TwinLineAmp Opto Compressor TLA2A

[rmaier]

Thanks, Ian. Glad I asked!
Ralph

 

[ruffrecords]

I have just been looking through this thread and I noticed a number of diagrams have gone missing. I have been back though my own posts and restored the missing ones.

The reason I revisited this thread is because it looks like I am going to have to bite the bullet and do a 'proper' tube compressor because I really do need one for the Mark III mixer. The idea at the moment  is it would be a 6U high 35mm wide module (the same size as the channel amps) and it would probably be labelled a bus amp because it might actually include  a (tube) VE bus amplifier. It would basically be Twin Line Amp, one stage used as the VE bus amp and the other one as a post fader output amp. In a small mixer, the buses are the obvious place to add a compressor so this is the module it should go in. The output amp is the obvious place to put it but there are no spare tubes so this would mean a (shock horror!) semiconductor side chain. The alternative is to have the VE amp elsewhere and just use the TLA as suggested earlier in this thread but in a 7Hp wide 6U high module. A nice edgewise meter or a series of LEDs could be used to indicate gain reduction and of course stereo linking would be a must.

Cheers

Ian

 

[scott2000]

;D

 

[pvision]

Ian, just make the transistor sidechain and mount it on a valve base. No-one need ever know

Nick Froome

 

[ruffrecords]

Ian, just make the transistor sidechain and mount it on a valve base. No-one need ever know

Nick Froome

Not a bad idea. I cold give it a sexy name like  ...... T4B

Cheers

Ian

 

[ruffrecords]

One of the issues with a semiconductor op amp side chain is you cannot use the simple technique of driving up to 20V rms through a 1K resistor to drive the LEDs because you would ned +-30V power rails. But you can do things with op amps you cannot easily do with tubes. For example a voltage to current convertor is easy to make:

This example is aimed at instrumentation but you can just replace Rload with a pair of back to back LEDs in Vactrols for a compressor. The 250R resistor sets the gain at 4mA per volt. In some ways this is ideal for driving LEDs because the op amp takes out their forward voltage drop. This allows us to set the threshold in some other way more under our control, using a regular diode for example. So maybe a couple more op amps for a precision rectifier followed by the diode and a cap like any other compressor, This would give us some control over attack and decay times and possibly simplify stereo linking. Maybe a couple of NE5532s would do the job.

Cheers

Ian

 

[ruffrecords]

I am still undecided if to use a semiconductor side chain so in the meantime I have been looking at the logic of linking two modules together. assuming each channel uses back to back Vactrols so it responds to both halves of the waveform, I decided to go with Nick Froomes  idea of driving one of the LEDs from the other channel so both channels have one LED driven by itself and one driven by the other.

If it is all in one box then it is easy to use a single switch labelled link to achieve this but in the Mark III mixer these will be separate modules. So you need to devise a method of linking the two together so they can swap LEDs but also so that a link switch on either unit can initiate the link. The attached schematic uses a DPDT relay in each channel to achieve this. Here is the schematic I came up with:

L1, L2, L3 and L4 are the four LEDs in the Vactrols. With the relays de-energised, each channel drives its pair of LEDs normally. Operating either S1 or S2 operates both relays  via the dc link and switches over one pair of LEDS. A spare contact on each relay operates a 'LINKED' LED on the front panel.

Cheers

Ian

 

[ruffrecords]

I treated myself to a new bench power supply the other day. The first thing I used it for was to make some measurements of a VTL5C3 vactrol. I fed the LED from my bench power supply via a 1K resistor. I measured the voltage across the resistor to get the current, the power supply told me the voltage being applied and a multi meter across the resistive element gave the actual resistance. Here is the raw data I acquired:

Notice that nothing really happens until the voltage reaches 1.5V which is the forward voltage of the LED. You will see I also got the spreadsheet to calculate the log of the current and the log of the resistance because we know one is supposed to be proportional to the other. Here is the graph you get when you plot log10(resistance) versus
log10(current):

A fairly straight line as expected. The next thing to do is to consider the best way to use this in a compressor.

Cheers

Ian

 

[ruffrecords]

Next thing I did is look at the actual resistance range that can be achieved. We need to attach a couple of pots in parallel with the LDR and unles we use pots in the megohms range we are not going to be able to use the really high resistance values. So I assumed the total parallel resistance of the pots is 100K. Then I worked out the attenuation we could get using these values  as the bottom leg of an attenuator using 100K or 22K as the top leg. Here are the results:

https://docs.google.com/spreadsheets/d/e/2PACX-1vT4asoFJ15eyaR825cd7rDeuX7tJq1KrxSS_vTmEnh_Am3QCJBQL0JiWv4psJvo0EN1n_hsFE2Eu_OZ/pubchart?oid=566785469&format=image

The blue line is with 100K and the red is with 22K. The 100K line has an initial attenuation of 6dB compared to 1.78dB for the 22K version so I subtracted 4 dB from the 100K figures so the two graphs would start at the same point. This is eqivalent to adding 4dB make up gain to the 100K plot.

As you can see the 100K plot is steeper than the 22K plot. The 100K plot reaches 20dB attenuation for about 11dBu control voltage whereas the 22K version needs about 16dBu. So these represent different ratios. This partly explains how the LA2a compress/limit switch works in the LA2A.

Next I need to work out how these translate into actual ratios once you close the loop.

Cheers

Ian

 

[rmaier]

Hi Ian,

As it happens, I've been working again (after a break - life seems to interfere far too much with DIY) on putting a pair of these into a single box, so I'm reading your posts with great interest. Are you thinking something along the lines of a stepped ratio switch?

I'm still early on in my build, so it might be possible to try to integrate some of your recent ideas.

Ralph

 

[ruffrecords]

Hi Ian,

As it happens, I've been working again (after a break - life seems to interfere far too much with DIY) on putting a pair of these into a single box, so I'm reading your posts with great interest. Are you thinking something along the lines of a stepped ratio switch?

I'm still early on in my build, so it might be possible to try to integrate some of your recent ideas.

Ralph

At this stage I am trying to understand what determines ratio but certainly with a view to including a ratio control.

Cheers

Ian

 

[rmaier]

Cool!

 

[Rob Flinn]

One of the issues with a semiconductor op amp side chain is you cannot use the simple technique of driving up to 20V rms through a 1K resistor to drive the LEDs because you would ned +-30V power rails. But you can do things with op amps you cannot easily do with tubes. For example a voltage to current convertor is easy to make:

This example is aimed at instrumentation but you can just replace Rload with a pair of back to back LEDs in Vactrols for a compressor. The 250R resistor sets the gain at 4mA per volt. In some ways this is ideal for driving LEDs because the op amp takes out their forward voltage drop. This allows us to set the threshold in some other way more under out control, using a regular diode for example. So maybe a couple more op amps for a precision rectifier followed by the diode and a cap like any other compressor, This would give us some control over attack and decay times and possibly simplify stereo linking. Maybe a couple of NE5532s would do the job.

Cheers

Ian

If you need +30v power rails, just wondering if you could use the MCI swinging op amp circuit ?  See attached.

 

[ruffrecords]

I have been struggling to get my head around the simple but of little circuit they use in the LA2A to switch from compress to limit. This involves tapping of the side chain feed slightly before the LDR rather than directly from it. I am getting strange results which may ex-plain why I have seen the extra resistor 'labelled select on test'. So I went back to basic and work out the maths for the network  so I can calculate Vin and Vout from the known LDR resistance versus control voltage reading I have already takes. 

At this stage I just want someone to check my maths (see attached jpg). Using these formulae still gives strange results. In the meantime I am going to try different resistor values to see what happens.

Cheers

Ian

 

[ruffrecords]

I have checked and rechecked the maths and it seems OK to me so I went ahead and tried it out some more in the spreadsheet. Turns out I was getting weird results because splitting the 100K into 78K and 22K was worng. You only need a few K, like 95K and 5K to get a very noticeable effect.  I worked out how to plot the input voltage versus the output voltage and plotted the results for various resistor combinations.

The 100K on its own (the normal configuration) gives a ratio of approximately 3:1.

97K/3K gives a ratio of about 4:1
96K/4K gives a ratio of about 5:1
95K/5K gives a ratio of about 7:1
94K/6K gives a ratio of about 8:1

Beyond that things start getting weird - by which I mean the output starts to decrease with increasing output.

I am not sure what is the mechanism that accounts for this behaviour. It may be a little feed-forward or it may be a little positive feedback. Whatever it is, it is very sensitive. With care it should allow us to come up with a design that allows a couple of rations higher than 3;1 to be included.

We already know that chainging the 100K series resistor to 22K also changes the ratio. I will next try to characterise that in the hope we may be able to get  some lower ratios as well.

Cheers

Ian

 

[ruffrecords]

I have carried out some more spreadsheet based  assessments using the data from the tests I ran on a VTL5C3.  I discovered that using a 22K series resistor instead of a 100K one gives a gentler compression slope of about 2:1. Also, by using the technique of tapping off the control voltage higher up the series resistor, I could could achieve higher ratios just like with the 100K. However, with 22K the results were a lot more stable. By varying the tap from 0 to 4K you can achieve compression ratios from about 2:1 to 8:1. So the next step is to knock up a circuit and try it. Here's what I drew in my notebook:

The 2K2 pot acts as a ratio control. No idea just yet what kind of law it needs but it is easier to prototype than a stepped switch. The rest is very similar to the original TLA2a design. (The 1200 resistor for the gain reduction meter should be 200 ohms).

Cheers

Ian

 

[rmaier]

Thanks, Ian, this is interesting. Yesterday afternoon, I finished a 3-month kitchen and dining room reno (happy wife, happy life!), and my mind has turned to more enjoyable pursuits like building compressors. I'm curious as to a few of the changes in resistor values, and wonder if you'd mind clarifying: why the change in values for the threshold and gain pots from 10K to 100K? I already made up a couple of 10K stepped switches, so I'm wondering if they're still usable.

Ralph

 

[ruffrecords]

Hi Ralph,

I have been giving some considerable thought to the alternative ways of arranging the various parts into a compressor. So I have backtracked this thread to the point you started talking about building one yourself.

Your first post was here:  https://groupdiy.com/index.php?topic=52847.msg870846#msg870846

Where you asked if the posted schematic was the best one to use and I said it was a good starting point. However, I have since carried out a lot of investigation into the Vactrol  gain reduction characteristics and in particular how you can implement a ratio control. This does mean returning to the original topology as illustrated in the very first post:

https://groupdiy.com/index.php?topic=52847.msg674382#msg674382

Unfortunately, this does require higher impedance pots in order not to unduly load the Vactrol so I am afraid the 10K pots do have to become 100K. I know you are using stepped pots so this means basically replacing every resistor with one ten times its value. Sorry.

With the incorporation of a ratio control the most up to date schematic is the one from  two posts ago. At the moment, since 2K2 pots seem to be rather scarce, I am using a 10K pot with a 4K3 resistor in parallel to make a 3K resistor. This, combined with an 8K2 series resistor should give a ratio range from 2:1 to 8:1 but I do not yet know the law required. As you suggested earlier it would probably be better to use a simple stepped pot for this - 6 positions would be plenty.

I have built most of the little breadboard circuit I posted and tested it out by driving it with 1KHz from a generator. The initial test was intended just to calibrate the attenuation in terms of side chain level in dBu. I will post the details another time but overall the threshold was at +3dBu and there seems to be a fairly gentle knee and the ultimate slope of the gain reduction cell is close to 1dB per dB. There was close to 13dB of attenuation at +19dBu input to the LED circuit. For the usual 20dB maximum gain reduction expected of a compressor we will need to feed the LED circuit a level 7dB higher i.e. +26dBu but this is not a problem for the TLA which can drive at least +30dBu into a 2K4 load. Presumably this would read about +5.5VU on the meter. This will have to wait until I build it into a TLA.

As part of this  test I attached a low cost vertical VU meter across the 200 ohm resistor. For 3.5dB gain reduction it read -15VU, for 8.8dB gain reduction it read -6.5VU and for 13dB of gain reduction it read -1.5VU.
It looks like the scale would be confined mostly to the upper half of a regular VU meter, perhaps due to the fact the it has its own built in threshold. May an LED bargraph GR meter would be better.

Cheers

Ian

 

[rmaier]

Hi Ian,

Thanks for the upadate - greatly appreciated, as always. I had a feeling I'd be changing out those pots! No worries, it's done easily enough. I'm looking forward to seeing what you find regarding the ratio pot law and metering.

Ralph

 

[ruffrecords]

Attached is a pdf of the results of the first tests of the circuit described above. Here is a neater version of the circuit:

The component values reflect the ones actually used. The Vactrols are VTL5C3.

Cheers

Ian