Lightspeed Attenuator a new passive preamp

Audio Precision Testing

I will test the Lightspeed but I am sure its the same. I just tested the Lighter Note on my Audio Precision. Bought one a short time ago and just hooked it up. My camera wont take pics of the screen because its bright vs the darker background. I will see if another camera in the house will do better. Most tests were done at 1V and 1kHz. There was a certain spot in attenuation that produced the least distortion. Interestingly its my favorite listening position on the dial. As long as the test is a sweep type test I went from 20-20kHz.
Anyway
Noise is 16uV no matter what voltage I am putting through it and is just a straight line.
THD+N has a few results
- at .5V=.007%
- at 1V = .0055%
- at 2.5V= .0154%
- at 10V=.146% which surprised me. I expected much higher as in several percent. It certainly sounds bad with high voltage through it even if not measuring bad.
I have a balance function on my remote and was curious how precise I could dial it in.
With 1V in at a bit less than half volume I dialed in 430.3mV on Left and 430.2mV on the right. Dead center if you ask me.
XTalk varies a lot. The best I got was -103.4db @782.5mV output, thats with 1V in. I have my volume limted so the kids cant screw it up plus the balance pot limits the volume. It seems that the higher in volume the lower the crosstalk.
Worst crosstalk was -64.27dB at 76.7mV which is limted by the 40Ohm rule of the shunt LDR. If it would go lower then the crosstalk would get worse.
IMD or intermodulation distortion is something I am not very familiar with at all. Reading the AP manual I am guessing that its supposed to give me a ratio of the low to the high frequency as it sends two frequencies through at once and I suppose measures distortion where they intermingle or something. Will read up on it. Anyway, I dont see a way to read or set the ratio yet. The settings are 1V in and with attenuation I am getting 365mV out. The low frequency was 250Hz for this reading and the high was 7k. IMD reading was .0108%.
Well, thats all I have got for now. This is with NSL32SR2. I think those of you wondering about SR3 at least can rest easy that SR2 is a good device.
Uriah
 
Last edited:
An interesting little snippet for those of you who own or have built a Lightspeed Attenuator, this will give you maybe some bragging rights.
Not only does the $28k DarTZeel NHB 18-NS preamp have a form of Lightspeed Attenuator controlling the volume (which they tried to patent).
But now there is a new player on the block, Constellation Audio with their $60k fet line stage preamp, the Altair which also uses a Lightspeed Attenuator form of volume control. (at least they didn't try to patent it)
Altair

Cheers George
 
Could be 48 1% trimming resistors to go with the quad unmatched ldr's, which in my opinion would detract from just a quad matched set, as the i/o impedances for both channels would not remain constant, and there would be another component in the signal path.
Or it could be 48 x ldr's in some paralleled setup for series and shunt ldr's, again this would definitely detract from a simple matched quad set.

Or it could be just a 48 position switched volume control instead of continuous rotary pot with a quad matched set, this would then sound the same no different than my production Lightspeed Attenuator.(this is the most logical explanation)

Cheers George
 
a series pair of series LDRs can easily be set up to give 50k each for a total 100k of series resistance.
A parallel pair of shunt LDRs can be set up for 40r each for an effective 20r of shunt resistance.

This set requires 8 LED/LDR for a stereo control.
-0.1dB to -74dB is within reach as the range from this doubled quantity of LED/LDR.

A few mores series and paralleling can be used to avoid the need to match and to give an even wider range of control. Avoiding matching releases a lot of cash to buy extra LED/LDRs.
 
Or it could be 48 x ldr's in some paralleled setup for series and shunt ldr's, again this would definitely detract from a simple matched quad set.

Paralleling LDR-s could average its characteristics... in this case we would get an LDR (array) that has lower impedance and characteristic that reflects more the average of a lot. But will paralleling resistors mean more distortion? They could use 12 LDRs paralleled and have 4 arrays ar for a SE setup is needed. If it is Balanced then 6 of them / array.
 
Paralleling LDR-s could average its characteristics... in this case we would get an LDR (array) that has lower impedance and characteristic that reflects more the average of a lot. But will paralleling resistors mean more distortion? They could use 12 LDRs paralleled and have 4 arrays ar for a SE setup is needed. If it is Balanced then 6 of them / array.


This could be the way, but if one thing the Lightspeed has taught me, it's the more is less principal. To have that many ldr's in the singnal path I don't know, something ring alarm bells to me.
Where you gain in lower labour cost (no matching) you may loose out on extra colourations of too many in the signal path.



Cheers George
 
The only downside to me is noise. Well, cost, board space, supporting circuitry, MTBF.... guess there could be a lot of downsides. Back to Johnson Noise. A resistor make its own noise the higher its resistance. I dont know if we parallel a bunch of these together and they start acting like one resistor with lower Johnson Noise or a bunch of resistors with a very high cumulative Johnson Noise.
 
Okay
Got a new batch of 1000 LDRs. The body style is different again. Almost everytime I order there is some sort of chang. This time I am enthused by what I see and feel. I am sure they sound great to of course. The thing I am most hoping for is better matching. The last batch was the second worst matching I ever got. They would match near perfect in the last three measurements but all over the place in the first measurement which means many never made it into customers hands. Lots of lost product there.
So whats different? They still look the same as far as size and being a small black tube with wires coming out of it. This time though the tube seems to be perfectly filled with black epoxy to the point that it seems that it is not a tube but a stick of black plastic. One piece is what it seems like. Of course its still filled with epoxy, but they just did a better manufacture job. The wires are not as thick as they have been but not as thin as last time. A decent thickness and this is encouraging. They are not frail. Writing on the sides is green. Never has been green before and the dot is white while in the past it has often been silver. SO, anyway I am filling the test jig. Will be a week or two before I have results as I dont like to speed through any of it. Would rather spend my time with my young uns! :)
Hope you guys all had a great summer and are heading into a perfect fall. Its been hot as you know what here. Well over 50 days of 100+ weather. I stopped counting.
Uriah
 
I am keen to make one of these attenuators, and have already made an Eagle PCB ready for when I order my parts. I have a couple of questions ... (These question may have already been answered already in this thread, but I have only read the first 50 or so pages, there's only so much you can read in an evening.)

1. Why are the resistors used 100 ohm? Surely when the pot is fully in one direction you have 5v/100 = 50mA, the Silonex devise LEDs are only rated at 25mA?

2. Do the devices have a logarithmic or linear response? Anyone got a sample plot of resistance against mA draw I can look at? (Basically I want to know whether I will need a log or a lin pot)

3. I noticed in a few posts, that some people said that their attenuation never reached quiet and was nearly full at about 1/4 turn. Surely, when the pot is turned right down, you would have voltage divider of 40ohms / 1M ohm, this would be silent, wouldn't it? As for max volume from 1/4 turn and up, could this be the 100ohm resistor allowing too much current to the LED?

4. Why are there only 2 trimmers used on 2 of the devices? Why are they not used on the other 2?

Again, sorry if these questions have been answered already in the previous 4233 posts :)
 
Last edited:
Another question if someone would be so kind to answer it....

If a 100K linear pot was used and it was set to halfway, that would be 50Kof resistance in series with the top 2 devices and 50K to the bottom 2 devices. 5v / 50K = 0.1mA ! Is 0.1mA enough to light the LED inside?

Is there a point at which the LED doesn't light, and if so, wouldn't this give a band on the pot where the volume doesn't appear to change?
 
Thanks. I'll keep reading on. I'm at post #630.

So far, it looks like a 100K dual log pot with 47uF caps across LEDs seems a good place to start.

There will be no need for the 47uf across each led, as if you look at the power supply for the production Lightspeed Attenuator, it already has a 470uf cap across the leds.
As far as using the 100k log pot goes with quad matched NSL32SR2S this is the best progressive control curve from min volume to 2 o'clock then it ramps up faster, which is the ideal curve to have.

Cheers George
 
Hi GeorgeHiFi. Thanks for your reply.

I've only just got to post #757, where I notice you have used only 1 trimmer and single gang pot. I suppose that's why you need quad matched NSL32SR2S now.

I should have read more posts before I spent 3 hours designing a board on Eagle which uses the old design with the 2 trimmers.

Do both these designs give the same quality, as the newer design has less components, but obviously needs quad matching. Also, you mention a 470uF cap across the LEDs....Is there a post with the schematic of this?

Also, where can I buy quad matched devices from?

Thanks again, and great work with this project :)
 
Last edited:
Hi GeorgeHiFi. Thanks for your reply.

I've only just got to post #757, where I notice you have used only 1 trimmer and single gang pot. I suppose that's why you need quad matched NSL32SR2S now.

I should have read more posts before I spent 3 hours designing a board on Eagle which uses the old design with the 2 trimmers.

Do both these designs give the same quality, as the newer design has less components, but obviously needs quad matching. Also, you mention a 470uF cap across the LEDs....Is there a post with the schematic of this?

Also, where can I buy quad matched devices from?

Thanks again, and great work with this project :)

Yes the trimmer only has to go on the louder of the to channels for final calibration. Quad matching gives you a better log feel to the volume control and more stable i/o impedances. Attached is the schematic it says 100uf but I put 470uf in, what's a few uf's between friends.
Cheers George
 

Attachments

  • Lightspeed-Passive-Attenuator-Schematic.png
    Lightspeed-Passive-Attenuator-Schematic.png
    59.7 KB · Views: 512
Last edited: