Vast majority (95%?) of amps out there are internally single-ended.
Is truly balanced, bridge, push-pull, architecture really advantageous for sound quality? We are not concerned here with power output.
Is truly balanced, bridge, push-pull, architecture really advantageous for sound quality? We are not concerned here with power output.
Every detail has its own purpose. Every problem has its own solution. Balanced can be useful if the cable is long and prone to noise. But the balancing circuitry itself brings its own noise. Bridging can be useful if you need the extra power. Of course with extra distortion. Push-pull also has its own merit. And of course issues.
My question is about internal architecture of amplifiers in the context of sound quality only.
So bridging is out because of extra distortion, right?
And cables are not the concern here.
So bridging is out because of extra distortion, right?
And cables are not the concern here.
Internal balance brings complications, such as the potential for non-minimum-phase behaviour. There are good reasons for using external balanced (on long cable runs) and internal unbalanced - so not surprising that this is what most systems do.
My question is very practical. I built single-ended amps and am dabbling in fully balanced internally. So far only problems with hum, buzz, noise. That amp is based on hypex ncore and integrated with attenuator and DAC. Single NC500 modules per channel, not bridged.
What are the engineering reasons to stay internally with single-ended and not go balanced? Can you also explain "non-minimum phase"-- will be very obliged. TIA.
What are the engineering reasons to stay internally with single-ended and not go balanced? Can you also explain "non-minimum phase"-- will be very obliged. TIA.
Last edited:
I doubt there is an answer to this question. Too vague, too many factors.
What said about bridging is wrong. Bridging brings some distortion cancellation.
The purpose of balancing in XLR interconnections is for noise immunity.
What said about bridging is wrong. Bridging brings some distortion cancellation.
The purpose of balancing in XLR interconnections is for noise immunity.
Sound quality within the amp has more to do with input stage topology and output stage topology with regards to 98% of all amps following an LTP Blameless style input. Some use a singleton input stage, and they can sound subjectively better to some people. It affects the harmonic distortion profile. A balanced input stage, in general, will cancel even orders and have higher odd orders, if you like that sort of thing. Output stages are sometimes SE Class A, SE push pull, Push Pull Class A, Push Pull Class AB, Quasi-Complimentary etc. they all affect the sound signature. Some amps are different like the DLH or the Aksa Lender which is an LTP but with signature of a singleton input stage and zero offset auto tracking. The topology is probably the biggest factor on how an amp sounds. With that said, complex balanced drive amps and circlotrons have their own sound - and being symmetric and balanced with have less even order harmonic distortion and more pronounced odd order distortion, again, some people like that as it makes speakers sound more crisp and have more bite - but it’s all what you like to hear and what you consider pleasing. For me, SE Class A is and remains my preferred sound signature. SE Class A can be made with fairly low THD in the 0.002% range - as opposed to SE triode amps in the 2% THD range. Finally, whether or not an amp has high negative global feedback changes the sound signature as well.
Here is an example of a topology that is SE Class A and has about 0.006% THD.
Aksa Lender P-mos Hybrid Aleph (ALPHA) Amplifier
Here is an example of a topology that is SE Class A and has about 0.006% THD.
Aksa Lender P-mos Hybrid Aleph (ALPHA) Amplifier
Last edited:
In most circuits there is one forward path, and this means that the amplitude and phase frequency responses are tied together - given one you can determine the other, this is known as minimum phase. In a balanced circuit there is a risk of there actually being two forward paths and this means that there could be extra phase shift not necessarily implied by the amplitude response.nugat said:
Note that you may need to distinguish between 'balanced' and 'biphase'. 'Balanced' means that the signal consists of the difference in voltage between two points. 'Biphase' means that the signal consists of the sum of two signals (typically antiphase) which are each the voltage from some point to some ground. 'Balanced' really only has an advantage for external cables. 'Biphase' means you have two forward paths to get right instead of one.
single ended transistor amps are not equal to SE triode amps.
there is no SS equivalent for plate feedback which is why I have a tube preamp in my SS pioneer gear.
there is no SS equivalent for plate feedback which is why I have a tube preamp in my SS pioneer gear.
OTOH .... symmetrical circuits seem to be better on average. Such as complimentary, symmetry, push-pull. Also, take a look at some of the new CFA topologies.
And, see John Curls topologies for symmetrical, balanced topologies which consistently sound good as well as measuring very well.
THx-RNMarsh
And, see John Curls topologies for symmetrical, balanced topologies which consistently sound good as well as measuring very well.
THx-RNMarsh
Nice to hear your explanation xrk971. That's more or less exactly what I was going to say.
Fully balanced amplifiers tend to have less 2:nd order harmonics. Also, since they are made up of a greater number of transistors, the complexity of the THD pattern will be higher.
This usually leads to a more crisp and bright sound. The high order harmonics will trigger your ears at higher frequencies.
Personally, I prefer as little balanced stages as possible - ultimately it will then become a wholly single ended amplifier. But there are exceptions.
Fully balanced amplifiers tend to have less 2:nd order harmonics. Also, since they are made up of a greater number of transistors, the complexity of the THD pattern will be higher.
This usually leads to a more crisp and bright sound. The high order harmonics will trigger your ears at higher frequencies.
Personally, I prefer as little balanced stages as possible - ultimately it will then become a wholly single ended amplifier. But there are exceptions.
Hi
In general, complementary-symmetry throughout sounds best to me and has many advantages when built carefully. There is a tendency in such designs to use better compensation methods than the ubiquitous "lin" amp uses, and this is where the biggest difference comes - not from the topology as much as from the cmpensation.
The basic "Lin" with a diff input, VAS and unity-gain current ampligfier is most often croppled by its Miller cap compensation. If you look at doug Self's work he has concentrated almost solely on achieving performance from this now ancient topology. he can get mid-band THD to reasonable-ish levels\, but sadly the THD20 is still 1960s performance. THD rises steeply past 8kHz, so his usual high-frequency test at 10kHz looks pretty bad.
I look at the Lin (or Thompson, or Blameless) as a single-ended amp despite the push-pull output stage. The VAS is nominally SE evenn though the current source load for it is active and DOES work actively. There is a simple measure to improve the CS by adding a speed-up cap - Doug shows this in the later version of his book.
Bob Cordell built a much more sophisticated version of this, with a single diff input, but a VAS that is push-pull and a mosfet current buffer with error correction. The basic THD of this amp is not really much better without the error correction, but... Bob implements a different compensation scheme and the THD20 is extremely low and slew-rate extremely high.
Most of the current feedback amps are symmetric and also have compensation that breaks the slew-rate limitation of Miller compensation.
I think the Lin has the misfortune of always being coupled to quite brutal Miller compensation, with the result that the 98% of all power amps built (which follow that pattern) all sound the same, with the same good and bad qualities to their sound. The "breathe of fresh air" topologies tend to be symmetric and have better compensation methods.
Back in the 1970s when Marshall Leech was presenting his Low-TIM PA, the first version had as benign Miller compensation as possible, so the amp had good bandwidth and a pleasant sound. He presented alternate compensation schemes as the amp evolved, and these methods have since acquired different names but are not any newer because of it.
Some of the more recent compensation schemes are tricky to stabilise and may cause start-up problems with the amp. So, for less experienced builders (like me) it is best to stay with simpler yet still effective methods and be free of the blameless-sound - which isn't blameless at all.
In general, complementary-symmetry throughout sounds best to me and has many advantages when built carefully. There is a tendency in such designs to use better compensation methods than the ubiquitous "lin" amp uses, and this is where the biggest difference comes - not from the topology as much as from the cmpensation.
The basic "Lin" with a diff input, VAS and unity-gain current ampligfier is most often croppled by its Miller cap compensation. If you look at doug Self's work he has concentrated almost solely on achieving performance from this now ancient topology. he can get mid-band THD to reasonable-ish levels\, but sadly the THD20 is still 1960s performance. THD rises steeply past 8kHz, so his usual high-frequency test at 10kHz looks pretty bad.
I look at the Lin (or Thompson, or Blameless) as a single-ended amp despite the push-pull output stage. The VAS is nominally SE evenn though the current source load for it is active and DOES work actively. There is a simple measure to improve the CS by adding a speed-up cap - Doug shows this in the later version of his book.
Bob Cordell built a much more sophisticated version of this, with a single diff input, but a VAS that is push-pull and a mosfet current buffer with error correction. The basic THD of this amp is not really much better without the error correction, but... Bob implements a different compensation scheme and the THD20 is extremely low and slew-rate extremely high.
Most of the current feedback amps are symmetric and also have compensation that breaks the slew-rate limitation of Miller compensation.
I think the Lin has the misfortune of always being coupled to quite brutal Miller compensation, with the result that the 98% of all power amps built (which follow that pattern) all sound the same, with the same good and bad qualities to their sound. The "breathe of fresh air" topologies tend to be symmetric and have better compensation methods.
Back in the 1970s when Marshall Leech was presenting his Low-TIM PA, the first version had as benign Miller compensation as possible, so the amp had good bandwidth and a pleasant sound. He presented alternate compensation schemes as the amp evolved, and these methods have since acquired different names but are not any newer because of it.
Some of the more recent compensation schemes are tricky to stabilise and may cause start-up problems with the amp. So, for less experienced builders (like me) it is best to stay with simpler yet still effective methods and be free of the blameless-sound - which isn't blameless at all.
You can try this type of compensation, easy to stabilize.www.diyaudio.com/forums/solid-state/317335-oitpc-output-inclusive-tpc-tmc.html
I designed and built amplifier monoblocks which were fully balanced internally -450W. The front end of the amp was ECC83. The power stage was bipolar driving Lateral FET. In balanced configuration the distortion was 0.005%. As an exercise I converted one monoblock amp into a stereo amp simply by rearranging the connectors and removed the external unbalanced to balanced converter. The distortion was now 0.12% because there was no longer any distortion cancellation.
The amps were auditioned with B&W 801 speakers and silver speaker cable. Since the actual circuitry hadn't changed apart from the unbalanced to balanced converter, which would not have added much sound signature, it was an interesting test.
The outcome of the test was that the two configurations sounded radically different. There is no point attempting to apply a 'better' attribute. The original monoblock configuration served its purpose as a studio amplifier. The unbalanced configuration produced one of the nicest sounds I have ever heard, but was not appropriate for the intended client.
The amps were auditioned with B&W 801 speakers and silver speaker cable. Since the actual circuitry hadn't changed apart from the unbalanced to balanced converter, which would not have added much sound signature, it was an interesting test.
The outcome of the test was that the two configurations sounded radically different. There is no point attempting to apply a 'better' attribute. The original monoblock configuration served its purpose as a studio amplifier. The unbalanced configuration produced one of the nicest sounds I have ever heard, but was not appropriate for the intended client.
THD 20khz is a pretty useless measure of sound quality. Most speakers won't produce even the first harmonic, and if they do no one will hear a 40khz harmonic even if it's at 100% of the signal level.
Within the bandwidth of most soundcard-based analyzers, though an IMD torture test at ~20kHz is about as good as we can get. Exposes the underlying linearity as loop gain falls off.
How so? Bridging doubles slew rate and potentially cancels H2 distortion. Of course the load impedance seen by each amp halves, but for a given output power, as long as the amps are designed around the load impedance, I can’t see why a pair of bridged amps wouldn’t have significantly lower distortion than a single amp.
20 KHz THD is more a measure of amplifier speed (and thus low IMD) than anything else. An amp with poor 20 KHz THD is likely to have slew rate problems - lots of intermods etc, which are in the audio range. Conversely an amplifier with good 20 KHz THD will also have good IMD.
Testing 20 KHz THD is more convenient (only one source needed) than testing IMD directly.
THD at 20KHz is not necessarily telling us anything about an amplifier's speed. I am currently working on a design for a small amp that has full power bandwidth to 2MHz. The THD 20KHz won't satisfy the low distortion crowd, but it is not a slow amp.
Amplifier speed can also mean ugbw, which 20k will tell us about 26 db better than 1k, provided a first order slope to the open loop integrator. 🙂
I am not sure my concept of speed is the same as everyone elses. UGBW and slew rate have caveats. How many so called 'fast amps' with high slew rates have input filters that prevent the design ringing with square waves? Seems to defeat the objective of a fast design.
In my view, a first base is to be able to cope with a 100nS rise time input without an input filter, but have no overshoot. This means UGBW and slew rate are secondary, at least to me. Following the bandwidth and slew rate doctrine doesn't result in stable amps. Getting a design stable is actually very hard. My previous efforts I can now see are flawed.
In my view, a first base is to be able to cope with a 100nS rise time input without an input filter, but have no overshoot. This means UGBW and slew rate are secondary, at least to me. Following the bandwidth and slew rate doctrine doesn't result in stable amps. Getting a design stable is actually very hard. My previous efforts I can now see are flawed.
- Status
- Not open for further replies.