Distortion Surface: old measurement, new approach

The power of narrative crafting and marketing!

To be fair, I can see how someone not familiar with the fact that components and systems can (and typically do) behave differently with different inputs, including frequency or power levels, to accept a single point measurement as a measuring stick of "goodness". Problem is when they continue to spread said wrong information.

 

In a sane world, this would shut down ASR. But it's not. They'll likely only grow and perhaps Amir will even run for president.

 

A better goal: Be the candle that sheds a light in the darkness. @atomicbob may be on to something here. I firmly believe there is some “Moffat Lattice” to be defined that can explain why some things sound like ass. That’d be cool - imagine Dan Clark and Jason and all those good industry folks building even better gear as a result — who cares if some website says their car’s exhaust measures better.

 

This reminds me of the PowerCube plots described by AP here: https://www.ap.com/technical-library/measuring-power-amplifiers-with-reactive-loads/

Is there an obvious relationship between these plots and PowerCube plots?

I think we need to know if that's before or after taking some deep whiffs....

 

Alternate visualizations as requested / suggested by @Biodegraded

Projection of 3D surface to contour map below




Display of projected contour map only in 2D image

 

Just being anal - can you flip the axis on the 2D versions? We're all used to level being vertical and frequency climbing from left to right.

And this is so f'ing cool.

 

At this time, the program I am using for projecting the 3D surface to a 2D contour does not allow me to choose axis orientation. My only option would be to import the graph into a photo editor, rotate and flip, which will screw up the labels.

 

Cool, no worries.

 

Can you add a colorbar so we know the magnitude for each shade?

 

Quick and dirty.

 

Still more to it. If the feedback applied amp stage is fast enough and this much NFB doesn't create TIM etc, the higher feedback version may be better able to drive an actual reactive load. Yes I know that 100db of feedback is a silly amount so these are both theoretical posts. IMO, ears ears still the best way to collate all the variables but you can't get to that level of goodness without a good technical execution 1st. I love that this thread is attempting to expand on the limited representations we normally see.

 

Contour maps have legends now.





Please keep in mind I am using a relatively inexpensive plotting program, PSI Plot, and am approximately two major versions behind the latest. For $300 it does a lot. But it doesn't do everything. Populating the legend automatically has a fixed numerical format of xx.xxxx. Each line may be edited manually in a text box to clean up the format at expense of time required and increased possibility of a typo.

The iDSD Diablo is on the bench for technical measurements now. When I have caught up on measurements and posts on which I am very behind, I may consider either an update to PSI Plot or looking at alternate plotting programs. I know Comsol Multiphysics could do what I want, at $4000 + 800/yr licensing. Not sure I want to go that far.

 

I've been looking at a few free graphics programs, all of which have their downsides - eg not allowing axes in log scale, not contouring well (or at all), not having decent visualization, or being a hell of a learning curve. Of the freebies GNUPlot looks like it'd do everything, but there'd be quite a learning curve involved even with a GUI front-end.

If anybody has any suggestions, all welcome.

 

It's amazing what the right info in the right visualization can do. Vastly different amps.

 

what I am noticing is the 3f performance is really good at -27.5 dBu and it isn't till about -10 that continues that performance.
interesting

 

@atomicbob: Are you open to sharing a dataset with me? If I have time next week, I can try to give you a GNUPlot equivalent so you can just plug in data and labels and have everything lined up…

 

The legend not being the same for both charts makes comparing the distortion figures a bit tricky, but that doesn't really matter for the most striking difference in the chart, the vertical bands on the Diablo vs. the horizontal bands on the DSHA.

If I'm reading it right, the Diablo's distortion falls as level increases, but doesn't vary across frequency at a given level. The DSHA's chart isn't as simple, but it's sort of the opposite - doesn't vary across level at a given frequency (with a couple exceptions, most notably a sharp fall around -10dBu around 1khz and up), but varies quite a lot with frequency (lots more distortion below about 200hz than above it).

Thinking about distortion patterns, my guess is that it could well be preferable, natural-sounding, etc, for distortion to vary proportionally with signal level, as opposed to, say, not varying such that it's a much bigger component of what you hear at low volumes than at high. This isn't based on much other than sketchy intuition though, especially since I haven't heard either amp.

I wonder if we'll see other terrifying band shapes with some of the other amps.

 

While running sweeps on the iFi Diablo an idea occurred to me about a work around for my plotting program. I believe the following graphs satisfy the collective requests.





Please note autoranging of color map levels between the two contours, which make them not exactly equivalent. This is new territory for me and I am relying considerably on application contour map defaults. Likely future maps will be constrained by the same color scales.

I have attached data sets for the DSHA-3F for anyone to play with.

 

All except that the word 'Legend' should go directly above your avatar

Most excellent!

 

One last request? Throw a crosshair at the point where the standard measurement is taken (1K, 2Vrms, etc.) for perspective.