Asking as one with no electrical engineering background- are there kinds of amplifier distortions, which speakers pass along, or emphasize more so than others?
Still wondering why a Rotel amp sounded different than a Linar amp with the same speakers, in the same room.
Both amps seemed to have excellent specifications.
Still wondering why a Rotel amp sounded different than a Linar amp with the same speakers, in the same room.
Both amps seemed to have excellent specifications.
Hi,
Problem with the printed specs that come with the amplifiers' marketing glossy paper is that they're hardly any indication of what the amp is going to sound like.
Under the circumstances as you describe the only time I heard amplifiers that sounded pretty much alike was when two (or more) of the same brand were compared, one say of higher or lower output wattage than the other.
I attribute the similarity in sound to the same topology used (which is often the case within a single brand).
Add to that the use of the same (brands of) passive components throughout the range and you can hear a house name "signature"...
Different brands of amps with similar specs can and IME almost invariably, do sound all different.
Cheers,
Problem with the printed specs that come with the amplifiers' marketing glossy paper is that they're hardly any indication of what the amp is going to sound like.
Under the circumstances as you describe the only time I heard amplifiers that sounded pretty much alike was when two (or more) of the same brand were compared, one say of higher or lower output wattage than the other.
I attribute the similarity in sound to the same topology used (which is often the case within a single brand).
Add to that the use of the same (brands of) passive components throughout the range and you can hear a house name "signature"...
Different brands of amps with similar specs can and IME almost invariably, do sound all different.
Cheers,
Some PNP bipolar transistors have very linear collector current vs. base current (from high-Z drive) characteristics. Some VFET MOS transistors have an extremely linear drain current vs. gate-source voltage (from lowZ drive) when they get to higher currents (usually where their power handling capability makes them hard to use, though). Much depends on the circuit configuration and signal source types.
So did I. To talk about the voltage linearity of the device, you put a load and an input voltage on it and measure the transfer function. You have three terminals to choose from for input, output, and common- one particular arrangement isn't "special" or "privileged" in some way.
Let's try again ...
I'll make this easy, ... suppose I design and build what I believe is a "faultless" amplifier, and I wish to convince an appropriate member of the scientific community of that fact, merely by sending him a set of measurements and graphs, what data would he have to see, as a minimum, and what would the numbers be - the sufficient case, IOW?
And I agree with that.
I'll make this easy,
... suppose I design and build what I believe is a "faultless" amplifier, and I wish to convince an appropriate member of the scientific community of that fact, merely by sending him a set of measurements and graphs, what data would he have to see, as a minimum, and what would the numbers be - the sufficient case, IOW?
Last edited:
She would have to see appropriate and relevant data with appropriate and relevant information about the test conditions that the data was collected under.
If we were talking about the electrical operation of the amp, the data would probably characterise the output of the device with reference to the input. Any number of characterising measures could be used, but they would have to show the amplification function has taken place with minimal loss of or addition to the source ie, low distortion.
All this ground has been covered before.
However, and more realistically, lets not pretend this would convince you. The data that would (or should) convince you is aggregated qualitative listening test results. Not one-offs, but listening tests carried out with enough subjects that the results are statistically meaningful. From this, any reasonably competent person with an engineering and/or science background would be able to identify if a statistically significant number of listeners found your amp to be truly flawless.
If we were talking about the electrical operation of the amp, the data would probably characterise the output of the device with reference to the input. Any number of characterising measures could be used, but they would have to show the amplification function has taken place with minimal loss of or addition to the source ie, low distortion.
All this ground has been covered before.
However, and more realistically, lets not pretend this would convince you. The data that would (or should) convince you is aggregated qualitative listening test results. Not one-offs, but listening tests carried out with enough subjects that the results are statistically meaningful. From this, any reasonably competent person with an engineering and/or science background would be able to identify if a statistically significant number of listeners found your amp to be truly flawless.
A few dozen times and extensively in the literature- including a few dozen papers I cited en masse earlier in this thread. Boxes of gain are easy. Coming up with a differentiator for marketing into the fashion niche takes far more creativity than the box of gain itself.
Groups, panels, and multiple listeners are fine and even necessary for certain questions, not for others. One size does not fit all.
But that's the tricky bit - what are the numbers? Without those, what you're saying is meaningless ..
No, I'm not going to worry about listening tests, we'll leave ears out of this entirely, for now. And mainly because you would need to characterise the ears: are they a group of people who are fanatical about the subtleties of playback distortion, or a group of "volunteers" pulled in "off the street", or somewhere inbetween?
Who said it was "a characteristic of the device"? It is a characteristic of a particular circuit fed a particular signal. It is thus a result of a method, but not necessarily an "easy" method; even it if were "easy", would this in itself diminish its value? It has been shown to correlate with listening quality - try listening to 50% distortion if you doubt this! In fact, some people prefer a little distortion (typically a few percent on peaks) and describe sound with much less distortion as 'sterile' or 'unmusical'.bwaslo said:
Avoiding feedback where it would be helpful is just incompetent engineering.cbdb said:
The balance between local and global feedback cannot be summed up in a simple rule. It requires planning gain distribution, assessing likely sources of distortion etc. This is known as circuit design. In some cases local positive feedback can reduce overall closed-loop distortion, but this technique has to be applied carefully!
The second sentence is not a consequence of the first sentence. 3/2 power gives all orders of distortion, up to infinity. 2 power just gives second order - generally considered more benign. Tubes are more linear for two reasons: anode feedback (in triodes), and the fact that even a small signal tube is really a power device so only a relatively small part of its characteristic is usually being used.rayma said:
When people want to design a very high dynamic range (i.e. very low distortion) small-signal preamplifier for RF they don't use FETs or valves. They use a power BJT, run it at high current with a heatsink, and add lots of NFB via a wideband transformer.
I have given numbers, but I can't remember if it was in this thread or another.fas42 said:
Numbers which would probably satisfy most people:
frequency response: -3dB first-order rolloffs at no higher than 20Hz and no lower than 20kHz (which implies a level response between).
noise: -80dB wrt maximum (maybe more needed for sensitive speakers in a small room)
distortion: harmonics and IMD below 0.5% at peak output, and monotonically decreasing for smaller signals, and also monotonically decreasing for higher orders (apart, perhaps, for some odd-even order imbalance).
hum/buzz etc: around or below noise level.
output impedance: damping factor greater than 20 across the full bandwidth, with no more than a 10% variation with signal level.
input impedance: greater than 50k across the full bandwidth, apart from a little shunt capacitance, with no more than a 10% variation with signal level. This implies a source impedance of no greater than about 2.5k.
envelope-related effects: less than 1% variation, apart from maximum power which can vary by up to 20%.
I may have omitted something important; others will chime in. Note that this is intended to satisfy people who wish to reproduce music; others may prefer some added distortion or a varying frequency response.
Of course, careful listening tests were required to arrive at most of the figures above.
Because audio has its own set of physics and rules, that the rest have not discovered yet, and they don't have their own FAS42 to point out that engineers are basically lazy throw it together and sell it fast people who rely on THD figures only...and all can be cured by using the correct combination of cables between components to tune the resultant sound field. Hell they don't even use QPs in the real world...never mind worrying about cable break-in and directivity of cables...
It that's the approach then there can be no absolute determination if something audibly distorts - everything, including the reference, may distort in the same manner; there is no differentiation.
But, if we listen to the input, and then to the output, with levels appropriately attentuated then we can say there is no audible difference. Have tests been rigorously carried out this way, where the amplifier is operating into different loads, at different power levels?
An example how to test:
In this case its an AD/DA converter, but you can do the same with any device.
As DF already has mentioned, there are lots of tests done to come up with some numbers, but they are a bit fuzzy. Engineer well below them, not that difficult, and its very likely (but not 100% certain!) your safe. Listening tests are always required to make sure.
Thanks very much, DF96, that's exactly what I was after ...
So, only very slowly dropping to say 0.4% at very low levels would be acceptable? And, similar levels being allowed, going to higher and higher orders?
Your input is much appreciated,
.
Thanks for that, Tattoo. What I would be particularly interested in, is where the DUT is a power amplifier operating at different levels of stress into a realistic load - what tests are around that have published results?
- Status
- Not open for further replies.