No, because in the example you provide I gets depleted before V and V gets entirely depleted by potentiometer, then leading to constructor problems requiring difficult matching of LDR devices to properly make LDR devices work with relevance to L pad audio circuits. This thread tries to assist constructors and invites designers to meet these difficulties. Also that THD can be lowered where voltage is also lowered across the LDR providing that it has sufficient current to still function... and that current has not been exhausted by resistance from use of a high value potentiometer.
Silonex
" It is better to drive the coupler LED from a constant current source, to minimize the effects of variations in LED forward voltage from device to device and temperature"
Silonex Inc.: Technical Reference: Audio Level Control with Resistive Optocouplers
Cheers
Ok, now my head hurts.
You mean that for example in the part below, pin4 (V-) is connected to the non-inverting input of TL072a?
Yes its a single rail connected op amp. as is common with single rail positive input is same potential as V- , in this case prior to both before the transistor connecting that potential in fashion of approx 0.692v lift. The design parameter of using the transistor being to contain the control circuitry in high impedance and stop possibility of LDR cathode interaction.
Cheers / Chris
Last edited:
Hi Tom;
You are taking this beyond linearizing the LED response to linearizing the whole LED/LDR. That's great, but as you say, a *lot* harder.
I think the benefits we refer to (i.e. loosely "active current control" vs. passive pot) are not inherently obvious unless you get your hands on these devices, build, test, and use them as volume controls in a a few different systems.
You probably have done this, so if it's been a while, to remind you, a pot controlled version has limited attenuation capability. In most situations, this is about 48dB or less. For some people that's not enough, and the "sweet spot" of really good control is in the wrong place. Also a problem for some, with a simple pot/resistor control, the volume never goes to zero.
For others, it can go from quiet to very loud in a very limited area of rotation control, depending on the efficiency of speakers used or gain of the power amp driving them.
By going to active current control, one can achieve closer to 90-100dB of attenuation, with a broader and more even control across the board.
Plus many other advantages, already stated by myself and others. The main ones I see are better stability and control at ultra low currents, which you need to get the large attenuation figures.
And also the ability then to operate better, in general, at lower currents which ultimately yields higher series and shunt resistances.
Which give a higher load impedance to upstream devices driving the LDR attenuator. Which results in lower source (device driving LDR) distortion.
Hi Chris,
Two things:
I get the part about it being better to use a current source, since the LEDs might have variations in their forward voltage, between devices. But do they really mean the SAME constant current, through the two corresponding devices in a stereo pair of attenuators (either the two series devices or the two shunt devices)?
For that to be true, we would need to believe that the two LEDs' brightnesses would be the same, for the same current, even if their forward voltage characteristics were different.
So it seems like the two LEDs wouold either need to be matched or there would need to be a trimmer for at least one of them.
You should definitely measure the resistances across the CdS cells, with an ohm meter, at various attenuation settings, to see how well the two channels' attenuations match (with no input or output devices connected).
Sorry. Not possible, unless you are simply using the lower end of the CdS's resistance ranges by using only the higher end of the range of LED currents.
Silonex is talking about the voltages across the CdS cells, themselves, not the voltages across the LEDs. The voltages across the passive-resistance CdS cells are from the music signals. So you would have to either use lower resistances, or lower music signal voltages, to lower the voltages across those.
To do that, you could either a) use two or more series CdS cells in place of each CdS cell that you currently use, or, b) lower the signal voltages just after the inputs and raise them again just before the outputs, which would involve using an amplifier at each output with a gain > 1 and probably a buffer amplifier and a fixed-value-resistors voltage divider at each input. Unfortunately, lowering the voltage level and then amplifying it would probably add noise to the resulting signal.
In case (a), you would want to also use a higher current through the also-in-series LEDs, to lower the resistances of each of the CdS cells, so that the total for the series cells would be about the same as it is now for one cell. That would cause the music signal voltages to induce a lower voltage across each CdS cell, by a factor of how many cells you used in series in place of each one cell you have now. OR, you could use even more cells in series but also not lower their resistances by as much, and get greater possible/maximum dynamic range.
In case (b), the main benefit would probably be the fact that there would be amplifiers at both the input and output, which would provide the ability to have a constant high input impedance and a very low output impedance. But(!), you could also implement case (a), between the amplifiers, using only unity-gain buffer amplifiers and not lowering the voltage, and get the best of both worlds!
Back when I was simulating the Lightspeed, the main problem I saw was that the input and output impedances changed very significantly, over the range of attenuation settings. Not only can that change the sound quality by loading the source differently and driving the amp differently, for different knob settings, it will also interact with cable capacitances and inductances and other parasitics (and possibly input filter components in some amplifiers), and will change the frequency response of the system, depending on the knob setting.
It's not a question of IF it will have those effects. It is only a question of how significant they might be, in different systems.
And I don't know how significant those effects might be, or if they are at all. But you should probably try to find out, or at least think some more about it.
Cheers,
Tom
As far as I can see, the only useful components in that circuit are the 7805 regulator, the 50K potentiometer and the cap multiplier based on Q1. Everything else is random nonsense.
edit: Correction, I guess the LM317 acts as a current limiter.
edit: Correction, I guess the LM317 acts as a current limiter.
Last edited:
Hi BFNY,
Yes, I have used these devices in various capacities, mostly in medium-precision instrumentation circuits that I designed. They seem to be handiest in feedback-based automatic control systems, where you don't have to worry about their exact characteristics too much because the feedback system just pushes them to exactly where they need to be, based on the error-signal level from the feedback subtractor.
I also used them as a current-controlled variable R, in RC filters. Once was to make an "adaptive envelope follower" that could automatically change its rate of convergence, i.e. its slew rate (especially handy for quickly diving down after a falling-amplitude sine signal while still having a big-enough capacitor to remove the ripple once the level stopped changing and the R was sent back to max), and once was to make a multi-stage "tracking filter" for a triangle-to-sine converter, which worked by using feedback of the output amplitude to automatically crank up the R (lowering the filters' cutoff frequencies) until the amplitude went down to a pre-determined level (compared to the known level of the triangle input, which didn't vary much for frequencies from 30 Hz to 22 kHz), which would always put the lowpass filter cutoffs just above the fundamental frequency and filter out most of the harmonics.
The same range of attenuation should be achievable with a potentiometer. And a zero-volume state could also be added to a pot setup. But I think I agree that it could be much better, and probably easier to implement well, with voltage-controlled current sources.
Again, I'm not sure that there's anything there that couldn't (theoretically) also be done with a pot setup. But I tend to believe that voltage-controlled current sources could result in a more-straightforward implementation of a better-feeling and better-behaved control knob, and also could provide much better low-current performance, with much higher resolution.
While I heartily agree with the desire for better (higher and less-variable) input impedance, I would be a little careful about using the higher end of the resistance range of the LDRs, for that reason, since that will invite more self-generated resistor noise as well as more potential for interference to be picked up, across the higher impedances, not to mention more distortion due to higher voltages being induced across the CdS cells' higher resistances, by the music signals.
The best way to do that is still to use buffer amplifiers. I would do that in a heartbeat, and would also use multiple LDR devices in series in place of each single device used now. "Minimum Components in the Signal Path" purists are a clueless lot.
You could also experiment with adding some fixed resistors, since that also tends to reduce the variation in impedance versus knob position (but probably has some cost in dynamic range; I forget; see my old posts in the Lightspeed thread; it can provide very signiicant benefits, in some cases).
If anyone wants LT-Spice simulation models of the Silonex devices (or the Vactrol devices), look for my old posts with the download links. You might have to measure their response speed and insert an appropriate capacitance. But I did the heavy lifting of entering all of the data points for the R vs I curve of representative devices (from the datasheets).
The latest VTL5C2 model and supporting files are also at Spice Component and Circuit Modeling and Simulation . For the Silonex models, I would have to search a backup disk from my older computer. So look for my download links in the Lightspeed and other threads, instead.
(Note that nothing at all is for sale, on any of my webpages. I just haven't updated them for years.)
Cheers,
Tom
Hi Tom
Re point 1 I have been through this( I give up ) use of trims, but vowed to progress beyond that, using trims to me is saying use a band aid solution, because its there... and its easy.
So where I am with this design, and suggesting and inviting other members to contribute. is trying to find a circuit that just works. on the speakers behind me playing at this very minute, is that very circuit, so its not just theory.
The same circuit with identical LDR's previously sounded not as good, but due to circuit revisions now sounds - as good as an attenuator can. There are no trims.
Which says its possible. Model 7 is close, but its different in voltage regulation, and as I have explained i will get to providing a schematic, and ensure it sounds the same as the earlier model, confirming before I publish,- a design is possible that just matches.
L/R balance is where I am putting most design effort into, to me its the top area to get good sounding audio. What I notice is that L does indeed behave differently to R, however compensates itself by adjusting in proportion the corresponding difference of series and shunt relative to the other channel ( when loaded ). L/R is by far the best myself and a friend have ever heard through electrostatic speakers. So to me this says it is correct. And just good reason rather than outright redesign, to define why it is correct with the parts being used.
Point 2 I agree with you that Silonex refer to lower voltage with reference to Signal side.
However regarding distortion in L Pad attenuation
What my schematic shows is perhaps new territory in that Voltage in particular, but current as well remains active throughout the entire attenuation range. What is interesting is that it still works as an attenuator. I believe what is being harnessed is a combination of the voltage that appears at negative in and op amp output when op amp is single rail connected. This I believe creates added ability for LDR to remain conductive on (and off concerning its purpose as an attenuator ). which is I consider is inversion across the feedback path occurring at the same time as attenuation.
Cheers / Chris
Last edited:
In the powering opamps thread I contributed to many posts, with no one trying a simple transistor and resistor.... yet I had many people thanking me because it sounded better.
Please do not create the same silly atmosphere of " Well it can't work because I won't try it and that is why it can't work, because I won't try it ".. . ... round and round.
Cheers / Chris
Last edited:
Well, agreed - the schematic - it may work, but certainly not obvious at a glance
*exactly* how. Maybe an explanation from a high level would help.
Last edited:
First question for "high level" would be: Does Q1 ever conduct any current, bearing in mind the only thing connected to it's base is a JFET gate? If not, is it just there for Feng Shui, or what?
p.s. By Q1, I mean the transistor directly above the LM317 in the picture.
Aaaargh! You can't even discuss this circuit in a normal way because there's duplicate component numbers all over the place.
When I mentioned Q1 in a previous post, I meant the one over on the right side of the picture, underneath the capacitor.
I am "extremely skeptical" about the possibility of using unmatched LED/LDR devices in any circuit and being able to make the left and right attenuation levels automatically match, without feedback from the CdS cells on the signal side.
Matching the LED currents won't do it.
I have already mentioned two ways that it could be done. But both of those would probably be most-easily implemented with the help of programmable digital processing of some sort, with appropriate interfacing.
If it can be done elegantly-enough without any sort of CPU/FPGA/PIC, or whatnot, that would be cool.
I guess it might not be too difficult, after all: Let's see... You'd need a current-sensing circuit for each CdS cell (probably either a current transformer, a hall-effect sensor, or, most likely, a low-value current-sense resistor in series with the CdS cell, with a differential amp's high-impedance inputs across the current-sense resistor), and another differential amplifier's high-impedance inputs across each CdS cell, to measure the voltage across it, then an analog multiplier chip (see analog.com) configured as a divider, to calculate the CdS resistance (V's voltage divided by I's voltage) and convert it to a voltage, then an error amplifier to subtract the resulting voltage (representing the actual measured resistance) from the desired/setpoint voltage (probably the appropriately-scaled voltage from one side of a user-controlled pot), with the "error amplifier" probably consisting of an opamp "differential integrator", and a low-pass post-filter, as in a DC Servo circuit (see my old posts about those), followed by a voltage-controlled current source circuit, feeding the LED. You need a setup like that for each LED/LDR device. And the left/right channels' series and shunt pairs would share the two control voltages from the user-controlled potentiometer.
QED.
Cheers,
Tom
Last edited:
[QUOTE You can't even discuss this circuit in a normal way because there's duplicate component numbers all over the place.
When I mentioned Q1 in a previous post, I meant the one over on the right side of the picture, underneath the capacitor.[/QUOTE]
Sorry I will correct numbering in new schematic, shortly 🙂
Cheers / Chris
When I mentioned Q1 in a previous post, I meant the one over on the right side of the picture, underneath the capacitor.[/QUOTE]
Sorry I will correct numbering in new schematic, shortly 🙂
Cheers / Chris
I can understand that current drive gives better control over light emission, and so reduces the need for matching the LEDs. What about matching the LDRs? I thought they have even wider tolerances than diodes. Maybe modern LDRs are better in this respect.
To reduce LDR distortion you need to reduce LDR signal voltage. How you illuminate it is completely irrelevant. The most you are going to get with better LED current control is a small improvement in channel balance. This may be worth having, but it is unlikely to fundamentally change the sound. Moving your head a few inches will change the channel balance by more in a typical room.
So yes, Chris, my hunch is that this thread is going to go the same way as your daft idea about feeding opamp supplies through a diode and a resistor and for exactly the same reason.
To reduce LDR distortion you need to reduce LDR signal voltage. How you illuminate it is completely irrelevant. The most you are going to get with better LED current control is a small improvement in channel balance. This may be worth having, but it is unlikely to fundamentally change the sound. Moving your head a few inches will change the channel balance by more in a typical room.
So yes, Chris, my hunch is that this thread is going to go the same way as your daft idea about feeding opamp supplies through a diode and a resistor and for exactly the same reason.
I am happy to report a second unit built to the same specification as the earlier model provides the same result , excellent L/R stereo separation, and pleasant attenuation.... should I build a third ?.
There are no trimpots, no set resistors and
LDR's have not required extensive matching.
Cheers / Chris
There are no trimpots, no set resistors and
LDR's have not required extensive matching.
Cheers / Chris
Last edited:
The reason now researched and demonstrated with excellent L/R balance 🙂 in order to provide onward consistency in models, is that the voltage regulator ground or adjustment pin following resistance to set voltage, is to be strictly returned to the transistor processing ground return/ lift exiting the positive input and v- point involving at that point the collector base connection, rather than ground. Earlier models prior to Revision 7 were done this way.... A LM317 is configured as a voltage regulator 1.25 x 520/220 + 1 and a 330n is placed from adj to this same transistor lifted ground. then onwards to current regulation 1.25/26 to connect to wiper... this provides measured 3.86v ref to gnd, at LM317 vOut to LM317 in.
Cheers / Chris
Last edited:
Are there any measurements of how the balance performs? It would also be nice to see a measured gain-position curve.
I agree, you still need to match LED/LDR devices. LED active current control helps, especially at low levels, but in my experience does not replace the need for matching.
And I think you need to provide some sort of front panel balance control. The better the match, the less you need to use it, if at all.
From the app notes Tom referenced in the other precision LED/LDR thread, it's clear these things age and change resistance over time. If Murphy has anything to say about that, that won't happen at the same rate for every device, meaning matched today could be unmatched down the road.
I was thinking more old school - many of my older tuners have a built in 400 Hz stereo reference tone. I could switch this in for a given volume position, and adjust the balance through centering the tone on speakers, headphones, or analog metering (or maybe LEDS). And just map the balance shifts needed, if any, vs. volume position.
I like the KISS principle.
Also wonder why no one is using the NSL-32SR3? It has lower distortion, and better low current control.
Oh wait, NP selected that one, didn't he? 🙂
- Status
- Not open for further replies.