DIYHiFi.org

All,

I'm looking to make a 5 channel volume control with remote control for my home theater amp (tweaked UCD chopper). The source is a PC and I have just upgraded to a Lynx 2B soundcard (after ditching an EMU 1820m - nice sound but unreliable) and the pots at the back of the amp or software volume control just don't cut it anymore.

So in my research I have looked at IC based volume (convenient to build but compromising in the sound department) or a resistor ladder controlled by relays (the number of relays gets out of hand for 6 channels) or a LDR based volume control. The LDR looks the most promising as it looks pretty easy to build and seems to be the "fad" volume solution for volume of the last 2 or so years (see the mega thread about 'lightspeed' on diyaudio). Apart from the usual concerns about source/amp impedances for a passive attenuator, I can't correlate the distortion specs for these devices with the rave reviews that LDR based pots get as they shouldn't sound great (distortion about 0.1% depending on configuration, spectra mostly 2nd harmonic).

My thinking is to use a 6 channel programmable potentiometer connected through the serial port of the PC for each channel, driving a voltage controlled current source for the LDR LED, and using a fixed resistor with the LDR in a shunt arrangement for attenuating the audio signal.

Reading through the material on diyaudio it seems there is lots of subjective assessment of LDRs for volume control ("best volume control ever", "so transparent", "gets out of the way" etc.) which makes me suspicious especially as folks over there prefer a higher distortion part and a configuration with 2 LDRs in a series / shunt arrangement when a single shunt LDR would be lower distortion - although Nelson Pass has posted some decent measurement data and a schematic.

As this forum is more focused on the engineering aspects of audio, I thought I'd post here asking what folks here think about LDRs as volume controls and any experiences that can be shared, especially to reconcile the rave reviews vs the distortion figures. I suspect the discussion will go along the lines of an exotic resistor comparison for a DAC I/V.....

If you want more information - the LDR part I'm thinking of using is the Silonex NSL-32SR23S.
Silonex application & tech info here: http://www1.silonex.com/audiohm/constants.html and http://www1.silonex.com/audiohm/levelcontrol.html
Nelson Pass assessment and comments on LDR volume controls: http://www.diyaudio.com/forums/showthread.php?postid=1520215#post1520215

Thanks for your feedback!

According to the distortion graphs posted by Pass 0.1% is the THD at 2V input - that's not the vol setting you listen at, is it. Figure this out & then check the graph - more in the range 0.003 to o.01% and the series/shunt ldr config has higher thd.

_________________
There is a crack in everything, that's how the light gets in!

Good point - typical listening levels would have lower distortion levels according to the graph, to the point where distortion in other parts of the system (especially speakers) will swamp the distortion of this device.

Is this approach something that you would use in your system for volume? I'm trying to understand why this solution gets rave reviews - is it audio voodoo at play or is there some real science behind why this solution apparently works well?

Your problem is going to be matching of LDRs for 6 channels - manufacturing tolerances are not good!

Don't think anybody has worked out what's at play but they do sound good

_________________
There is a crack in everything, that's how the light gets in!

Yes - however one advantage of PC control and why I plan to have 6 x individual programmable potentiometers DS1806 and per channel current sources is that I can measure the LDR resistance vs LED current performance of each LDR, and setup a mapping table in the PC to align the response per LDR to a standard curve. Basically the volume control on the PC is planned to be a dotnet application that tunes the per channel volume according to the specific LDR characteristics (measured offline) and sends the volume word to the DS1806 by bit-bashing the RTS and DTR RS-232 control lines fed directly into the DS1806 (avoiding having to use a micro for interfacing). So with a bit more complexity with a VCCS per channel (an opamp + a couple of resistors per channel - pretty simple) and some coding I can tune out the device matching issues. Selecting devices which match reasonably well to start with will help. Also for 6 channels, the L+R balance is the most important, the other channels don't need to track as closely.

Does this sound workable?

I've had a chance to put my hands onto one LDR unit. It had LDR both in series and parallel though. No way to talk about any breakthrough here, but it sounds good indeed. Especially taking into account very poor distortion figures. For me this means that this concept may work but more linear parts than those offered by Silonex are surely needed. IC/silicon resistors or so called digital potentiometers are by far more linear (and I didn't really found them really sonically compromising, so long as they don't include opamp at signal path, and especially not when you consider their price).

As for the LDR distortion, it increases as the signal level increases, and it increases even more as you "turn the wiper" down. Thus the worst case with 2V RMS input level and -80 dB attenuation bring more than 10% THD, and the string of odd harmonics dominate, with significant second harmonic too. That case is surely of no practical interest but it may give you some wider view angle. Some more typical case with 0.5V input level and -12dB attenuation produces 0.007% THD (almost solely second and third harmonic), but at -40dB attenuation THD increases to 0.46% (again the string of odd harmonics plus second harmonic).

Thanks Pedja for the feedback.

I looked into LDR volume 15 years ago and the silonex part although not perfect is more linear than the old LDR setup I tested (which had >1% distortion at normal listening levels) - so the silonex NSL-32SR23S part is quite good compared, and if its around 0.07% for typical levels that should be fine. I intend using a single LDR in a shunt arrangement which should be lower distortion than the LDR series/shunt arrangement that you tested.

I will build this and report how it turns out, today I prototyped the USB/Serial code to drive a DS1806 digital pot for the LED control, and used LTSpice to fine tune the VCCS, so will order the parts next week. If folks are interested I can post a schematic & protel board layout (double sided with lots of groundplane for shielding) that I'll do on Sunday.

Anybody care to surmise why the LDRs sound so good even with this THD. Is the THD mostly 2nd harmonic, is the THD cancelled in some way in use, does Georgehifi's theory of avoiding the capacitive behaviour of micro-contacts at high frequencies in pots hold?

_________________
There is a crack in everything, that's how the light gets in!

I think its euphonics, 60's 70's studio gear they use for vocal levelling like LA2A limiter has LDR element too.

Well, you want someone to seriously consider these LDRs, and do the required steps to put this stuff into its own place, correct?

Taking into account distortion performance of currently available parts, I don't consider LDRs worth further effort, but if you belive they do, or you have found someone else that does, then ask them to put one LDR into the signal path and to compare it against some known resistor. Characterization done this way will tell you most everything.

As for distortion, no it is not 2nd harmonic. Nor do I know why and how it may cancel. So here we go:

1. 2V RMS input, no attenuation, THD=0.02%




2. 2V RMS input, -20dB attenuation, THD=0.74%





3. 0.5V RMS input, -20dB attenuation, THD=0.007% (mistakenly I typed above "-12dB", it should read -20dB)





4. 0.5V RMS input, -40dB attenuation, THD=0.46%





Nice signature btw, and the thought that also runs through my head and is actual these days... As Cohen will be in Belgrade on 2 September.

I misunderstood you - I thought you liked their sound?

BTW what would be the sound level at the speakers through a UCD400 with these settings?

Do you have a link for the graphs (can't find them on Dr Jordan's site) - no details are given - Silonex ldr ? Device?

Some different results to your post - link will be supplied if requested but it's to diya

Images attached

Are you going to Leonard? Best concert I've ever been to. Beats Pink Floyd in Montreal's Olympic Stadium 1976 (that puts a date on me I guess)

_________________
There is a crack in everything, that's how the light gets in!

Nelson's Measurements for a series/shunt version:

_________________
There is a crack in everything, that's how the light gets in!

Final set of images at mid attenuation from above post


Edit: What I meant to say in above post: BTW what would be the sound level at the speakers through a UCD400 with the vol settings in your graph?

_________________
There is a crack in everything, that's how the light gets in!

Pedja,

What LDR device are you measuring, the NSL32SR2 or the NSL32SR3? The SR3 although harder to match (wider resistance range) is a lower distortion part.

I have a couple of thoughts on why it sounds good. Firstly the LDR series+shunt configuration has a -6db distortion null where the resistance of both LDRs are equal as the 3rd harmonic components cancel each other. Secondly the distortion spectra is dominated by the 2nd harmonic especially around the null, and 2nd harmonic sounds "nicer" (compared to distortions where more of the distortion is in higher harmonics) - so setting your normal listening level to be around this attenuation would be optimal. As Nelson has pointed out below, listening above -6db has an "expansive" compression characteristic which may give a bit more life to the sound.

Regards,
Dean

I do consider them good, but subjective performance doesn't make me swallow their distortion performance. And such an inconsistency along the wiper road is a bit frustrating to me.


Not familiar with UCD, but it should be easy to multiply resulting voltage with amp gain?


Those are my graphs, hosted at my space, and DrJordan is only the software I use. Silonex, yes, NSL-32SR3.


Sure, no way to miss something like this. I will let you know.