On further inspection, the power supplies used to supply the input differential pairs apparently IS ALREADY REGULATED, as they are +/- 15V. Therefore we have the equivalent of a current source, for all practical purposes. This complaint is all nonsense, IN THIS EXAMPLE.
If I read this right:
Since the +/- 15VDC PS for the input differential pairs are already regulated,
there would be little need for a CCS?
I probably won't modify this circuit beyond bias and input filters.
I have really learned some of the theories behind amplifier design (no tests please).
I had dreams of going into amplifier design. . .I'll leave it to the pros and real DIYers.
Since the +/- 15VDC PS for the input differential pairs are already regulated,
there would be little need for a CCS?
I probably won't modify this circuit beyond bias and input filters.
I have really learned some of the theories behind amplifier design (no tests please).
I had dreams of going into amplifier design. . .I'll leave it to the pros and real DIYers.
There is NOTHING wrong with modifying the unit, it is just the actual advantage of doing so. The biggest problem with this circuit is the output quiescent current. Fix that, and you are more than half way. The second problem will be passive parts, especially cheap ceramic caps. Change these out for guaranteed NPO or mica caps, and you will be OK. Most aluminum caps are tolerable, although there are better quality units available.
That helps with the poor PSRR, but does nothing for the variation in quiescent current with input voltage. This means very poor common-mode rejection ratio.john curl said:
gni --
Don't give up yet. Your amp is similar in principle to the Leach amp, although the Leach has a few improvements and more power. You might pick up some hints by reading through Marshall Leach's website, in particular the section on his amp:
http://users.ece.gatech.edu/~mleach/lowtim/
There are a lot of posts on the Leach amp in this thread under Group Buys:
"Jens Rasmussen Leach clone group buy"
I agree with Mr. Curl that the first thing you should do is increase the idle current in the output, then replace the cheap ceramic caps (I like mica myself). Selective replacement of some resistors is inexpensive and might be useful. Then, if you want to explore this topology more thoroughly, build the Leach amp and use it for investigating circuit improvements. The Leach amp would be a good jumping off point: it is simple and reliable; parts are easy to get; and it is documented extensively. And if you haven't subscribed already to audioXpress, I recommend that.
Caps and types:
C101,201 10 µF,25V Electro [Input cap]
C102,202 2n7J 100 Yellow Rectangle [2.7nF? 5% input bypass to gnd parallel to R201]
C103,203 470J 50V Silver [47pF? 5% NFB bypass R213]
C104,204 220 µF,25V Electro [NFB to ground]
C105,205 47 Disc [47pF bc of VAS Q205]
C106.206 47 Disc [47pF bc of VAS Q206]
C107,207 blue [be of Q207]
C108,208 blue [be of Q208]
C109,209 473J 1H-P Silver Rectangle [47nF? 5% bias ce Q211]
C110,210 100nJ 100 [Yellow Rectangle Zobel 100nF? 5%]
C111,211 1.0 µF,50V Electro [Q212 Protection]
Not sure which ones to change. . . or with what?
I have changed R218 right (R118 left) from 2.7K to 2.0K taking the bias
from 0V to 9mV (41mA at 71 degrees F) and around 5mV at 61 degrees
and 28mA. The heatsinks are now warm. At 30mV bias (137mV) the heatsink
is hot to the touch longer than 5 seconds. That's why I set it to 9mV. . . and
it was easy to drop two 1K carbons since I had them.
Almost all resistors are carbon 1/8 watt. Output base Qs are 1/4 watt and are
blue (look like metal film).
Going to take a look at Leach amp. . .
C101,201 10 µF,25V Electro [Input cap]
C102,202 2n7J 100 Yellow Rectangle [2.7nF? 5% input bypass to gnd parallel to R201]
C103,203 470J 50V Silver [47pF? 5% NFB bypass R213]
C104,204 220 µF,25V Electro [NFB to ground]
C105,205 47 Disc [47pF bc of VAS Q205]
C106.206 47 Disc [47pF bc of VAS Q206]
C107,207 blue [be of Q207]
C108,208 blue [be of Q208]
C109,209 473J 1H-P Silver Rectangle [47nF? 5% bias ce Q211]
C110,210 100nJ 100 [Yellow Rectangle Zobel 100nF? 5%]
C111,211 1.0 µF,50V Electro [Q212 Protection]
Not sure which ones to change. . . or with what?
I have changed R218 right (R118 left) from 2.7K to 2.0K taking the bias
from 0V to 9mV (41mA at 71 degrees F) and around 5mV at 61 degrees
and 28mA. The heatsinks are now warm. At 30mV bias (137mV) the heatsink
is hot to the touch longer than 5 seconds. That's why I set it to 9mV. . . and
it was easy to drop two 1K carbons since I had them.
Almost all resistors are carbon 1/8 watt. Output base Qs are 1/4 watt and are
blue (look like metal film).
Going to take a look at Leach amp. . .
I included the codes with my reading of them. . . did I get them that wrong?
The tiny blue ones? I didn't find much about that color coding.
The tiny blue ones? I didn't find much about that color coding.
Can you find any codes on those blue ones at all? It will probably be just a 3-digit number. Also double-check how they are connected around Q207/Q208.
And C103/203, do you mean that is silver in color? Is it a little round one?
John Curl has said which to change -- the cheap ceramic ones and you show those in your list. And he said what kind of cap to change them with. Stick with same value of C, don't go lower in voltage, higher is OK as long as the cap isn't too big physically.
Keeping with the "we can't do everything" theme....... Go back through your list and circuit board. Find the cheap ceramic caps and then go to Mouser or Digi-Key and find what you think are appropriate replacements. Then post as a specific solution rather than a broad question. I think that approach might better received.
A first cut on resistors would be those on input and feedback. You can use inexpensive 1/4w metal films from Digi-Key or Mouser.
You're still under-biased but sounds like you're stuck with it and have done the best you can there.
And C103/203, do you mean that is silver in color? Is it a little round one?
John Curl has said which to change -- the cheap ceramic ones and you show those in your list. And he said what kind of cap to change them with. Stick with same value of C, don't go lower in voltage, higher is OK as long as the cap isn't too big physically.
Keeping with the "we can't do everything" theme....... Go back through your list and circuit board. Find the cheap ceramic caps and then go to Mouser or Digi-Key and find what you think are appropriate replacements. Then post as a specific solution rather than a broad question. I think that approach might better received.
A first cut on resistors would be those on input and feedback. You can use inexpensive 1/4w metal films from Digi-Key or Mouser.
You're still under-biased but sounds like you're stuck with it and have done the best you can there.
It's also a matter of priorities. As JC mentioned, changing those caps *may* bring about audible (if not measurable) differences - there's always the X-factor!
OTOH, putting in that CCS in the input stage will markedly drop your common mode distortion of the input stage (spot on Mr. Evil) and that *is* measurable and audible.
Jan Didden
OTOH, putting in that CCS in the input stage will markedly drop your common mode distortion of the input stage (spot on Mr. Evil) and that *is* measurable and audible.
Jan Didden
Mr Evil said:
Bzzzt. Perhaps if you are using JFETs and MOSFETs to make your input stage and VAS (huge gm differences) but if you are using degenerated BJT's with adequate collector current is is relatively easy to make small signal stages that exhibit next to no even order distortion at all.
Cheers,
Glen
G.Kleinschmidt said:
Hi Glen,
I think it is not so much the small signal stages in themselves. Even if you use a very well balanced symmetrical diff input stage, at some point you need to get from their collectors or whatever to the equally symmetrical Vas stage. So you need a PNP and an NPN device which come together at the Vas. It is those two devices whose collector (or drain) currents will be very difficult to make equal, and the current difference will need to be balanced out and one would most probably use somekind of CM control loop to unbalance the diff input structure to balance the Vas.
So it's the *principle* of synmmetrical diff input stgae that cuases the issue not the implementation itself per se.
Jan Didden
You worry too much. IF this problem was so great, we would have thrown out comp diff decades ago. IF you tried to do this amp with a single diff, THEN you would have to have a diff second stage with a current mirror. THIS circuit is just as difficult to keep balanced, BECAUSE of beta mismatching AND static thermal offset. Make a few REAL DESIGNS and see for yourself. EVEN if you put 4 complementary, BETA matched devices in ONE package and use it for the 2'd stage and MATCHED input parts in the single differential, there will still be 2'nd harmonic generation due to thermal mismatch of the second stage due to individual thermal differences. I have made both types of design, both comp diff and single diff, both bipolar and fet over the last 40 years in many dozen designs, and have found that to be realistic.
gni
If you want to pursue the constant current source idea, you could use a current regulator diode in place of R209 and R210. There are better and cheaper ways to implement a constant current source but this is the simplest way to do it here.
Calculate the current through R209 and 210, then replace them with the appropriate current regulator diodes. Do a search at mouser.com for "current regulator diode" then limit your search to the diode with the right current (go to next higher current if you need to) and in a DO-35 case. You'll find a drop-in current source to replace those resistors.
Try replacing the few caps and resistors first. Make sure everything works ok. Listen. Then replace R209 and 201 (and their counterparts in the other channel) with the current regulators and listen again.
If you want to pursue the constant current source idea, you could use a current regulator diode in place of R209 and R210. There are better and cheaper ways to implement a constant current source but this is the simplest way to do it here.
Calculate the current through R209 and 210, then replace them with the appropriate current regulator diodes. Do a search at mouser.com for "current regulator diode" then limit your search to the diode with the right current (go to next higher current if you need to) and in a DO-35 case. You'll find a drop-in current source to replace those resistors.
Try replacing the few caps and resistors first. Make sure everything works ok. Listen. Then replace R209 and 201 (and their counterparts in the other channel) with the current regulators and listen again.
I have built plenty of real designs, INCLUDING complementary ones where all four input devices are on the same die. For simple amps like this one, they come out worse than using the same number of devices in an asymmetrical topology. It's really only worth going symmetrical if you already have a good amp design i.e. one with current source tails.john curl said:
- Status
- Not open for further replies.