"REAL ENGINEERING" starts with investigating the effects of parameter spreading and, for example, deviation from the optimal bias setting . Have you done this John? I'm sure you did, so what was the outcome?
Yes I did, Pavel! So?
PS: Read this again: http://www.diyaudio.com/forums/showthread.php?postid=1757759#post1757759
edit : Or did you have something else in mind?
PS: Read this again: http://www.diyaudio.com/forums/showthread.php?postid=1757759#post1757759
edit : Or did you have something else in mind?
Aging may affect things like beta in matched pairs, I often find large differences in production amplifiers that are years old. I am assuming that transistors from the same batch should be close. This directly affects the diff pair and hence the effectiveness of feedback I would think. We always assume that the feedback always gives us the difference between the scaled output and the input. With mis-matched transistors in that diff. pair, this isn't happening as well as it may be assumed.
I do find better stability using matched outputs and drivers. When the 2SA968B and 2SC2238B were discontinued, I was very unhappy. I could get close NPN and PNP compliments from these, and the Motorola (at the time) devices had too large a spread between NPN and PNP for my tastes. Not only that, but On Semi came up with a figure of a 10:1 reduction in THD by using matched output devices, before feedback of course. Buy if John is worried about -140 dB distortion with NPO ceramic caps, this would tend to swamp out those issues.
Just some observations made over years of servicing on my part.
-Chris
I do find better stability using matched outputs and drivers. When the 2SA968B and 2SC2238B were discontinued, I was very unhappy. I could get close NPN and PNP compliments from these, and the Motorola (at the time) devices had too large a spread between NPN and PNP for my tastes. Not only that, but On Semi came up with a figure of a 10:1 reduction in THD by using matched output devices, before feedback of course. Buy if John is worried about -140 dB distortion with NPO ceramic caps, this would tend to swamp out those issues.
Just some observations made over years of servicing on my part.
-Chris
Actually, a slight deviation in Iq in EACH output device can be an asset, as the transistion from class A to class B is spread out. Over time, one can always re-bias the circuit, if they really care, just like someone can tune up a car.
I used to do WORST CASE analysis for Friden in 1966, when we were making military circuits on an IBM mainframe computer using ECAP. It generated a 'sensitivity' matrix and put each part in the 'worst case' position to create maximum deviation in output. It worked a little too well. It actually FAILED some circuits that would not actually fail, and this could be shown by slide rule calculation. It created a big problem with the military.
However, we did put in extreme changes in BETA over time and temperature, as well as significant changes in other components such as fixed resistors.
Today, we use more precise resistors, and I don't don't think that even a measurable change in the beta of the output transistors would change things very much. It should be obvious by now that the Iq set-point is a recommended range, not an absolute number. It IS important to be somewhere near the recommended range, but not that important just to make a listenable amplifier.
However, we did put in extreme changes in BETA over time and temperature, as well as significant changes in other components such as fixed resistors.
Today, we use more precise resistors, and I don't don't think that even a measurable change in the beta of the output transistors would change things very much. It should be obvious by now that the Iq set-point is a recommended range, not an absolute number. It IS important to be somewhere near the recommended range, but not that important just to make a listenable amplifier.
Hi John,
-Chris
That's why when I build prototypes, one gets normal devices and the other gets the worst, mis-matched and used stuff I can find. Except for any diff pairs, we know what happens when those aren't matched. Anyway, I run both types and do measurements also.
True, and it's good the setting isn't that critical. Bias tends to wander about. For new circuits, I use the THD meter at various levels. That's get me in the ballpark - and it let's me know how touchy things are.
-Chris
Hi John,
Thanks.
I'm one of those guys who has no sim experience, so I draw it out on paper. Then I assign some component values and lay out the PCB. Most of the time, the circuits work and I can optimize them by changing components. The circuit additions are easy by tacking the parts on the foil side.
I found that using the worst case parts in a prototype shows me where the weaknesses are, and what real world performance will exceed.
It's also a good indicator of what some of the "hack" service shops will release from service to a customer. Now there is a thought that should keep you up at night John.
-Chris
Thanks.
I'm one of those guys who has no sim experience, so I draw it out on paper. Then I assign some component values and lay out the PCB. Most of the time, the circuits work and I can optimize them by changing components. The circuit additions are easy by tacking the parts on the foil side.
I found that using the worst case parts in a prototype shows me where the weaknesses are, and what real world performance will exceed.
It's also a good indicator of what some of the "hack" service shops will release from service to a customer. Now there is a thought that should keep you up at night John.
-Chris
I don't sim much myself, but I have done it over the decades. It is just too much work, at this time in my life.
are we static matching or sitting for half a day in front of a curve tracer with a pile of devices batching and sub-batching, or do we have a fancy computer operated curve matching/printout unit?
_-_-bear
_-_-bear
Solid state lifetime was briefly mentioned here. Back in the old days, we had an American military handbook that we used to calculate estimated part lifetimes. In those days, more than 40 years ago, we would assign a product lifetime of 1000's to millions of hours to various devices. Generally, the worst parts were capacitors, especially tantalum and aluminum. This should be no surprise, and they sometimes failed completely, but usually they just changed value and changed their internal resistive impedance.
Working with test equipment 35 years old or more, I find that CAPACITORS are the biggest problem, and you sometimes can't tell much difference from the outside. You just have to replace it, to get the equipment to work again.
In audio, with the exception of real equipment breakdown, electrolytic caps do not change the actual practical performance much, as they are usually not used for critical functions, such as oscillators or tuned notch networks. As a result, the audio products just get 'tired', and distort a little bit more than if they were new.
Often a little clean-up of dust, connectors, and re-biasing is enough to really get pretty good performance out of an amp again. For example, I once bought an original Electrocompaniet power amp (the original Otala design) for $200 from an audio dealer from his used shelf.
I took it home, cleaned it up and re-biased it, and now it is the amp I use in my office, and it is DARN GOOD sounding. Better than some of my cheaper stuff. It is about 33 years old now, but it is still going strong. However, before I re-adjusted it, it had drifted enough, due to cheap passive parts, etc, to be just about useless. This beats Circuit City, any day.
Working with test equipment 35 years old or more, I find that CAPACITORS are the biggest problem, and you sometimes can't tell much difference from the outside. You just have to replace it, to get the equipment to work again.
In audio, with the exception of real equipment breakdown, electrolytic caps do not change the actual practical performance much, as they are usually not used for critical functions, such as oscillators or tuned notch networks. As a result, the audio products just get 'tired', and distort a little bit more than if they were new.
Often a little clean-up of dust, connectors, and re-biasing is enough to really get pretty good performance out of an amp again. For example, I once bought an original Electrocompaniet power amp (the original Otala design) for $200 from an audio dealer from his used shelf.
I took it home, cleaned it up and re-biased it, and now it is the amp I use in my office, and it is DARN GOOD sounding. Better than some of my cheaper stuff. It is about 33 years old now, but it is still going strong. However, before I re-adjusted it, it had drifted enough, due to cheap passive parts, etc, to be just about useless. This beats Circuit City, any day.
john curl said:
that brought a smile
An externally hosted image should be here but it was not working when we last tested it.
Circuit City would like to thank all of the customers who have shopped with us over the past 60 years.
Unfortunately, we announced on January 16, 2009, that we are going out of business.
For the most current information regarding Circuit City and our store policies,
please refer to the links below:
» Read the press release about Circuit City's liquidation
» For the latest answers to questions regarding current store policies,
visit the Breaking News section of the Circuit City Investor Relations website.
The FAQs available there are updated as we receive new information,
so feel free to keep checking back for the latest
» Circuit City is currently hiring temporary labor to assist in the store liquidation process.
If interested, visit the Circuit City jobs site
http://www.circuitcity.com/closed.html
928432
Why don't you elaborate on temperature, PMA? Of course, much of the reliability in parts is proportional in a big way to the INTERNAL operating temperature of the device.
mpe?john curl said:
Out of topic
Is Friden a maker of minicomputer for which Philips was the distributor in europe ?
JP
Hi John, Pavel,
My guess is the well known failure rate that increases exponentially with temperature. I can't remember the exact relationship numerically, but it's easy to look up if you need the number. It's just important to remember that device temperature should be kept as low as reasonably possible or you have this reliability penalty.
That's another thing that was ignored as audio equipment got worse in the 80's and forward. The Revox company seemed to forget the most with their B251 integrated amplifier.
-Chris
Edit:
Hey Jacco,
You lucky guy! I miss those, only fakes these days. Where have all those good parts gone? ... and why?
My guess is the well known failure rate that increases exponentially with temperature. I can't remember the exact relationship numerically, but it's easy to look up if you need the number. It's just important to remember that device temperature should be kept as low as reasonably possible or you have this reliability penalty.
That's another thing that was ignored as audio equipment got worse in the 80's and forward. The Revox company seemed to forget the most with their B251 integrated amplifier.
-Chris
Edit:
Hey Jacco,
You lucky guy! I miss those, only fakes these days. Where have all those good parts gone? ... and why?
- Status
- Not open for further replies.