Yes. You may remember that I queried this a while back, even before this thread.Unfair!...my Holy methods on evil TMC
OK, I confess. This is probably useful in any amp with a VAS emitter resistor.
Mike does have some point. A small VAS emitter resistor does not alter the VAS behaviour in the way sometimes claimed.
And the added capacitor in Toni's TMC variant also does not do the obvious.
The phase is improved but not by an added zero, instead it seems to help neutralise an RHP zero. This is a bit unexpected.
I made my earlier enquiry because I was uncertain about these interactions and my doubt seems to have been justified. It needs more study.
Best wishes
David
Yes. But not for his naive reasons.Mike does have some point. A small VAS emitter resistor does not alter the VAS behaviour in the way sometimes claimed.
And the added capacitor in Toni's TMC variant also does not do the obvious.
The phase is improved but not by an added zero, instead it seems to help neutralise an RHP zero.
When Bob posts his EKV FET models, I'm going to do a little series in my http://www.diyaudio.com/forums/solid-state/235188-tpc-vs-tmc-vs-pure-cherry.html thread showing where things are 'exactly' as Cherry predicted and where I've been taking liberties.
I need the EKV models cos I want to show good performance vaguely similar to what I've actually built. ie something that someone might actually want to build.
As I said, enhanced VAS + EF2 is 4 devices in the forward path so stability is ALWAYS conditional. [deleted: pedalogical rant on 'conditional' & 'stability']
Ever since I did my 'linear' circuit analyzer (80's) I've stopped talking about poles & zeros except to teach. I look at all these dodges as 'kinks in Nyquist which may or may not improve stability.'
BTW, the 'Bode step' is just a kink in a specific place. I doubt if Hendrik would have considered it as anything other than 'just another kink'. Him, Nyquist & Harold Black were all mates.
_____________
I hope Toni, tests my latest tweak to the Cherry version of his amp. If it works, I'll go into pontificating verbal diarrhea mode.
Of course I will do another "cherry" test. But may take some days...
My sim with disabled output zobels shows you have now achieved about 35 degree PM and 12 dB GM.
I have added 2x28p caps to reflect the long ribbon cable capacitance.
Please check attached sim if your tweaks are correctly inserted.
BR, Toni
My sim with disabled output zobels shows you have now achieved about 35 degree PM and 12 dB GM.
I have added 2x28p caps to reflect the long ribbon cable capacitance.
Please check attached sim if your tweaks are correctly inserted.
BR, Toni
Attachments
Last edited:
A quick look at Toni's plot for "Cherry" does not show conditional stability, in the Nyquist sense that the term is usually employed.
I doubt if you are correct. The whole 500 pp. book is pretty much a build up to solve the maximum feedback problem. Do you have some special information about Bode's opinions?
Best wishes
David
Oh, really? Pray, demonstrate how it is possible to have a stable output stage inclusive Miller compensation loop without having capacitive shunt compensation at the output of the second stage.
Err.rrh! It may have escaped your notice that this thread presents 'real life' evidence of this amazing factPray, demonstrate how it is possible to have a stable output stage inclusive Miller compensation loop without having capacitive shunt compensation at the output of the second stage.
Most recently in #519 & 521. There are remaining artifacts similar to those seen in #233 but again the 'real life' evidence suggests these are due to 'parasitics' [1] rather than naive interpretations of PM, GM etc. Rest assured we will not rest until these have been laid to rest .. and it will be done without evil 'shunt compensation at the output of the stage'.[2]
There's also the 'real life' evidence in my tpc-vs-tmc-vs-pure-cherry.html thread but I quite understand if you discount this. They are after all, only Jurassic claims by a beach bum.
You'll note the Cherry versions of Toni's amp also have rather good 'real life' THD performance which we are hoping to improve further.
In contrast, some pseudo gurus have posted circuits which, even in SPICE world, are sadly lacking in stability & other performance parameters. I'm not sure further study of these or their pontificating is of practical use.
__________________
[1] .. for da pedants. This is strictly untrue as ALL stability issues are interrelated. For da 200+ pgs of pontificating on this subject, please search this website for 'parasitics'.
[2] Again strictly impossible. BJTs etc have evil bits with fancy names like Cob, Ccb etc which all add capacitance to the output of a VAS. And there are 'parasitics' like Toni's ribbon cable too.
But the pseudo gurus who advocate even more evil capacitance and other bits on the VAS are only proving their poor understanding of stability and how to achieve it. Their naive efforts usually only increase THD by giarnomous amounts.
But often, their circuits also have poor stability on real loads ... which they have missed in their obsession with 'linear' SPICE measures like GM, PM bla bla. Some of these pseuds have rather low standards re stability & THD performance .. so this isn't too surprising. And 'real life' seems quite irrelevant to them
Last edited:
I did put 'conditional' & 'stability' in quotes in my 200 pg rant
Only that he was a practical man besides being a headbanging mathematician.
How else would you describe a 'Bode step' but as a fudge/tweak/dodge. You can always do the maths and pontificate on poles & zeros after you've got it to work.
I'm still hoping to see your approximation to his mythical (?!) max feedback network. Probably easiest to show it on an evil TPC amp.
BTW, I'm impressed you still have a copy. In da old days, I even pre10ded 2 unnerstan his maths but those days are long gone
Last edited:
Has anyone noticed these unbelievably amazing perfect driver transistors available on Mouser lately?
2SC4027T
2SA1552T
http://www.mouser.com/ds/2/308/EN2262D-D-96030.pdf
Maybe we don't need the C5171/A1930 anymore? I don't know when the long-lead packages will be available.
2SC4027T
2SA1552T
http://www.mouser.com/ds/2/308/EN2262D-D-96030.pdf
Maybe we don't need the C5171/A1930 anymore? I don't know when the long-lead packages will be available.
Yes. They were found while looking for SMT medium power devices. Very good specs although SOA doesn't look amazing.
Thanks for bringing these to our attention. Sanyo had a plethora of awesome BJTs that were severely culled in the last few years
Fortunately these ones seem to have survived and ON Semi (who bought Sanyo Semiconductor) are continuing to make them.
They have lower Cob and better SOA than the 2SA1930/2SC5171, but the NPN part has substantial variation of beta versus Ic. The 1930/5171 are both ruler flat in this regard (i.e., perfect).
These Sanyo/ON Semi parts are very similar to the Fairchild KSA1220A/KSC2690A pair. The Fairchild parts have higher Cob but better SOA.
Those are just 2sa1507/2sc3902 in a different package. They have been available for some 30 years. Rotel used these in many of their amps and they are still in use by Accuphase. Onsemi are still producing these too.
While they seem nearly identical, the curves aren't the same so I don't think they use the same die. Also those are 10W, the C4027/A1552 are 15W. They do appear to share a die with the C4614/A1770.
http://www.mouser.com/ds/2/308/EN2101-D-117693.pdf
The C3902/A1507 seem only slightly more useful above 40V. Hmm. The C4793/A1837 have twice the SOA above 40V. However both the Sanyo pairs beat the C5171/A1930 in SOA, and beat the C4793/A1837 in Hfe knee current. So in driving over 100mA, the Sanyos might have less distortion.
In any case, the 4027/1552 are the ones available on Mouser right now. There are some other Sanyo BJTs available, but I couldn't find a way to list them all.
I actually read the curves wrong the first time. I thought they were linear to 5A, not 500mA! Oh well. I knew that didn't make sense.
... and here the result with added 10R as explained ...
Stability betters significant but is not enough for an amplifier to go into production! There is still tendency for oscillation. (See attached picture of 20kHz square wave test using long ribbon cable. Short ribbon cable shows a bit more oscillation).
Distortion level compared to TPC/TMC not so bad but not better as predicted.
long ribbon cable:
THD21k@200W@8R
0.00186% (bw 80 kHz)
0.00229% (no bw limit)
short ribbon cable:0.00186% (bw 80 kHz)
0.00229% (no bw limit)
THD21k@200W@8R
0.00197% (bw 80 kHz)
0.00239% (no bw limit)
==> THD betters with longer cable!0.00197% (bw 80 kHz)
0.00239% (no bw limit)
Attachments
Great! Does anybody know if there are good ltspice models for these devices available?
BR, Toni
I doubt it!
Do you have Bob Cordell's book? It has a useful introduction to building spice models, but there are some extra things to bear in mind. See post #60 in the finding spice models thread. I assume you have already downloaded my zip file and therefore have the spreadsheets. If not the link to the zip is in that post.
For extracting data-sheet information, check out Engauge Digitizer.
Last edited:
Did you see my earlier post? (#551)
One thing that the data-sheets don't tell you is the early voltage. It's surprisingly low for the 2SA1930. Perhaps the Sanyo parts are better?
Err.rrh! It may have escaped your notice that this thread presents 'real life' evidence of this amazing fact
Most recently in #519 & 521. There are remaining artifacts similar to those seen in #233 but again the 'real life' evidence suggests these are due to 'parasitics' [1] rather than naive interpretations of PM, GM etc. Rest assured we will not rest until these have been laid to rest .. and it will be done without evil 'shunt compensation at the output of the stage'.[2]
There's also the 'real life' evidence in my tpc-vs-tmc-vs-pure-cherry.html thread but I quite understand if you discount this. They are after all, only Jurassic claims by a beach bum.
You'll note the Cherry versions of Toni's amp also have rather good 'real life' THD performance which we are hoping to improve further.
In contrast, some pseudo gurus have posted circuits which, even in SPICE world, are sadly lacking in stability & other performance parameters. I'm not sure further study of these or their pontificating is of practical use.
__________________
[1] .. for da pedants. This is strictly untrue as ALL stability issues are interrelated. For da 200+ pgs of pontificating on this subject, please search this website for 'parasitics'.
[2] Again strictly impossible. BJTs etc have evil bits with fancy names like Cob, Ccb etc which all add capacitance to the output of a VAS. And there are 'parasitics' like Toni's ribbon cable too.
But the pseudo gurus who advocate even more evil capacitance and other bits on the VAS are only proving their poor understanding of stability and how to achieve it. Their naive efforts usually only increase THD by giarnomous amounts.
But often, their circuits also have poor stability on real loads ... which they have missed in their obsession with 'linear' SPICE measures like GM, PM bla bla. Some of these pseuds have rather low standards re stability & THD performance .. so this isn't too surprising. And 'real life' seems quite irrelevant to them
Your naiveté and ignorance of elementary analogue electronics is simply shocking.