LDR attenuators - Better balance with DC bias

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all

At the risk of being shot......
I think it may be useful to add a bit of DC bias to the photocells handling the audio signal.

``````````````````````````````````````````````

The idea is that the photocell's resistance is the same at DC and audio frequencies, so by measuring the DC voltages you can find out exactly what the attenuation at audio frequencies is.

More importantly, it becomes fairly easy to implement a servo which compares the DC voltages of the left and right channels, and adjusts the LED currents to give (almost) perfect channel balance, without any need for trimmers or matching of parts.

If DC offset is considered undesirable, one can use a symmetrical bridge type circuit to cancel any DC at the audio input and output.

OTOH, in AC coupled circuits where DC offset isn't a problem, there's another advantage to be had; Nelson Pass's measurements showed the distortion to be mainly 3'rd harmonic, but with suitable DC bias the distortion characteristic can be changed to being predominantly 2'nd harmonic, which some people may prefer.

Also, according to PerkinElmer:
The minimum distortion or threshold distortion shown
in Figure 3 is a second harmonic of the fundamental frequency. The
actual source of this distortion is unknown, but may be due to some
type of crossover nonlinearity at the original of the I-V curve of the
photocell.
If there is indeed some kind of zero-crossing distortion in their devices, it can be avoided by biasing the device away from zero.

Cheers - Godfrey
 
Last edited:
Member
Joined 2004
Paid Member
A few days ago I had same idea if it would be possible to use DC in a LDR circiut and have possibility of resistance measurements in operating conditions.
This would allow to adjust resistance fine enough to use LDRs in a balanced circuit. The 2nd harmonic distortion of LDR should be canceled out in balanced circiut.
If it would be possible also, as you state, to change 3rd harmonic by DC bias to be predominantly 2nd harmonic most of LDR induced distortion could be eliminated.
1543
 
godfrey said:
The idea is that the photocell's resistance is the same at DC and audio frequencies
Is this known to be the case? LDRs introduce distortion. Therefore they are (to some extent) non-linear. Therefore their AC and DC resistance will not be the same, although it may be close. Is it close enough that perfect DC balance implies good enough AC balance?
 
Is this known to be the case? LDRs introduce distortion. Therefore they are (to some extent) non-linear. Therefore their AC and DC resistance will not be the same...

Is that indeed the case? If the current versus voltage characteristic were (say) linear plus a small square law term, the device would be nonlinear, but could have identical AC and DC resistances. All it needs is sufficiently low reactance (stray cap or inductance).
 
@DF96: Close enough, I think. The distortion would have to be pretty horrific to get even 1% difference.

Below are a couple of "simplest possible" circuits that I'm thinking about to start with.

In the first one, the DC voltage at X is directly proportional to the gain (i.e. 1 / attenuation) of the circuit. For control purposes, it will either have to be filtered or compared with a reference and integrated. If an output buffer is used, that could be configured as a 2'nd order VCVS filter to minimize the VLF output when the user twiddles the volume knob.

In the second circuit, The DC voltage at X is also proportional to gain, but there will be minimal signal voltage imposed (if the DC offset is zero). The DC voltage at Y can be monitored (via an integrator) by a second servo to set the DC offset to zero.

I take back what I said about the servos being easy. One awkward bit is choosing time constants due to the LDRs having very different rise and fall times, both of which vary wildly with attenuation level.

There's also the question of whether the control system should simply be linear or if it would be better e.g. to take the log of the measured voltage (to get a control proportional to dB), and use an exponential function for the LED current to compensate. Maybe some overall non-linearity is called for? Nelson's measurements suffest that, but looking at some datasheets, it seems to depend on the selected device.

And then there's dynamic range vs noise and topology......
Much head-scratching ahead.
 

Attachments

  • simple.GIF
    simple.GIF
    2.3 KB · Views: 429
  • bridge.GIF
    bridge.GIF
    5.5 KB · Views: 425
Last edited:
SY said:
f the current versus voltage characteristic were (say) linear plus a small square law term, the device would be nonlinear, but could have identical AC and DC resistances.
I understand DC resistance to be V/I and AC resistance to be dV/dI. A non-linear curve could have these equal at certain points, but in general they will be different. If it is only weakly non-linear then the difference will be small enough to ignore.
 
It could be done with a micro, and memory for the measured LDR curves. But this way it could be done with no memory, or processor, which could also be accomplished by using pilot tones that were outside of the audio range.

I think it might be difficult to filter out the DC, well-enough, especially since music usually skews one way or the other, for a short time, continuously, meaning the filter time constant might have to be too long, in order to ignore those meanderings.
 
Last edited:
It could be done with a micro, and memory for the measured LDR curves. But this way it could be done with no memory, or processor, which could also be accomplished by using pilot tones that were outside of the audio range.

Yep, but on the other hand, is it really a real time problem? It would be also possible to set volume, hit a "cal" button, and have the out of band tone do it's thing for 1-2 secs, then turn off. No one could really complain about that, because it's normally off.
 
The devil is in the details. Have you played with LDRs? Some points to consider:

1. At higher resistances, it takes quite a while for resistance to completely stabilize. At high resistances, it can take a minute to completely stabilize and then it wanders around a central value +/- a few hundred ohms.

2. Whatever control mechanism you use, it must be capable of accurate control over a range of maybe 15ma at 40 ohms to .005ma at 10K. There is a wide range of sensitivities, I'll attach a list I compiled using my present current control arrangement. At the high end, a change of .001ma can change the resistance by 1K ohms down, or 10K ohms up (a range of 9K~20K). With some of these devices at 10K ohms, a slight change in current will make the unit act like it fell off the edge of the world, resistance-wise.

I believe that I've read that udailey has limited the upper end of his Lighter Note to 8k ohms to avoid the drop-off at 10K.

3. What measurable DC voltage will you use in a device whose resistance will vary across a range of 40 ohms to 10K ohms, always keeping in mind that power is limited to 50mW?
 

Attachments

  • ScreenShot006.jpg
    ScreenShot006.jpg
    137.3 KB · Views: 271
  • ScreenShot009.jpg
    ScreenShot009.jpg
    77 KB · Views: 260
Last edited:
I would expect that the monitoring circuit would compare the DC output voltage from the two sides. Make an adjustment and then compare again. Alternatively, a VERY SLOW time constant could be used to compare the two DC outputs.
One could put "cal" on a timer, 2seconds or 2minutes or whatever. Then it holds and disconnects itself.
 
It's not just to avoid a pot, is it?

1. The LDRs wouldn't need to be matched.
2. LDRs can't really be matched, anyway. They don't just differ by an offset. Their R vs LED Current curves all have different shapes.

Some of the possibilities:

1. Use digital memory and auto-calibrate once, or at every startup, or whenever the Cal button is pressed. Auto-cal would result in some type of lookup table for each LDR/LED device, so that the two channels could be controlled to give accurate resistances for each of the four devices, for every setting of the attenuation control.

2. Don't use digital memory and use an out-of-band signal (DC or HF) and a automatic feedback control system to continuously either a) set all four resistances for correct and matching attenuation, or, b) cause the slave channel's two LDR/LED devices to track the resistances of the master channel's devices, for whatever resistances happened to result in the master channel's devices, from the attenuator pot setting.

3. Similar to 2, above, except that the out-of-band signal would only be engaged when the attenuation control was changed. Then the resulting settings would need to be remembered, somehow, maybe with digital memory or digital pots, or sample-and-holds.

It might have been nice if the music signals could have been used, to constantly auto-cal, by using current- and voltage-sensing instrumentation amplifiers, RMS-to-DC convertors, and analog dividers, to calculate a voltage corresponding to each resistance, continuously, which would have been fed to differential integrators that would have also used setpoint voltages derived from a Lightspeed-style pot's output voltages and would have then provided control voltages to voltage-controlled current sources for the LEDs.

That might have worked, somewhat, if the time-constant was extremely large/slow (except when there wasn't any music!). But it seems like it would be far better, anyway, to hold the resistances completely constant, except when the attenuation is being changed by the user, which probably rules-out 2a, above. But, the same sort of system might work as part of the auto-calibrator implementation for number 3, above. Maybe only one such measurement-and-control system would actually be needed, with switching to let it look at each of the four LDR/LED devices sequentially during auto-cal.

But I have to say that it's looking like wapo54001's approach (see his thread), which is basically like number 1, above, I think, is looking better and better, although 3b, above (like 2b) might be simple-enough to make it reasonable to attempt.
 
Last edited:
They seem to just replace one set of problems with another (worse) set.

I wonder if one couldn't get better results using JFETS instead of LDRs as the variable resistance elements. Here's links to a couple of relevant documents:
http://www.eecg.toronto.edu/~kphang/papers/2001/martin_AGC.pdf
http://www.edn.com/contents/images/120601di.pdf

It should be easy to get just as much distortion as with the LDRs, for those that like that sort of thing. Ahem, let me rephrase that.... Those that like the sound of LDRs despite their distortion may also like the JFET version despite it's distortion.

Sample circuits below.
 

Attachments

  • jfet.GIF
    jfet.GIF
    6.7 KB · Views: 187
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.