|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
209.97.232.148
In Reply to: Did Charles HansEn say what the lessons learned were? posted by andy_c on April 7, 2007 at 21:39:56:
Let's just say that some of the discussions from DIYAudio a few years ago regarding cascodes paid some dividends in this case. It started me thinking in some new directions regarding the front-end circuitry. Thanks for your input back then.I also got some inspiration from peufeu's analysis of the Lavardin "memory distortion" phenomenon. That ultimately turned out to be a dead-end (for me at least), but it opened a few new doors.
On-Semi's ThermalTrak output devices are something that the audio community has been waiting decades for (and, no, I don't like the Sankens with built-in bias diodes because they have the wrong internal topology). It took a *lot* of fiddling to get them to work right (nothing's ever as easy as it seems), but in the end they also provided a nice improvement.
But perhaps the biggest single thing was that I had the luxury of fiddling with them for four months before releasing them to production. Just playing around with different component values and fine-tuning the circuit by ear brought the performance up to an entirely new level. I spent hundreds of hours listening and tweaking, listening and tweaking, listening and tweaking.
Follow Ups:
such as the SAP15N/SAP15P, I'd love to know why the internal "circuit is wrong".
These were developed by Kenwood and Sanken and seem like a good inexpensive design to me. I have two amps that use them and the bias is rock steady and there is no need to worry about driver/output temperature tracking, compensation lag and emitter resistor matching and mounting among other things.
I'm no expert but I find it hard to believe that the internal "circuit is wrong".
Not to criticize, but I'd be more worried about nesting the MXR pc boards capacitors and magnetics in what is a fairly compact thermal mass. I have not seen the bottom cover and so do not know about your ventilation scheme, but the fact that it became so hot during the mandated FCC test is not what I would expect for $16,000, CNC machining brilliance aside.
I like Ayre products overall and for a few years have had your power amps on my short list.
The Sanken SAP series have two problems:1) They are only an emitter follower double. In my experience an emitter follower triple will dramatically decrease the distortion compared to a double. This is because you are decreasing the loading on the front end by a factor of roughly 100x.
2) The way that the driver transistor is wired up, it will only operate in class AB and not class A. This prevents the driver from draining off the charge from the reverse-biased output devices, also increasing the distortion.
You can read more about this on the Marshall Leach website. He tried all of the topologies and found that the triple emitter follower gave the best performance. (This was invented by Bart Locanthi in the mid-'60s for JBL and called the "T-Circuit".) I found the exact same thing in my experiments.
< < developed by Kenwood and Sanken and seem like a good inexpensive design > >
Maybe so. You can hide a lot of the problems with this output device by applying a lot of negative feedback. And I guess that if you can save $0.50 or so by getting rid of the pre-drivers, then that might make sense for an "inexpensive design". However, it is far from optimum...
Interesting!
Maybe that is why they have not caught on. They have been available for five or six years now.
I know Cambridge is using them, but I'm not sure if anyone else is.
I guess they are fine for inexpensive designs.
Thanks for responding.
The "memory distortion" thing is interesting. I'm sure I haven't seen all the literature on this, but what I have seen seems somewhat speculative.There are some new simulation models for BJTs that support the concept of the instantaneous junction temperature changing within a simulation, even being different from one time step to the next in a transient simulation. One of these is VBIC and the other is the Philips Mextram. Parameter extraction for these looks like a nightmare though. None of the free or reasonably priced simulators have full support of these models either. LTSpice supports VBIC but not the idea of the junction temperature changing within a sim.
Still it would be interesting to see someone who doesn't have an agenda do a thorough study of the phenomena using such tools. Maybe some college guy will step up to the plate :-).
There is a guy in France with an audio website that has the most information on "memory distortion" that I have seen (link below). He has recently started participating on the DIY Hi-Fi and DIY Audio forums.But like I said, I think that it is a dead end. If you read Peufeu's stuff, he comes up with a "magic" resistor that helps stabilize the operating point of the active device. His theory is sound, as I established with simulations. But the problem is that the circuit both measures and sounds better without the "magic" resistor.
The conclusion I reached is that there is something else going on that is more important than the instantaneous temperature fluctuations of the "memory distortion". But I could be wrong. I've never heard the Lavardin amps myself. Maybe they sound even better than the MX-Rs!
The junctions that form all types of transistors: bipolar, MOSFET, and JFET, have capacitances that vary with the instantaneous voltages across them. This means the signal modulates the open-loop bandwidth of a transistor amplifier stage.Running a gain stage from a high supply voltage minimizes this variation. Running a high bias current reduces the effective resistance in series with the capacitance (as does shining laser light on it, ala Edge...).
Tubes don't have this particular problem, but all of them are microphonic to some extent.
< < ... minimizes this variation....as does shining laser light on it, ala Edge... > >Let's just say I'm from Missouri on this one. In the first place, Edge doesn't even make this claim. Instead they claim that they somehow use the laser to bias the output stage.
However, I'd love to be proven wrong. Do you have any links to anything that shows that shining a laser on a semiconductor junction minimizes the capacitance variations?
(And even if you did, would it really be a good idea to cut open the hermetically-sealed package so that you could shine the laser on it?)
"However, I'd love to be proven wrong. Do you have any links to anything that shows that shining a laser on a semiconductor junction minimizes the capacitance variations?"I once had a transistor that was light sensitive. Seriously! When a customer would open up a telecom cabinent that contained the rectifier that it was in, the voltage out of a rectifier would mysteriously go up. Opening and closing, shading... led to that one transistor. Perhaps edge got a hold of a batch of them.
The capacitance still varies, but there is less R-C time delay associated with the variation.I spoke with the designer about it at a CES/THE show. It is a way of increasing the carrier concentration without high electrical bias.
The package is there to keep the outside world away from the silicon die. For example, EPROM memory chip packages have expensive little windows in them to allow the UV light, but nothing else, to penetrate. Dust, dirt, polluted air, and humidity will combine to contaminate the exposed surfaces of the silicon die, create leakage paths, and encourage corrosion of the metal leads and bonds. Cutting open a transistor package is not something I would do.
< < Tubes don't have this particular problem, but all of them are microphonic to some extent. > >Plus most of them use steel pins and steel internal parts and are prone to hysteresis distortion from the magnetic fields.
Not to mention the fact that they are wearing out every time you turn them on...
Well, it’s a good thing then that being a high voltage device, the currents are exceedingly low and so are the magnetic fields produced internally. A tube like a 6SN7 can produce VERY low distortion with a spectrum that is low order. Hysteresys distortion is a problem in output transformers associated with Tubes however and they are to varying degrees microphonic.
For that the tap test will tell.Yes they wear out a little bit each time but finding either at a garage sale or attic, the odds of a 20 year old solid state amplifier working are smaller than that of a 30 or 40 year old tube amp working. Plastic transistors are not “air tight” like Pyrex and frit glass, they usually get leaky and fail. Tubes can leak too when the repeated thermal expansion causes micro fractures around the frit and pins.
SS and Tubes are entirely different ways of doing the job, each has totally different strengths and weaknesses from an engineering standpoint.As for the “memory”, the thermal shift is real, I have no idea if you can hear it.
On the other hand, for a Voltage amplifier the cascade approach already minimizes the collector swing on the onput, making the Voltage between the two transistors larger would further reduce the pdis swing and increased the avg pdis.
Heat them inputs up with static bias
Best,
Tom Danley
So right. A tube amp is simple for anyone to fix. It's easier to replace six or eight tubes than it is to find out which of the 40 or 50 transistors inside has failed.
Provided the tubes are good, they can last much longer than a ss amp. Replace the handful of electrolytic caps. every ten or fifteen years and you are all set for another decade. Or you can do the same with a ss amp, although replacing 40 electrolytics in it might be a hassle...
I literally spent 30 years buying one ss amp after another trying to find one that had the sound I wanted. None did. Most of them sounded the same to me, regardless of how much I spent.
"None did. Most of them sounded the same to me, regardless of how much I spent.
"
!
And so are most of us......... :)
| ||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: