of the two LDR resistances does not stay constant, as we would like.
The manufacturer of the LDRs that are used in the Lightspeed actually published a circuit that they said would linearize the LDR's R vs I curve. But I couldn't get it to work, even in the LTspice simulator. I did then design a linearization circuit that works. But it requires a matched LDR in addition to the LDR being linearized, and quite a few other components.
Simply adding linearization resistors to the Lightspeed circuit, as I reported in my posts in the Lightspeed thread, does help a whole lot. And if the source and load impedances are known, the linearization can be tailored to fit them and can be very effective. But, as I have said, I don't know if it would be a significant improvement, or not. But there is guaranteed to be SOME amount of variation of the frequency response of the total system, versus attenuation level, the way it works now. I am pretty sure that it is insignificant for most systems, just based on listening reports in the Lightspeed thread. But there might also be cases where it is significant. However, for me it's just mostly an academic exercise, so far.
And of COURSE the shunt R (of the shunt LDR) also affects the input and output resistances/impedances. As far as the signal is concerned, it is just a two-port network containing a (variable) resistive divider.
One problem is that the LDRs' Rs' resistances do not change linearly, versus the LED currents. It is more like logarithmic. So, when the LED currents are varied "inversely" (to each other), the sum of the two LDR resistances does not stay constant, as we would like.
The manufacturer of the LDRs that are used in the Lightspeed actually published a circuit that they said would linearize the LDR's R vs I curve. But I couldn't get it to work, even in the LTspice simulator. I did then design a linearization circuit that works. But it requires a matched LDR in addition to the LDR being linearized, and quite a few other components.
Simply adding linearization resistors to the Lightspeed circuit, as I reported in my posts in the Lightspeed thread, does help a whole lot. And if the source and load impedances are known, the linearization can be tailored to fit them and can be very effective. But, as I have said, I don't know if it would be a significant improvement, or not. But there is guaranteed to be SOME amount of variation of the frequency response of the total system, versus attenuation level, the way it works now. I am pretty sure that it is insignificant for most systems, just based on listening reports in the Lightspeed thread. But there might also be cases where it is significant. However, for me it's just mostly an academic exercise, so far.
Last edited:
as far as my LDR attenuators experience goes, well, i was quite a careful listener for several things: 1) attenuation vs SQ 2) output impedance vs SQ
i tried it with a 10k solid state amp and a 100k tube amp (SET).
my findings are that it really all depends if you have buffers or not. buffers do mask a final touch of transparency (i've heard JFET as well as AD827 (yakk)) but OTOH they quite drastically stabilize the sound and there's a big benefit in bass drive. so it's really to be chosen by a user - what fits better in his system, there's no universal recipe yet. with the above amp impedances, LDR sounded supremely transparent, a class above any transformer or autoformer i've heard but it lacked body and weigh which i would attribute to far to high output impedance and had a bit of a nervous, glassy sound. unfortunately i couldn't borrow any higher imp power amp to see if there would be some input impedance that would be set as a realistic minimum for power amp that wouldn't be affected with LDR impedance.
i've also find LDRs to sound fine at high attenuation levels which makes it almost ideal solution for low listening. for example both S&B TVC and slagle AVC sounds much worse and it's audible.
i think it needs to be carefuly used - like paul hynes does it in his best active tube preamps, an ideal volume control that needs a perfect surrounding (that in his case act as a buffer).... not really sure yet as a standalone design before a more complete solution to impedance issue is sorted out.
i tried it with a 10k solid state amp and a 100k tube amp (SET).
my findings are that it really all depends if you have buffers or not. buffers do mask a final touch of transparency (i've heard JFET as well as AD827 (yakk)) but OTOH they quite drastically stabilize the sound and there's a big benefit in bass drive. so it's really to be chosen by a user - what fits better in his system, there's no universal recipe yet. with the above amp impedances, LDR sounded supremely transparent, a class above any transformer or autoformer i've heard but it lacked body and weigh which i would attribute to far to high output impedance and had a bit of a nervous, glassy sound. unfortunately i couldn't borrow any higher imp power amp to see if there would be some input impedance that would be set as a realistic minimum for power amp that wouldn't be affected with LDR impedance.
i've also find LDRs to sound fine at high attenuation levels which makes it almost ideal solution for low listening. for example both S&B TVC and slagle AVC sounds much worse and it's audible.
i think it needs to be carefuly used - like paul hynes does it in his best active tube preamps, an ideal volume control that needs a perfect surrounding (that in his case act as a buffer).... not really sure yet as a standalone design before a more complete solution to impedance issue is sorted out.
is there a link for how this is measured?
i'm also wondering how is possible to have a single impedance figure in a LDR case.... i mean it would be great, i just don't see how's that possible.
George's definition and some discussion of it are in thje vicinity of post number 1048, at the following link.
http://www.diyaudio.com/forums/anal...uator-new-passive-preamp-105.html#post1395130
As far as getting a single number for the input impedance, if you add the correct resistances in parallel with the LDR resistances, you can get the input impedance to vary within about a 1 or 2 percent band (as the attenuation varies over its full range), around an input impedance that you have chosen.
Someone asked me, a couple of weeks ago, to find the parallel resistance values to give a 10k input impedance, with minimal variation versus attenuation level, with a 1 Meg load. Using my LTspice simulation of the Lightspeed circuit, I found that 23.2k in parallel with the series LDR's R and 23.2k + 750 Ohms in parallel with the shunt LDR's R gave 10.037 kOhms average input impedance, with the input impedance varying between 10.1417k and 9.9261k. Without any added parallel resistances, i.e. with the original Lightspeed circuit, and with the same 1 Meg load, the input impedance varied between 17.859k and 12.875k, with an average of 14.495k.
Tom
hornperfect,
The way I measure the input impedance, within an LTspice simulation, is the standard method: I put a fixed current source in series with 50 Ohms (estimated source impedance) across the input/ground terminals of one channel and plot the voltage across the input/ground terminals divided by the current into the input terminal, as the attenuation is varied from one extreme to the other. That gives the input impedance as seen by a 50-Ohm source. (More generally, for other circuits, an AC current source could be used, instead, to plot the input impedance versus frequency.)
Sorry, the link that I gave, in a previous post, was to how George measured OUTPUT impedance, not input impedance. I believe he was wrong, by the way. But I don't think that output impedance was what you were asking about. So the question is moot. At any rate, output impedance should be measured similarly to input impedance, when using a simulator.
Another thing, which I should have mentioned in my previous post: The power supply voltage was changed to 4.65 Volts, for the 10k input impedance case. Also, the maximum input impedance that you can choose depends on the power supply voltage and on the LDRs themselves. I guess that to go above a certain limit might require also using series resistances (in series with the LDRs' Rs). But I have not investigated that.
One more thing: For anyone who has downloaded my LTspice simulation files for the Lightspeed circuit, note that the "sweeptime" parameter, which sets how many seconds to take to sweep the attenuation potentiometer from one extreme to the other, must be increased to 10 seconds. Otherwise, the time-constants in the LDRs will cause them to not react nor follow/track quickly-enough, causing somewhat-incorrect results to be plotted.
Tom Gootee
The way I measure the input impedance, within an LTspice simulation, is the standard method: I put a fixed current source in series with 50 Ohms (estimated source impedance) across the input/ground terminals of one channel and plot the voltage across the input/ground terminals divided by the current into the input terminal, as the attenuation is varied from one extreme to the other. That gives the input impedance as seen by a 50-Ohm source. (More generally, for other circuits, an AC current source could be used, instead, to plot the input impedance versus frequency.)
Sorry, the link that I gave, in a previous post, was to how George measured OUTPUT impedance, not input impedance. I believe he was wrong, by the way. But I don't think that output impedance was what you were asking about. So the question is moot. At any rate, output impedance should be measured similarly to input impedance, when using a simulator.
Another thing, which I should have mentioned in my previous post: The power supply voltage was changed to 4.65 Volts, for the 10k input impedance case. Also, the maximum input impedance that you can choose depends on the power supply voltage and on the LDRs themselves. I guess that to go above a certain limit might require also using series resistances (in series with the LDRs' Rs). But I have not investigated that.
One more thing: For anyone who has downloaded my LTspice simulation files for the Lightspeed circuit, note that the "sweeptime" parameter, which sets how many seconds to take to sweep the attenuation potentiometer from one extreme to the other, must be increased to 10 seconds. Otherwise, the time-constants in the LDRs will cause them to not react nor follow/track quickly-enough, causing somewhat-incorrect results to be plotted.
Tom Gootee
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.