TwinLineAmp Opto Compressor TLA2A

[rmaier]

Hi Ian,

I've been working on my build a bit more over the past little while. I've added another output on my external PSU that I built for my Classic preamps (have I said they're my most prized piece of gear  ?) to accommodate the compressor. I've also done some of the internal wiring and revisited my breadboard layout to see if it will work with the latest revisions. Additionally, I've ordered a pair of inexpensive LED meter kits, in case the VU meters I already have are unsuitable.

I noticed that your most recent schematic omits the 10K VU meter series trimmer... I'm assuming it's still an option?

Also, I'm at the point in my build where I'm ready to order a few parts for the stepped pots. Do you have any news on the law for the ratio pot? 

Thanks!

Ralph

 

[ruffrecords]

Hi Ralph, The law for the ratio is still a bit of a mystery. Try as I might I cannot get my head around the math although the spreadsheet does appear to work.

So I decided to build it and measure rather than speculate and simulate. Coincidentally I finished the first set of measurements a few hours ago. The circuit I am using is very similar to the one I posted in reply #99. The two 100K pots have been replaced by a 47K resistor across the two opto resistors and the signals are fed direct to the gain make up and  side chain amplifiers. (two halves of a TLA). The gain make up amp has a 10K gain resistor fitted so its gain is about 10dB. The side chain amp has a 4K3 gain resistor fitted so its gain is about 20dB. I measured the output for an input level from -20dBu to +17dBu in 1dB steps and noted the output level with the ratio pot fully clockwise (high ratio) and fully anti clockwise (low  ratio). The results are in the graph below:

With a -20dBu input the output is about -12dBu so there is a gain of 8dB. 10dB of this is in the TLA and there is a loss in the gain cell of about 2dBdue to the 11K series resistor and the 47K that emulates the two pots. The threshold appears to be at an input level of about -15dBu from which point both outputs curve gently. The lo ratio curve (blue) is asymptotic to a ratio of 2:1 (it has a long straight bit). The high ratio curve I expected  to reach 8 or 10:1 but its ratio gradually increases. It seems to be close to 3:1 for the first half climbing to about 5:1 for the second half. I suspect this is a direct effect of the method of altering ratio. I have used a linear pot so I will try it at the mid position to see what ratio that gives.

The technique clearly allows ratio to be adjusted but the law needed of the pot is not yet known. The actual input/output curve and ratio achieved does not agree with the simulations and is lower than expected. I need to think about this some more but my inclination is to try a couple of other tapping points to try to characterise it.

Cheers

Ian

 

[shabtek]

Ian, what is the signal you used? Steady sine ?

 

[ruffrecords]

Ian, what is the signal you used? Steady sine ?

Yes 1KHz sine.

Cheers

Ian

 

[ruffrecords]

I have carried out another test on the basic opto compressor circuit. Rather than just ry a different resistor setting for the ratio control I though I would check the extreme just so we know what kind of range of ratios is possible. So I connected the side chain amp directly to the input so it effectively acts as a feed forward compressor. Since we already know the law of the gain cell is close to 1dB gain reduction for every 1dB level increase above the threshold, this should in theory give a near infinite ratio limiter. In practice the gain cell has some variation throughout its range so we don't quite get this but it turns out to be pretty close. Here is the graph of input versus output:

As you can see, the gain changes rapidly above and input of -20dBu and the output is essentially unchanging from an input level of -10dBu right the way up to an input of +10dBu. Above this there is less gain reduction. I suspect both the gain cell is running out of available attenuation (it is already at 25dB) and the side chain amplifier with an input of +17dBu and a gain of 20dB is struggling to output the required +37dBu ( about 55V rms !!!).

So it seems perfectly stable even in feed forward mode so I propose to tweak the design to allow the side chain tapping point to roam over most of the range of the input resistor. So I plan to change the input resistor from 8K2 to 1K and change the 10K with 4K3 in parallel with it to just a 10K pot. So we will still have the basic 11K resistor at the top of the gain cell but the threshold tap will be able to roam over about 90% of the range. Fortunately this is an easy mod to do so I will try it and report back.

Cheers

Ian

 

[ruffrecords]

I have repeated the tests with the 1K input resistor followed by the 10K side chain pot as the top arm of the gain cell. With the pot at the 1K resistor end the curve is almost identical to the limit curve I got last time. This looks still to be mostly feed forward as evidenced by the dip in output level during the limit part of the curve. With the pot turned so the side chain take off is at the signal output gives the same 2:1 curve we got before. So far nothing new. So, at the maximum input point I adjusted the ratio pot for 0dB output to put the ratio somewhere between limit and 2:1 ratio. I still do not have a handle on how ratio changes with pot position so this was just a stab in the dark. Here are the results:

I have tweaked the input scale so that all the lines go through the same origin at -10., -10 which make it easirer to compare them with 1:1. The green line represents 1:1, the red line the 2:1 ratio and the blue is the near feed forward still with the dip in the output and running out of gain reduction or headroom at the highest input levels. The yellow line is the new shot in the dark curve which is remarkably asymptotic to a ratio of 10:1 and looks to be very well controlled.. So I quickly turned off and measured the actual resistance values with the pot in this position. The values were as shown in the graph; 7K2 from input to tap off point and 4K7 from tap off point to output. The total resistance is higher than the previous 11K2 because the pot on its own is actually over 10K.

We know from prior test that 8K2 and 3K gives a ratio approaching 5:1 so it looks like we have three points on the graph describing ratio versus pot position:

A graph with three points on it is not very convincing unless they happened to be in a straight line which they are not, Given that all of the curves have a fairly gentle knee such that the ratio gradually climbs to the final value, there does not seem to be a lot of point having lots of selectable ratios. For the moment I am happy to have 10:1, 5:1 and 2:1.

Cheers

Ian

 

[rmaier]

Thanks, Ian, this is great. I'm just trying to visualize how  to implement this on a rotary switch.
Ralph

 

[ruffrecords]

Thanks, Ian, this is great. I'm just trying to visualize how  to implement this on a rotary switch.
Ralph

I will draw a schematic and post it.

Here it is:

Cheers

Ian

 

[rmaier]

It's been a while, but I've  built my compressor pair and they pass audio. How should I go about calibrating gain levels (GAIN 1 and GAIN 2) on the boards?

Thanks!

Ralph

 

[ruffrecords]

It's been a while, but I've  built my compressor pair and they pass audio. How should I go about calibrating gain levels (GAIN 1 and GAIN 2) on the boards?

Thanks!

Ralph

For the gain make up you do not really need much more gain than the maximum gain reduction you expect to use. 20dB is a sensible value for the maximum gain reduction. When the compressor is not operating there is about 2dB loss via the input resistors and the the two pots so you need to add this on. Also if you are using a 2:1 output transformer you need to add another 6dB to offset the 6dB step down. That makes 28dB on my calculator. Add a couple for luck and call it 30dB.

For the side chain gain it depends on how low you want the threshold to be. The knee is not far away from 0dBu out of the side chain. so the lowest threshold will be 0dBu minus the side chain amplifier gain. Let's try 30dB for the side chain gain. With the threshold set to -30dBu a signal that peaks at 0dBu will be compressed between -30dBu and 0dBu. If the ratio is set to 2:1 there should be 15dB of gain reduction and the output will be -15dBu. So we set the make up to 15dB to get the peak back to 0dBu. Hence the -30dBu to 0dBu range has been compressed into -15dBu to 0dBu which seems about right. For higher ratios we would normally need a higher threshold so less gain in the side chain. So 30dB for the side chain gain seems like a good starting point.

I notice in the earlier diagram the threshold pot has a resistor in its grounded leg. This means compression is never completely turned off. It might be best to short this initially so you know you can always turn compression off.

To set the gain make up gain I would suggest turning the threshold off so there is no compression and set the gain make up to minimum. Then feed in a -20dBu signal and set the gain make stage gain to give 0dBu out of the transformer.

Cheers

Ian

 

[ruffrecords]

I know it has been a while since this thread has been active but I still get asked questions about tube based opto compressors. So I have been through this entire thread and extracted all the information I consider to be releveant. I have gathered it all into a single pdf file that hopefully gives anyone all the info they need to build themselves a tube based mono or stereo optical compressor.

Cheers

Ian

 

[scott2000]

Wow!
Thanks for this!

Looks like your low on the TLA boards... Maybe you could run a few TLA2A? ;D

 

[Heikki]

I wish ruffrecords would make the sacrifice and write a book about designing pro audio tube equipment. There are 17 million books about tube guitar amps, few on tube hifi amps and zero tube pro audio books.

 

[ruffrecords]

Wow!
Thanks for this!

Looks like your low on the TLA boards... Maybe you could run a few TLA2A? ;D

Actually I have just one left. I just tweaked the board layout a bit. Unfortunately,  the new fast spreading Covid strain here in the UK, my age and my underlying health conditions means  I have been told to shield. So I can only go out to go to the doctors or a for a bit of exercise. Otherwise I am confined to barracks. So no more trips to the post office for me for a couple months. With that in mind I may well open source the twin line amp pcb so people can have their own made.

Cheers

Ian

 

[Holger]

I know it has been a while since this thread has been active but I still get asked questions about tube based opto compressors. So I have been through this entire thread and extracted all the information I consider to be releveant. I have gathered it all into a single pdf file that hopefully gives anyone all the info they need to build themselves a tube based mono or stereo optical compressor.

Cheers

Ian

Very interesting paper, thanks Ian.

 

[ruffrecords]

I wish ruffrecords would make the sacrifice and write a book about designing pro audio tube equipment. There are 17 million books about tube guitar amps, few on tube hifi amps and zero tube pro audio books.

Hmmm. The thought has crossed my mind but I suspect it is an even bigger job than building a mixer.

Cheers

Ian

 

[rock soderstrom]

I know it has been a while since this thread has been active but I still get asked questions about tube based opto compressors. So I have been through this entire thread and extracted all the information I consider to be releveant. I have gathered it all into a single pdf file that hopefully gives anyone all the info they need to build themselves a tube based mono or stereo optical compressor.

Cheers

Ian

Interesting document with a lot of information. Thanks for your effort!

 

[vitopower]

Thanks, Ian!  Really looking forward to this one.

 

[L´Andratté]

Ruffrecords, thanks a lot for publishing!
All your write-ups are printed out in folder in my workshop.
Excellent stuff, thoughtful, informed and no rambling whatsoever !

PS I always listened the vtl5c1 as MUCH faster than vtl5c3 (as both specs state similar time constants)?!?
the vtl5c10 is not well known, but I have used it with success! It´s in the Demeter Compulator e.g....

 

[ruffrecords]

PS I always listened the vtl5c1 as MUCH faster than vtl5c3 (as both specs state similar time constants)?!?
the vtl5c10 is not well known, but I have used it with success! It´s in the Demeter Compulator e.g....

I think there is still  fair bit of work to do in selecting the right Vactrol device. I have been corresponding for quite a while with on DIYer who has now pretty much built the compressor described in the document - with one exception - he made his own Vactrols. He used an NSL 19M51 light dependent resistor which has a peak spectral response at 550nm which is somewhere between green and yellow. This is close to the spectral response of Perkin Elmer type 3 material which is used in the VTL5C3. He placed the NSL along with a green LED in some heat shrink tubing and cooked it. He was much happier with the sound, especially at high signal levels.

Cheers

Ian