Member
Joined 2003
lineup said:
Only 44? I've an early version of something like that with, IIRC, 200 741's. About 10 square feet of PCB. Awesome! Don't know how it sounded, it was defect.
Jan Didden
Not wanting to hijack this thread, but Nat Semi used to make an op-amp that was capable of delivering 100 watts of output.
I remember a tech I used to work with back in the early 90s made a power amp using these devices, but I don't think he ever got it to work properly. There was some kind of unresolvable problem.
I remember a tech I used to work with back in the early 90s made a power amp using these devices, but I don't think he ever got it to work properly. There was some kind of unresolvable problem.
janneman said:
So you have been there, Jan.
Probably before I even understood how one transistor works ..
But you, as usual? miss my point with my post.
Which was not really about paralleling many op-amps, 44/100.
44 or 100 or 150 .. mine is Bigger than yours syndrome 😀
My post was a comment to your, in my opinion, intelligent way of dealing (or not deal with)
'imperfect' matching of output devices.
Paralleled power transistors, in your case.
Lineup 🙂 trying to make them old audio gurus to understand
lineup said:
Ahh, yes, sometimes I miss the point. And that 200 opamp thing wasn't mine, so it's more a 'his is bigger than yours' syndrome 😉
lineup said:
Well, in theory you can have say, 5 parallel devices that are EXACTLY the same, but then you'r no worse of, as far as matching is concerned, than a single device. But in 99.999% of the cases there are differences between devices and chances are good that you improve matching by paralleling. I think.
Jan Didden
(Sometimes trying to get the point even if there is none).
janneman said:
You're thinking in terms of a statistical average but practically speaking it is of a detrimental effect to have mismatched devices. Apart from the obvious mismatch in bias current you will have a mismatch in current sharing between devices as some devices will tend to hog the current more than others which will lead to overheating and potential breaches of the SOAR with catastrophic results 🙁
I believe that the Japanese used matched devices in their amps from the same batch, particularly when parallel devices were used.
snoopy said:
That's why we have emitter resistors...😉
Yopu maybe think about the old days with SONY VFET's that literally had so much spread in Vgs-on that it was impossible to share them even with 1 ohm emitter R. You had to select them in the same range, the same family. But not match them.
Jan Didden
janneman said:
You wouldn't use 1 ohm resistors with just two parallel devices because there would be two much losses in the resistors. Maybe with 10 devices in parallel but not two 😉 In any event it might help to even out current sharing but you would still have the issue of mismatch in bias currents.
Hi Andrew T
Try simulating the effect of a 47 uF /12k input resistor on a 20 Hz square wave, and compare with 4.7 uF.
It is surprising how big the input capacitor needs to be to support good LF performance. I tend to use DC coupling, normally, but where offset voltages are of a concern, at least 22 uF!
I have noticed that MJL's need higher current to reduce crossover distortion in simulations, (but not measured significant differences using higher bias currents). This seems to be related to the drop of fT at low currents. gm doubling isn't an issue if a current source drives the VAS. The Miller cap of course reduces the impedance (increasing the problem of gm doubling) at higher frequencies.
I'm surprised that this Self clone amp still uses such a large miller capacitor with output transistors 10 times faster than the MJ802 etc. which Self wrote about in Electronics World. One would have thought that other improvements might have been done ...
cheers
John
Try simulating the effect of a 47 uF /12k input resistor on a 20 Hz square wave, and compare with 4.7 uF.
It is surprising how big the input capacitor needs to be to support good LF performance. I tend to use DC coupling, normally, but where offset voltages are of a concern, at least 22 uF!
I have noticed that MJL's need higher current to reduce crossover distortion in simulations, (but not measured significant differences using higher bias currents). This seems to be related to the drop of fT at low currents. gm doubling isn't an issue if a current source drives the VAS. The Miller cap of course reduces the impedance (increasing the problem of gm doubling) at higher frequencies.
I'm surprised that this Self clone amp still uses such a large miller capacitor with output transistors 10 times faster than the MJ802 etc. which Self wrote about in Electronics World. One would have thought that other improvements might have been done ...
cheers
John
If I had to guess almost none of the high volume production amps incorporate matching, just due to the extra time it takes.... and yet we don't have blocks of flats going up in flames due to failure.
Nordic said:
I think you will find that the Japanese amplifiers used matched devices. They had access to matched devices long before anyone else did 😉
john_ellis said:
The Douglas Self circuit has a potentially fatal flaw in it. Because there is no current limiting on the driver or VAS transistor it is possible to destroy it if it goes into saturation and the emitter follower transistor before it feeds excessive current into the base. I have been told that this is indeed what happens to practical implementations of this circuit. Normally the base current is limited by the diff pair but as soon as an emitter follower is used then the base current can be much larger and potentially destructive to the VAS transistor.
Simple fix is to use an emitter resistor of say 10 ohms for the VAS transistor and a current shunt transistor to the base of the emitter follower 😉 In actual fact if you look on page 134 figure 5.24 of his 3rd edition this other circuit does actually use this protection 😉
Jaycar KC-5470 and KC-5471 "new in 2008"
I was wondering if anyone has actually tried one of these amps yet? Jaycar has the kits on their "new in 2008" flyer priced at $89.95 for the amp and $54.95 for the power supply. Which seems to me quite a reasonable price if they are any good.
col.
I was wondering if anyone has actually tried one of these amps yet? Jaycar has the kits on their "new in 2008" flyer priced at $89.95 for the amp and $54.95 for the power supply. Which seems to me quite a reasonable price if they are any good.
col.
Wow, I recently read D. Selfs book and am in the process of
building/designing a 4 OP device main amp based on "blameless".
To my astonishment the amp in this thread is almost the
"clone" of my endeavors except for the thermaltraks
and undercompensated vbe
that Roender commented on.
I am using the same drivers (mje15030/1) ,almost the same outputs,
(mjl3281/1302) and to get the amp thermally stable I had
to mount a real Vbe,(bd139) with a thermal insulator on one of the outputs
(2 mica pads) to get a +- 20mv bias tracking between 25-60C.
Bias wise, I noticed switching distortion below 15ma so
I settled for 35Ma per device.
Otherwise, my amp is the same except for no VAS buffer
and a LED CCS.
Since they are selling this as a kit it must
not have any real safety issues and it's nice
to see my device choices commercialized.
I do have a question about a EF amp with this topology.
Which driver arrangement is better:
A.Drivers with emitter resistors tied to output(like this amp),or
B.Drivers with "floating" single resistor between the emitters.
Why do different EF designs use both topologies and what
are the advantages and/or disadvantages of them.
(I left both possibilities open on my PCB's)
Any comment on this would be appreciated,
Thanks ,OS.
BTW , Its nearly impossible to get thermaltrak's here...,
but even if I could I'd opt for the more "traditional"
output choice.
building/designing a 4 OP device main amp based on "blameless".
To my astonishment the amp in this thread is almost the
"clone" of my endeavors except for the thermaltraks
and undercompensated vbe
that Roender commented on.
I am using the same drivers (mje15030/1) ,almost the same outputs,
(mjl3281/1302) and to get the amp thermally stable I had
to mount a real Vbe,(bd139) with a thermal insulator on one of the outputs
(2 mica pads) to get a +- 20mv bias tracking between 25-60C.
Bias wise, I noticed switching distortion below 15ma so
I settled for 35Ma per device.
Otherwise, my amp is the same except for no VAS buffer
and a LED CCS.
Since they are selling this as a kit it must
not have any real safety issues and it's nice
to see my device choices commercialized.
I do have a question about a EF amp with this topology.
Which driver arrangement is better:
A.Drivers with emitter resistors tied to output(like this amp),or
B.Drivers with "floating" single resistor between the emitters.
Why do different EF designs use both topologies and what
are the advantages and/or disadvantages of them.
(I left both possibilities open on my PCB's)
Any comment on this would be appreciated,
Thanks ,OS.
BTW , Its nearly impossible to get thermaltrak's here...,
but even if I could I'd opt for the more "traditional"
output choice.
Be very careful about fake mje15030/1 , I suspect the ones I got with the modules are just that , changed them with new ones and low and behold the bias voltage is exactly what they said it would be with a 68 ohm resistor in series with the bf470, must now do some testing before it will be used
- Status
- Not open for further replies.