You're certainly entitled to your opinion, in my opinion this is a waste of time. TIM, PIM or whatever get over it.
Ok - why don't you let rip and let's see what the assembled expertise has to say?
Member
Joined 2009
Paid Member
There's BS all over audio, commercial and non-commercial. I think most of us can smell it pretty easily when we see it. What I find quite fascinating is how the ego is presented through people's posts. From some posters I'm used to seeing it, from some it's really quite entertaining, but I'm a bit disappointed by some - I thought they were mostly above this kind of stuff based on their contributions in the past, but sure is fascinating what you can learn about human behaviour through the microcosm of internet forums.
I like gnf, from both an engineering perspective and from my experience of listening to amps that use it. I really do think it's a huge benefit to audio. But there are great sounding amplifiers that don't use it. In other words, I haven't found any correlation between gnf and poor sound. it's not compulsory to use it, but the Question is - do you know how to use it well ?
I like gnf, from both an engineering perspective and from my experience of listening to amps that use it. I really do think it's a huge benefit to audio. But there are great sounding amplifiers that don't use it. In other words, I haven't found any correlation between gnf and poor sound. it's not compulsory to use it, but the Question is - do you know how to use it well ?
Last edited:
My hunch is that what the subjectivists are hearing with GNFB is insufficient PSRR from their output stages. Seeing as the vast majority of amps are classAB the power rails get corrupted from having to deliver haversine currents with high levels of HF. This happens with opamps too - even though the loads they drive are at least a couple of orders of magnitude higher impedance than speakers.
But PRRR is much lower without GNFB in competently designed amplifiers - even in class AB. Or are you saying lower PSRR is actually the sound they like?
80 dB at 1 KHz is quite normal for a SS amp with GNFB.
(I am nog sun expert - just expressing an opinion here)
How was the PSRR measured? I've found that (for example) PSRR of chipamps based on DS plots doesn't match up with what's going on in real life.
Have you got an 80dB 1kHz PSRR figure from your own designs and if so, how did you measure?
Member
Joined 2009
Paid Member
I think PSRR is the wrong hunch too, unless 'lazy' designers are relying on gnf to cover up inadequate power supply design. Most of the differences I have heard between gnf amps and non-gnf amps is, I suspect, down to interaction between the amplifier with the reactive load of real world speakers.
Member
Joined 2009
Paid Member
I wasn't planning any tests for this notion, but offer my thoughts in case it resonates with others who have much more experience that I. All I have to offer is a purely subjective opinion based on the messing about I've done with various amplifiers and speakers this past few years.
The explanation might be rather boring though - higher output impedance affecting the bass is one aspect. Another aspect is the treble, generally sounds different but my non GNF amps are Class A whereas my gnf amps are class AB so hard to say I've isolated just the one variable here.
I suppose it's possible to design an amplifier with adjustable feedback, say one that varies how much feedback comes after the OPS verses how much comes after the VAS - preserving the overall gain. Anyhow, I ramble.
The explanation might be rather boring though - higher output impedance affecting the bass is one aspect. Another aspect is the treble, generally sounds different but my non GNF amps are Class A whereas my gnf amps are class AB so hard to say I've isolated just the one variable here.
I suppose it's possible to design an amplifier with adjustable feedback, say one that varies how much feedback comes after the OPS verses how much comes after the VAS - preserving the overall gain. Anyhow, I ramble.
This sounds plausible to me. I've managed to get some opamp based preamps to sound quite good just by changing the decoupling/bypassing arrangement and this is one thing that could explain it.
Excuse possible stupid question: Do you mean not using GNFB will prevent stuff on the rails from getting in to the signal node via the output stage? Does this extrapolate to any improvement by using less GNFB rather than none?
On a sim, you move the AC source to the PSU voltage source - you get the PSRR plot then directly at the output.
I have not measured it directly - not hard to do, but it would need an hour or so of set-up work.
See Harry Dymond's TPC comp paper for some useful input on PSRR.
For CFA 50 dB at 1 KHz is easily achievable and with larger filter networks on the rail, 70 DB. For VFA, count on being about 20 dB better.
Yes I'm aware of that. But its not very close to a real-life situation and I suspect this might be one reason PSRR plots don't hold too much water in my experience. There are two aspects missing from this way of doing a PSRR in a sim - first is there's no signal present, second is there's no load.
If you have a spare hour to do it, I'd be interested in the results and methodology.
I'll be interested to read that.
Thanks for your input owdeo - I also have the same experience with improving the power supplies to opamp circuitry. The same mechanism at work I believe.
No this is a pertinent question - I think this is the issue yes, keeping the power rail induced noise out of the loop. But at this stage its only handwaving - open loop a typical EF output stage might have only 60dB PSRR, maybe less at HF.
I suspect using less GNFB means having more attenuation of the PSU hash before it reaches the LTP? I've not looked into this in any detail so far - my focus at the moment is on designing better output stages with improved PSRR.
The sonic signatures guy had no hard science people on his committee, I noticed.
Member
Joined 2009
Paid Member
A PSRR plot is done with no input because you want to remove that as a variable and look at it in isolation.
In a real world situation, you are going to get lots of confounding factors, so I don't expect the sim results to match reality. However, you can quickly see how ccircuit hanges affect PSRR and investigate 1st order effects.
On my e-Amp, the no input signal mains hum and harmonics are all below -105 dB. At 180W out into 8 ohms, some of the peaks are at -85 dB as measured with sound card.
In a real world situation, you are going to get lots of confounding factors, so I don't expect the sim results to match reality. However, you can quickly see how ccircuit hanges affect PSRR and investigate 1st order effects.
On my e-Amp, the no input signal mains hum and harmonics are all below -105 dB. At 180W out into 8 ohms, some of the peaks are at -85 dB as measured with sound card.
He doesn't quite say that.
When I got a bit into the piece I was reminded of a really sophomoric paper I wrote for a philosophy class. In my defense I was about a sophomore at the time 😀
But the professor wrote in the margin: You move fast here.
The numbers of unsupported conclusive leaps in Cheever are most worrisome. There are some very odd mistakes, and a plethora of typos and misspellings.
I guess Total Aural Disconsonance didn't quite catch on.
What is also interesting to me: both Cheever and the sonic signatures guy were employees of universities.
Oh well.
Last edited:
A circuit with 100% NFB will have the output stage directly connected to the input stage i.e. 0dB voltage gain. Having more signal gain means less of the 'dirty' signal from the output stage appears at the LTP because there's an attenuator in place between the two.
You can also insert gain in the return path, although few do. The gain can be frequency-dependent and of course phase-compensated.
Just sayin'