I never faced with that problem. I use 2 pairs, with +/-55V, and it rock stable even with 2ohm load (not resistor, but 4 pcs of 8ohm speakers parallel).
It was strange to me as well. I have some of that drivers. They are working well, but totally kill the sound to me.
I use them to push 3 pairs of NJW0xxx with same conditions as above. The only issue was the oscillation, so I have to add 100ohm series base resistor, and 22-33pF B-C capacitor.
2690/1220 are great devices. Used them in smaller amp, as driver, and now I use the npn in the Vbe multiplier.
Sajti
Wow those are especially weak. The 100V devices - 2SA2205 & C6099 - I am using now are stronger than them.
(A2205 and C6099 are not actually proper complements and are pretty situational picks. I use them in a CFP style output and they work very well. PHPT6100* is also a fun pick for the adventurous)
Nice to hear that, Sajti.
How about the sound of 2SC6145/2SA2223?
I'm looking for the best sounding BJT for my ear. I've tried big Sanken LAPT 2SC2922. It's sound is awesome for bass & treble, very details, but less mid and unclear.
Now i'm using MJL3281A/MJL1302A, much better.
It's hard to tell. I feel the good clarity, powerful and detailed. Unfortunately I never compared it to the 2922/1216.
Sajti
Thanks Sajti,
It's very hard to say what better. If you want, you must compared by A-B test, by same front-end.
I've tested many devices and i'm happy with ON Semi devices, although it request to must be biased at high quiescent current >=250mA (lower your sound will sharp and hard).
Regard,
It's very hard to say what better. If you want, you must compared by A-B test, by same front-end.
I've tested many devices and i'm happy with ON Semi devices, although it request to must be biased at high quiescent current >=250mA (lower your sound will sharp and hard).
Regard,
Last edited:
Bob just wanted to let you know that your Cordell-Models.txt file has
duplicate definitions, reported in the SPICE error log.
These are the lines right out of the file:
* Z8_2 8.2V Zener Diode Diodes Inc. 1N5237B
* created January 21, 2011 copyright Cordell Audio
.model Z8_2 D(Is=1.5n Rs=.5 Cjo=185p nbv=12 bv=8.2 Ibv=10m Vpk=8.2 mfg=CA012111)
*
*
* 1N5237B 8.2V Zener Diode Diodes Inc. 1N5237B
* created January 21, 2011 copyright Cordell Audio
.model Z8_2 D(Is=1.5n Rs=.5 Cjo=185p nbv=12 bv=8.2 Ibv=10m Vpk=8.2 mfg=CA012111)
duplicate definitions, reported in the SPICE error log.
These are the lines right out of the file:
* Z8_2 8.2V Zener Diode Diodes Inc. 1N5237B
* created January 21, 2011 copyright Cordell Audio
.model Z8_2 D(Is=1.5n Rs=.5 Cjo=185p nbv=12 bv=8.2 Ibv=10m Vpk=8.2 mfg=CA012111)
*
*
* 1N5237B 8.2V Zener Diode Diodes Inc. 1N5237B
* created January 21, 2011 copyright Cordell Audio
.model Z8_2 D(Is=1.5n Rs=.5 Cjo=185p nbv=12 bv=8.2 Ibv=10m Vpk=8.2 mfg=CA012111)
Just to be clear here, I started the discussion of Bryston commenting about what I jokingly
call the can't make up our mind output stage (CMOM) in the 3B model. I have the book
now, a member generously sent me his that he is not using, and I don't see anything like
the 3B in Figure 7.35. My comment referred to the fact that they use both NPN and PNP
final devices on both rails. In fact Figure 7.35 has two schematics A and B. B is a triple
EF with a cascode (Q2) on the first EF (Q1) and a current source, probably to make it
non-switching - have not done a full analysis. The author refers to R1 and R2 as setting
the gain, however there are no R1 or R2 to be found in any part A or B of Figure 7.35.
The resistors are not labeled but none of those shown could possibly set the gain of the
output stage. Follow the emitters Q1 - Q4 -Q5 all emitter followers thus there is no way
for it to have gain. And there are no paired NPN-PNP finals on both rails.
7.35a is also an EF triple with bootstrapping on the first EF, probably for voltage headroom
not for more gain since it is unity gain. The only unique thing there is that Q1 and Q2 form
a CFP first stage, BUT again there is no way for any of the resistors to set the gain. Again
follow the emitters Q1, Q3, Q4 thus there is no way for it to have voltage gain.
The 3B uses a triple for the final 3 transistors but they are wired as common emitter not
EF as shown.
Perhaps the author would like to correct the diagram, he probably needs to show more
of the circuit, and correctly identify R1 and R2.
If 7.35 is anything like another Bryston model I'd like to see the schematic if anyone knows.
I was hoping to learn something from this book but nothing regarding the Bryson's and
I'm going to stick to my CMOM view regarding the 3B and similar models for now. I was
actually hoping to learn a trick or two but probably not without simulating it myself. I
have a feeling that current sharing between the NPN and PNP finals is not going to be
very good. I could simulate the output stage in isolation but not motivated enough at
this point.
Interesting that I had the bootstrapping idea many years ago that is shown in 7.35a.
I have also recently been looking at simplifying the Bryston output stage for use as an
upgrade to the Tigers and came up with something very similar to 7.35a but of course
the CFP part has gain in my concept.
Thanks for the reference I will take a closer look at the rest of the book and I would like
to discuss the Bryston output stage but not until we know exactly what the author was
talking about.
Good thing I didn't waste my time looking for another Bryston schematic, Figure 7.35
provides reference number 241 which is the 1987 3B schematic in the author's list of
references.
I simulated the 3B without protection, and have attached a snip of the output stage here.
It can be seen that Q13 is the common emitter stage with R28 and R30 setting the gain.
Q15, Q11 and Q20 complete the CFP as a triple for current gain but they are also common
emitter with their collectors connecting to the output. Q3 is the only single EF and it is the
CMOM device. Kolinummi's diagrams are nothing like this but claimed to be from the 3B.
I have to wonder if this is the wrong figure or if he got the protection mixed up with the
signal path in the 3B - makes no sense to me. Both of Kolinummi's figures show a triple
EF and there is none to be found in the 3B.
I've also attached the full 3B schematic in case there is any interest.
Attachments
Last edited:
Hi Pete,
I'm sorry to be so slow in getting back to you on this. I must have somehow missed a couple of posts.
I developed those procedures myself out of necessity because I could not find anything online. However, my adventures in modeling were inspired by Andy Connor's modeling work on power output BJTs. His observations of how bad the published models were caused him to get involved in modeling them as well. I think a couple of my procedures were very similar to what he did in his modeling.
Cheers,
Bob
I've always been interested in the Bryston output stage. It is not at first intuitive, at least to me. I've written a second chapter on output stages for my second edition, and about 2 years ago I wrote the section that describes the Bryston output stage. I'll have to go back and review it in light of the material on it that Mark and you have brought up.
I seem to recall that there are some significant beta matching requirements between the PNP and NPN devices, so SPICE results may have to be viewed carefully if the SPICE models inadvertantly make the NPN and PNP look too much alike.
The stage can be viewed 2 ways. First as a CFP with an EF "helper" and second as an EF with a PNP "helper". One of the neat things about it is that the driver current is used twice, so that the driver needs only to deliver half the current for a given amount of output current for the two output transistors (as compared to driving 2 EF devices of the same sex in parallel).
I don't recall seeing an improvement in crossover distortion as compared to an EF output stage, but perhaps I did not look hard enough. I don't remember if I did a wingspread on it. If I didn't, I should have, and should have done it with a few different kinds of mismatch between the NPN and PNP devices. I also don't recall observing or describing a non-switching aspect to it.
As an output stage with gain, it does give the VAS more headroom to work with, which can be nice.
As I said, I'll have to go back and look at what I wrote
.Cheers,
Bob
You say it is not intuitive so I'll offer some of my thoughts and a question for members.
If you short R26 and delete Q3 you have an almost conventional CFP but with a triple
for the output device and gain in exactly the same way as was done in the Tigers.
You make an interesting observation that Ie is used twice as it comes in through Ic in
driving Q3 the EF helper when you put the schematic back to normal. I do agree that
is rather clever.
I'm not a fan of using both types of finals so I'm looking at ways to simplify and if
possible improve that output stage but keeping it with gain. One way is to just wire
Q13 and Q15 as CFP with gain and then go 2 EF from the output of that CFP stage.
This would be keeping the "evil" CFP to as simple as possible.
The second way is as already stated, short R26 and delete Q3.
The third is to short R24, R39 and delete Q20.
Each case keeps the same number of current gain stages.
Advantages and disadvantages of each one, which is best?
If you short R26 and delete Q3 you have an almost conventional CFP but with a triple
for the output device and gain in exactly the same way as was done in the Tigers.
You make an interesting observation that Ie is used twice as it comes in through Ic in
driving Q3 the EF helper when you put the schematic back to normal. I do agree that
is rather clever.
I'm not a fan of using both types of finals so I'm looking at ways to simplify and if
possible improve that output stage but keeping it with gain. One way is to just wire
Q13 and Q15 as CFP with gain and then go 2 EF from the output of that CFP stage.
This would be keeping the "evil" CFP to as simple as possible.
The second way is as already stated, short R26 and delete Q3.
The third is to short R24, R39 and delete Q20.
Each case keeps the same number of current gain stages.
Advantages and disadvantages of each one, which is best?
Thank you for responding and answering questions, no problem on how long it takes.
I'll try to compare your methods with the Intusoft paper that I've linked here before
just to see if there are any new ideas that might crop up.
A few more things about how I look at that output stage.
Let's call shorting R26 and removing Q3 as the CFP-3 with gain and one might ask
how does it change when R26 is not shorted and Q3 is back in. Q11's collector is
just one Vbe (Q3) drop above the output so it really doesn't change much but now
most of Q11's collector current, minus that through R26, goes into Q3's base.
Similarly Q11's emitter current, minus that through R24 goes into Q20's base. For
high beta in Q11 Ic is approximately equal to Ie, and if Q3 and Q20 are Vbe matched
and beta matched then they will share current.
I've thought a bit about the current sharing and given how different MJ21193/4 are
as far a beta droop, beta vs. temp, and Vce, etc. there is little hope of matching
them. Which raises the question do we need close matching? - probably not. I'd
shoot to match them at the max dissipation point so that they roughly power share.
Match is not so important when Vce is small as long as the peak current does not
exceed one device plus whatever percentage mismatch of the other device's rating.
At small Vce's the output stage is not power limited. At high Vce with with small Ic
one device can handle the current and again the power is low so similarly sharing is
not so important. Shoot for the best current split at the max dissipation point.
Consider also, that when buying in large quantities it will be easy to get very good
matches among groups of NPN and PNP since there will be many from the same
wafer. So the NPN on the top rail should be a good match to the NPN on the bottom
rail and the same for the PNPs. The top and bottom should then have a very good
match as long as Q11 and Q10 have high beta so their Ic and Ie are approximately
equal. I remember that the beta droop at 10A is very different between the NPN and
PNP types of the MJ21193/4 and by having one of each type on both sides even if they
don't share there will be symmetry top to bottom reducing even order distortion.
An observation is that good same type matching turns into good top to bottom matching
with this topology.
Let's call shorting R26 and removing Q3 as the CFP-3 with gain and one might ask
how does it change when R26 is not shorted and Q3 is back in. Q11's collector is
just one Vbe (Q3) drop above the output so it really doesn't change much but now
most of Q11's collector current, minus that through R26, goes into Q3's base.
Similarly Q11's emitter current, minus that through R24 goes into Q20's base. For
high beta in Q11 Ic is approximately equal to Ie, and if Q3 and Q20 are Vbe matched
and beta matched then they will share current.
I've thought a bit about the current sharing and given how different MJ21193/4 are
as far a beta droop, beta vs. temp, and Vce, etc. there is little hope of matching
them. Which raises the question do we need close matching? - probably not. I'd
shoot to match them at the max dissipation point so that they roughly power share.
Match is not so important when Vce is small as long as the peak current does not
exceed one device plus whatever percentage mismatch of the other device's rating.
At small Vce's the output stage is not power limited. At high Vce with with small Ic
one device can handle the current and again the power is low so similarly sharing is
not so important. Shoot for the best current split at the max dissipation point.
Consider also, that when buying in large quantities it will be easy to get very good
matches among groups of NPN and PNP since there will be many from the same
wafer. So the NPN on the top rail should be a good match to the NPN on the bottom
rail and the same for the PNPs. The top and bottom should then have a very good
match as long as Q11 and Q10 have high beta so their Ic and Ie are approximately
equal. I remember that the beta droop at 10A is very different between the NPN and
PNP types of the MJ21193/4 and by having one of each type on both sides even if they
don't share there will be symmetry top to bottom reducing even order distortion.
An observation is that good same type matching turns into good top to bottom matching
with this topology.
Last edited:
Pete, there is an error in that figure https://linearaudio.net/oops
The correct figure 7.35 is here: https://linearaudio.net/sites/linearaudio.net/files/arto ch7 fig 7.35 corrected.png
Apologies,
Jan
The correct figure 7.35 is here: https://linearaudio.net/sites/linearaudio.net/files/arto ch7 fig 7.35 corrected.png
Apologies,
Jan
