etheory - a bunch of projects in progress

[ruffrecords]

I agree, it is good to thrash these things out in a frank but friendly way.

OK, I think we are converging. Notice I said if you add 10dB of gain before the knee the threshold is reduced by 10dB but the ratio is unaffected. If you look at the THAT app note out posted you will see that the ratio control adjusts the gain after the knee. If you look at this THAT app note:

http://www.thatcorp.com/datashts/dn00A.pdf

you will see exactly the same thing. And in this one too:

http://www.thatcorp.com/datashts/dn113.pdf

The point I am making is that gain before the knee, only changes threshold. Gain after the knee only changes ratio.

Cheers

Ian

 

[etheory]

The point I am making is that gain before the knee, only changes threshold. Gain after the knee only changes ratio.

Cheers

Ian

Here is an interactive Desmos on-line spreadsheet I made to illustrate the point I am making:
https://www.desmos.com/calculator/pp7bio0adr

This is a full simulation of a compressor, using an exponential VCA, a simple soft-knee computation, and simulates feed-forward and feed-back compression using the knee before and after the gain in the side-chain (labelled R).

If you play with the R value you'll see the ratio change as the side-chain gain changes.
The effect is different if the threshold is before or after the side-chain gain, but in both cases the ratio is clearly effected directly by the side-chain gain.

The math here is quite simple, y = x * V, where V is whatever the vca gain is. This is either a function of the output (or y, for feedback) or the input (or x, for feed forward).

cheers,
L

Edit - a modified version with a logarithmic envelope detector - which gives a slightly different response: https://www.desmos.com/calculator/b1myyas4lu

 

[ruffrecords]

I an afraid that apeadsheet makes ;itt;e sense to me. I am sure there is a falw in your nath somewhere. I will put together my own spreadsheet to convince myself one way or the other and post it her. In the meantime ley me ask you this. How would you implement a threshold control that does not alter ratio?

Cheers

Ian

 

[etheory]

In the meantime ley me ask you this. How would you implement a threshold control that does not alter ratio?

I think this question is best asked in the context of an actual schematic.
If you give me a specific example I can create a Desmos spreadsheet to show the curve of a solution that could be come up with.
In most cases, you just want to DC-offset the signal down into the knee device, be that a diode, VCA, tube, or something else.

In the 1176 this is achieved by offsetting the control signal into the rectifying diodes, in a THAT Corp comp, by offsetting using an input offset current to an inverting virtual earth opamp amplifier, or, in the case of something else, you might compensate the offset manually in a stepped fashion for each different ratio option (like the 1176 does).

Having said all that, as you can see by the Desmos graph, if the ratio comes after the threshold, it all gets a lot easier to manage.

 

[ruffrecords]

In the meantime ley me ask you this. How would you implement a threshold control that does not alter ratio?

I think this question is best asked in the context of an actual schematic.
If you give me a specific example I can create a Desmos spreadsheet to show the curve of a solution that could be come up with.
In most cases, you just want to DC-offset the signal down into the knee device, be that a diode, VCA, tube, or something else.

I think that would be instructive. I choose the LA2A.

Cheers

Ian

 

[etheory]

OK, so....

Based on this schematic: http://www.recproaudio.com/diy_pro_audio/diy_files/la2a/la2a_1968.jpg

The threshold and knee functions and fixed based on the distortion components of the tubes, the luminescent-panel nonlinearities and a bit the most-likely exponential response curve of the opto-resistors with regards to resistance.

Based on the circuit as presented, the limit/comp control is instructive. My guess without spending time simulating this (which would be harder than for the average circuit since I'd need a decent model for the opto-resistors and luminescent panel, but I'd be happy for someone to help me with one) would be as follows. People, please disagree with me when I go wrong, but....

1.) The threshold point could be changed independently of the ratio, by inserting a diode from V4, after C11, to the luminescent panel. That diode would form a new knee. By driving that point below ground, you could cut off the signal. Adjusting the DC bias here would change the knee, independently of ratio BUT, that's kinda cheating. So, you could simply DC-offset the signal going into the luminescent panel down. Without knowing the precise properties of those panels I assume that driving them backwards (or negative) makes them behave somewhat like a diode?

2.) Changing the ratio without changing threshold. I think you could change the ratio by simply adjusting R7. But since the knee is applied after the ratio, your threshold will move, as per my Desmos sheet. So to compensate, you'd need to use the above DC-offset trick to keep the knee constant for a set of chosen values of ratio using a 2-pole rotary switch, or similar (the 1176 uses an identical trick).

The main issue I can see with fine-tuning the LA-2A is the fact that the overall function of the threshold is going to be a very complicated shape of many different non-linear elements. But somehow finding the curve-response of those elements and DC-shifting the signal through that portion of the circuit down towards GND should do the trick.

 

[opacheco]

OK, so....

Based on this schematic: http://www.recproaudio.com/diy_pro_audio/diy_files/la2a/la2a_1968.jpg

The threshold and knee functions and fixed based on the distortion components of the tubes, the luminescent-panel nonlinearities and a bit the most-likely exponential response curve of the opto-resistors with regards to resistance.

Based on the circuit as presented, the limit/comp control is instructive. My guess without spending time simulating this (which would be harder than for the average circuit since I'd need a decent model for the opto-resistors and luminescent panel, but I'd be happy for someone to help me with one) would be as follows. People, please disagree with me when I go wrong, but....

1.) The threshold point could be changed independently of the ratio, by inserting a diode from V4, after C11, to the luminescent panel. That diode would form a new knee. By driving that point below ground, you could cut off the signal. Adjusting the DC bias here would change the knee, independently of ratio BUT, that's kinda cheating. So, you could simply DC-offset the signal going into the luminescent panel down. Without knowing the precise properties of those panels I assume that driving them backwards (or negative) makes them behave somewhat like a diode?

2.) Changing the ratio without changing threshold. I think you could change the ratio by simply adjusting R7. But since the knee is applied after the ratio, your threshold will move, as per my Desmos sheet. So to compensate, you'd need to use the above DC-offset trick to keep the knee constant for a set of chosen values of ratio using a 2-pole rotary switch, or similar (the 1176 uses an identical trick).

The main issue I can see with fine-tuning the LA-2A is the fact that the overall function of the threshold is going to be a very complicated shape of many different non-linear elements. But somehow finding the curve-response of those elements and DC-shifting the signal through that portion of the circuit down towards GND should do the trick.

Dear eTheory,

Very interesting analysis!.....The idea to change the Threshold independently of the ratio by using a Diode (DC biased Diode)  in the Side-Chain output path, is very similar to THAT use with its designs inserting a Diode (Biased Half Wave Rectifier thru an Opamp) with  a pot threshold control (variable DC Bias for the Diode) "before" the Ratio amp!!....in this LA-2A case your proposed Diode is installed "after" of the Side-Chain preamp ......similar principle!!

This same principle is I want to apply in the U273 in order to have a variable Threshold control pot using the D2 Diode with add DC Biased (positively)......Did you check my schematic for this mod???  my doubt is if this DC Bias flow through of Side-Chain Preamp Transformer winding???......

But the adjust o R7 (or R2???) in order to change the ratio without changing threshold point, is like to change the gain of the Ratio Relation (http://www.thatcorp.com/datashts/dn01A.pdf)  in a Feedforward Compressor described according to THAT note; that's changing the G gain.......Here 0<G<1 for compressor processing, Am I rigth?.

Very illustrative and clear your analysis, I have learn so much in this days about!!

Opacheco

 

[ruffrecords]

OK, so....

Based on this schematic: http://www.recproaudio.com/diy_pro_audio/diy_files/la2a/la2a_1968.jpg

The threshold and knee functions and fixed based on the distortion components of the tubes, the luminescent-panel nonlinearities and a bit the most-likely exponential response curve of the opto-resistors with regards to resistance.

What about the effect of the peak reduction control? It is clearly in the side chain and affects the threshold but does it also affect the ratio?

Cheers

Ian

 

[opacheco]

OK, so....

Based on this schematic: http://www.recproaudio.com/diy_pro_audio/diy_files/la2a/la2a_1968.jpg

The threshold and knee functions and fixed based on the distortion components of the tubes, the luminescent-panel nonlinearities and a bit the most-likely exponential response curve of the opto-resistors with regards to resistance.

What about the effect of the peak reduction control? It is clearly in the side chain and affects the threshold but does it also affect the ratio?

Cheers

Ian

Ian and eTheroy,

In my humble opinion this kind of processors have 3 components, these are:

1- A detector circuit (in this case the signal (coming from the R6/R7 input stage)
2- A discriminator (or threshold control) of the Control Voltage (it can or can't exist in some processors like this)
3- A voltage controllable device with gain stage (or not !) this is THE RATIO Network! (this is the side-chain preamp in the LA-2A)

In this LA-2A case the discriminator doesn't exist and its lack allow any change in RATIO Network induce (like side effect)  a Threshold change due to gain change itself in the voltage controllable network (more voltage for the same input signal level).....More gain in the RATIO = More RATIO + Threshold change due to the discriminator of voltage control level doesn't exist.

Correct to me if I'm wrong please!!

Opacheco.

 

[opacheco]

OK, so....

Based on this schematic: http://www.recproaudio.com/diy_pro_audio/diy_files/la2a/la2a_1968.jpg

The threshold and knee functions and fixed based on the distortion components of the tubes, the luminescent-panel nonlinearities and a bit the most-likely exponential response curve of the opto-resistors with regards to resistance.

What about the effect of the peak reduction control? It is clearly in the side chain and affects the threshold but does it also affect the ratio?

Cheers

Ian

Ian and eTheroy,

In my humble opinion this kind of processors have 3 components, these are:

1- A detector circuit (in this case the signal (coming from the R6/R7 input stage)
2- A discriminator (or threshold control) of Control Voltage (it can't be doesn't exist like some processors like this)
3- A voltage controllable device with gain stage or not. THE RATIO Network! (in the LA-2Athe this is the side-chain preamp )

In this LA-2A case the discriminator doesn't exist and this lack allow any change in RATIO circuit induce (like side effect)  a Threshold change due to gain change in the voltage controllable network. More gain in the RATIO = More RATIO + Threshold change due to the discriminator of level dosen't exist.

Correct to me if I'm wrong please!!

Opacheco.

I forgot!!.....The "Discriminator (or threshold control) of the Control Voltage (it can or can't exist in some processors like this)" can be installed AFTER or BEFORE of RATIO Network..

Opacheco.

 

[ruffrecords]

I have put together a simple spreadsheet that demonstrates how gain before the threshold only affects the threshold and not the ratio and how the only thing that affects the ratio is the characteristic of the gain control element. It is attached as a .txt file so you will need to change the extension to .xlsx to load it into a spreadsheet application.

At the top left are the basic parameters that can be varied:

1. VCA gain (loss) in dB per volts. Intitially set at -10dB/V
2. Side chain gain befroe threshold, initially set at 0dB
3. Threshold in volts. This is a simple representation of a sharp knee. It passes nothing untill the threshold then passes the input voltage less the threshold. It is set initially to 0.775 volts to set the threshold conveniently at 0dBu at the output.

If you run the spreadsheet as it stands, the graph will show the knee at 0dBu in and 0dBu out. Above the knee, the output increases 6 dB for the next 14dB of increase in input level. This is a ratio of  2.33:1

If you change the side chain gain to 6dB, you will immediately see the threshold drops to an input level of -6dbu. For the next 14db of input leevele increase (from -6dBu to +8dBu, the output increases by 6dB from -6dBu to 0dBu. Again this is a ratio of 2.33:1.

It does not matter what value you set the side chain gain to before the threshold, all it does is alter the threshold. The ratio of the next 14dB increase input to the corresponding increase in output range is always 2.33:1.

However, if you change the VCA gain to -20dB/V with 0dB side chain gain, the threshold remains at 0dBu but the ratio changes. As the input rises from 0dBu to +21dBu, the output changes by 6dB which is a ratio of 3.5:1. Change the side chain gain to 6dB and the threshold still drops to -6dBu. The ratio remains 3.5:1.

So as I said originally, changing the gain before the threshold only affect the threshold, changing the gain after the threshold only alters the ratio.

Cheers

Ian

 

[etheory]

Hi Ian!

Very much looking forward to analysing this when I get home later today.

Is this an excel spreadsheet? It's extension is .txt what would I rename it to?

Thanks,
L

 

[ruffrecords]

Hi Ian!

Very much looking forward to analysing this when I get home later today.

Is this an excel spreadsheet? It's extension is .txt what would I rename it to?

Thanks,
L

It was created with Google Sheets and exported in Microsoft Excel format. You need to change the extension to .xlsx. I can export it in other formats if you wish.

It repesents a feedback compressor with a discontinuity (the threshold) in the feedback loop. It is hard to model this  with the input level as the variable but quite simple if you model it with the output as the variable (calculating the input required to produce it) which is what I have done.

Cheers

ian

 

[opacheco]

Well look like anybody want to swim in these complicated area topic!!......

I would like to get a serious analysis (from mathematically or electronically point of view) of the Threshold before or after the Ratio section because isn't too clear if this Threshold network section location impact on the Threshold Value???? It isn't sufficiently clear!

Any comment?

Opacheco.

 

[ruffrecords]

Well look like anybody want to swim in these complicated area topic!!......

I would like to get a serious analysis (from mathematically or electronically point of view) of the Threshold before or after the Ratio section because isn't too clear if this Threshold network section location impact on the Threshold Value???? It isn't sufficiently clear!

Any comment?

Opacheco.

I have been thinking about compressors on and off for a long time. My final year university project back in 1973 was a compressor. A few years back I did a lot of tests with a feedback opto compressor trying to work out what affected ratio and what affected  threshold. I was very surprised to find that changing the gain before the threshold (rectifier) only affected the threshold - the ratio stayed exactly the same. The only thing that affected the ratio was the control law of the opto.

The problem is a feedback  compressor has a non-linearity in the feedback loop which makes it hard to visualise. If you think about what happens up to the threshold (rectifier) any gain change simply alters the signal level at which the threshold occurs. After the threshold the ratio depends on the gain cell characteristic in terms of dB/V. The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the ratio; it still just affects the threshold. It seems counterintuitive but  every test and simulation I have done confirms this is the case.

Cheers

Ian

 

[opacheco]

Well look like anybody want to swim in these complicated area topic!!......

I would like to get a serious analysis (from mathematically or electronically point of view) of the Threshold before or after the Ratio section because isn't too clear if this Threshold network section location impact on the Threshold Value???? It isn't sufficiently clear!

Any comment?

Opacheco.

I have been thinking about compressors on and off for a long time. My final year university project back in 1973 was a compressor. A few years back I did a lot of tests with a feedback opto compressor trying to work out what affected ratio and what affected  threshold. I was very surprised to find that changing the gain before the threshold (rectifier) only affected the threshold - the ratio stayed exactly the same. The only thing that affected the ratio was the control law of the opto.

The problem is a feedback  compressor has a non-linearity in the feedback loop which makes it hard to visualise. If you think about what happens up to the threshold (rectifier) any gain change simply alters the signal level at which the threshold occurs. After the threshold the ratio depends on the gain cell characteristic in terms of dB/V. The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the ratio; it still just affects the threshold. It seems counterintuitive but  every test and simulation I have done confirms this is the case.

Cheers

Ian

Dear Ian,
Very interesting! and I believe you too but did you include any work about this in your thesis?.......I think must exist a pure Math/Electronic Analysis sufficiently clear for this conclusion.

THAT Corp. have a very well done analysis (the better than I could find in anywhere! http://www.thatcorp.com/datashts/dn00A.pdf http://www.thatcorp.com/datashts/dn01A.pdf) but in this analysis the Rectifier (the threshold network) is BEFORE the Ratio network and they don't mention to any dynamic processor using gain BEFORE the Threshold Network.

Opacheco.

 

[ruffrecords]

Well look like anybody want to swim in these complicated area topic!!......

I would like to get a serious analysis (from mathematically or electronically point of view) of the Threshold before or after the Ratio section because isn't too clear if this Threshold network section location impact on the Threshold Value???? It isn't sufficiently clear!

Any comment?

Opacheco.

I have been thinking about compressors on and off for a long time. My final year university project back in 1973 was a compressor. A few years back I did a lot of tests with a feedback opto compressor trying to work out what affected ratio and what affected  threshold. I was very surprised to find that changing the gain before the threshold (rectifier) only affected the threshold - the ratio stayed exactly the same. The only thing that affected the ratio was the control law of the opto.

The problem is a feedback  compressor has a non-linearity in the feedback loop which makes it hard to visualise. If you think about what happens up to the threshold (rectifier) any gain change simply alters the signal level at which the threshold occurs. After the threshold the ratio depends on the gain cell characteristic in terms of dB/V. The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the ratio; it still just affects the threshold. It seems counterintuitive but  every test and simulation I have done confirms this is the case.

Cheers

Ian

Dear Ian,
Very interesting! and I believe you too but did you include any work about this in your thesis?.......I think must exist a pure Math/Electronic Analysis sufficiently clear for this conclusion.

THAT Corp. have a very well done analysis (the better than I could find in anywhere! http://www.thatcorp.com/datashts/dn00A.pdf http://www.thatcorp.com/datashts/dn01A.pdf) but in this analysis the Rectifier (the threshold network) is BEFORE the Ratio network and they don't mention to any dynamic processor using gain BEFORE the Threshold Network.

Opacheco.

Oops, I just realised I wrote the last part incorrectly. When I said :

The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the ratio; it still just affects the threshold.

I should have said:

The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the threshold; it just affects the ratio.

I did not include this in my thesis because I did not realise this at that time!!

The THAT app note is correct with the ratio control after the recitfier.

Cheers

ian

 

[opacheco]

Well look like anybody want to swim in these complicated area topic!!......

I would like to get a serious analysis (from mathematically or electronically point of view) of the Threshold before or after the Ratio section because isn't too clear if this Threshold network section location impact on the Threshold Value???? It isn't sufficiently clear!

Any comment?

Opacheco.

I have been thinking about compressors on and off for a long time. My final year university project back in 1973 was a compressor. A few years back I did a lot of tests with a feedback opto compressor trying to work out what affected ratio and what affected  threshold. I was very surprised to find that changing the gain before the threshold (rectifier) only affected the threshold - the ratio stayed exactly the same. The only thing that affected the ratio was the control law of the opto.

The problem is a feedback  compressor has a non-linearity in the feedback loop which makes it hard to visualise. If you think about what happens up to the threshold (rectifier) any gain change simply alters the signal level at which the threshold occurs. After the threshold the ratio depends on the gain cell characteristic in terms of dB/V. The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the ratio; it still just affects the threshold. It seems counterintuitive but  every test and simulation I have done confirms this is the case.

Cheers

Ian

Dear Ian,
Very interesting! and I believe you too but did you include any work about this in your thesis?.......I think must exist a pure Math/Electronic Analysis sufficiently clear for this conclusion.

THAT Corp. have a very well done analysis (the better than I could find in anywhere! http://www.thatcorp.com/datashts/dn00A.pdf http://www.thatcorp.com/datashts/dn01A.pdf) but in this analysis the Rectifier (the threshold network) is BEFORE the Ratio network and they don't mention to any dynamic processor using gain BEFORE the Threshold Network.

Opacheco.

Oops, I just realised I wrote the last part incorrectly. When I said :

The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the ratio; it still just affects the threshold.

I should have said:

The hard bit to grasp is that after the threshold, the gain in the audio path has no effect on the threshold; it just affects the ratio.

I did not include this in my thesis because I did not realise this at that time!!

The THAT app note is correct with the ratio control after the recitfier.

Cheers

ian

I understood so!!.....after the threshold, the gain have no effect on the threshold; it just affects the ratio.......

But well, we need a more math accurate analysis until. I asked for that to THAT Corp. by e-mail maybe they have a light about this issue.

Opacheco.

 

[ruffrecords]

I understood so!!.....after the threshold, the gain have no effect on the threshold; it just affects the ratio.......

But well, we need a more math accurate analysis until. I asked for that to THAT Corp. by e-mail maybe they have a light about this issue.

Opacheco.

It will be interesting to hear what THAT says.

Cheers

ian

 

[opacheco]

Well, I sent two messages to THAT Corp. but look like they don't have any comment about our issue!!......Would be a good idea someone try to contact them again with the same question!!

I am very intrigued for a analysis for that analysis......

Opacheco.