This is well documented. There are SPICE parameters to model this. RBM and IRB.
I answered a question about this previously. The thread should be easy to find.
Best wishes
David.
I recommended Hawksford to answer your last query and Edmond even took the time to post the link.
Was it helpful?
Last edited:
Toshiba 2SA1316 and 2SC1329. Rbb = 2 ohms.
I haven't soldered them in yet but I may stockpile a few more😉
Best wishes
David
Thanks to you & Edmond. Unfortunately, unlike Guru Baxandall & Prof. Cherry, Hawksford is of the 'obfuscate everything so da unwashed masses will be kept in the dark' school.
I've been staring at his paper for the last few days but its difficult cos kunt reed en rite. 😱
Have you got some of the 2sa1316? I discovered futurelec in various countries still had stock of 2sc1329 about a year ago but the PNP seemed to be Unobtainium. This was for an active ribbon mike application.
I've got a cryptic note in my handwriting that though these were the among lowest rbb devices I'd found, my own tests suggested they weren't quite rbb=2R.
The best devices I've tested between 0.5mA & 3mA (in da previous Millenium) were
Hitachi 2sa108x / 2sc545x range rbb=5R
Rohm 2sb737 about 2R and 2sd786 about 4R5
Sadly, all Unobtainium 😡
Last edited:
2SC3329
2SA1316, 2SC3329 I believe is intended.
Around 2004 a friend bought a couple 1k lots of each of those, from which I got about 200 pieces of each. Haven't used them yet either. He said they didn't have particularly low current noise, but he was a bit hazy on what to expect I think (as there is no reason to expect anything special there). According to the databook they have pretty high beta (BL bracket 350-700 at 2mA). rbb' is stated as 2 ohms typical at 1mA. They are not very fast parts compared to comparable ones with otherwise-similar characteristics , and at that only a single number is given for gain-bandwidth.
It would be fascinating to know how construction differs to entail these differences. I remember Scott W. said some conventional design rules were violated to get the low rbb' parts in the AD797.
Zetex (now subsumed under Diodes Inc.) claims some medium-power parts to have low rbb': NPN ZTX650 orZTX690B series, PNP ZTX750 or ZTX790A series. See their Design Note 11 Issue 2 June 1995. Last time I looked they were still making them. They say "typically a few ohms".
For some following this but not well-versed in the details, remember that sufficient collector current must be used to get down to the lowest voltage noise. The shot noise in collector current flowing through the equivalent emitter resistance is one term, mostly independent of rbb' and its variations. Because the shot noise goes up as the square root of current but the re goes down about directly, the voltage noise term goes down with the square root of the current. So it takes some current to make that term comparable or small compared to the contribution of rbb' when it's 2 ohms.
I used to scour transistor manuals but stopped paying that much attention at some point. I was surprised to find the 2SA1316, which until then I didn't know about, in the phono section of an NAD product! I bought a demo unit, the Monitor Series Stereo Preamplifier 1000, for which I've never seen a schematic, and there were a couple inside. A not-great choice for MM because of the parallel noise, and if you happen to have the cartridge disconnected (particularly) in MC position, you can get quite a nice blast of noise! I still hope someday Atkinson and company will test preamps with open or simulated-cartridge-load inputs, as well as his practice of measuring with the input shorted.
2SA1316, 2SC3329 I believe is intended.
Around 2004 a friend bought a couple 1k lots of each of those, from which I got about 200 pieces of each. Haven't used them yet either. He said they didn't have particularly low current noise, but he was a bit hazy on what to expect I think (as there is no reason to expect anything special there). According to the databook they have pretty high beta (BL bracket 350-700 at 2mA). rbb' is stated as 2 ohms typical at 1mA. They are not very fast parts compared to comparable ones with otherwise-similar characteristics , and at that only a single number is given for gain-bandwidth.
It would be fascinating to know how construction differs to entail these differences. I remember Scott W. said some conventional design rules were violated to get the low rbb' parts in the AD797.
Zetex (now subsumed under Diodes Inc.) claims some medium-power parts to have low rbb': NPN ZTX650 orZTX690B series, PNP ZTX750 or ZTX790A series. See their Design Note 11 Issue 2 June 1995. Last time I looked they were still making them. They say "typically a few ohms".
For some following this but not well-versed in the details, remember that sufficient collector current must be used to get down to the lowest voltage noise. The shot noise in collector current flowing through the equivalent emitter resistance is one term, mostly independent of rbb' and its variations. Because the shot noise goes up as the square root of current but the re goes down about directly, the voltage noise term goes down with the square root of the current. So it takes some current to make that term comparable or small compared to the contribution of rbb' when it's 2 ohms.
I used to scour transistor manuals but stopped paying that much attention at some point. I was surprised to find the 2SA1316, which until then I didn't know about, in the phono section of an NAD product! I bought a demo unit, the Monitor Series Stereo Preamplifier 1000, for which I've never seen a schematic, and there were a couple inside. A not-great choice for MM because of the parallel noise, and if you happen to have the cartridge disconnected (particularly) in MC position, you can get quite a nice blast of noise! I still hope someday Atkinson and company will test preamps with open or simulated-cartridge-load inputs, as well as his practice of measuring with the input shorted.
Accuracy my policy, eventually
I'm looking at some output stage enhancements now, and things get messy in a hurry. I can appreciate the simplicity of Malcolm Omar's approach. Of course this is all pretty academic, as there's plenty of loop gain to reduce the output stage distortion as it is, which is running around 0.15% with 50W into 4 ohms resistive in one version I lifted out of another design exercise. I probably should plug in some of those favored DMOS parts instead of nasty ones designed for switching, particularly as the former have a nice ~zero-tempco-of-drain-current bias point at sane quiescent dissipations.
Having just read two subjective reviews in Stereophile of an integrated amp product, solicited by JA because it measured so poorly, lots of distortion, huge RIAA errors on phono --- and with both reviews utterly loving the piece --- I am figuratively shaking my head. But I still pursue clean, quiet, and fast.
Should have said an old Sumo schematic. Rich May showed one to me when discussing HEC circa 1991, when he was my boss at Harman. Rich had been at Sumo.
I'm looking at some output stage enhancements now, and things get messy in a hurry. I can appreciate the simplicity of Malcolm Omar's approach. Of course this is all pretty academic, as there's plenty of loop gain to reduce the output stage distortion as it is, which is running around 0.15% with 50W into 4 ohms resistive in one version I lifted out of another design exercise. I probably should plug in some of those favored DMOS parts instead of nasty ones designed for switching, particularly as the former have a nice ~zero-tempco-of-drain-current bias point at sane quiescent dissipations.
Having just read two subjective reviews in Stereophile of an integrated amp product, solicited by JA because it measured so poorly, lots of distortion, huge RIAA errors on phono --- and with both reviews utterly loving the piece --- I am figuratively shaking my head. But I still pursue clean, quiet, and fast.
Yes.
And there are some helpful tips from another in the Manufacturer's Comments section. Who knew that elevating things with other than myrtlewood blocks could be almost as good?
It's indeed confusing. Fig. 5 most likely shows not the base spreading resistance, but the resistor equivalent noise. This is the sum of spreading resistance noise (bias independent) and the shot noise (decreases with Ic). Hence, the total equivalent noise resistance decreases with Ic.
Very high base impurity concentration (to make up for very low spreading resistance) and very thin base region (to make for beta). The higher the doping gradient and thinner the base, the more difficult to control the process. These devices were likely double ion implanted (base/emitter), which made them very expensive as discretes.
Du.uh! Kunt cont I the 😱
I've tested loadsa medium power devices, including the Zetex ones and so did Great Guru Baxandall. Indeed some have good rbb. But they had to be selected for 1/f & popcorn noise and the yield is poor. Gave up looking once I found the Hitachi devices as these were better than any of the medium power devices and far more consistent.
This was circa 1980 so things may be better in da 21st century. Low noise is a lot about 'clean' manufacture.
The medium power devices had much lower hfe so current noise was poorer too (for common emitter amps).
A better option to replace the Unobtainium devices might be paralleling
2sa970 2sc2240 about rbb=25R
2n4401 2n4403 about 40R ... IIRC
Be careful of RB in SPICE models. It dun usually reflect rbb for noise. The default value is RB=10R. Pity Mouser & Digikey are out of stock of the default LTspice NPN & PNP devices 🙂
This is clearly explained by Baxandall in MEH Chapter 8 (23 meg file) http://www.leonaudio.com.au/cohen.htm
These were his last published words on LN design and it includes an Unobtainium subtext on transformer design. Much more readable than his earlier Wireless World Nov 1968 article. But that explains zillion G LN design so wannabe LN gurus need it too.
There's a copy in my Yahoo MicBuilders Group Files directory under Ricardo. You have to join.
LOL. The psychology of it all!
There are two views to this whole debate:-
View1
Reviewers are a bunch of dirty rotten scoundrels who claim that bad performing amplifiers sound good. And, its proven by JA who measures these things, and often they are TERRIBLE
View2
Your hearing system actually doesn't give a hoot once you get below about 1 % distortion, and that's proven by all those reviews of bad measuring amplifiers that apparently sound wonderful according to the reviewers.
I think it was JA who wrote an article in Stereophile entitled 'Subjective Reality vs Objective Fantasy'
Last edited:
Subjective Fact or Objective Fantasy? | Stereophile.com
is a bit heated, defensive, trots out many of the usual "subjectivist" bloody shirt memes, distortions of the considered positions of those actually familiar with psychoacoustic science and audio engineering, EE theory
is a bit heated, defensive, trots out many of the usual "subjectivist" bloody shirt memes, distortions of the considered positions of those actually familiar with psychoacoustic science and audio engineering, EE theory
The best way to understand it is to trace base currents and to assume that all base currents are equal so if you sink one away from an Ie you have effectively compensated for that transistor's base current. And vice versa.
jan
One of the things that Zetex seemed to be good at was transistors, but no one I know has surpassed the Japanese for consistency and clean processes --- but that's based on observed behavior by many, and we can't get inside top-notch ICs from ADI et al. to know how well they are managing*. I had an "interesting" discussion once with Pease about Japanese quality, which he either found convenient to disbelieve, or simply was furious about their business practices. Bob really hated to be wrong about anything, I discovered, after I followed up on a challenge he made in his column about JFET behavior at tiny drain currents by one of his favorite people to hate, Keats Pullen. My bench results vindicated Pullen's disputed claims, but only with some very old Teledyne Crystalonics parts and Toshiba 2SK364s (same die as 2SK170).
Richard, speaking of Baxandall, though I have yet to see it, Walt Jung said he found a followup in Baxandall's letter to the editor, about his initial Baxandall-Shallow article describing his "superpair", in which LTE he emphasizes the effective collector output capacitance reduction. Perhaps his explanation will be more straightforward of how Boxall's feedback compound transistors work to reduce distortion, than trying to reason it out from the Hawksford slope distortion paper, which I think mostly just deals with the Aldridge cascade circuit (the bootstrapped cascode that Hawksford thought he invented).
As far as I know, no one has as yet done a comprehensive frequency-dependent analysis of the Boxall. It shouldn't really be all that difficult, starting with one that omits parasitic inductances, then adds them in later.
EDIT
*from overall performance one supposes quite well!
Last edited:
Thanks for that. I think I will make this a minor pet project to hunt down what may have been published.
Someone at THAT told me that they had tried to make higher-beta versions of their arrays but the yields were terrible, with a lot of breakdown voltage issues. This was candid and not spoken of as confidential, although accompanied by the line that beta didn't matter much for most of the intended applications, particularly involving log conformance.
Precisely, and note that at high frequencies the most important component is the "displacement current" due to Ccb, which is variable with voltage and hence a distortion mechanism.
Another benefit is that the noise and drift in the main Q base current is recycled, so that the noise and drift in it is also reduced accordingly, leaving you with the auxiliary transistor's base current noise and drift (but the disadvantage of the voltage noise at the main emitter of two parts in series). There is a crude temperature compensation due to the opposing polarity parts as well, although the desire to use a smaller device at lower currents for the auxiliary transistor makes trying to match the curvatures (as speculated on by Richard) tricky, and one also wants the impedance in the main emitter to be high so that the recycling is almost 100% effective. That is why used as a compound cascading device the pair is really effective. It's also possible to contrive a circuit with slightly more than unity current gain, but one courts disaster in a hurry.
And another bother is the need for the extra bias current for the aux transistor, although in many cases that can be just a resistor.
Common-base stages such as the auxiliary transistor have power gain if voltage gain can be developed at the collector. Inductance in the collector may provide this needed impedance. It's no surprise that the stage can be difficult to stabilize, as the mechanism of correction is positive feedback. There is a thread in here about it somewhere, which has "Baxandall" in the subject line. Unfortunately there wasn't anything quite definitive, iirc.
Last edited:
There is a third 'View' except it isn't a view but experimental fact ..
Most HiFi Reviewers are Deaf. The average HiFi reviewer does much worse in Double Blind Listening Tests bla bla than the 'man in the street' whose performance is slightly worse than the 'woman in the street'. I've quite a lot of reliable data on this.
There are (very few) notable exceptions. JA is on my very select list of true golden pinnae. J Gordon Holt said 'High End' went down the tubes when reviewers refused to conduct Double Blind Listening Tests bla bla.
I design stuff for those who can tell the difference .. true golden pinnae and the man & woman in the street. For HiFi reviewers, I point out my stuff is hand carved from solid Unobtainium by Yorkshire virgins. 🙂
This is corroborated as well by people I know. If they were to publicize it more they would probably get less cooperation from reviewers.
I too have found it amusing since long time ago that JA for many times wrote in his speaker performance measurement result that goes as part of a speaker review about his fellow reviewer, something like "this strong resonant mode might color the mid range tone with a nasal quality and contribute a congested sound when excited, however, xxx never complained in the subjective review..... I can only conclude this resonance appears much worse than it sounds......" 🙂