I had read it before, but nothing jumped out at me. Perhaps I should read it again. My kit here has been modded to the point where I think my speakers speakers are creating more coloration in sound from the tweeter than the amp. So I
I hope so, Andrew.
When I've started this thread I was hoping that you could help too.
As I've told many times I respect you, your opinions and your knowledge.
I repeat myself: If you give me informations with a positive attitude I'm happy to learn something new
IMHO, Andrew the problem it's not what you say but how...
If you don't state clearly in your posts what is the problem you're pointing out, maybe with some hints on how it can be solved, if you don't argument people possibly don't get it.
It happened several times in this thread and in the RevO one.
But when I realized what you were talking about I gave you credit everytime and thanked you for your help.
I'm always open to other people opinions, obviously including yours.
Fine
It seems that there are multiple issues that need to be addressed. From current listening, it seems that the transient characteristics is dominant, meaning how fast the current can onset, and how fast it can stop. The bandwidth shows that it seems fast enough, but how fast can it stop? The damping measurements I have conducted mainly shows how little it will be effected by feedback through speaker motion. The positive thing is that it attenuates feedback where the impedance curve rises. So it will be less sensitive to speaker impedance variations.
The second dominant would be distortion, this would be the way the circuit is biased/compensated. With LM3886 grounded to the power ground, if we take care of how ground current flows, then the bias can be constant such that it will not create interaction inducing additional harmonics. This is one main reason I am in the beta board group, I am interesting in seeing how this will make a difference.
The NS paper talks about output capacitance, which I am not sure how it relates with these issues, or if it just relates with the gain/phase margins.
Another thing that seems unclear is how non-symmetric source impedance will effect performance in measureable form? Distortion?
The second dominant would be distortion, this would be the way the circuit is biased/compensated. With LM3886 grounded to the power ground, if we take care of how ground current flows, then the bias can be constant such that it will not create interaction inducing additional harmonics. This is one main reason I am in the beta board group, I am interesting in seeing how this will make a difference.
The NS paper talks about output capacitance, which I am not sure how it relates with these issues, or if it just relates with the gain/phase margins.
Another thing that seems unclear is how non-symmetric source impedance will effect performance in measureable form? Distortion?
It is one of the most important feature of the My_Ref...
I'm curious too to read your results.
I think the relevant section is Appendix A, where is described how matching affects output impedance.
Probably, maybe Andrew can tell us some more about it.
BTW I think this extreme matching thing is useful to achieve the best possible performance not for the correct behaviour of the amp.
Thousands of working kits around the world testify it
Just a note that may or may not apply here. The perceived quickness of the response of Siva's V1.3 was favorably enhanced by both the Mundorf AG caps and the LFO1 in my listening tests. In fact that was actually the prominent change I heard. If you gurus can implement more of that at the development stage - all the better. With my limited technical understanding, I'm not sure how much if any, of Siva's concept is included in this design.
While waiting for PCBs...
I've tried today on my breadboard version of the FE PS the OnSemi LM3x7TG (TO220) specified in the BOM in place of the NatSemi LMx7L (TO92) I've used so far...
Wow!
They make quite a big difference in timbre (fuller), refinement and soundstaging...
The only (minor) regret is that the TO92 NatSemis were 'snappier' but probably the TO92s require less current so I expect to regain that snappy sound adjusting the 68R resistors to 50R.
Maybe it depends from the package, we will see next week when their NatSemi counterparts will arrive...
I've tried today on my breadboard version of the FE PS the OnSemi LM3x7TG (TO220) specified in the BOM in place of the NatSemi LMx7L (TO92) I've used so far...
Wow!
They make quite a big difference in timbre (fuller), refinement and soundstaging...
The only (minor) regret is that the TO92 NatSemis were 'snappier' but probably the TO92s require less current so I expect to regain that snappy sound adjusting the 68R resistors to 50R.
Maybe it depends from the package, we will see next week when their NatSemi counterparts will arrive...
There are people that like a certain type of sound rather than a more realistic sound. The difference sometimes is confusing. I think I have not been in enough concert halls to know how each would sound, but if the playback can reveal differences in hall characteristics significantly, including studio recording, then I think one can have a better handle on how to control the sonics.
I suppose this is the case, with the OnSemi, after warm-up, sound became more 3D, richer in harmonics and decay.
And I would be asking "why".
Why has the sound changed?
What is causing that effect?
Has something electrically changed that is affecting the sound (waveform)?
Hi Andrew,
reading the datasheets my hypothesis on minimum current seems reasonable.
The LM3x7L (TO92 100mA) seems to need at least 5mA of output current while LM3x7T (TO220 1,5A) seems to need at least 10mA.
With the 68R value a current of circa 18mA is generated so the margin on minimum current is pretty different.
With 50R we can have 25mA, I suppose that it will fill the gap.
Maybe also the different origin (Onsemi vs NatSemi) could have a role since the specs are very similar but not identical.
A more fair comparison will be the one between OnSemi and NatSemi TO220 1.5A LM3x7s.
Last edited:
If the harmonics and decay are higher in both dead studio recordings and live hall recordings, then it may mean less fidelity. However, it may be more to one's liking. This is where it gradually becomes difficult to make decisions.
When we will build the beta boards we will compare our results.
It shouldn't be difficult to source some NatSemi to compare...right?
What I am going to do with the beta boards first is to give it the same +/- 27VDC that I use now, try to keep as close to your LM318 DC power as possible but put the LM3886 C4 mod in, use the original MyRef values in other components and see how they compare. I just increased the cap size parallel to C4 to 3300uF, further increase triggers the power supply protection circuit on the MyRef but not my other amp (yet). Increased differentiation between hall reflections and instruments make instruments stand out more.
Now you are thinking this through.
You have asked yourself "does the change in regulator type, lead to a change in the quality of the regulation?"
That question needs an answer and not by ear ! The ear is the evidence that something has changed.
The next question should be "what type of regulation parameters suit the amplifier?" and the supplementary question could be "what regulation parameters make the amplifier misbehave?"
BTW,
the 317 (5mA) and the 337 (10mA) often have different minimum current for specified performance.
You have asked yourself "does the change in regulator type, lead to a change in the quality of the regulation?"
That question needs an answer and not by ear ! The ear is the evidence that something has changed.
The next question should be "what type of regulation parameters suit the amplifier?" and the supplementary question could be "what regulation parameters make the amplifier misbehave?"
BTW,
the 317 (5mA) and the 337 (10mA) often have different minimum current for specified performance.
Last edited:
Spot-on questions that, sadly, I'm only in part able to answer (remember that my electronics knowledge is very limited, I've asked help also for this reason).
My goals (in part inspired by Mauro's recommendations on a private e-mail) were:
- isolate LM318 from LM3886's PS pollution (mostly current, I suppose)
- isolate LM3886 PS lines from LM318's PS pollution (mostlly voltage, I suppose)
- simple with low count parts
- Shunt type
Since I've got no help for the shunt regulator here I've found it on an italian forum and the FE's LM318 PS is the result.
Any help, suggestion or instructive post about it is welcome.
Yes, I've seen.
Do you think I should expect problems from this difference, it should be compensated and how?
If I've understood correctly the shunt topology should take care of it by design:
The LM318 use the needed current, any excess is shunted by both the transistor and the zener.
This is not perfect, though, since the possible different amount of current shunted would change a bit the output impedance of the regulator.
Any suggestion?
Yes, it's probably so but, nevertheless, I must learn a lot of things about it and any instructional post would be appreciated...
For the Fremen Edition II (one to two years in the future...
) also a completely regulated PS could be interesting. In the meanwhile it would be interesting to find out if a bigger current for the negative rail would improve performance in an audible way.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.