Douglas Self Line Input compared to Mackie line input

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Yes. But Thor's 'thing' seems to be that CMRR is not so important.

Not quite.

It is that CMRR is not universally important.

Where large common mode noise voltages exist (suggesting something is wrong with power/earth/ground routing anyway) CMRR is our defense.

But if make sure to keep those common mode voltage to a minimum anyway, what do we gain by designing for high CMRR?

I mentioned TANSTAAFL ( "There Ain't No Such Thing As A Free Lunch") to refer to what is called opportunity cost. If we needlessly design for unnecessary high CMRR, we pay a cost, be it excess noise, excess circuit complexity, both.

And I mention "industry standard practice" to indicate that it should be analysed, robustly questioned and if we can see alternatives, we should consider.

Example, commonly the audio band output impedance of a "common" balanced output is 100 Ohm (which matches appx. many twisted pairs), but why should we apply this in a context of our own studio? Equally the 10-20k balanced input impedance (bridging load) is meant not alter the termination of a terminated twisted pair, long enough that we encounter transmission line effects at audio frequencies (which is nowadays extremely rare).

In reality these two "industry practice 'standards'" are outdated, unnecessary and counterproductive.

In a modern EMI environment a simple build out resistor is suboptimal.
An LCL "T" filter is much more likely to EMI at bay and will produce a buildout network with next to zero Ohm audio band impedance. When combined with quality cables we will likely see lower CMRR degradation in practice with (say) 1.2kOhm input impedance using the simplest possible balanced receiver circuit, with excellent noise performance and a very simple circuit.

Driving both wires in a balanced line is again an outdated concept.

Use an impedance balanced drive using Rail To Rail Op-Amp's on +/-18V gives us in excess of +22dBu clipping levels, a balanced line receiver with unity gain and using 600 Ohm for all four resistors (2k4 * 4 in parallel) using OPA1611/12 we get -118dBu noise with 22dB headroom, or 140dB Dynamic Range. Using NE5532/34 degrades SNR by ~5dB to 135dB DNR.

CMRR is likely in practice significantly better than 60dB at low frequencies (using multiple 0.1% tolerance resistors and the law of averages to in effect get 0.025% tolerance) , so more than ~ -60dBu common mode noise will be needed to degrade this figure, more for 5532.

Hard to beat on complexity vs. cost vs. performance and likely unachievable if we design for unnecessarily high CMRR.

Of course, if your application is likely to routinely experience large levels of common mode noise and large source impedance imbalances, by all means design in a suitable circuit.

Again, TANSTAAFL, be sure you know what is needed and what is priority and design accordingly, do not follow "received wisdom" blindly, no matter who dispenses it.

1703761066303.png

Thor
 
Not quite.

It is that CMRR is not universally important.

Where large common mode noise voltages exist (suggesting something is wrong with power/earth/ground routing anyway) CMRR is our defense.

But if make sure to keep those common mode voltage to a minimum anyway, what do we gain by designing for high CMRR?

I mentioned TANSTAAFL ( "There Ain't No Such Thing As A Free Lunch") to refer to what is called opportunity cost. If we needlessly design for unnecessary high CMRR, we pay a cost, be it excess noise, excess circuit complexity, both.

Agreed
And I mention "industry standard practice" to indicate that it should be analysed, robustly questioned and if we can see alternatives, we should consider.

Agreed
Example, commonly the audio band output impedance of a "common" balanced output is 100 Ohm (which matches appx. many twisted pairs), but why should we apply this in a context of our own studio?

We don't. The common 100 ohm output impedance is created by series build out resistors designed to counter the effect of the cable capacitance on the op amp output stage stability. Nothing at all to do with the characteristic impedance of the cables used. TANSTAAFL as you say.
Equally the 10-20k balanced input impedance (bridging load) is meant not alter the termination of a terminated twisted pair, long enough that we encounter transmission line effects at audio frequencies (which is nowadays extremely rare).
It is nothing to do with the termination of the twisted pair which is immaterial at audio frequencies. It is designed to avoid the 6dB drop in signal level that occurs with a 600 ohm source and load. No ore, no less.

Cheers

Ian
 
At my last day job involving audio product design we had customers all around the world dealing with dodgy mains power. The customers don't want excuses they want products that work even under difficult circumstances.
Industry standard practices often have some experiential justification but good engineering is to question everything. During my stint managing an engineering group, part of the job was keeping my engineers from reinventing already round wheels, as is the nature of design engineering.
We don't. The common 100 ohm output impedance is created by series build out resistors designed to counter the effect of the cable capacitance on the op amp output stage stability. Nothing at all to do with the characteristic impedance of the cables used. TANSTAAFL as you say.
+1 but conveniently the characteristic impedance of most audio cables is around 50 ohms, so it's possible to satisfy multiple constraints.
It is nothing to do with the termination of the twisted pair which is immaterial at audio frequencies. It is designed to avoid the 6dB drop in signal level that occurs with a 600 ohm source and load. No ore, no less.
yup there is still 600 ohm I/O legacy gear out in the world. Customers don't want excuses , they just want good audio.

JR
Cheers

Ian
 
Ian,

We don't. The common 100 ohm output impedance is created by series build out resistors designed to counter the effect of the cable capacitance on the op amp output stage stability. Nothing at all to do with the characteristic impedance of the cables used. TANSTAAFL as you say.

It is nothing to do with the termination of the twisted pair which is immaterial at audio frequencies. It is designed to avoid the 6dB drop in signal level that occurs with a 600 ohm source and load. No ore, no less.

Back in the analogue age, when black round sheets of vinyl were used to store acoustic information and radio waves were modulated in amplitude or frequency, distribution of radio programmes was often via telephone wiring (as opposed to telephone lines), analogue. The cables were long enough between repeaters to cause transmission lines effects to appear at audio frequencies.

The typical impedances we use today go back to those systems in the 1950's to 1970's (before digital system like Nicam/aptX took over).

Yes. we use build out DEVICES to stabilise the Op-Amp against a capacitive load causing oscillation, but there is no reason to have a specific impedance.

Anyway - it's ancient history.

In 2023 it is much better to use this kind of build-out network:

1703782673504.png

Preferably with a pair of very high frequency ferrite beads on the output side.

Thor
 
For DIY, YOU are the customer...

Thor
we actually agree on that point.

I do several things in my DIY audio paths that I could never get away with in even "value" SKUs while designing for Peavey. I will gladly tolerate a few switch clicks for the benefit of avoiding multiple electrolytic capacitors in my audio path. Also I will live with turn on/off transients that Peavey customers would disproportionately judge audio path quality poorly from.

Customers tend to overvalue cosmetic sound issues (like clicks, pops, and turn on thumps that they can hear, more than path linearity that they can't. Since the customer is always right it helps sales to meet their expectations. ;)

JR
 
The common 100 ohm output impedance is created by series build out resistors designed to counter the effect of the cable capacitance on the op amp output stage stability

See the reference I posted in the thread on interfacing modern and antique designs: post with link to Richard Hess paper.
That paper was written in the context of a broadcast facility, so needed to drive some long lines. Too low output impedance was found to result in resonance due to bulk inductance effects in the cable. An output impedance of 60 Ohms differential was found to give a good balance between damping the inductive peaking and causing high frequency roll-off due to effects of bulk capacitance.
 
Ian,



Back in the analogue age, when black round sheets of vinyl were used to store acoustic information and radio waves were modulated in amplitude or frequency, distribution of radio programmes was often via telephone wiring (as opposed to telephone lines), analogue. The cables were long enough between repeaters to cause transmission lines effects to appear at audio frequencies.

The typical impedances we use today go back to those systems in the 1950's to 1970's (before digital system like Nicam/aptX took over).

Yes. we use build out DEVICES to stabilise the Op-Amp against a capacitive load causing oscillation, but there is no reason to have a specific impedance.

Anyway - it's ancient history.

In 2023 it is much better to use this kind of build-out network:

View attachment 119533

Preferably with a pair of very high frequency ferrite beads on the output side.

Thor

Interesting stuff. Although it does seem rather component heavy and costly compared to a simple series build out resistor. The Powerpad of the THS6012 might also be a problem for DIY if no access to paste/oven assembly.
Still, looks like it will drive some serious cable load 🙂
What are the load assumptions wrt the RC snubber on the output ?
 
Interesting stuff. Although it does seem rather component heavy and costly compared to a simple series build out resistor.

The build out resistor will do jack against EMI ingress. If you don't care about that, you can simplify to just an inductor in parallel with the build out resistor. Given the cost of SMD inductors the cost is not material.

The Powerpad of the THS6012 might also be a problem for DIY if no access to paste/oven assembly.

Paste and hot air is fine.

Still, looks like it will drive some serious cable load 🙂

Yup, +22dBu into 100 Ohm.

What are the load assumptions wrt the RC snubber on the output ?

> 100R to open open circuit.

Thor
 
The build out resistor will do jack against EMI ingress. If you don't care about that, you can simplify to just an inductor in parallel with the build out resistor. Given the cost of SMD inductors the cost is not material.
Agreed. And I was simplifying tbh. I practice I would put some rfi protection in there.

Paste and hot air is fine.
I should have mentioned hot air too. Depending on how 'DIY' someone is then hot air might not be practicable. Regardless of that it's an interesting application of that device.


Yup, +22dBu into 100 Ohm.
That should cover all the bases 😊

> 100R to open open circuit.
Thanks
 
That should cover all the bases 😊

When combined with a "generic" balanced input with all resistors adjusted to 600 Ohm (2.4k /0.1% 4pcs parallel and using a sufficiently low noise Op-Amp, the system should be able to offer end to -115 dBu noise with +22dBu maximum levels and very little HD at any level.

It does retain a certain degree of backwards compatibility with generic equipment, though CMRR may degrade.

The TPA6120/THS6012 is really a luxury, a OPA1656 could cover 2 channels just fine, with less brawny drive.

Thor
 
While looking at the schematics for the Mackie Onyx 1220 mixer, I noticed they implemented a line input with +/-20db trim using only two opamps stages.



I'm wondering how this approach compares to the praised Douglas Self line input proposed on his book. The Mackie looks more useful since unity gain should be at center of the pot, and there is more gain range.. but what is the tradeoff compared to the Self approach (below)?

View attachment 117412
Mackie i using a classic design that was deployed in a lot of mixers before they were in existance and a standard circuit in application guides.

What they call 'Douglas Self' is what we called servo feedback amplifier in the mid 80s and was featured in TI's op amp application guide back then.
 
The build out resistor will do jack against EMI ingress. If you don't care about that, you can simplify to just an inductor in parallel with the build out resistor. Given the cost of SMD inductors the cost is not material.
While the modern RF environment is worse than I was last active in these trenches, the output build out resistor creates a voltage divider with the active output driver's very low output impedance. This output impedance is reduced by negative feedback (working with Aol open loop gain). Since stability compensation rolls off this open loop gain at high frequency, source impedance typically rises with frequency.

At the cost of somewhat higher source impedance two resistors in series with a good quality HF cap to ground from the middle node can provide a passive pole that keeps working. RF coming back into the output is/was a known problem with audio power amps. Something about long unshielded speaker cables making effective antennas. 🤔 Power amps often used series inductors in their output with real poles in the very outputs. We have seen examples of RFI feeding all the way back into the input stage LTP and causing rectification if not fast enough.

Caveat Lector... my hands on experience with this was decades ago.

JR
Paste and hot air is fine.



Yup, +22dBu into 100 Ohm.



> 100R to open open circuit.

Thor
 
while browsing some old schematics, I noticed that the Soundcraft 8000 uses the same "double inverted" configuration (at least to my novice eyes) that we were discussing before, although with a slight more complex gain control network (well, more complex for me)


Screenshot 2024-01-13 at 12.27.31.png

I also have another question..what would it be the best approach (noise-wise) to add a direct output after this receiver? the goal is to send this direct output to an external audio interface for recording. I think impedance balanced output with a pair of 75R resistors after the opamp would be enough (I don't think I need fully balanced output, the connection between the console and the audio interface are extremely short, so I think I can sacrifice CMRR here) but looking at other consoles I often see this approach:

Untitled.png
 
That output stage is a "Ground Sensing" configuration. It provides CMRR into an unbalanced input by adding the remote "Ground" into the signal transmitted. Well covered in Self's "Small Signal..." book. That version is also configured as a "Zero Impedance" output.
fwiw you don't need to use 75R resistors specifically. 75R (like 50R) comes from commonly used RF termination / characteristic impedance considerations eg with coax cable. But they are not so easily available - see E ranges/values. Any standard value in range say 47R to 100R should be fine.
 
I think impedance balanced output with a pair of 75R resistors after the opamp would be enough (I don't think I need fully balanced output, the connection between the console and the audio interface are extremely short, so I think I can sacrifice CMRR here)

If I understand your question correctly, you seem to believe that impedance balanced output reduces CMRR compared to a symmetrically driven output.
That is not the case for a true impedance balanced output, meaning that you take care to actually match impedance between the driven side and the undriven side. I have seen some outputs which appeared as if the designer only partially understood the principle, such as having a matching resistor, but not including the output capacitor on the undriven side. One such design had a low enough value coupling cap on the driven side that the impedance mismatch between the driven and non-driven legs at 60Hz was close to 50%. A 50% mismatch is not "balanced" for any reasonable interpretation of the word.

So to your original question, yes, adding an impedance balanced output after the op-amp is probably the best way to do what you want, and if you do it properly there will be no CMRR penalty.
 
If I understand your question correctly, you seem to believe that impedance balanced output reduces CMRR compared to a symmetrically driven output.
That is not the case for a true impedance balanced output, meaning that you take care to actually match impedance between the driven side and the undriven side. I have seen some outputs which appeared as if the designer only partially understood the principle, such as having a matching resistor, but not including the output capacitor on the undriven side. One such design had a low enough value coupling cap on the driven side that the impedance mismatch between the driven and non-driven legs at 60Hz was close to 50%. A 50% mismatch is not "balanced" for any reasonable interpretation of the word.

So to your original question, yes, adding an impedance balanced output after the op-amp is probably the best way to do what you want, and if you do it properly there will be no CMRR penalty.
thank you for the clarification,
so now my question is, what are the advantages of using "active" impedance balanced output with an additional opamp (like the one I showed above) versus the simpler approach of using just the series build out resistor + its matched resistor on the undriven side?
 
Back
Top