Official Forssell Limiter Build thread

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Does anyone 'get' the ratio control on this circuit? I see how it works by padding down how hard you drive the LED, but here's my quandry:

-The harder you drive the LED, the more gain reduction you get,
-The higher the series resistance (R28), you get a lower slope
-If the ratio control is wired as drawn on the schematic, when the pot is all the way left (wiper on 1), you get the lowest value of resistance, and should get the highest slope, conversely when the wiper is at the other end (5k) it should be the most gentle slope.

Is the ratio control drawn in backwards? Do we really want a reverse log pot there, or do we want more of a log pot wired backwards?

Just curious because I'm having some problems with my ratio switch right now.

Cheers,

Kris
 
I just checked it on my scope (in XY mode), and the ratio pot is drawn in backwards on the schematic (or the pot just works backwards). Lower resistances in the ratio control result in higher compression ratios, and higher resistances prevent current from getting to the LED thereby lowering compression ratios.

I'm trying to come up with a chart with some resistor values for the ratio control versus compression ratio. Unfortunately, the scope that I'm working on doesn't display in dB, so it's a somewhat laborious process (well, not so hard for the low ratios, but telling the difference between a 4:1, 6:1, and 10:1 slope on a linear display is pretty difficult, as they all look like flat lines).

I'm also now having a bit of trouble implementing the makeup gain control. The output amp is configured (as per the schematic) with a gain of 11 (a little over 20dB), and I want to put an attenuator just after the opto element to pad the signal down to compensate for the output stage gain. The problem I'm now having is that it seems that there may be some interraction between the attenuator and the feedback path into the sidechain of the compressor. I have tried wiring up a 24 dB attenuator (100 k total resistance) wired as a pot between the non inverting input of the amp and ground, but haven't had much success with this approach. I'll try just switching a resistor shunted to ground just aft of the LDR...maybe that will work better, without the series reistance that my faked 'pot' approach was adding.

Stay tuned....

Oh, and if anyone has any tips for getting a scope to display log, or to measure compression ratios, plz lemme know!

Cheers,

Kris
 
I don't think a 5k log pot backwards will work too well...the taper needs to to give you more resolution on the pot at the higher resistances rather than the lower ones.

Really, what will work best is a rotary switch with some hand chosen resistors. I'm working on selecting those resistors today...I hope to write something about by this afternoon.

Cheers,

Kris
 
hmm maybe i am confused then because i thought a rev log was just a log pot with the resistance 180deg from the normal log, meaning that it was "backwards". many state CW or CCW for log and rev log.

so does a rev log mean that the Log slope is 90deg from normal and not 180deg?

:shock:
 
A reverse log pot has a slightly different taper than a log pot wired in revers....yeah, it's alot closer to reverse log than a linear pot is, but it's still not the same.

Check out http://www.geofex.com/Article_Folders/potsecrets/potscret.htm

for some plots of the various tapers, and a way to approximate a rev log pot.

I honestly believe that for this compressor that a rotary switch is the best way to go. It'll keep the ratio more repeatable, and give a good range. Currently I'm trying to set up the proper resistances to obtain the following ratios:

1.2:1, 1.5:1, 2:1, 3:1, 4:1, 6:1, 8:1, 10:1

I figure that should be more than adequate for everything from subtle compression to heavy limiting (there's no infinity, but the slope of the opto element gets steeper with increasing current drive anyways, so the slopes won't be perfectly linear anyways.

When I said that the schematic is backwards, what I mean is that the low ratios will occur for a high value of resistance, and high ratios for a low value. If you want the ratio pot to go from low to high while turning clockwise you need make sure that you get 5k on the wiper at the full CCW position.



Cheers,

Kris
 
Okay, I've done some testing (really quite laborious)....

I hooked up volt meters to my signal generator (kinda noisy, but still works), and to my output. I replaced the resistor to ground in the feedback loop of the output op-amp with a 100k resistor in an attempt to get the gain down. However, there was still 2dB gain coming through the unit. Nevertheless that was enough to keep my test signals from saturating the output.

I also hooked up a scope in X-Y mode (x on the input, y on the output), and turned 'persist' on the waveforms. This let me dial in a consistant threshold value as I could easily see the compressor 'knee' in the XY traces.

With the volt meters set to read in dB I stepped up the amplitude of my test signal (1k sine wave, comp attack and release set to as fast as they could go, and various test ratio resistors just resting in the PCB). I recorded the input and output dBV readings to an excel spreadsheet.

The curves are very smooth...by this I mean the knee of the compressor is very gentle. This is a nice trait for an opto compressor (for me its one of those 'defining' traits for optos), but it makes calculating the ratio somewhat of a challenge. I brought the data into Matlab, and did linear regression on all points above the threshold to obtain the ratios.

Here's the results:

Ratio resistance value -> Ratio
10.0k -> 1.97:1
6.19k -> 2.2:1
4.02k -> 2.5:1
3k -> 2.75:1 (not sure of the exact value I used here, but it's close to 3k)
2.59k -> 3:1
1k -> 5:1

My best guess is that 4:1 lies at around 1.8k or so....


I'll continue testing another time....I need to get more results in the sub 1k range.

I was a little surprised though that the comp couldn't do a ratio lower than 2:1.

Cheers,

Kris
 
wow that is great. I am eagerly awaiting the conclusion of your findings as i populate my boards..

I am also thinking about the rev log pot.. Although "perfect" would be nice, I'm looking more for usefulness. Would using a reverse wired log pot really screw the usability up that badly?
 
Well, after having done my experiments, I think that a log pot in reverse might work out quite well after all, it will give you about a 1/4 turn to get from 2:1 to 4:1, and the rest will get you from 4:1 to around infinity:1.

I don't think that the ratio control terribly repeatable as a result of the properties of the vactrol itself. In my tests its been a little different from one application of power to the next. I don't know if there's a burn in issue, or if this is just memory effects in the vactrol (looking at the curves for the VTL5C1 you can see a pretty big difference between a '24 hour dark' opto, and one thats been driven with some current).

I hope to have the ratio control sorted out within the next couple of days.

Cheers,

Kris
 
I wouldn't waste the time and money to find some suitable revlog pot. A 6 or 12 position Lorlin switch like in the Pultec or GSSL would cover all the values you need. They are also cheap, repeatable (in case of mastering application), and you can find them everywhere.

chrissugar
 
More ratio resitance measurements:

825 Ohms = 6.4:1
619 Ohms = 9.24:1 (with an odd kink at the end, where at around 20dB of reduction or so it can't hold the high ratio anymore and turns into about 2:1)
215 Ohms and 100 Ohms resulted in negative compression ratios, where the output gets quieter as the input increases. 100 ohms gave -8.6:1 and 215 ohms gave -23:1

All these tests were with a 100 ohm series resistor (R28 on the schematic I think).

I think it's best to stick with a minimum resistance of around 500 ohms to avoid the negative ratios.

I hope to get a few more test points in, then I'll plot up a graph of the results. Maybe the DIY factory can host it....

Cheers,

Kris
 
Okay the final set of resistance/ratios are in:

1k78 = 3.8:1
1k33 = 4.3:1
750 = 10:1 (I know, I got the same value for 619 ohms too...but this time no wiggle at the end)

So, the compiled list, with a 100 ohm series resistor (plus the adjust put set to around 250 ohm) is:

10K = 1.97:1
6k19 = 2.2:1
4k02 = 2.5:1
3k01 = 2.75:1
2k59 = 3:1
1k78 = 3.8:1
1k33 = 4.3:1
1k = 5:1
825 = 6.4:1
750 = 10:1
619 = 9.2:1

Plot these up and you'll see the logaritmic trend.

Hope this helps everyone on this project.

cheers,

Kris

<edit ....fixed a typo>
 
Just when I thought I had a handle on this comp, I think I went and learned something.

I had been puzzling over why I had a 2.2dB gain on my build (with the discrete input and output...JEFET992s). I replaced R11 (was 1k) with a 100k resisor, which according to my back of the envelope calculations should have given me a gain of 1.1, or +0.8 dB. So, why was I 1.4 dB over that?

Well, I think I may be able to explain it....please someone let me know if I'm wrong in my explanation. When R11 is 1k, it acts to pad down the output of the servo circuit, and set the gain for the output stage. When I swapped this resistor out for a 100k resistor, now there is actually a lower impedance path back through R12 (49.9k) into the servo amp IC. This acts to increase the overall gain.

Basically, the formula for noninverting op amp gain is 1 + Rf/Rg, which in our schematic is 1+ R7/R11. But, with R12 being small relative to R11, you effectively need to include it's effect in the calculation of Rg. So, considering R11 and R12 in parallel gives Rg = 33k. Going back to the op-amp gain formula this gives 1+10/33 = 1.3 = 2.27 dB gain.

I think that explains the gain that I was seeing. Also this brings up a possible issue with makeup gain control. If you do gain control by varying R11, you are changing the authority of the servo stage, as well as the gain of the output amp (and changing the bandwidth)...lots of interraction going on there. At low gains, the servo will have a high authority, and could generate more DC errors due to the servo IC's own internal offset.

I think it's best to use a fixed gain (ie a fixed value of R11), and attenuate prior to the output stage. I'm trying to figure out how best to do this. My gut tells me that putting a 100k log pot in place of R6 and taking the wiper to the non inverting input of the opamp is the way to go, but I haven't tried it out yet. More on this in a few days....

Cheers,

Kris
 
Okay, just got the makeup gain working. I made it variable in approximately 2dB steps from -2 to +20 dB using a 12 position Graywill switch.

What I did was remove R6 (100 k shunt to ground just after the opto), and replace it with a 12 position switch wired as follows:

Pos Resistance Gain (at output)
12-11 21.5k
11 -10 16.91k
10-9 13.3k
9-8 10.5k
8-7 8.25k
7-6 6.49k
6-5 4.57k
5-4 3.87k
4-3 3.01k
3-2 2.37k
2-1 1.87k
1-GnD 7.8k

The 21k resistor at the top of the switch is connected to the trace connecting to the non inverting input of the output opamp.

This gives the following gains (measured at the output relative to a 0dBV input)

1 = -1.82dB
2 = 0.02 dB
3 = 1.92 dB
4 = 3.84 dB
5 = 5.81 dB
6 = 7.79 dB
7 = 9.88 dB
8 = 11.98 dB
9 = 14.07 dB
10 = 16.15 dB
11 = 18.23 dB
12 = 20.31 dB

The switch presents an approximately 100k impedance to the opto network preceding it, basically it mimics a pot with a minimum value.

Hope this helps someone (other than just me).

Cheers,

Kris
 
Back
Top