Yes, as usual the people who manage to get some good work happening then ruin it, botch it badly - because the primary drive is that good ol' fella, $$$$$$$$$$$$$$$$$ ...
DiffMaker is a question mark for me, so far I'm not impressed: the software itself is flakey, it crashes at a drop of a hat -- I'm struggling to get it to do a job anywhere as well as someone playing manually with Audacity can achieve, I note that there are other comments about in this regard ...
There is probably scope for someone to get really get stuck in, and do a far more decent job than DiffMaker currently can do.
DiffMaker is a question mark for me, so far I'm not impressed: the software itself is flakey, it crashes at a drop of a hat -- I'm struggling to get it to do a job anywhere as well as someone playing manually with Audacity can achieve, I note that there are other comments about in this regard ...
There is probably scope for someone to get really get stuck in, and do a far more decent job than DiffMaker currently can do.
One of DiffMaker's claims is that it can handle slight timing shifts, and I haven't yet been able to 'force' it to deal with such. By deliberately creating a copy of a track with effectively a slight clock misalignment, it grinds and grinds, and eventually crashes when asked to deal with doing a comparison.
I've spent too many years in the software game, haven't got the patience any more to baby poorly written programs to get them to behave themselves ...
I've spent too many years in the software game, haven't got the patience any more to baby poorly written programs to get them to behave themselves ...
Yes, the fact that the companies then indulge in the OTT propaganda, and marketing, once they get a sniff of a solution badly undermines their credibility ... but, that doesn't mean that there isn't something of value in the initial, underlying work done ...
Re. DiffMaker, apart from a DIY effort are you aware of any other software that is useful for this sort of analysis?
Re. DiffMaker, apart from a DIY effort are you aware of any other software that is useful for this sort of analysis?
Thank you anyway, Michael. The text above gives me something to mull over. If I get stuck, I will aks you via PM, if that's all right.
Last edited:
In my view there is little, if any, merit in these tests. They don't seem to have taken account of the fact that sampling seems to be asynchronous, so much of what they are seeing is clock jitter in either the CD player or their sound card. They also seem to take no account of the effect of anti-alias and reconstruction filters. Instead, they seem to treat the whole things as a data sampling problem - presumably this fits with their previous experience in defence work.Robert F said:
If I took digital audio data from a CD, and also digitised (asynchronously) the analogue output, I would expect to see major differences. This is what they see. Careful alignment will reduce, but never eliminate these differences. To get rid of them you would need synchronised clocks and exact knowledge of all the filtering (analogue and digital) which has taken place in both the CD player and the sound card.
To me, all these tests demonstrate is that defence contractors may be ignorant about digital audio. It may suggest that defence contractors are not as smart as they think they are (even when doing defence work) but hide behind racks of impressive and expensive test equipment.
But, if the testing methodology was sufficiently refined such that dynamic tracking of clock drift and misalignment was involved, and upsampling was applied intelligently to precisely align the tracks at any point of the sample then the results would be meaningful. This after all is what Diffmaker promises - albeit implemented rather poorly from what I've seen so far. Processing power is cheap these days, there's no reason that it can't be done - the key requirement is that the programmers really know what they're doing ...
Firstly, we're interested whether an effect takes place which is measurable, and whether it's consistent, repeatable. In that paper they claim that it is, and I see no reason to doubt that. Then, one can argue what is the cause of that, and especially whether totally conventional and reasonably priced methods can achieve the same results.
Our big headache in this is nailing the 'problem' occurring, actually being able to point to a clear 'reading' that some abberation of the signal is taking place. If statistical or other relatively complex analysis has to be done to confidently point the finger, so be it - the point is to be able to corner the mongrel ...
Our big headache in this is nailing the 'problem' occurring, actually being able to point to a clear 'reading' that some abberation of the signal is taking place. If statistical or other relatively complex analysis has to be done to confidently point the finger, so be it - the point is to be able to corner the mongrel ...
Anything on bootstrapping is pure gold to me be they for or against . I never witnessed walking on water so can not give an opinion . Bootstrapping I have seen and love it .
This is the hazard of not having any real experience in setting up and running proper experiments.
A question . I have built a very simple mostly pentode amplifier which is pleasing against my friends 300 B design . Where I might have an advantage over some is that I tried to define the problems before even starting . To my advantage I have a big bucket of valves to try . Some are close to death and others above spec . I was given serious warnings by others I can not expect the performance of an amplifier like this to be repeatable . Well whatever it is I got wrong that is not true .
I have no loop nor plate to plate feedback . I have no transistor current source or sinks . The design is modest in it's aims . The speakers suitable to use with it .
Here is the question . No valve measures very differently to the next . Before anyone says I have lousy equipment I do have access to an Audio Precision test set . Results I get and from AP are not too different if a valve design . The question is why do various EL 34's sound different ? To sum it up it is like this . It is like you wake up having slept 5 years . Everything is subtly different . Same newsreader looking older . Cars are a littler different . It is the same , yet it is different . I stress all are full spec devices here . Bias is fine if asking .
I feel as my amplifier is so simple it should have no mysteries . The differences are real . Almost like Invasion of the Body Snatchers real . One guess is how back EMF is handled ? I slightly discount impedance differences as this seems not to have a strong correlation . If a journalist it would be 5 or 3 stars differences .
One thing my amp does do is help OK recordings sound a bit more detailed . I value that , it was my belief it would . I don't like euphoniuc additions . I accept a little .
I have no loop nor plate to plate feedback . I have no transistor current source or sinks . The design is modest in it's aims . The speakers suitable to use with it .
Here is the question . No valve measures very differently to the next . Before anyone says I have lousy equipment I do have access to an Audio Precision test set . Results I get and from AP are not too different if a valve design . The question is why do various EL 34's sound different ? To sum it up it is like this . It is like you wake up having slept 5 years . Everything is subtly different . Same newsreader looking older . Cars are a littler different . It is the same , yet it is different . I stress all are full spec devices here . Bias is fine if asking .
I feel as my amplifier is so simple it should have no mysteries . The differences are real . Almost like Invasion of the Body Snatchers real . One guess is how back EMF is handled ? I slightly discount impedance differences as this seems not to have a strong correlation . If a journalist it would be 5 or 3 stars differences .
One thing my amp does do is help OK recordings sound a bit more detailed . I value that , it was my belief it would . I don't like euphoniuc additions . I accept a little .
I have no reason to doubt that they sincerely believe that. Given the flaws in their method, I have every reason to doubt it. Essentially, they are measuring the time domain response of two sets of filters (CD out, sound card in) yet without synchronising clocks. They are doing the inverse of Stuart's digital insertion test: they are inserting an analogue channel into a digital path and then measuring the result. The large differences they initially see expose the flaws in their method. Careful alignment eliminates most of these; much of what is left is just noise.fas42 said:
The exact mix of distortion harmonics and IM depends on bias point and signal level, which in turn depend on details of valve response. With no feedback it would be surprising if there were not small differences in sound when you swap valves. That is partly why feedback was invented: to make circuit behaviour more independent of device details.nigel pearson said:
Back EMF and impedance are just two different ways of talking about the same thing. People who don't realise this, or hope others don't realise this, can bamboozle with 'science'.
As indicated in the document, it is a briefing paper, a commentary on work done to date, a progress report - a quote at the end sets the perspective:
What I'm taking from it is that at least they've had the courage to try and gain some greater understanding of a 'bigger picture'. Whether they're doing more, to get decent meat on the bones, is another question, and there's every possibility it's bogged down into an "it's all too hard!" stalemate - to be just used, as is, for marketing propaganda ...
Last edited:
I feel I took that into account . I would say it still surprises me . I am so pleased the amplifier is almost as predictable as a loop feedback amp on measurements .
One guy advocates putting a 1N4007 diode from UL tap to G2 of the EL 34 to overcome back EMF problems . I was warned off as a friend says it then becomes a voltage doubler . Mine is an SE amp at 450 V . I suspect it would be OK . It must happen in PP versions even without the diode ? What a minefield .
The opposite is true if motorcars . In a darkened street these days I have no idea of the make . Wasn't true in the past . So sickening to have a copycat world . The wind tunnel said NO is I suspect the reason . The little Fiat 500 is easy to spot so it is not required by science to be identical . The 500 measures well . Fiat and Ferrari , great .
To me the implication in the paper is that the differences are reversible - that is, after alignment to minimise differences they make a change to the environment, by changing a cord say - and the results alter. To have reasonable confidence in their results they would need to reverse that change and repeat the experiment, which should cause the first set of results to be replicated. Even though not precisely stated as having happened the tone of the article implies those sort of procedures would have been followed.
- Status
- Not open for further replies.