In my studies, to deal with the real world and ugly speakers, one pair of bjt outputs is not any good for more than about 30W. Am I expecting too much for an amplifier to be stable and well behaved when clipping into 2 Ohms? Not easy or cheap, but it makes sense to me with what I know about real speakers. It seems like a reasonable tradeoff to give up some numbers game of zeros to get clipping behavior.
I had a different experience. Both my B&K and some of my Haflers came biased at under 100. When I raised them to 135, classical guitar was cleaner and less metallic.
Are therma-tracks still in production? I thought they were killed a year ago.
The .1's seemed harder to stabilize in my reasonable project and I did not like the standing power dissipation so I went back to .22's. It happens to be the value that came with the MX-50 Chinese junk kit with CFP output ( single pair) Maybe the better thermal stability of the CFP makes them viable.
The .1's seemed harder to stabilize in my reasonable project and I did not like the standing power dissipation so I went back to .22's. It happens to be the value that came with the MX-50 Chinese junk kit with CFP output ( single pair) Maybe the better thermal stability of the CFP makes them viable.
I noticed that I could set bias as low as 15mA using vertical fets, but this was with a HEC circuit. The significant increase in dv/dt of the error signal indicated that it need quite a bit of Vgs adjustment in order to cancel out the crossover notch. With higher bias the error signal was smaller and with lesser slop. When I disconnected the error correction so that the circuit was just a follower, I got very different results under heavy complex loads. It took at least 100mA to begin to get rid of the crossover notch. Better results were more around 200mA.
Hi Bob,
Soulution 700/710 power amp used Sanken LAPT 2SC3519A/2SA1386A and 3 3W/0.33R resistors paralled to emitter resistors.
Attachments
I used 0.15 RE in my TT amp and it's thermally stable and sound excellent.
The published and "intended" application for semiconductors has more to do with the major application (read $$ purchasers) than with anything else. It's a good idea to consider the actual specs and run some curves, then decide if it would be suitable.
Any device that is capable of high frequency operation is susceptible to parasitic oscillation. Mosfets are not particularly so in my experience. Careful layout and circuit design are required for any wide bandwidth high gain amplifier, mosfet or not. Plenty of BJT amps have been built and manufactured that were/are marginally stable and prone to parasitics.
Imo.
_-_-
Bear,
Yea, I had a C&M labs amp once. After the third time it self destructed they gave up. All BJT as far as I know. I have never tracked down any information on them other than some hero-worship bio of the company.
I still suggest that the physicists who design chips may know a few things we don't. As an example, the hot-spot issue in trench MOSFETS. You won't see that on a curve. It does not mean don't go there, it means pay attention to what they say as a hint. (Well in the case of trench fets, it does mean don't go there)
I have somewhat limited experience with only small J-fets and Lateral MOSFETS in amps, a few V-Vets and Lateral fets back in the old lab days at a computer company. But they do seem more susceptible to parasitic oscillation. Of course, layout and design are important. I do things like fly the gate stopper right off the pin, not on the board. Easy for DIY, bad for production. Haflers seemed to really need the small bypass cap on the heat sink. Long leads may account for part of that.
In other words, sure we all use screw drivers to open paint cans. Who has not used pliers on a nut? They have their limitations and the correct tool still works better.
.1 vs .22:
For every bit of good advice, someone will always challenge it as they got away with doing something else. That does not invalidate good advice. Just because 40 or so years ago in my miss-spent youth I might have driven a bit impaired once and got home safe does not mean one should ignore good advice like driving sober!
BITCH: Looking for SPICE models for ALFET. sales@semilab-tt.com is the address their corporate page lists for a contact. Bad address. Come on folks, maintain your WEB page! I like the way my Reasonable deals with clipping in the lateral fet driver/BJT output version. I am thinking the single die version may make a reasonable driver stage. I suspect they will be very close to the Exicons.
Yea, I had a C&M labs amp once. After the third time it self destructed they gave up. All BJT as far as I know. I have never tracked down any information on them other than some hero-worship bio of the company.
I still suggest that the physicists who design chips may know a few things we don't. As an example, the hot-spot issue in trench MOSFETS. You won't see that on a curve. It does not mean don't go there, it means pay attention to what they say as a hint. (Well in the case of trench fets, it does mean don't go there)
I have somewhat limited experience with only small J-fets and Lateral MOSFETS in amps, a few V-Vets and Lateral fets back in the old lab days at a computer company. But they do seem more susceptible to parasitic oscillation. Of course, layout and design are important. I do things like fly the gate stopper right off the pin, not on the board. Easy for DIY, bad for production. Haflers seemed to really need the small bypass cap on the heat sink. Long leads may account for part of that.
In other words, sure we all use screw drivers to open paint cans. Who has not used pliers on a nut? They have their limitations and the correct tool still works better.
.1 vs .22:
For every bit of good advice, someone will always challenge it as they got away with doing something else. That does not invalidate good advice. Just because 40 or so years ago in my miss-spent youth I might have driven a bit impaired once and got home safe does not mean one should ignore good advice like driving sober!
BITCH: Looking for SPICE models for ALFET. sales@semilab-tt.com is the address their corporate page lists for a contact. Bad address. Come on folks, maintain your WEB page! I like the way my Reasonable deals with clipping in the lateral fet driver/BJT output version. I am thinking the single die version may make a reasonable driver stage. I suspect they will be very close to the Exicons.
0.15 is probably OK with a good design with fairly conservative heat-sinking.
Connect a DVM across your emitter resistors and measure the bias voltage when the amp is cold. Let warm up for about an hour with no signal and measure it again. Now run the amp at 1/3 power into 8 ohms for 15 minutes. Kill the signal and measure the bias. Watch how much the bias changes over time as the bias gradually returns to the value that it was when the amplifier was initially warmed up. This test can be very revealing.
Cheers,
Bob
I find the distortion ( crossover distortion? ) from a mosfet output stage less objectionable than the distortion from a bjt output stage. I seem to remember that Ben Duncan made a similar comment some time ago.
I try to set the bias to the most satisfactory level from a listening point of view.
On lateral mosfet amps I find this to be around 75 to 100 ma per pair. ( I normally use Hitachi lateral mosfets )
Don
I try to set the bias to the most satisfactory level from a listening point of view.
On lateral mosfet amps I find this to be around 75 to 100 ma per pair. ( I normally use Hitachi lateral mosfets )
Don
After it was brought to may attention Renesas still makes the 2SK1056/J160 I revisited their use as drivers for two pair BJT's in my Reasonable project. Based solely on the simulation, this combination seems to behave the best at 2 Ohms and at clipping.
On paper, ya' just choose your poison. Based on simulations, with all other things being as equal as they can be, the only differences is how they behave at clipping and into low loads.
Applying this to what my wife's critical hearing has issues with, horn and string crescendo's, I have to go with the behavior as the critical parameter over crossover issues. Lateral drivers, BJT outputs.
On paper, ya' just choose your poison. Based on simulations, with all other things being as equal as they can be, the only differences is how they behave at clipping and into low loads.
Applying this to what my wife's critical hearing has issues with, horn and string crescendo's, I have to go with the behavior as the critical parameter over crossover issues. Lateral drivers, BJT outputs.
As far as clipping and clipping behavior go, the output transistors should never be allowed to clip at the rails. Baker clamps can be used to keep the entire output stage and the VAS transistors from saturating. Giving up a few volts of swing at the output can improve matters quite a bit. Moreover, most amplifiers will clip occasionally on well-recorded uncompressed music. When they clip, you want it to be clean.
Of course, a more extreme approach is to put an adaptive soft clip circuit ahead of the main amplifier circuit so that none of the amplifier ever clips. Here, of course, the controversy is that you are really giving up a lot of zeros in the THD department at modest to high power levels. In principle, of course, you can make the clipping as soft or sharp as you want with such circuits.
Cheers,
Bob
In sim, the baker clamp on the VAS was not very clean, but flying clamps across the bias servo seems to work very well. Getting it symmetrical requires a little bit of work. The clamp transistor on the darlington VAS also helps a surprising amount. The difference between positive and negative clipping is a bit eye-opening. Again, maybe a reason to look at a symmetrical VAS eventually. I am not sure I can trust many of the part models, but it seems logical that I would want fast soft recovery diodes such as BAT54.
Rally, the best solution is an amp so big it never clips. I have no idea what "obviously big enough" is. 250 on my living room. Somehow that shoots the 10W class A club in the foot.
Top see if I ever actually clip, I would compare the instantaneous peak of the VAS to the rails?
Rally, the best solution is an amp so big it never clips. I have no idea what "obviously big enough" is. 250 on my living room. Somehow that shoots the 10W class A club in the foot.
Top see if I ever actually clip, I would compare the instantaneous peak of the VAS to the rails?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.