LDR as a digital pot

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Matador

Well-known member
Joined
Feb 25, 2011
Messages
3,063
Location
Bay Area, California
I was looking over the schematics to the Mesa Boogie Triaxis, which was one of the first all-tube preamps that could be controlled via MIDI.

They accomplished this by replacing the traditional controls in the tube circuits with dual LDR's, and wrapping one side of each LDR within the feedback path of an op-amp to linearize the IV curve.  Mesa's patent just expired last month, so I don't feel bad about posting the circuit here!

images


So the opamp adjusts the drive of the LED until the non-inverting terminal equals 2.5V (the adjustment of the pot at the inverting terminal), which means the resistance of the fixed resistors at the analog switches is equal to the resistance of the LDR at that brightness (and since the LDR is a dual LDR, the resistance of the 'other side' should be equal as well).

However dual LDR's are extremely hard to find:  is there any reason why the LDR can't be replaced with a dual JFET or MOSFET?  The op-amp could drive the gate and adjust the gate drive to make the resistances between the two devices match as before. 

The one limitation I can think of:  most JFET's can only have 40V or so across the SD junction, and MOSFETS have a parasitic diode that prevents the source from rising above the drain.  Hence these devices may not work in cases like the midrange of a Fender tone stack.

Is there nay reason this can't work with regular dual FET transistors?
 
Matador said:
I was looking over the schematics to the Mesa Boogie Triaxis, which was one of the first all-tube preamps that could be controlled via MIDI.

They accomplished this by replacing the traditional controls in the tube circuits with dual LDR's, and wrapping one side of each LDR within the feedback path of an op-amp to linearize the IV curve.

I believe the reason for th dual LDR / compensation might in fact have been thermal drift. LDRs are sensitive for infrared light (heat) as well as other, more useable parts of the spectrum. Use LDRs without that compensation and you basically build an acoustic thermometer - the sound changes with tempaerature. Particuarly joyful in a tube amp ;-)

I owned 2 different programmable preamps / amps in the late 80s. The sound was always changing, despite my efforts to use fans and all kind of stuff. Incredibly useless but expensive gear. It took me decades to really understand why. The dual LDR seem to be a smart solution.

This obviously doesn't answer your question for a replacement part for the LDR, but maybe it helps putting the circuit into perspective. Marshall had a programmable preamp as well and they didn't use LDRs. Maybe someting to look into. I believe that you don't need any compensation if the part you use is thermally stable.

Michael
 
Matador said:
I was looking over the schematics to the Mesa Boogie Triaxis, which was one of the first all-tube preamps that could be controlled via MIDI.

They accomplished this by replacing the traditional controls in the tube circuits with dual LDR's, and wrapping one side of each LDR within the feedback path of an op-amp to linearize the IV curve.  Mesa's patent just expired last month, so I don't feel bad about posting the circuit here!

images


So the opamp adjusts the drive of the LED until the non-inverting terminal equals 2.5V (the adjustment of the pot at the inverting terminal), which means the resistance of the fixed resistors at the analog switches is equal to the resistance of the LDR at that brightness (and since the LDR is a dual LDR, the resistance of the 'other side' should be equal as well).

However dual LDR's are extremely hard to find:  is there any reason why the LDR can't be replaced with a dual JFET or MOSFET?  The op-amp could drive the gate and adjust the gate drive to make the resistances between the two devices match as before. 

The one limitation I can think of:  most JFET's can only have 40V or so across the SD junction, and MOSFETS have a parasitic diode that prevents the source from rising above the drain.  Hence these devices may not work in cases like the midrange of a Fender tone stack.

Is there nay reason this can't work with regular dual FET transistors?
The ancients are stealing our ideas again... I tried that on the bench back in the 70s and dismissed it.  I grounded the common connection between two cells of a dual LDR. One half was placed into an op amp NF circuit with a resistor connected to a positive voltage. The op amp varied the current feeding the two LDR light source to effect a variable resistance. Using a dual should insure decent tracking between halves.

The half not in an op amp NF path was connected as a simple attenuator. This worked well enough for a one off, but was not ready for prime time in console automation, even the old early VCA s were far more practical for that.

FWIW I also messed with dual JFETs for that same pursuit of investigating practical VCA replacements.

JR

PS: do you happen to have that dual LDR patent number handy? I am curious since I did something very similar on my bench 40+ years ago... 
 
Matador said:
However dual LDR's are extremely hard to find:
Even with factory-matched dual LDR's there is a huge dispersion, which makes tracking between LDR's difficult to optimize.

  is there any reason why the LDR can't be replaced with a dual JFET or MOSFET?  The op-amp could drive the gate and adjust the gate drive to make the resistances between the two devices match as before. 

The one limitation I can think of:  most JFET's can only have 40V or so across the SD junction, and MOSFETS have a parasitic diode that prevents the source from rising above the drain.  Hence these devices may not work in cases like the midrange of a Fender tone stack.
This is not too difficult to overcome, but requires additional circuitry, particularly because the FET and all its control circuitry must be floated, except actually the midrange pot in the Fender stack, which is referenced to ground.
A serious issue is the huge distortion when the audio signal exceeds about 100mV. This results in a compromised noise performance and needs rearranging the level diagram, which can be damaging to the sound signature.
Read that article https://www.onsemi.com/pub/Collateral/AN-6603.pdf.pdf
 
Matador said:
Mesa's patent just expired last month,
I wonder what they had actually patented, since the idea of reciprocating was not new then; when I learned that at school in 1968, it was old hat. Reciprocating bipolar transistors for current mirrors was usual.
 
I wonder how well matched are the parameters of dual FETs that are important for gain reduction applications? I wonder if a push pull arrangement could be made that cancels a lot of the distortion FET gain cells produce?

Cheers

Ian
 
ruffrecords said:
I wonder how well matched are the parameters of dual FETs that are important for gain reduction applications? I wonder if a push pull arrangement could be made that cancels a lot of the distortion FET gain cells produce?

Cheers

Ian
Push-pull arrangement is good at reducing 2nd-order distortion. FET's, being symetrical, and with the help of the standard half-voltage feed to the gate, already does it.
There would be a theoretical advantage in using paralleled FET's, with the 3dB-per-doublement law; however, the issue there is that each FET's control voltage must be adjusted separately, which turns out to be a practical nightmare.
 
ruffrecords said:
I wonder how well matched are the parameters of dual FETs that are important for gain reduction applications? I wonder if a push pull arrangement could be made that cancels a lot of the distortion FET gain cells produce?

Cheers

Ian
I experimented with dual JFETs that were a single piece of silicon so should have similar substrate parameters.. still meh.

FWIW this is pretty mature technology and crazy to think some simple end run is available to better the massive process investments and design advances that THAT corp made in their later generation VCA designs. 

Many smarter people then me killed a lot of brain cells trying. But whatever,,,

JR

 
Here was the patent: application in 1992.

https://patents.google.com/patent/US5208548A/en?oq=5%2c208%2c548

Mesa's part was supposed to be equivalent to a Vactrol VTL5C4, except with dual LDR elements with a single LED.

https://www.thonk.co.uk/wp-content/uploads/2013/08/vtl5c3c4.pdf

It was apparently manufactured by Hamamatsu in Japan.
 
If you're talking about replacing the LDR in that amp with a JFET circuit I would say that is very doubtful. The LDR provides opto isolation. I wouldn't be surprised if there were a little DC between the LDR driver and tube side. And AC signal on the tube side is potentially very high. No, an LDR is perfect. Just watch Ebay maybe. Check all the little boutique guitar parts stores like small bear, thonk, tayda, amplified parts, etc.

UPDATE: Just searched Ebay for "dual vactrol" and got this:

The Xvive VTL5C3/2 is a dual-element optical isolator/coupler. It features high isolation, steep slope, good dynamic range, low drive current, and small light history memory. Ideal for use in audio limiting, compression, remote gain control and many other applications.
This device is functionally compatible to the PerkinElmer VTL5C3/2.


UPDATE 2:

Actually it looks like VTL5C4 is much slower (1.5s vs 35ms turn off time) so that part won't be quite right:

VTL5C3    30kΩ @ 1mA 5Ω @ 10mA 1.5Ω @ 40mA 10MΩ 20 (typ) 75db (typ) 2.5ms 35ms
*VTL5C4    1.2kΩ @ 1mA 125Ω @ 10mA 75Ω @ 40mA 400MΩ 18.7 (typ) 72db (typ) 6.0ms 1.5sec


There is an Xvive VTL5C4/2 but not on Ebay.
 
The parts are definitely drying up:  with the cadmium sulfide being not kosher with RoHS they are becoming harder and harder to find.  Mesa wants nearly $15 each, and a single Triaxes preamps has almost 30 of them. :(

I suppose they may be sortable like other devices when ordered in the 100's:  then two LED's can be run in series and you hope each track each other pretty well?
 
Matador said:
Here was the patent: application in 1992.

https://patents.google.com/patent/US5208548A/en?oq=5%2c208%2c548
So it looks like what they've patented is slowing down the NFB loop in order to damp the response (the capacitor between output and inverting input)!
 
..makes sense - it's the calming of control-loop hysterics that is the tricky part here (C* in pic), but I never expected such to be patentable.

I spent quite some time in late 80'es trying to get opto VCA's behaving like VCA's or OTA's regarding control curves, so that predictable results could be had. Only later I learned to love the unpredictable.

pic related

Jakob E.
 

Attachments

  • Ratiometric_opto-vca_idea_1990 - upd2014.JPG
    Ratiometric_opto-vca_idea_1990 - upd2014.JPG
    321.3 KB · Views: 39
abbey road d enfer said:
Push-pull arrangement is good at reducing 2nd-order distortion. FET's, being symetrical, and with the help of the standard half-voltage feed to the gate, already does it.
Except that technique still only works over a limited signal level range. Push pull cancellation shuld be level independent?


Cheers

Ian
 
ruffrecords said:
Except that technique still only works over a limited signal level range.
Really? As far as I can tell, distortion increases continuously with level, so the "limited range" is set by how much is tolerable. I don't see any reason why it would be different in a push-pull configuration.
Similarly to a tube amp, once the push-pull arrangement has decreased 2nd-order distortion, the remaining products increase continuously. That's why NFB is necessary; unfortunately it doesn't apply to voltage-controlled resistors.
 
So although the dual LDR's are rare, single LDR's are still obtainable and common at reasonable prices:

https://www.digikey.com/product-detail/en/advanced-photonix/NSL-32/NSL-32-ND/5039800

Which are $2.88 each in quantities of 10.

So LDR's specify a maximum on resistance (40 ohms for the part above), and the minimum off resistance (500k for the part).  If I were to run a constant current of 10mA through both diodes from two random LDR parts at the same time, I wonder how much the absolute difference in resistance would be?  And even if resistance differed by 25%, wouldn't this be trim-able with an external pot in the "second" LDR?  And in common guitar amp tone controls, would the difference even be audible?  8)

I guess I don't have a feeling for a) how much different LDR's differ in maximum resistance at a given input LED current, and b) how different the IV curves looks from part to part as a function of LED current.
 
Matador said:
So although the dual LDR's are rare, single LDR's are still obtainable and common at reasonable prices:

https://www.digikey.com/product-detail/en/advanced-photonix/NSL-32/NSL-32-ND/5039800

Which are $2.88 each in quantities of 10.

So LDR's specify a maximum on resistance (40 ohms for the part above), and the minimum off resistance (500k for the part).  If I were to run a constant current of 10mA through both diodes from two random LDR parts at the same time, I wonder how much the absolute difference in resistance would be?  And even if resistance differed by 25%, wouldn't this be trim-able with an external pot in the "second" LDR?  And in common guitar amp tone controls, would the difference even be audible?  8)

I guess I don't have a feeling for a) how much different LDR's differ in maximum resistance at a given input LED current, and b) how different the IV curves looks from part to part as a function of LED current.
It has been a number of decades since I looked at this seriously. I was bench testing with a dual LDR that should have tracked with itself pretty well.. I abandoned it for a number of different reasons.

Of course you can trim anything to match at one specific level.

You could also teach a microprocessor to learn the actual resistance curve vs drive current for individual device variations, making this approach even more practical (again not to be too encouraging). This could even eliminate using the second LDR and  allow the micro to only operate in the side chain as much or as little as desired. Providing a pure old school analog path, with optimized ideal LDR, plus all kinds of modern bells and whistles, not to understate benefit of gain accuracy.

This might even be patentable but I give it to the world for free... 8) Of course do a search since ancients often steal my better ideas.

JR
 
squarewave said:
That's not the problem. The problem is that VTL5C4 has a very slow release at 1.5 seconds.
The inherent attack/release characteristics of the original technology was accidentally responsible for the success of some simple early designs. We like to think that all popular legacy designs are the result of masterful engineering. Sometimes it was just chance.

or not...  ::)

JR
 
Back
Top