Nigel, this has been thought out thoroughly for the last 50 years. IF you increase the current in the differential pair too much, you get increased current noise contribution, and it is not quiet unless you drive with 50 ohms or less. That is one good reason to not go above 2ma with bipolar transistors. Also, bias current goes up, almost proportionally with increased current. However slew rate will remain the SAME if the gain-bandwidth of the amp remains the same, because the increase in differential stage current increases the Gm proportionally, and then your compensation cap increases proportionally. Only degeneration using emitter resistors or increasing the gain bandwidth of the amp will get you increased slew rate. That is the 'KEY' in Solomon's paper. Unfortunately, emitter resistors WILL add input noise. So, very low noise is out.
The rational alternative is the low noise jfet pair, with inherently lower Gm, so you don't need any source degeneration resistors, and immunity from current noise. NOW you can run any source current you want, mostly, because the Gm does NOT track the current, so you are not 'chasing your tail' so to speak. I hope this helps.
In any case, please look at successful working examples of amps and preamps and note that they are often pretty much the same, as far as operating current is concerned.
The rational alternative is the low noise jfet pair, with inherently lower Gm, so you don't need any source degeneration resistors, and immunity from current noise. NOW you can run any source current you want, mostly, because the Gm does NOT track the current, so you are not 'chasing your tail' so to speak. I hope this helps.
In any case, please look at successful working examples of amps and preamps and note that they are often pretty much the same, as far as operating current is concerned.
+10 ,
Basic skills are lost due to such , best to start with basics before moving on to technology..
Thank is for that . Good explanation . I had supposed that gain of 1 was a priority and determined the whole reason d'etre of the design , maybe so ? NE5534 gives us the choice with external compensation option . I sometimes wonder if 5534 is disliked because people are so use to other op amps being unity gain stable that when asked to tweak they don't want to to ?
bjt diff pair noise with higher bias isn't a major issue in power amp front ends where we are expecting line level signal V - today we expect low Z drive at up to 2 Vrms from digital sources/SS preamps - can use lower Z volume pot - or can skip the volume pot for digital
where input current noise isn't limiting then we can use higher bias, degeneration R to make bjt front ends more linear than jfets: http://www.diyaudio.com/forums/soli...ning-not-whats-your-belief-64.html#post501789
where input current noise isn't limiting then we can use higher bias, degeneration R to make bjt front ends more linear than jfets: http://www.diyaudio.com/forums/soli...ning-not-whats-your-belief-64.html#post501789
Last edited:
I didn't succumb to simulators until I was nearly 50. I think I waited too long. I did use calculators fairly early on, although the first I acquired was an HP-25, a machine I still miss although it didn't run all that long on the battery pack.
When Brad Plunkett found out I was using simulators he told me he was very disappointed, mostly joking. When he finally began using them a bit, only a few years ago, he presented a circuit to me that he'd come up with, having "unbelievable" performance. It was a good idea nevertheless, and was built, incorporated into a design that very much needed it, and went into mass production.
I pointed out that the shortfall between its performance, which was wholly adequate to the task at hand, and what the simulator predicted was mostly due to his use of a default transistor model, and that he could pull down a list of real parts including a 2N3904 and greatly improve the accuracy of his results.
I then THOUGHT about the mechanisms that still limited the performance and conjectured ways to correct for them. This led to a series of additional designs, verfied by simulation AND construction.
Simulators must not become a substitute for thought, although there are factions within the communities that advocate things like "genetic programming", where actual topologies are created "at random" and then "prove" their robustness in a sort of Darwinian fashion. It's somewhat reminiscent of quality assurance people who believe various methodologies can replace people with real insight, experience, and expertise with a lot of duller or at least uninformed folks doing experiments and using statistics. Pease of course loved to skewer the more egregiously silly examples, like the voltage regulator that worked better after enhancement via Taguchi approaches because it eliminated line regulation, something that hadn't been included among the measurables.
But one does need to be open to about everything. It's easy for an old guard to get complacent, nay even defensive. When Planck reluctantly presented his empirical formula for solving the "ultraviolet catastrophe" a number of then-traditional physicists said they could fix things without such a preposterous notion by using various "devices" with which they felt comfortable (Nernst in particular had some ideas which seem absurdly hidebound to us today, but must have seemed quite plausible at the time).
Despite its antiquity the 5534 remains an amazingly good part. Besides the external compensation access, those same terminals allow the use of alternative input stages, including JFETs. Pretty soon the added parts may rather dwarf the chip that was to be so convenient and space-saving, and one can be pushed closer to an all-discrete design, but such is life.
"In Plato's Phaedrus there is the story of king Thamus who reigned over a city in upper Egypt. He had a friend Thot who was an inventer, amonst others inventing script. After Thot has exalted all the virtues of writibg ("This is an invention that will make Egyptians wiser and stronger - my creation is beneficial for memory wisdom").
Thamus: This will bring forgetfullness in the soul of those who learned because they no longer exercise memory, and by relying on written texts, they will not find wisdom from within but from strange symbols from outside. Your pupils will only find the appearance of wisdom, not wisdom itself.They will read much without learning, but will be awkward to deal with because their so called wisdom is only imagined."
Apologies for the slight OT, but I was just reading Neil Postman's "Technopolie". L'histoire se repète, as the French say.
jan
Yet I think in the end he was wrong. One central theme in Postman's writing(s) is that technological developments not just improve the way we do things. Cell phones not just make it easier to stay in touch, they actually change our culture, the way we as a society communicate and live together. Cars not just make it easier to get around, they have fundamentally changed our culture.
In Naples I've seen streets where the young gather in their cars at night, and they literally past up the windows with newspapers and that is their world where they talk, fight, court each other and make the occasional baby. The car, in their society, radically changed the way they court and select mates and prepare for family life.
And take TV, another game changer if there ever was one, way beyond just being a medium to bring programs in the living room.
Simulators are not just an easier way to 'test' circuits, they changed the game. People with no real understanding of circuits can now design a circuit that on the face of it works quite well and might even do so in real life. Question is, where is the harm in that - pursuit of happiness and all that.
BTW Postman hates TV, you should read his thought-provoking "Amusing ourselves to death".
jan
Last edited:
Even Self, one of the 5534 stalwarts if ever there was one, has acknowledged the 4562 as an improvement.
A tube of 25 LME49710 in DIP arrived yesterday. $1.29 from the distributor in that quantity.
The reason to get them: I lack a reliable model for the part, and want to see how well a modified linuxguru-like buffer, or a buffer with a little gain to increase the output swing, will work if inside the loop. I used the dual parts in SM in the front end of a switchmode power amp and got very good results, so I don't doubt the part itself.
About the only thing one could wish from the 49710 is a voltage swing a bit closer to the rails. Of course it is a bipolar input part, and has as well a little more current noise than the 5534.
Jan, 26 dB is a heueristic value for me. The best amps I have ever heard were almost all around that value, give or take 3 dB, with the only exception of H/K models, which range from 12 dB (HK 680) to 25 dB (usually limited to their cheapest model).
However, there were a few exceptions, for example, one being the LAS power amp I highlighted here some time ago, which is a total turnabout, it has VERY little local degeneration, it's almost 100% turned towards global NFB, yet manages to sound wonderful.
As for the top limit, to be frank, I never investigated that point, I never felt the urge to. I suppose there should be no problem, but I only assume this. The reason is that I find that if an amp has been done well, NGB over 26 dB is not likely to bring any notable improvements, except on paper. And I don't listen to paper.
So, in conclusion, I feel the amp simply should not need more than 26 dB to achieve what it can do. Other than that personal feeling, no other reason.
Personally, I find that a standard bipolar transistor pair in a differential circuit will usually do what it can by the time your PER TRANSISTOR bias value gets to 1 ... 1.4 mA, or 2 ... 2.8 mA for both.
However, I must add that I generally use up to 2k collector load resistors, typically 1.5 or 1.8 k. For a reason I cannot explain, I have a slew of second thoughts about more than that, my conscience (if you believe I still have one at this age
) bothers me when I go above 2k. FETs are different matter, In my view, the only FET I ever used to any measure is the venerable old 2SK170, and in my experience, it does really well when the total bias current (for both in a pair) is around 3.65 mA, at a voltage of 8.2, 9.1 or 10 V. Always in a cascode configuration, typically BC546B, 2N5551 or BF422 above them.
You can imagine my surprise when I saw the schematics of the H/K 870 power amp from 1985, which also used exactly the same value. Well, even a blind chicken gets the seed every now and then.
JCX, I am not going to argue the point, I'll take your words at face value.
However, my ears say otherwise. Not always, or by default, of course, but I find it's possible to have just 17 dB of global NFB and literally GREAT sound, as demonstrated by my H/K 6550 (and that's a SEPP design, too).
Just goes to show that despite the fact that we have learnt a hell of a lot since the wild 70ies, we still don't know it all. Far from it.
I can't speak for Nige, but it sure has made me rethink some things I thought were done and over with.
I have always agreed with the Italian designer Bartolomeo Aloia, a man with impeccable credentials and numerous projects ranging from simple ans cheap to complex but damn ood, who said one that if you want lots of current at the output, you have to have decent current at the input.
It seems I have been going a bit overboard with that thought, meaning that I have probably overbiased some differential stages to a suboptimal level. Kind of shot myself in the foot.
I just realized that from your comment above. To me, it makes operfect sense, and why I didn't see it myself I really don't know. Now that you said it, it seems so perfectly obvious I feel like banging my head against a wall for not seeing it myself.
On the subject of noise - while I agree with what you said, I would still point out that a bit of that added noise could be done away with if you use electronically regulated power supplies, which are assumed to be quite a lot cleaner than unregulated ones (at least, they should be).
One more trade-off to take into account.
I second Jan's comment - a truly GREAT post.
It made me think back to a text I read sometime in the past, somewhere, and was shortsighted enough not to save it, from somebody in Harman/Kardon, who more or less said the same thing. He claimed that a classic bipolar differential pair should have an overall gain of no more than 11:1, and no less than 5:1, and the in-between values would show up the bipoars at their very best, assuming the design has no other flaws, such as a poorly judged bias current.
H/K themselves did 5:1 a few times - for example, the input bipolar differential stage of my 6550 integrated amp has 2x220 for degeneration, and 2x1.2k for load. In the 870 power amp, it's 2x220 for degeneration, and 2x1.5k for load, and that for a FET (2SK170) cascode (2SC2240) input stage. And in their top model from 1998, HK680, it's 2x100 for degeneration and 2x1k for load in a straight 2SC2240 differential stage (bias 2x1.2 mA).
I think not.
The marketing gimmick of "zero feedback" simply meant no overt GLOBAL NFB, but it also probably means that the feedback has all been relegated to individual stages. In other words, it has been made all local.
I have heard a power amp which claimed that, from a Danish manufacturer, Dennsen. While it no doubt worked, it left me with an impression of a loose sound, not quite all together now. The bass seemed a bit bloated, not too much so, but definitely lacking some control and definition.
All very subjective, of course. I can't help it if I banged on a drum kit in my wild and wanton youth, so I instinctively listen to the bass lines first. The friend I did the banging with, while he strummed a classic guitar was much luckier than I was, he went to to eventually become one of the greatest local rock legends of out time, with a career from 1976 until this day, and still going strong. The price he paid for it is that he's now an alcoholic, while I drink a glass of something on special occasions only (birthdays, new year, etc).
"much higher feedback is needed to have the new harmonics all fall with increasing order as we are told is necessary"
One of the things I noticed ( about 6000 posts ago) was the Rotel designs did this where the DH-120 when stock did not. After the mods, it did behave as suggested.
I have not seen this mentioned before, and would appreciate a source so I can read some more. Does it matter what the initial level of distortion is, or just the general profile of harmonics? What is the characteristic sound of one or the other?
One of the things I noticed ( about 6000 posts ago) was the Rotel designs did this where the DH-120 when stock did not. After the mods, it did behave as suggested.
I have not seen this mentioned before, and would appreciate a source so I can read some more. Does it matter what the initial level of distortion is, or just the general profile of harmonics? What is the characteristic sound of one or the other?
It might be . I recently built a zero feedback amplifier which has 0.03 % THD at 0.63W . It was better than that down to 50 uW . It used the opposite distortion of a pentode and triode . Pre distrotion I suppose . If all triode it had considerably more distortion ( 5 % full power ) . The amplifier made the Old DIN4550 just . If I remember correctly 0.1% THD at 1.6 W . It doesn't seem much to shout about , however for a zero feedback SE design it is very good . 2 valves and 0.56V in for 8V out .
I am not comfortable that it is a low distortion design . It looks like low distortion . The sound is rather good . I notice multi-tracked voices like never before , especailly if the same voice . I have been offered some 211 valves to see if a few more watts can be had . Bass is not the greatest . I hope to try it with Quad ESL 63's soon as they do not require high damping factor . I certainly feel the all triode version was inferior . Along the way I tried Schade feedback , I have doubts about it . I would liken the effect of my design to that of Ultra Linear feedback although this is not feedback . Have I become a valve convert ? No , if anything the other way . I certainly have enjoyed doing it .
- Status
- Not open for further replies.