AndrewT said:
It's similar in concept I guess - taking the output stage out of a feedback loop at high frequencies. But Leach's idea works on the global feedback loop, and ends up increasing distortion by reducing feedback around the output stage, instead taking global feedback from the VAS at high frequencies. TMC reduces distortion by increasing feedback around the output stage (by putting it inside the Miller loop at audio frequencies in addition to the global feedback loop).
Ok, I get it now! I guess I already knew what TPC was after all! For TMC, I have no experience. I tried TPC in my amp but couldn't improve the phase margin the way I wanted, the -3dB point was too low. With a simple miller, it is not too good of a shape but I liked the resulting bandwidth better. Maybe I just didn't calculate the TPC in an accurate manner but anyway without feedback I guess it is not a big problem.
Regards
Regards
darian said:
Well Darian, between 0.003% and 0.0001% there's a x30 gap. Now, I'm not talking if those ppms are audible, but certainly technical performance comes to a price.
However, if you pull from the YAP 2.1 all the frills (protections, decouplings, servo, etc...) what's left is not that scary anymore:
An externally hosted image should be here but it was not working when we last tested it.
If you don't count the output devices (which are required anyway for these power levels, what's left is 24 small signal transistors (out of which some are still required for bias purposes only) and an opamp. I wouldn't say that's very much, even when comparing with the Leach old designs.
Now, I struggled for quite some time to simplify as much as possible these designs (starting with the PGP, which is an absolute monster) and I started to believe that's about all it can be done if it is to go into this kind of performance. The VSOP amp is simpler (it replaces the entire input stage with a simple opamp), but, comparing to YAP, it's not quite there in terms of performance and stability.
Again I'm not saying that this kind of performance is required to reach the ultimate listening experience. As unfortunately it's already visible, people are happy with and will continue building Self, JLH, JC, and a few other 30 years old classics clones, to Kingdom Come. That's about the sound quality from which there is subjectively nothing more to add.
Syn08,
You can notice that your distortion levels could be quite hard to measure in real life (noise level of the PSU?). Moreover, I use a discrete input buffer and not a high feedback opamp. This is because of my choice to make it totally global feedback free and I wanted a "thermal distortion" free input style. It has a cost. I also use fets because I love them, the distortion is worse for sure. I would gain a magnitude in distortion with bjt's at the input. And lastly, as you know, just by multiplying the output mosfets you can gain a lot in linearity! And most of the distortion comes from the output stage in my case. My design is also aimed at being reasonable in term of idling power and cost. That's something that is to take into account. Earlier versions of it were better in term of distortion but just too crowded to my taste.
Regards
You can notice that your distortion levels could be quite hard to measure in real life (noise level of the PSU?). Moreover, I use a discrete input buffer and not a high feedback opamp. This is because of my choice to make it totally global feedback free and I wanted a "thermal distortion" free input style. It has a cost. I also use fets because I love them, the distortion is worse for sure. I would gain a magnitude in distortion with bjt's at the input. And lastly, as you know, just by multiplying the output mosfets you can gain a lot in linearity! And most of the distortion comes from the output stage in my case. My design is also aimed at being reasonable in term of idling power and cost. That's something that is to take into account. Earlier versions of it were better in term of distortion but just too crowded to my taste.
Regards
Bob Cordell said:
Hi Bob,
Thanks,
Yes, 4MHz is the loop gain unity gain frequency. It could look high, but if you recall the first YAP, that one had 8 MHz and was rock solid as well. A good layout with no missed details (like the point to pick the FB, long traces only at low impedance nodes, etc...) and a RF style ground plane helped to achieve such high values.
Gopher said:
You are correct and here's my stance on this:
http://www.diyaudio.com/forums/showthread.php?postid=1591779#post1591779
So do you have a better target for DIY? Or should we all build Self amps, with an 100pF Cdom around the BD139 VAS, forever?
Gopher said:
This is the way I look at it.
Speakers produce mostly 2nd- and 3rd-order distortions, but amps, even some reasonably well-designed ones, can produce significant amounts of high-order distortions. We'd like to get rid of as much high-order stuff as possible. One solution might be to start with an amp having almost no high-order distortion terms at all and use no feedback. This seems reasonable for a preamp, where operating all stages in class A and being careful with the topology can give a practical solution.
But for a power amp, class A is only practical for low-power designs, unless you want to waste some serious energy. So one ends up starting with a class AB design, which by its very nature has lots of high-order distortions in the absence of feedback. Further, it's been shown by Baxandall and others that feedback around an amp having moderate distortion does a much better job reducing benign low-order distortions than it does the high-order ones. So one must work extra-hard to reduce the high-order terms using feedback, by minimizing the open-loop distortion, using multiple feedback loops and so on. If one reaches a THD figure of around 1 ppm or so, it becomes very difficult for obnoxious high-order distortion terms to "hide". So the low THD may not always be an end in itself, but the byproduct of using high feedback, multiple feedback loops and other techniques in an attempt to wipe out high-order distortions of class AB power amps.
Mr Evil said:
- Material costs: given a certain output power, it's the power devices, power supply (xformer plus smoothing caps), heatsink, PCBs and chassis driving the cost. Low power electronics costs today almost nothing. The parts for YAP 2.1 are less than $35/channel (not including the power devices, which you would need anyway), as ordered from DigiKey.
- Ease of manufacturing: this is DIY, not an OEM.
- Reability: has almost nothing to do with performance.
- Size: have you noticed the size of YAP 2.1 installed on the heatsink? http://www.diyaudio.com/forums/showthread.php?postid=1883844#post1883844
- Efficiency: 67% at max power for Class B, be it Self, YAP or whatever. Nothing to do with performance. Unless you mean low efficiency, then Class A rules.
Any other "goals"?
THD itself of 0.0x% is unhearable, no audio input can cause a slew of more than several V/us, audio range is only 20Hz-20kHz and speaker/cable design takes no advantage of damping factor more than 20.
These are facts.
However all these parameters are a benchmark, which indicates technical perfection of the circuit. We don't directly need better parameters then quoted above, but we need a good design, because surprisingly amplifiers which should not, can and do sound different.
These are facts.
However all these parameters are a benchmark, which indicates technical perfection of the circuit. We don't directly need better parameters then quoted above, but we need a good design, because surprisingly amplifiers which should not, can and do sound different.
If the price is >0, then there is room for improvement. It's possible to reduce the cost of the heatsinks and power supply by increasing efficiency, for example.syn08 said:
Sorry, I should have said "construction" instead of "manufacturing". Some designs are easier to construct than others; ones that are simpler and less sensitive to layout (both electrical and thermal) are easier.syn08 said:
Reliability is something that can be improved with better design, and judging by the number of posts on this forum from people with busted amps, it does need improving.syn08 said:
It doesn't matter what size it is, it's always possible to do better.syn08 said:
And do real amps acheive that efficency? At all power levels? What about class G, H or even D? Class D in particular promises vast improvements in many areas eventually.syn08 said:
I'm not sure what you mean by "nothing to do with performance"; efficiency is a property of a working amplifier, and everything here is something about audio amplifiers that can be improved instead of distortion.
as per the VSOP amp?
JPV said:
- For general current feedback principles, this Texas Instruments document
- For the input stage, the Alexander current feedback amp
- For the output stage, the previous YAP amp version and Andy's suggestion for an improved input stage
- For the TMC compensation, Edmond Stuart's work on this forum (and the following hot debate) and on his website www.data-odyssey.nl
- For the active clamp, the PGP amp and the VSOP amp (scroll to the end of the thread).
- For the overload protection, the previous YAP amp version
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.