Stability
I just looked at the LM3875 datasheet. Since it is not unity gain stable, it will almost certainly oscillate with the 100pF capacitor. I think it has to be reduced to 3.9pF or less or removed altogether, or a 47kohm resistor has to be put directly in series with it. Oscillations may also cause the funny hum Luke noticed; if the output is switching between two unregulated supplies with a switching frequency of a few hundreds of kilohertz, hum is to be expected.
Regarding the stability without the 100pF, the following things may or may not help:
A. adding decoupling capacitors very close to the IC supply pins and to the collectors of the transistors,
B. adding small resistors or ferrite beads in series with the bases of the transistors (to eliminate parasitic Collpits oscillations due to base wire inductances),
C. putting an RC series network (2.7 ohm and 100nF) in parallel with the load or between the LM3875 output and ground,
D. shunting a 100pF or 220pF capacitor across the input to ensure stability when the signal source is not connected.
By the way, a +/- 45V supply slightly exceeds the maximum rating, which is 84V between the positive and negative supply.
Luke, if you don't see any oscillations on your oscilloscope but you see the DC output voltage changing when you put your hand closer to the circuit or when you touch a node with an isolated screw driver, it is probably oscillating at a frequency which is higher than your scopecan handle.
I just looked at the LM3875 datasheet. Since it is not unity gain stable, it will almost certainly oscillate with the 100pF capacitor. I think it has to be reduced to 3.9pF or less or removed altogether, or a 47kohm resistor has to be put directly in series with it. Oscillations may also cause the funny hum Luke noticed; if the output is switching between two unregulated supplies with a switching frequency of a few hundreds of kilohertz, hum is to be expected.
Regarding the stability without the 100pF, the following things may or may not help:
A. adding decoupling capacitors very close to the IC supply pins and to the collectors of the transistors,
B. adding small resistors or ferrite beads in series with the bases of the transistors (to eliminate parasitic Collpits oscillations due to base wire inductances),
C. putting an RC series network (2.7 ohm and 100nF) in parallel with the load or between the LM3875 output and ground,
D. shunting a 100pF or 220pF capacitor across the input to ensure stability when the signal source is not connected.
By the way, a +/- 45V supply slightly exceeds the maximum rating, which is 84V between the positive and negative supply.
Luke, if you don't see any oscillations on your oscilloscope but you see the DC output voltage changing when you put your hand closer to the circuit or when you touch a node with an isolated screw driver, it is probably oscillating at a frequency which is higher than your scopecan handle.
Hi all,
have been really busy lately and working on several projects, but i got to work on this again and found no oscillatons with my scope (techtronix 200Mhz) and thought I would connect one channel as no dc and oscillations looked promising and music signal looked musical on the scope
. Then disaster struck my humble little amp 
It smoked and hissed and a thick cloud of black smoke (vapour from my transistors) filled the room.
Were talking flames here, and legs blown clean away from the transistor.
Not sure If I want to continue as my fiance is a little afraid of my hobby and doubts my competency
I might try this again on varoboard and hope it works but I have doubts.
I did hear music at low volumes for about 10 secs and all was good minus any modifications.
Ill keep u posted when I start up again.
have been really busy lately and working on several projects, but i got to work on this again and found no oscillatons with my scope (techtronix 200Mhz) and thought I would connect one channel as no dc and oscillations looked promising and music signal looked musical on the scope
. Then disaster struck my humble little amp It smoked and hissed and a thick cloud of black smoke (vapour from my transistors) filled the room. Were talking flames here, and legs blown clean away from the transistor.
Not sure If I want to continue as my fiance is a little afraid of my hobby and doubts my competency
I might try this again on varoboard and hope it works but I have doubts.
I did hear music at low volumes for about 10 secs and all was good minus any modifications.
Ill keep u posted when I start up again.
Perhaps an interseting modification to the circuit in the first post of the thread would be to use a capacitor of 100uF or so from the opamp output to the load instead of the resistor, or maybe the cap in parallel with the resistor of somewhat higher value than the resistor is now. The higher value the resistor (with no cap) the further the opamp has to swing from less than one gate threshold *toward* the other during the crossover region, the resistor supplying the load current during this crossover duration. If there was a capacitor there instead, it cold losslessly supply the load during this period.
Maybe it's not much of an idea... dunno.
Edit-> One thing a cap would probably do, especially in a mosfet version with no resistor, is to control the rate of handover of current from the opamp to the big fets. Suddenly turning current off in part of an amplifier cct is not a good idea at all. Also, at higher frequencies the mosfets would be mostly bypassed, keeping an extra pole out of the cct. As the frequencies get lower and the power generally requirements get larger then the "afterburner" really starts working. Hmmm....
Maybe it's not much of an idea... dunno.
Edit-> One thing a cap would probably do, especially in a mosfet version with no resistor, is to control the rate of handover of current from the opamp to the big fets. Suddenly turning current off in part of an amplifier cct is not a good idea at all. Also, at higher frequencies the mosfets would be mostly bypassed, keeping an extra pole out of the cct. As the frequencies get lower and the power generally requirements get larger then the "afterburner" really starts working. Hmmm....
http://www.dirkbaumann.de/assets/images/150Wuebl.jpg
I did something close to this but used TDA7294 and MOs-FET`s (is that ok ???) IRF640/9640
It worked very good
I had 24Vac at the output with 2.4ohm load that is about 240W after that the protection kicked in ..... the mos-fets were warm so as for the tda ....
Then as i was switching some cables or something it blew !!!! TDA is dead and IRF640 damn !!! So my question is did i hit something and did a short or i shouldn`t use MOS-FET`s ?
I think i`m going to try tomorrow again
I did something close to this but used TDA7294 and MOs-FET`s (is that ok ???) IRF640/9640
It worked very good
I had 24Vac at the output with 2.4ohm load that is about 240W after that the protection kicked in ..... the mos-fets were warm so as for the tda ....Then as i was switching some cables or something it blew !!!! TDA is dead and IRF640 damn !!!
So my question is did i hit something and did a short or i shouldn`t use MOS-FET`s ? I think i`m going to try tomorrow again
I've built amplifiers like this before using both normal and power op-amps, with both bipolar and MOSFET output stages. In fact I always try to include a resistor across the output stage if it's class B, even in discrete amps where possible. It's a good way to get the efficiency and simplicity of an unbiased class B output stage, without the high distortion. Adding some bias to the output transistors (but still much less than class AB) helps reduce the remaining distortion (you can still measure crossover distortion with the resistor - it doesn't get rid of it completely).
You can see examples of it in datasheets for quite a few different op-amps. As MarcelvdG said, it's not a true current-dumping topology, although the principle is similar.
You can see examples of it in datasheets for quite a few different op-amps. As MarcelvdG said, it's not a true current-dumping topology, although the principle is similar.
Now that you mention it, this type of thing *is* pretty common in datasheets and appnotes. I have been doing a great deal of reading since this thread was started!
I have been toying with the idea of a general-purpose shop amplifier with this general type of setup...Thanks to Tim__X for bumping it back to mind!
I have been toying with the idea of a general-purpose shop amplifier with this general type of setup...Thanks to Tim__X for bumping it back to mind!
Has anyone ever tried bootstrapping a current-dumping gainclone (see http://www.edn.com/article/CA45890.html)
How about a bootstrapped, current-dumping, rail-to-rail controled gainclone http://www.edn.com/archives/1994/092994/20di5.htm
How about a bootstrapped, current-dumping, rail-to-rail controled gainclone http://www.edn.com/archives/1994/092994/20di5.htm
That stuff won't work.
This does:
http://www.national.com/an/AN/AN-446.pdf
To further confuse the issue, how to run a LM3886 off ±100V and get 1KW at 2 ohms (see fig. 11)
This is called suspended supply operation. To drive 1KW at 2 ohms you would need ten LM3886 in parallel for A1, three pair of MJ21193/94 for Q1 and Q3, one pair of MJE15030/31 for Q2 and Q4, 35V zeners for D3 and D4.
This is a very advanced project and requires 0.1% resistors to set the gain around A1.
http://www.diyaudio.com/forums/showthread.php?threadid=36751&perpage=10&highlight=&pagenumber=3
This does:
http://www.national.com/an/AN/AN-446.pdf
To further confuse the issue, how to run a LM3886 off ±100V and get 1KW at 2 ohms (see fig. 11)
This is called suspended supply operation. To drive 1KW at 2 ohms you would need ten LM3886 in parallel for A1, three pair of MJ21193/94 for Q1 and Q3, one pair of MJE15030/31 for Q2 and Q4, 35V zeners for D3 and D4.
This is a very advanced project and requires 0.1% resistors to set the gain around A1.
http://www.diyaudio.com/forums/showthread.php?threadid=36751&perpage=10&highlight=&pagenumber=3
I can see why there might be problems with latch-up on bootstrapped designs, and overvoltage on that rail-to-rail design, but surely there are ways around those problems? After all, that design in the natioanl app note was bootstrapped, just in a more complicated way.
P.S. thanks for the link, it's an interesting app note.
P.S. thanks for the link, it's an interesting app note.
The word is 'suspended supply', it is not the same thing as a bootstrap.
The rail to rail thing will not work with a gainclone chip without so more work that its not worth the effort.
If you need to drive a pair of FETs rail to rail from an opamp (which is what that note was about), that is still a bad schematic.
There are much better ways of doing it.
The rail to rail thing will not work with a gainclone chip without so more work that its not worth the effort.
If you need to drive a pair of FETs rail to rail from an opamp (which is what that note was about), that is still a bad schematic.
There are much better ways of doing it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.