Groveling is right out, and I'm fresh out of pearls. But washing is encouraged.
I like to pay attention to thermals as they are usually neglected, and are not straightforward to model with most off-the-shelf simulators. In many situations the effects are small enough to be neglected, but in others they can be spoilers, although typically at performance levels that may have little audibility. When they can be almost eliminated and without compromising other things (except maybe low parts count) I like to do it. In this case the strategies for one thing, lower parallel "current" noise, play nicely into ones for reducing thermals.
It bothered me that Ramus in the TI piece says that "the inverting input current noise is always much higher than VFB equivalents". Although this seems intuitively plausible --- certainly there's more current flowing in those emitters --- I'm not fully persuaded that it is a fundamental limitation for bipolar CFA versus bipolar VFA. I think the issue is obscured by other circuit details. So I've been looking at the problem and came up with some approaches that seem to pan out.
I'm not using VSSA but my #500 offering. The DC conditions are sorta copied from VSSA though.
I'm still only halfway through the huge VSSA thread but my impression is that DC output stability is good.
Not too sure about VAS current stability which is why I've asked about the setup procedure several times.
I'm hampered by never having burnt solder on anything similar ... while I've been playing with Blameless types for 30+ yrs so have a 'real life' handle on most VFA features.
But even with these caveats, I think I can take a Diamond IPS CFA and improve its THD, at least in SPICE world by reverting to simple.
Can you suggest a suitable candidate to try? Not too complex please ... else my small brain locks up. And BJT rather than MOSFET OPS unless you have EKV SPICE models for them.
I'm not being confrontational .. just want to try stuff & learn. Only 30+ yrs to catch up. I only worsen my naive CFA efforts when I convert to Diamond IPS but that's probably cos I'm incompetent.
The reason IC manufacturers don't use 'simple' is the one Bonsai brought up and I acknowledged. DC coupling.
_______________
This is my best CFA effort. Don't laugh please. It has yucky compensation cos I still dunno how to do a proper Tian probe with these 'symmetrical' amps.
The 'fair' comparison is with my #502 circuit. 13 devices. Same OPS. Same IPS currents. I'll claim #502 is an 'optimised' VFA within these constraints. If anyone can do a better VFA here, please speak up.
The CFA is certainly not optimised as its by a CFA newby but still appears to be better all round technically.
THD is mainly 2nd and 'flat' with freq. which suggests PNP/NPN matching between the 2 sides sets the limit. Dunno what to do about that without a lot more complexity.
Loadsa stability stuff still to do.
I'm still playing with Triples and 'advanced' compensation inspired by Bonsai & others but without much success.
________________
Brad, I've resolved to wash at least twice a day. 🙂bcarso said:
Attachments
I was surprised when my MF80 project measured better than it simmed haha. It did 10ppm @ 10K. Na I'm not expecting anything special but given my previous experience with simmed/built amps it should end up more than adequate.
Did you post your VAS in my VAS thread? If it runs circles around the THD I achieve I'd be interested to learn more. So yeah I'm looking forward to your sim 🙂
So untrue.
Output bias current is fixed to 160 mA from cold start to hot condition, no variations at all.
VAS bias current varies from 12-16 mA from cold start to hot condition.
Output DC offset is in a range 0 +/-1 mV max.
No power on/off bump, click, ... whatsoever.
What members mentioned is trimmer rotation sensitivity since these are 5 turns only SMD Bourns trimmers. But there is no problem to calibrate VSSA properly according installation instructions, I can do it in 60 s.
I did. A TIM distortions reduction benefit related to audio is nowhere to be found. TIM for DSL signal conditioning applications (where frequencies (read: input signal slopes) are 2 orders of magnitude higher) is in a completely different league. Again, there is no problem to build either a VFA or a CFA with TIM distortions buried in the noise ground.
Really, since when?
You don't seem to understand how TIM distortions are generated, see Walt Jung articles.
Well, my last day at work is this Friday. Yes, retiring from the Day Job (undersea fiber optic communications). I should then have more time to brush off the dust from the EKV models and see if I can get them out there. Of course, my wife Angela is busily preparing a "Honey-do" list that is the length of her arm, now that I will have all of this time on my hands.
Cheers,
Bob
Bob,
Those honey due lists are usually just a way to keep us out of their hair and hopefully if you can make her think you now have other things to pursue she will give you some time for yourself! It would be real nice to see another book from you and I would look forward to your contribution to the science. This cfa amplifier section could use a real working over that could answer many questions I imagine that is longer than your other arm. Also a real few chapters on how to use the monolithic power amp chips in bridged and parallel applications for stable high powered amplifiers would be appreciated. Good luck on the retirement, I am sure you deserve it.
Steven
Those honey due lists are usually just a way to keep us out of their hair and hopefully if you can make her think you now have other things to pursue she will give you some time for yourself! It would be real nice to see another book from you and I would look forward to your contribution to the science. This cfa amplifier section could use a real working over that could answer many questions I imagine that is longer than your other arm. Also a real few chapters on how to use the monolithic power amp chips in bridged and parallel applications for stable high powered amplifiers would be appreciated. Good luck on the retirement, I am sure you deserve it.
Steven
I wish op amps were class A, but, alas, most are not. Their output stages are class AB in most of them. Of course, the results of this depend on how much current they must deliver into the load.
Cheers,
Bob
Congratulations! And I'm sure you will not suffer from a lack of things to do to keep your mind active.
Brad Wood
What I would like to see from someone (and as Mr. Natural says to Flaky Foont, not me) is a comprehensive book on switchmode amplifiers. However, I fear it would require a few volumes at least to cover all of the material to date, and probably swiftly become out-of-date upon publication.
The topic is also a veritable mine field of controversy, particularly in terms of the priorities of various inventions. One could be assured of at least some negative reviews by various partisans, no matter how carefully one did the research. I recall one of the rare times I attended an AES convention, in 1997, and in one paper session I was concerned that there might be physical altercations during the question period after some of the presentations. There seems to be something about amplifiers, and particularly class D etc. ones, that fosters belligerence 😱
Brad,
I have looked at the N-core Amplifiers and the cost is just to much to use them commercially as there is a determined effort not to do that it seems. Trying at this time to design one with the limited information I think as far as I am concerned would be in vain. Today I see the monolithic chips as the only viable alternative given the ever smaller availability of discrete devices and the costs of building that type of amplifier.
Watching a bunch of EE's at an AES conference getting that belligerent would have been something to see. It only shows how fractured the professional side has become. When I look at the claims for amplifier output that most pro-audio amplifiers are claiming into as low as 2 ohms it is impossible to think of anything but a switch-mode amplifier being able to produce those numbers. Almost seems like they produce more power than is coming from the AC line input side!
I have looked at the N-core Amplifiers and the cost is just to much to use them commercially as there is a determined effort not to do that it seems. Trying at this time to design one with the limited information I think as far as I am concerned would be in vain. Today I see the monolithic chips as the only viable alternative given the ever smaller availability of discrete devices and the costs of building that type of amplifier.
Watching a bunch of EE's at an AES conference getting that belligerent would have been something to see. It only shows how fractured the professional side has become. When I look at the claims for amplifier output that most pro-audio amplifiers are claiming into as low as 2 ohms it is impossible to think of anything but a switch-mode amplifier being able to produce those numbers. Almost seems like they produce more power than is coming from the AC line input side!
If you look at how effective and efficient some sealed-box overmotored woofers are with lots of bass boost and driven from really efficient amps (that return reactive energy efficiently back to the power supply), the argument for switchmode is pretty strong.
I'm not interested in them for sound quality in the midrange though, although they have improved enormously. But if you really have to move a lot of air...
The problem i think, to some extent, with various rivalries is that a lot of people went to the same school in Denmark, and everyone was pretty much getting along over beers. Then one of them decided to launch a commercial venture, and the accusations of stealing ideas began to fly. I don't think it has stopped.
One of the people who could write a definitive book on switchmode is Gerald Stanley. But I gather he's not predisposed to writing very much, and they still keep him quite busy. Although I haven't auditioned it, I'm still puzzled at what reviewers didn't like about the recent big Mark Levinson amp, which employs Gerald's invention of two opposed buck converters and both avoids shoot-through currents that plague conventional half-bridges, and facilitates interleaving for additional power and lowering output high-frequency ripple. It does however use a minimum of two high-quality inductors.
As far as low cost, Philips (who still are the patent holders I think despite spinning off NXP) supposedly don't care if you use Bruno Putzeys' UcD invention, provided you use NXP semiconductors. And that design was developed primarily to see what would be the cheapest switchmode topology! Of course Hypex will be pleased to sell you their version too.
Brad,
There is the rub, I am working on a full range system and the agreement so far from what I have followed is that the class-D amplifiers are still not the answer in full range systems. If I was only looking for a solution to a sub-woofer application or even the bass section only I agree that this looks like the answer. Since we are so close to each other perhaps I can invite you to lunch some day and show you what it is that I am attempting to achieve.
Steven
There is the rub, I am working on a full range system and the agreement so far from what I have followed is that the class-D amplifiers are still not the answer in full range systems. If I was only looking for a solution to a sub-woofer application or even the bass section only I agree that this looks like the answer. Since we are so close to each other perhaps I can invite you to lunch some day and show you what it is that I am attempting to achieve.
Steven
Not too yucky, dont be afraid, your example has good performance, you need to temp compensate the 2 input transistors, can be done via the CCS. Its these that cause bias instabilty in the vas which in turn affect the outputstage. Take a look at audiolabor schematics, here a different method was used to stabilize bias, the same way JLH first did in the 60s. These amps were highly acclaimed by the press in the 1980s. See how elegantly the diamond takes care of this problem. The posts about the diamond sounding worse is pure speculation and fiction, complexity is the same. Instead of 2 actives for CCS use them for the diamond. In the diamond one can use diodes instead of the current setting resistors and doing so you can get rid of the need for servo and have offsets close to that of VFA. This will affect bandwith negatively but in a power amp this is not a issue as the outputstage already does that.
I notice you used miller comp, I still stick to shunt as being better for CFA topology, you can obtain higher loop gain at high frequency. Also using shunt youll notice a Triple EF works optimal with this topology. No driver or other measures need to be taken for outputstage stability like one needs to take when coupling Triple with VFA. See in the SSA thread how Keantoken suggested compensation, he was on to something as well.
Ive shown my diamond schematic and performance was confirmed by Wahab, some think that the ULGF of 7 MHZ is too high, it may well be lower in practice. The amp has been in production for 2 different manufacturers for over 15 years and no complaints recieved about stability but Ill leave it at that.
Youre not going to beat 1 ppm THD20 though, Im afraid, for now.
If Scott Wurcer decides to show some of Gosner s innovative CFA circuits than we talking. The differences in THD performance between VFA and CFA are minuscule.
Complexity for both is the same, the vssa input is limited in its scope for higher performance possibilities though. 4mA is a lot though, one needs a stiff vbe multiplier for BJT use.
Perfs were indeed good for THD10 but there s also others numbers ,
i ll come back soon with a few more sims to give a broader picture
but i could already say that CFA and VFA are not equal linearity wise ,
at least using your schematic as a base to compare with.
Brad,
You don't like this weather, I think we are shooting for 104 later this week with humidity to go with it! I will follow up on my invite when our shoes won't melt if we have to walk a few yards and it doesn't seem like we are in Florida getting wet in a few minutes time. This can't last long I hope.
ps. I was the vendor for the baffles for the XPL series at JBL back when that was a top level product.
You don't like this weather, I think we are shooting for 104 later this week with humidity to go with it! I will follow up on my invite when our shoes won't melt if we have to walk a few yards and it doesn't seem like we are in Florida getting wet in a few minutes time. This can't last long I hope.
ps. I was the vendor for the baffles for the XPL series at JBL back when that was a top level product.
Yes I feel like Calif. people are among the biggest weather wimps when I recall how bad it tends to be in so many other parts of the country. Although this latest blast is not unprecedented, it feels a bit cruel after the very mild August. It will pass. I'm also getting old enough now that I'm not as bothered by heat as I used to be. In the meantime I may move at least the laptop on which I'm typing and the modem to the bedroom, where a portable keeps up better in the much smaller space.
I can do some schematic drafting on the laptop for the particular CFA gm core thread I said I would start, although my simulator program chokes with Windows 7. Some have sent me some other simulators and I appreciate their kindness, but don't have time to get up to speed at the moment. Maybe I will break down and try LTSpice, as it seems to be becoming something of a de facto standard.