NSL 5910 (LA-2A, LA-3A photocell), Dynamical standardization, Tools, Measurement

[miquel]

In first time excuse my wrong English and this loooong text.

The objective is NSL 5910 the dynamic characteristics standardization.

One: Tools

I have built an impulse controllable generator to illuminate a high brightness white led, (10 mA current), and an aluminium dispositive to support the LDR and the led. Also there is a 1000 Hz, variable voltage (up to 60Vrms) generator, to study a real T4B. For the present analysis I need an 10,0 V cc. external source.

Image. Pulse generator and tool photos in operation:

http://i749.photobucket.com/albums/xx138/miquel_bucket/DSC01346.jpg


Two: The resistive compression net in LA-2A and LA-3A.

It’s easy to describe that resistive net; with Excel you can calculate the dB attenuation for an every LDR resistance. For example: with 1.000 Ohms (habitual value in strong light LDR situation) the attenuation is 29,4 dB, and with 1 MOhm (close to darkness) the attenuation is 0,2 dB (close to no attenuation…), see next image.

Image:

http://i749.photobucket.com/albums/xx138/miquel_bucket/TheLA-2Anet.jpg


Three: resistance measure and time registration.

We need a datalogger (I have a datalogger from a cheap digital oscilloscope).

I built a resistive measure net that it is showing in the below image:

http://i749.photobucket.com/albums/xx138/miquel_bucket/Resistancemeasurewithoscilloscopedatalogger.jpg

I chose the voltage measurement across the LDR because I like that the voltage rise when resistance rise. The dark voltage is in the top of the datalogger. It’s easy to write an EXCEL sheet to calculate every resistance for every read voltage. It’s important to read the voltage with the oscilloscope probe in 1/10, in order to the input impedance to be 10 MOhms. The voltage reading in oscilloscope is, obviously, 1/10 of the real voltage.


Forth: Chose a representative value

I chose 0,6 V measured, (that is 6 V real), because is “round” number, corresponds a resistance of 150 KOhm, and that resistance correspond an 1,5 dB attenuation (more or less). I think -1,5 dB is a good value to see the end of compression release. Of course can be another value, as -1dB or less, but than then the time will be longer.

Next I show the datalogger image from a Real NSL 9010 dynamics measure:

http://i749.photobucket.com/albums/xx138/miquel_bucket/01analisi900ms.jpg


Five: What we can see?

I see in left the line in the 1V value (real 10V), that is the voltage across the LDR when its resistance it’s very high (more than 50 MOhm). Next, the LED starts bright and, in short time, the voltage dawn to very low values (remember: LDR light resistance is less than 1000 Ohms…!). At the end of the light impulse (at the moment 250 ms constant) the voltage rise and this slope shape shows the real behaviour.

Next we can put the datalogger axis on the values: 1.- When the voltage start rise (the led turn off), and 2.- when the voltage arrive at 0,6 V (6 V real). The datalogger shows the milliseconds, in this case 800 ms.

The standard value for this LDR (number one) is 800 ms.


Six: Comparative between six unities LDR

Remember: unities are milliseconds to go from full compression to -1,5 dB, (with an 250 milliseconds impulse).

1 – 800 ms
2 – 825 ms
3 – 675 ms
4 – 1.175 ms
5 – 1.750 ms
6 – 1.000 ms


Conclusions

Now I have only six NSL 5910 unities not connected in circuit, that number is not sufficient for establish statistical conclusions.

There are big differences between unities, in really I can see three families, first is when the time is between 600 and 800 ms, second family time may be around 1 and 1,2 seconds, and the third family is number five, with 1,7 seconds.

Another analysis is the static behaviour, not described here, but I have measured the resistance in dark (well… at 30 seconds dark) and all LDR has more than 10 MOhms resistance except number 5 that it has only 2,9 MOhms. In really, at 60 seconds dark, all LDR has more than 20 MOhms and number 5 has only 3,7 MOhms. I thing this LDR has not quality sufficient for compression uses.

Anyone have experiences to share?

Cooming soon: static analysis… weah…..

 

[abbey road d enfer]

Nice job, but I think you should start with the static parameters, against which you could compare you dynamic measurements.

 

[miquel]

Today I started the static measurements....

Miquel

 

[nielsk]

I suggest a frequency sweep as well, especially if you are wanting to match cells. I think you may find different response across the audio band...

 

[CJ]

use to have the orig  urei  test method given to the maker

somebody wanted the vacuumbrain.com name real bad so i got leveraged out,

all you really need is on and off resistance ,

then put a pair into the t4b and listen.

some cells sound better, you can actually hear it,

 

[miquel]

Thank you

Now I am working in the static measurements with light patterns. Results coming soon. Very interesting.

But in this moment I can gift the full light and dark resistance for every LDR (remember, I have only 6 unities without installing into a T4B.

The first number is the LDR number (the same number that is shows in first post), the second is the light resistance in Ohms, (using the LED source from the experiment described in past post), and in third is the dark resistance at 30 seconds in dark.

(We need standardise the time to measure the dark value, because that MOhms value don’t stabilise until loooong after)

1    48 Ohms  23,0 M Ohms
2    36          10,6
3    41          17,0
4    50            6,5 
5    54            2,1 
6    40            8,6 

Notes:

The dark resistances has a repetitive index test very bad, I think is because I use a long light impulse (2 o 4 or more seconds) but I don’t control that exactly. When the resistance is very high it shows a very flat asymptote and it’s difficult to read a repetitive good value.

Conclusions:

The main differences between unities are in dark values (from 2,1 M Ohms to 23 M Ohms), but if you calculate the attenuation in real LA-2A net, the attenuation with 2,1 MOhm is 0,12 dB and with 23 MOhms is near 0. I think every dark resistance value higher to 2 MOhms (0,12 dB attenuation in dark) is good, in theory….!

The light resistance is every time very low, lower than NSL 5910 in a real T4B, (is known that are some hundreds Ohms). That is because de LED light is may be stronger than the Panel light.

In really I think that analysis is not very important, and may be it will be better the analysis with two light patterns to measure “two points” compression comparative between different LDR, and measure the different depth of compression with the same signal variation.

It’s important to remember that I would like to achieve a single numbers to compare different LDR unities, of course, the better information will be in a curves or, sure, in a real audition…. Any better than that….!

 

[CJ]

you could do a graph using 10 different light levels

use a rotary switch with resistors

if the cell has a higher dark resistance, then it will probably be "faster" than other cells,

so the compressor will have a quicker response time

i do not know what this does to the sound, maybe a supper fast cell will sound harsh,

maybe it will sound better,

you can buy other cell models that are close to the standard nsl, try them out.

you might come up with a better t4b, samples are usually free if you pay shipping

 

[Gus]

C.J.
What happened to vaccumbrain?

 

[CJ]

it got sucked into a vacuum, or a black hole, something to do with high speed neutrinos,  ;D

some company paid ipower big bucks for the domain name, and i could not match up,

then the company went under,

i bet rafa backed it all up, remeber the "takeover" ?
he had a ton off the old place,

good to see you Gus!!!

 

[miquel]

Thank you

Today… Static measurement.

I've been thinking how to get a single value that represents the compression depth for each cell NSL 5910.

My first idea was to create a light pattern that represent a specific compression, for example, -5 dB, and compare the value obtained for each cell. But the problem is that the value obtained at one point is not representative, the LA-2A has the Peak-reduction setting and internal adjustment (adjust stereo) and it was not representative of the compression action of each cell.

I have therefore set up two light pattern fixed:

1 .- lighting equivalent to a 10dB average compression
2 .- lighting equivalent to a 5 dB average compression

These light patterns were obtained with a potentiometer in series with the LED (the same dispositive of the previous experiments) and the reference is the average of several units LDR. Also I put a switch to easily switch between -10dB and -5dB hypothetical attenuation.

These patterns will be fixed and it will not matter what is their real absolute exact value, because the important to consider is the difference read between both of them. For example, if in the reading position-10dB the measured attenuation is -9.8 dB and in -5dB position is -4.9, the depth of real measured compression is 4.9 dB.

Conversely, if another unit shown -11.3 dB at -10dB position, and -5,8 at 5dB position, the depth of compression will be 5.5 dB.

Both cases are real; first is the LDR number 1 and the second is number 6.

Next I show the table with the results for my 6 NSL 5910, first is the LDR number, second the dB attenuation equivalent at 10dB hypothetical, third the dB attenuation at 5dB hypothetical, and forth the subtraction or equivalent depth compression, the important unity in this moment.

1    9.8  4.9  4.9
2    9.2  4.3  4.9
3    9.3  4.5  4.8
4  10.0  5.2  4.8
5  10.1  5.3  4.8
6  11.3  5.8  5.5

All unities are dB attenuation (therefore has a positive sign).

Conclusions

1.- The differences between units are not large.

2.- The unities 1 to 5 sows the same compression capacity (4,8 to 4,9) in relation to theoretical 5,0 dB, only the unity 6 shows a high value. (Compression larger).

3.- The observed differences in the temporal analysis (post 1) were much more important than this static measurement.

Miquel

 

[CJ]

you can use your computer as a data logger and plot the decay curve on a graph.

or use a storage scope and take a pic of the scope pattern with your digital camera

or put a 9 v battery in series with the cell, use a dropping resistor to scale a vu meter that will read +16 db at min resistance and -4 db for max resistance,

kill the light and watch the vu meter

every cell will produce a slightly different needle decay,

 

[PRR]

Yes, use an exact model of LA-2 network but drive it with 1V DC for easy measurement.  Drive LED to get specific interesting db reductions. 2db, 6db, 12db, 18db, and 24 and 30db for extreme sound-slam. Log the LED current needed for each reduction.

Because you will, in real use, NOT "apply 5mA LED current", you will adjust for "about 6db average gain reduction". You do not really care what the LED current is while you are using the limiter. You do need to note the LED current when selecting "similar" LED-LDR units.

Apply the "6db" current to the LED suddenly and measure the time it takes to fall 5db; that's approximately the Attack time for small reduction. Measure again for "20db" current to cause 17db reduction, the large reduction attack time.

Yes, the time with-light affects the release time. Will you run your limiter "easy" or "hard"? When I use limiting on live recordings it may be 20 minutes of no-limiting (dark) and 5 seconds of limiting (light). But when I smash sound for posting on the Web I might limit 95% of the track, lots of light-time.

I do NOT think the "dark" resistance matters. As you say, 2 Meg is "no" compression, and 200K is hardly audible. I would call "release" the time from 5db or 20db to 1db.

Personally, what I did with very mis-matched LDRs, I adjusted the fixed resistor in the network. In my case, 22K in left 19K in right gave a good match in the 1db to 3db reduction zone. This was good when attempting to record live (to tape) without limiting, but limiting lightly when needed, and without stereo-shift. This was not a perfect fix because the 20db limiting levels did not match and BIG drum solos shifted in the stereo image.

 

[CJ]

the best t4b i ever made used radio shack ldr's and an el panel out of an alarm clock.

unfortunately, they figured out that it is much cheaper to use a blue led instead of trying to generate megavolts from a pair of aa batts thru a crude chineese switcher,
so no more cheap el panels, that would probaly not last as long as the good stuff.

once you master LDR's you can customize your motion detector to give no false trip when some one is breaking into your weed garden, which shold be ready to harvest next week,  watch for green catipillars 

 

[shabtek]

ahh, good ole CJ


ever burn a florescent orange worm,
they look like seeds at first til they start squirming and break through the pod

 

[CJ]

you cool too.

 

[miquel]

The dynamic study, for example the release time, was exposed in the first post.

I said that I used a digital data logger, but the images were not included, I included only the results. Now include two images of release time datalogger, the unit's faster and the unity slower.

Release time, unity faster (675 ms from dark to -1,5 dB)

http://i749.photobucket.com/albums/xx138/miquel_bucket/03analisi675ms.jpg

Release time, unity slower (1750 ms from dark to -1,5 dB)

http://i749.photobucket.com/albums/xx138/miquel_bucket/05analisi1750ms.jpg

The datalogger shows the evolution of the attenuation from the moment just lighting tourn off (impulse noise ceases) until the gain recovers a certain value, in this case 1.5 dB below the value of no compression.

It should be remember, my intention is to achieve a single numbers, to characterize each unit LDR with a little numbers.

 

[PRR]

You will also want to know the Attack Time.

I see about 25mS for the "675" part. However in actual use you will not SLAM the LED as hard as your test does. Trim the LED current for 20db (0.1) attenuation, let it recover, then apply that much current and log the fall.

> my intention is to achieve a single numbers

There are two curves, Attack and Release. Neither one is an "ideal" curve; approximately exponential but really several exponentials (from the several layers in the LDR) superimposed.

One-Number works only a little better than using one number to grade the curves on a horse, car, or woman.

We usually use several numbers. My Molly, grading 0-10, her shoulders are a 10, her butt is an 8, her ears are 13 and her tail is a 0. (Molly is a Corgi dog, show-quality except her ears are silly-big).

I think you want Attack (say 10db), Small Release (3dB), and Long Release (12dB). Three numbers is not TOO many; easier than judging show-dogs.

 

[squib]

then of course you need to measure distortion! This is a factor that varies widely from cell to cell!