Dear forum:
Due to the ever increasing number of power amp's, integrated amps and preamp's advertised as "balanced" and or "fully balanced" designs, which have gained popularity primarily due to home theater, I'd like your opinions on the following please:
1. The majority of those designs, as in pro-audio also, feature some type of circuit at the XLR inputs, but from there, continue as only single-ended types.
They vary considerably regarding their CMRR performance.
Also, most of them, offer RCA inputs as well, except in pro-audio where the alternate input types are 1/4" jacks (instead of RCA) and often, Tip-Ring-Sleeve types instead of XLR.
What is interesting is that the common types of specifications given for each design usually show better specifications when using the RCA (or unbalanced) ins rather than the so-called balanced inputs.
In pro-audio, the quasi balanced ins usually offer (as the only advantage) a higher operating level at the inputs.
2. My main interest concerns the actual design of fully balanced, mirror imaged circuitry types. That is to say: separate circuitry for the positive and negative signals for each channel from input to output.
Since there are many manufacturers designs available, there will be-of course-differences in excellence in those designs.
When a manufacturer offers the common basic specifications, one can not usually figure out the quality of the unit.
Regarding CMRR, I assume that it refers to the ability of the circuitry to reject as much of the unwanted signal as possible across a given bandwidth and at various levels, etc.
What should one look or inquire about regarding these designs?
What sort of "Hype" or marketing language do you usually come across?
3. Besides CMRR, what other types of specifications should show as an example of the advantages of fully balanced designs compared to single ended types?
4. Will there always be an advantage to balanced versus unbalanced in properly designed systems of each type?
Thank you,
Due to the ever increasing number of power amp's, integrated amps and preamp's advertised as "balanced" and or "fully balanced" designs, which have gained popularity primarily due to home theater, I'd like your opinions on the following please:
1. The majority of those designs, as in pro-audio also, feature some type of circuit at the XLR inputs, but from there, continue as only single-ended types.
They vary considerably regarding their CMRR performance.
Also, most of them, offer RCA inputs as well, except in pro-audio where the alternate input types are 1/4" jacks (instead of RCA) and often, Tip-Ring-Sleeve types instead of XLR.
What is interesting is that the common types of specifications given for each design usually show better specifications when using the RCA (or unbalanced) ins rather than the so-called balanced inputs.
In pro-audio, the quasi balanced ins usually offer (as the only advantage) a higher operating level at the inputs.
2. My main interest concerns the actual design of fully balanced, mirror imaged circuitry types. That is to say: separate circuitry for the positive and negative signals for each channel from input to output.
Since there are many manufacturers designs available, there will be-of course-differences in excellence in those designs.
When a manufacturer offers the common basic specifications, one can not usually figure out the quality of the unit.
Regarding CMRR, I assume that it refers to the ability of the circuitry to reject as much of the unwanted signal as possible across a given bandwidth and at various levels, etc.
What should one look or inquire about regarding these designs?
What sort of "Hype" or marketing language do you usually come across?
3. Besides CMRR, what other types of specifications should show as an example of the advantages of fully balanced designs compared to single ended types?
4. Will there always be an advantage to balanced versus unbalanced in properly designed systems of each type?
Thank you,
I don't think so. I think this popularity mainly comes from professional audio.
What do you want to do with them?
"balanced makes your home theatre system sound better"
Linearity performance might be slightly better, since the signal amplitude is divided over two phases, meaning that single-ended amplitude is 2x smaller when compared to the supply voltage. But in a well-designed system, this really shouldn't matter.
Yes, you can run miles of cable without worrying about picking up interference from other electric appliances. That's about it in terms of real-world advantages. The disadvantage is that at some point inside the signal source, an amplifier stage sees twice the load, meaning performance might be degraded compared to an unbalanced system. If you don't risk picking up interferers and you're not turned on by XLR connectors, I recommend you go with unbalanced (in whatever form, RCA or jack or pretty much whatever).
--
Greetz,
MatchASM
The point of differential signalling is to get high rejection of interference on the signal pair. This is useful if the cable between the source and the amp is very long (maybe 100's of feet or more). However, note that the rejection only works if the interference is actually common mode -- i.e. the same amount of signal in injected on both signal wires. For a tightly twisted pair, this is generally the case.
An ideal differential amp will only "see" the voltage difference between the two inputs and amplify just that. It will ignore any common-mode voltage present on the two inputs. In other words, the two inputs could be 1 kV above amp ground and (assuming the amp had enough common mode range to handle this) the amp would still only see the differential voltage between the two inputs.
In the real world, the amp does not completely reject the common-mode signal. Its ability to reject the common-mode signal is the CMRR - Common-Mode Rejection Ratio. That's the money spec.
Consumer audio being consumer audio is really more about marketing than about specs. Most people do not have the test equipment needed to verify the specs. So it's easy for manufacturers to claim that their designs are "fully differential" or "fully balanced" (what does balanced really mean in this context?!) and few people would be able to challenge this.
~Tom
An ideal differential amp will only "see" the voltage difference between the two inputs and amplify just that. It will ignore any common-mode voltage present on the two inputs. In other words, the two inputs could be 1 kV above amp ground and (assuming the amp had enough common mode range to handle this) the amp would still only see the differential voltage between the two inputs.
In the real world, the amp does not completely reject the common-mode signal. Its ability to reject the common-mode signal is the CMRR - Common-Mode Rejection Ratio. That's the money spec.
Consumer audio being consumer audio is really more about marketing than about specs. Most people do not have the test equipment needed to verify the specs. So it's easy for manufacturers to claim that their designs are "fully differential" or "fully balanced" (what does balanced really mean in this context?!) and few people would be able to challenge this.
~Tom
One thing to watch out for is the degradation of the CMMR of an input due to mismatches in the impedance of the circuit feeding it, including the wire. Bill Whitlock of Jensen has done a lot of work on this. He has shown that some circuits can lose 20 db CMMR from a small imballance (which is typical) in the capacitance of the interconect cable.
Hi
There are advantages to using a balanced topology, but there are some disadvantages as well. The design of the circuit is important in minimizing common mode distortions. If the intent is to have a balanced output, the IPS and VAS should be designed as a single balanced circuit not two seperate single end amplifers simply driven with oposite phase signals. Using a single balanced topology instead of two SE circuits will help reduce common mode distortions as a difference between the two outputs. Common mode PS distortions are not fed to the speaker either. Ground return current of the speaker is eliminated as well as there is no reference to GND.
On the flip side a balanced amp usually requires a balanced input so this limits the program sources available. For a single end input to balanced output many folks use OP-amps, one inverting the other non-inverting, to create a balanced signal as transformers have thier own issues. Unfourtunately this usually leads to phase shifting at higher frequencies because an inverting OP-amp has different properties than a non-inverting circuit. Phase differences at the balanced output create distortion. The circuit and PCB layout is more complicated because you now have two output stages along with all the other extra parts.
But why use balanced output anyway? A lower voltage power supply is an advantage even if the current taken from it is two fold and more uf's of filter caps are required. But then, lower voltage filter caps are cheaper and smaller per uf.🙂 Lower voltage transistors can be used which opens the door to better device selection.
The balanced amp I created can crank ~120Wrms into 8R using +/- 24V. The output transistors are 60V devices. The VAS is a balanced bridge topology so one side works off of the other. The pre-amp circuit can accept either XLR or SE and requires a simple jumper to select the input source. In SE mode the pre-amp circuit converts the signal to balanced. If you feed the + input to the bridge, the two outputs will be balanced in 180 degree phase but the - output voltage will be smaller. The circuit uses a deliberately misbalanced bridge topology and a common mode error correction scheme to balanced the signal voltage, esentially it converts the difference in voltage amplitude at the two outputs to a difference of output Z but with equal output voltages. At 2MHz, of course gain drops a bit but phase alignment is perfect. (circuit's pretty fast😎) The input Z of the next amplifier stage is very high so the difference in the two signals become eliminated from the transfer function resulting in a much better SE to BAL conversion than one could achieve from using op-amps. I once posted a discription of the circuit here but no-one was really interested.🙄 Oh well. Perhaps I will do a write up after the final design is complete, which could be a while because of lack of funding.
Anyway, IMHO the pros out weigh the cons and balanced topology is the way to go, if you have a good solid balanced signal and proper PCB layout.😉
There are advantages to using a balanced topology, but there are some disadvantages as well. The design of the circuit is important in minimizing common mode distortions. If the intent is to have a balanced output, the IPS and VAS should be designed as a single balanced circuit not two seperate single end amplifers simply driven with oposite phase signals. Using a single balanced topology instead of two SE circuits will help reduce common mode distortions as a difference between the two outputs. Common mode PS distortions are not fed to the speaker either. Ground return current of the speaker is eliminated as well as there is no reference to GND.
On the flip side a balanced amp usually requires a balanced input so this limits the program sources available. For a single end input to balanced output many folks use OP-amps, one inverting the other non-inverting, to create a balanced signal as transformers have thier own issues. Unfourtunately this usually leads to phase shifting at higher frequencies because an inverting OP-amp has different properties than a non-inverting circuit. Phase differences at the balanced output create distortion. The circuit and PCB layout is more complicated because you now have two output stages along with all the other extra parts.
But why use balanced output anyway? A lower voltage power supply is an advantage even if the current taken from it is two fold and more uf's of filter caps are required. But then, lower voltage filter caps are cheaper and smaller per uf.🙂 Lower voltage transistors can be used which opens the door to better device selection.
The balanced amp I created can crank ~120Wrms into 8R using +/- 24V. The output transistors are 60V devices. The VAS is a balanced bridge topology so one side works off of the other. The pre-amp circuit can accept either XLR or SE and requires a simple jumper to select the input source. In SE mode the pre-amp circuit converts the signal to balanced. If you feed the + input to the bridge, the two outputs will be balanced in 180 degree phase but the - output voltage will be smaller. The circuit uses a deliberately misbalanced bridge topology and a common mode error correction scheme to balanced the signal voltage, esentially it converts the difference in voltage amplitude at the two outputs to a difference of output Z but with equal output voltages. At 2MHz, of course gain drops a bit but phase alignment is perfect. (circuit's pretty fast😎) The input Z of the next amplifier stage is very high so the difference in the two signals become eliminated from the transfer function resulting in a much better SE to BAL conversion than one could achieve from using op-amps. I once posted a discription of the circuit here but no-one was really interested.🙄 Oh well. Perhaps I will do a write up after the final design is complete, which could be a while because of lack of funding.
Anyway, IMHO the pros out weigh the cons and balanced topology is the way to go, if you have a good solid balanced signal and proper PCB layout.😉
Last edited:
Such designs, while technically "balanced" in that the impedances of each line is balanced with respect to ground, which is what "balanced" refers to, has no common-mode rejection.
To reject common-mode noise, the input needs to be differential. And having separate circuits on each side like that is not differential. It will pass common-mode noise as if it were signal.
se
Hi:
Thanks everyone for replying.
As usual, here at diY, I new that I would get in-depth answers.
I definitely want to continue with this thread, but have to be absent from my computer for a few hours...just wanted to let everyone know that I did not just fade away from the topic.
I'll return in a little while as this is a topic of great interest.
Thank you.
Thanks everyone for replying.
As usual, here at diY, I new that I would get in-depth answers.
I definitely want to continue with this thread, but have to be absent from my computer for a few hours...just wanted to let everyone know that I did not just fade away from the topic.
I'll return in a little while as this is a topic of great interest.
Thank you.
If I can toss in my own prejudices:
Balanced cables between equipment have advantages if they are long or in a noisy environment, provided the two ends are done properly. In a normal domestic environment this is usually unnecessary, and is just 'marketing'.
Balanced signals within equipment are often a bad idea, unless you really don't have a signal ground. You have to either maintain high CMRR all the way through, or ensure very low common-mode distortion - either adds complication when remembered or problems when forgotten. You can't actually get both paths through the circuit to have exactly the same gain/phase characteristics so there are problems of HF stability.
If for some reason I wanted to have balanced input and output from an amp I would probably still use unbalanced signals within.
The most useful place for balanced input is a phono preamp but the RIAA network would be almost impossible to get right in balanced form.
Balanced cables between equipment have advantages if they are long or in a noisy environment, provided the two ends are done properly. In a normal domestic environment this is usually unnecessary, and is just 'marketing'.
Balanced signals within equipment are often a bad idea, unless you really don't have a signal ground. You have to either maintain high CMRR all the way through, or ensure very low common-mode distortion - either adds complication when remembered or problems when forgotten. You can't actually get both paths through the circuit to have exactly the same gain/phase characteristics so there are problems of HF stability.
If for some reason I wanted to have balanced input and output from an amp I would probably still use unbalanced signals within.
The most useful place for balanced input is a phono preamp but the RIAA network would be almost impossible to get right in balanced form.
Thanks DF96 and everyone else.
My experience in pro audio has proven that many issues have occurred by using balanced interfaces. As many have pointed-out, their is a difference between a balanced interface, running single-ended and various designs using balanced circuitry through the entire circuits.
I can remember maybe one device that was labeled as fully balanced end to end. Although it performed well within the particular network, I can not say anything more about those design types.
As for those respondents who implied that RCA or other standard related types of I/O can usually run for long distances, it all depends on the design of the associated equipment, or that miles (so to speak) of cables are for balanced interfaces to avoid noise, or other, that too is a big maybe, having been witness to very difficult to solve noise/hum/buzzes issues.Many times associated professionals doing contract work for install have had to cut the screens at one end or the other and many other tricks as well.
I had a conversation with Bill Whitlock about this subject and as you may know, Bill is very smart, but I was unable to retain all that he explained.
DF96... I read you reply above and wish to ask some fundamental questions please:
I can say that balanced equipment has worked well and out performed single-ended in those situations where it was intended, keeping in mind that some times it was worse.
About fully bal versus single ended. If as you say...
... " You have to either maintain high CMRR all the way through, or ensure very low common-mode distortion - either adds complication when remembered or problems when forgotten. You can't actually get both paths through the circuit to have exactly the same gain/phase characteristics so there are problems of HF stability.
If for some reason I wanted to have balanced input and output from an amp I would probably still use unbalanced signals within."
... then my question is how does single ended handle the positive and negative halves of the signal if as I assume the differences in signal path would be much less exact than a differential mirror imaged balanced circuit?
If trying to make precise signal paths through duplicate but opposite circuitry is hard, why is it more exact when one part of the wave form is amplified through a much less matched circuit?
Not in my opinion. The input differential may not have a reference to ground, but the common mode voltage will be referred to ground. It will have to stay within some range of ground. In addition, the output common mode will have to be controlled -- assuming that it's a differential-in, differential-out circuit. CM feedback circuits are notorious for issues and good ones are difficult to design. That said, it's a problem that's been solved in many integrated (IC) designs.
Opinions of this will differ. What's meant by "fully balanced" is up to the marketing person who writes the marketing spiel.
~Tom
MC and many well rated DAC ICs provide differential output. So it is possible to build fully differential path up to the speaker drivers were hot and cold mixed together while canceling all noise accumulated since the signal had quit DAC IC. It requires all components must be fully differential internally however.
Most HiEnd equipment provides such capability (let's recall Krell FPB for example) so such equipment would benefit from XLR connectivity despite the cable length.
Audiophiles who likes Single Ended Tango-Tamura (etc) equipment would be better stay with RCA.
Most HiEnd equipment provides such capability (let's recall Krell FPB for example) so such equipment would benefit from XLR connectivity despite the cable length.
Audiophiles who likes Single Ended Tango-Tamura (etc) equipment would be better stay with RCA.
Last edited:
If there is only one path through the circuit instead of two parallel paths then it does not need to be matched - as there is no other half to match to! It merely needs to be reasonably accurate.
The "positive and negative halves" of the signal are handled as a single signal by normal Class A circuitry. The signal is actually the voltage difference between a given point and ground. This is easier to handle for amplification and feedback than a pair of signals, where the actual wanted signal is the difference between them. You can sample a voltage (for feedback purposes) with a resistor or two; to sample a voltage difference with high CMRR needs a transformer or differential amp just to get the feedback signal.
The "positive and negative halves" of the signal are handled as a single signal by normal Class A circuitry. The signal is actually the voltage difference between a given point and ground. This is easier to handle for amplification and feedback than a pair of signals, where the actual wanted signal is the difference between them. You can sample a voltage (for feedback purposes) with a resistor or two; to sample a voltage difference with high CMRR needs a transformer or differential amp just to get the feedback signal.
I think that before this stuff can be meaningfully discussed some terminology needs sorting out.....
A Balanced output refers to a circuit where each leg is 'driven' by the SAME SOURCE IMPEDANCE relative to the local common mode reference, it says NOTHING about opposite polarities or anything of that sort (but that is one way to do it).
A balanced input measures the difference between the two signal legs while presenting the same (and ideally very high) common mode impedance to each.
You can think of a balanced line as operating as a Wheatstone bridge where the imposed common mode noise is the exciting voltage, if you think of it in those terms then it becomes apparent that for good behaviour in the presence of real world unbalance due to component tolerance and real cables and connectors, the common mode input impedance should be very much greater then the source impedance so that small imbalances in source impedance do not seriously unbalance the bridge.
This is where the common 4 * 10K around a 5532 input stage falls down as the common mode impedance is just too low given the typical 100r or so of the build out resistors.
Due to the fact that a properly done balanced input stage should work with any signal source presenting the same lowish impedance to both legs, this whole 'fully differential' thing is at best confused thinking as it will typically not work well where one leg is driven via a 100r buildout and the other is 'driven' by a hundred ohm buildout to the sources audio reference (0v) rail... This is a perfectly valid balanced line drive arrangement and works perfectly if your receiver handles the conversion to single ended at that point.
What does NOT work is building two power amps and feeding one from the non inverting signal and the other from the inverting one, as for some perfectly valid sources you will find you can only get 1/4 of full output power (because one amp will be holding its output damn close to 0v).
The right way to handle a differential input is to convert it to single ended referenced to the local audio reference (NOT the screen connection, that goes directly to chassis right at the connector) then do whatever you are going to do (Possibly making it differential again, whatever).
Balanced lines have one major advantage as well as a rather nasty noise problem....
The win is that there is no longer any need to ever export the audio reference rail from any equipment (it can be purely local), so not only is there far better rejection of magnetic fields in the cables (twisting does that), but currents flowing in screen connections due to common mode differences cannot impose IR voltages on the audio (Unless someone has stuffed up and connected the screen to the audio reference rail instead of the case, Major German microphone maker, looking at you!).
This is a HUGE win once cables are long, possibly cross earth domains and can have significant common mode differences between the ends.
Consider a class II double insulated CD player connected to a class I (earthed) amp, not an uncommon situation, if the CD player has a class Y cap leaking say 0.25mA to the case from the power line (High, but IIRC on the limit), then if the phono cables are say 0.1 ohm screen and connector resistance, that means that we are inducing 25uV of noise due to the leakage current flowing to earth down the screens.
If the CD is putting out 1V at full scale that means the induced noise is ~ the same level as the LSB of the CD.
Things get much more ugly if you have a class II preamp with multiple class II sources as ALL the leakage can end up on the screen connections of the preamp outputs where that 1V of CD player output may have been turned down 25+db to make a comfortable listening level.... All of a sudden that leakage is only 60db below the peak audio level, add a nice dynamic classical CD to the mix......
None of this is an issue with a balanced interface because that screen current does not (to a first order) get anywhere near the audio.
The downside:
Noise! A simple 4 * 10K + opamp (which is a poor balanced RX anyway) input stage is significantly noisier then a single ended input (The noise gain is higher and there extra resistors thermal noise does not help), but even so this noise contribution is normally less then the contributions from the processing electronics, and if you really care you can design a differential stage to be as quiet as a single ended one (But it is a pain and takes board area).
Hope someone finds the engineering tradeoffs interesting.
It is probably my broadcast engineer background showing but I tend to stick balanced IO on most things (and run it so full scale digital is at least +20dbu), your mileage may vary.
Regards, Dan.
A Balanced output refers to a circuit where each leg is 'driven' by the SAME SOURCE IMPEDANCE relative to the local common mode reference, it says NOTHING about opposite polarities or anything of that sort (but that is one way to do it).
A balanced input measures the difference between the two signal legs while presenting the same (and ideally very high) common mode impedance to each.
You can think of a balanced line as operating as a Wheatstone bridge where the imposed common mode noise is the exciting voltage, if you think of it in those terms then it becomes apparent that for good behaviour in the presence of real world unbalance due to component tolerance and real cables and connectors, the common mode input impedance should be very much greater then the source impedance so that small imbalances in source impedance do not seriously unbalance the bridge.
This is where the common 4 * 10K around a 5532 input stage falls down as the common mode impedance is just too low given the typical 100r or so of the build out resistors.
Due to the fact that a properly done balanced input stage should work with any signal source presenting the same lowish impedance to both legs, this whole 'fully differential' thing is at best confused thinking as it will typically not work well where one leg is driven via a 100r buildout and the other is 'driven' by a hundred ohm buildout to the sources audio reference (0v) rail... This is a perfectly valid balanced line drive arrangement and works perfectly if your receiver handles the conversion to single ended at that point.
What does NOT work is building two power amps and feeding one from the non inverting signal and the other from the inverting one, as for some perfectly valid sources you will find you can only get 1/4 of full output power (because one amp will be holding its output damn close to 0v).
The right way to handle a differential input is to convert it to single ended referenced to the local audio reference (NOT the screen connection, that goes directly to chassis right at the connector) then do whatever you are going to do (Possibly making it differential again, whatever).
Balanced lines have one major advantage as well as a rather nasty noise problem....
The win is that there is no longer any need to ever export the audio reference rail from any equipment (it can be purely local), so not only is there far better rejection of magnetic fields in the cables (twisting does that), but currents flowing in screen connections due to common mode differences cannot impose IR voltages on the audio (Unless someone has stuffed up and connected the screen to the audio reference rail instead of the case, Major German microphone maker, looking at you!).
This is a HUGE win once cables are long, possibly cross earth domains and can have significant common mode differences between the ends.
Consider a class II double insulated CD player connected to a class I (earthed) amp, not an uncommon situation, if the CD player has a class Y cap leaking say 0.25mA to the case from the power line (High, but IIRC on the limit), then if the phono cables are say 0.1 ohm screen and connector resistance, that means that we are inducing 25uV of noise due to the leakage current flowing to earth down the screens.
If the CD is putting out 1V at full scale that means the induced noise is ~ the same level as the LSB of the CD.
Things get much more ugly if you have a class II preamp with multiple class II sources as ALL the leakage can end up on the screen connections of the preamp outputs where that 1V of CD player output may have been turned down 25+db to make a comfortable listening level.... All of a sudden that leakage is only 60db below the peak audio level, add a nice dynamic classical CD to the mix......
None of this is an issue with a balanced interface because that screen current does not (to a first order) get anywhere near the audio.
The downside:
Noise! A simple 4 * 10K + opamp (which is a poor balanced RX anyway) input stage is significantly noisier then a single ended input (The noise gain is higher and there extra resistors thermal noise does not help), but even so this noise contribution is normally less then the contributions from the processing electronics, and if you really care you can design a differential stage to be as quiet as a single ended one (But it is a pain and takes board area).
Hope someone finds the engineering tradeoffs interesting.
It is probably my broadcast engineer background showing but I tend to stick balanced IO on most things (and run it so full scale digital is at least +20dbu), your mileage may vary.
Regards, Dan.
Good point about temperature noise. Thank you Dan!
Differential path cannot eliminate it and might be noisier. Fully differential op amps like LME49724 could be more convenient to deal with differential signal.
Differential path cannot eliminate it and might be noisier. Fully differential op amps like LME49724 could be more convenient to deal with differential signal.
Not really, apart from things like mix buses (that are a can of worms all on there own) balanced lines are a win for external audio, but generally far less of a win inside gear (There are exceptions, the aforementioned mix buses and the immediate surroundings of ADC and DAC sand spring to mind), but for everything else inside the box you are typically better off going single ended and being careful with the reference connection layout.
Some of the Crystal semi ADC data sheets are wonderful examples of what NOT to do in this respect.
Most of the noise problem is down to the noise gain of the differential receiver more then thermal noise, but in practise this noise contribution is usually small compared to the rest of the system (especially given that differential inputs usually run much hotter then the unbalanced world does).
Consider something like a simple RC highpass filter, utterly trivial single ended (resistor, cap, possibly a buffer or two, trivial...), but if you try a differential version (two caps, resistor,....) then in real life the CMRR goes to pot an octave either side of the 3db point (manly due to cap tolerances), it does not take much differential phase shift to totally screw up the CMRR.
Far better to convert to single ended then do the processing then convert back to differential before leaving the board then to try to go differential all the way and be constantly fighting a loosing battle with real component symmetry (Also generally lower parts count and less noise).
The input circuitry around the differential to single ended conversion is also often a good place to apply some bandwidth limiting to ensure that radio stays out, and that what gets in is not going to push any later stages into slew rate or other intermod producing issues, power amps can be good at this if precautions are not taken.
It is worth being a little careful however as the input circuitry will usually be located before the attenuator that serves as the volume control so it is sometimes possible to clip the input electronics on something like a power amp even with the gain backed right off.
As ever the whole thing comes down to careful consideration of the engineering tradeoffs.
Regards, Dan.
Some of the Crystal semi ADC data sheets are wonderful examples of what NOT to do in this respect.
Most of the noise problem is down to the noise gain of the differential receiver more then thermal noise, but in practise this noise contribution is usually small compared to the rest of the system (especially given that differential inputs usually run much hotter then the unbalanced world does).
Consider something like a simple RC highpass filter, utterly trivial single ended (resistor, cap, possibly a buffer or two, trivial...), but if you try a differential version (two caps, resistor,....) then in real life the CMRR goes to pot an octave either side of the 3db point (manly due to cap tolerances), it does not take much differential phase shift to totally screw up the CMRR.
Far better to convert to single ended then do the processing then convert back to differential before leaving the board then to try to go differential all the way and be constantly fighting a loosing battle with real component symmetry (Also generally lower parts count and less noise).
The input circuitry around the differential to single ended conversion is also often a good place to apply some bandwidth limiting to ensure that radio stays out, and that what gets in is not going to push any later stages into slew rate or other intermod producing issues, power amps can be good at this if precautions are not taken.
It is worth being a little careful however as the input circuitry will usually be located before the attenuator that serves as the volume control so it is sometimes possible to clip the input electronics on something like a power amp even with the gain backed right off.
As ever the whole thing comes down to careful consideration of the engineering tradeoffs.
Regards, Dan.
I agree. Common mode has to be referenced to GND. In my circuit there are two DC servo amps that acomplish this via seperate input floating GND references. CM feedback is the best way to keep common mode distortions low. The complementary bridge circuit is quite good for this and can be controled by CM FB loops to make sure the current on each side of the bridge is always equal. This is important in reducing CM distortions. But it has to be driven by an acurate balanced input signal. SE to XLR can be tricky if you intend high accuracy and understand the drawbacks of using a simple transformer over a range of frequencies. It can be done very well electroniclly but I have found it is not as straight forward as it seems.
Having a true balanced signal within a circuit is not a bad thing if it is done properly. There are significant advantages....such as 2X SW for realitivily nothing amongst other advantages.🙂Bridging two seperate SE amps is certainly not the way to produce a quality balanced output, even though many people recommend that route. As stated differential phase shift create terrible results in the common mode realm. This is why I do not use OP-amps to convert SE to XLR. Common mode feedforward error correction is a much better solution.
BTW, although more expensive, matched component devices are your friend.😉
Last edited:
SE to XLR output is trivial, pin 1 goes directly to chassis, pin two goes to the single ended output and pin 3 goes to the audio reference (0v) rail on the pcb via a resistor equal to the build out resistor on the single ended output, job done.
This makes a perfectly valid balanced line output from a single ended circuit at the cost of a single resistor (and the rather bigger cost of an expensive connector), if you don't need the extra 6db of swing (and usually you don't), this is perfectly acceptable practise.
For balanced input stages (where common mode range is known to be reasonable) the THAT Corp parts are nice, where common mode may become extreme (say between buildings or if overhead cables are in play), then Lars Lundahl make the correct answer (As do Jensen, Sowter et all). The trade off with the iron is that it is worse in all ways except isolation and common mode impedance then a good sand based solution.
Regards, Dan.
This makes a perfectly valid balanced line output from a single ended circuit at the cost of a single resistor (and the rather bigger cost of an expensive connector), if you don't need the extra 6db of swing (and usually you don't), this is perfectly acceptable practise.
For balanced input stages (where common mode range is known to be reasonable) the THAT Corp parts are nice, where common mode may become extreme (say between buildings or if overhead cables are in play), then Lars Lundahl make the correct answer (As do Jensen, Sowter et all). The trade off with the iron is that it is worse in all ways except isolation and common mode impedance then a good sand based solution.
Regards, Dan.
- Status
- Not open for further replies.