Lumba Ogir said:
Totally incorrect, on all counts. I am about to believe you know how it really is, and twist the truth just to stir the pot and have some fun.
I've never seen such an amount of gross errors per line posted. Details are available, but I suspect you won't even bother to debate the facts.
Iain McNeill said:
syn08
on topic please.
no more personal attacks.
Ok.
Paralelling devices, in particular BJTs, destroys the synergetic relation between the electrons and the holes. Therefore, I/Gm^3.14 decreases to the point of no return, where the sonic blue marshmellows meet the acoustic signature of the pn junctions.
JFETs are better, because marshmellows are green and silent. The BJTs cat scratching noise is absent and also the phonons are sparse, so electrons have a longer average path before knocking their buts against the phonon quantum wells.
OK, I`m just saying it does not sound good, mainly because of mismatch in signal processing and increased capacitances
hitsware said:
Bipolars poison themselves too. For example, they destroy themselves by own radiation. I gave an example already about a negative voltage generated on base-collector junction by breaking-out base-emitter junction by a reverse potential. Now, do that and compare noise figures before and after.
Lumba Ogir said:
It depend on concentration of free blue marshmellows on surfaces of semiconductors, how is it's grow relates to decrease in the signal source dynamic resistance.
Paralleling can be relatively useless, unless you have reasonably matched devices or significant source or emitter resistance, and this extra resistance tends to make things noisier.
Good beta or Idss matching is what I use. It is true that paralleling devices adds to significant non-linear capacitance on the input and this is a bad thing. In fact, even a single device, especially a fet can have too much non-linear capacitance for many situations, including directly after a volume control.
For example, before the dual 2SK170 or 2SK240 was available, I designed a line stage with 2SK146 and 2SJ73 dual pairs. Even with a 10K pot, it created some extra distortion due to the nonlinear capacitance. Changing to the 2SK240 and 2SJ75 pair reduced the distortion, significantly. Today, I use the 2SK389 and the 2SK109 for this position. However, none of these devices are available to anyone except Erno , so you will have to make do with what you can find, AND many of you have been lax in this.
Good beta or Idss matching is what I use. It is true that paralleling devices adds to significant non-linear capacitance on the input and this is a bad thing. In fact, even a single device, especially a fet can have too much non-linear capacitance for many situations, including directly after a volume control.
For example, before the dual 2SK170 or 2SK240 was available, I designed a line stage with 2SK146 and 2SJ73 dual pairs. Even with a 10K pot, it created some extra distortion due to the nonlinear capacitance. Changing to the 2SK240 and 2SJ75 pair reduced the distortion, significantly. Today, I use the 2SK389 and the 2SK109 for this position. However, none of these devices are available to anyone except Erno , so you will have to make do with what you can find, AND many of you have been lax in this.
john curl said:
Nope.
There is very little reason why one should match devices (especially JFETs) for low noise aplications. Matching JFETs will do good for cancelling distortions and minimize the offset in complementary stages.
As for bipolars, carefully MISmatching devices in parallel will actually linearize the stage - see the Schmook/Gilbert multi-tanh stage.
stinius said:lineup said:
Thanks stinius.
I have always like your sensible posts. 😉
So what you are saying is that the VALUE in Ohm, is the crusial thing here.
What I pointed out, regarding my good circuit
is that when we start to concern about resistors contributed noise
then the rest of the design
must be of awfully good quality.
Ot is similar to this hypotetic review of Schumacher Ferrari F1 car:😀 😀 God! How much sense in such a comment ...
A lower resistance higher conductance resistors
and my circuit is good enough.
Maybe to be prefered over syn08 TOPIC construction?
Now what type of resistors we use, have a greater influence, still
Is there any better choice here, tha ordinary 0.25 Watt metalfilm 😕
I mean, being a non-yet-resorted-to-tubes-man,
I dio not suggest wirewound, cabon or carbon composite resistances.
Unless I am a noise-is-an-overrated parameter-in-audio thinker.
/halojoy - foemerly the Great Carl Gromanswe of Sweden
I would like to see more people try stuff rather than scaring them off. Saying you need to buy a bag of 1000 2SK170's to get enough for a few matched quads is absurd. Noise on a FET goes roughly as the fourth root of current so there is a HUGE smoothing of the effect of current sharing. The noise variance on ANY four GR grade 2SK170's paralleled is insignificant. This is provable mathematically and practically.
OTOH I AM scared off by Allan Wright's and Lynn Olsen's electronics. The thought of sourcing all those sometimes very obscure valves and all those transformers is too much for me. More power to those that go that route. I was facinated by Lynn's interstage coupled push-pull triode amp. My friend's grandfather, who worked for Simplex Wire and Cable on the first trans-Atlantic phone lines said nothing could beat that for sound (pre-WWII). It probably does still compete favorably.
OTOH I AM scared off by Allan Wright's and Lynn Olsen's electronics. The thought of sourcing all those sometimes very obscure valves and all those transformers is too much for me. More power to those that go that route. I was facinated by Lynn's interstage coupled push-pull triode amp. My friend's grandfather, who worked for Simplex Wire and Cable on the first trans-Atlantic phone lines said nothing could beat that for sound (pre-WWII). It probably does still compete favorably.
scott wurcer said:
Absolutely agree. People tend to go on like a pack of old women
WRT matching jfets.
I get a batch of a few hundred, a sheet of A4 and put a few strips
of double sided tape on it. Underneath that two rows, one with
current IDSS and one with ID at 20R inserted in series with source.
You can get a system working, wait an exact time for the
measurement where the current settles, say 5 to 10 sec and
quickly jot the numbers and stick the fet on its head above them.
It doesn't take too long to plough through a few hundred. For
most people here this will last quite a while.
I have found this to work extremely well and can quickly select
same sex or opposite sex pairs/multiples that bias up within a
mV.
cheers
Terry
I usually do the same with bipolars
when I order big batch of them
sort them accoding to vbe or hfe
with a little note beside each of the transistors
when I order big batch of them
sort them accoding to vbe or hfe
with a little note beside each of the transistors
Who said 1000's? NOT ME! But 25-100 of a certain device can be helpful for a good match.
In commercial production we sometimes have access to 1000's of devices, after all, we have to make 100's of amps, preamps, etc. to make it worthwhile to do so.
SINCE we have to use 1000's of devices, it makes sense to hire someone to presort them to some level in advance. Usually 1ma Idss for fets. Is this so difficult?
Now, you amateurs, how many parts do you need? 25, 100? It is best to buy a reasonable number of same part, all at once and sort them. Then when you need a pair, you can pick them out of the sort. You may not need pairs always, so some odd sort devices might be used for something else, besides a parallel pair or differential pair. This is not difficult, and no worse than matching output tubes.
What I was referring to was paralleling NON-SORTED parts and hoping for optimum results. Usually, some devices will current-hog and basicially turn off the others. Kind of a waste, isn't it?
Now, someone here provided a link that SHOWED that 2SK170's and 2SJ74's were still for sale at a reasonable price in Europe. Then why the whining that you can't get 'exotic' parts? You haven't even tried.
In commercial production we sometimes have access to 1000's of devices, after all, we have to make 100's of amps, preamps, etc. to make it worthwhile to do so.
SINCE we have to use 1000's of devices, it makes sense to hire someone to presort them to some level in advance. Usually 1ma Idss for fets. Is this so difficult?
Now, you amateurs, how many parts do you need? 25, 100? It is best to buy a reasonable number of same part, all at once and sort them. Then when you need a pair, you can pick them out of the sort. You may not need pairs always, so some odd sort devices might be used for something else, besides a parallel pair or differential pair. This is not difficult, and no worse than matching output tubes.
What I was referring to was paralleling NON-SORTED parts and hoping for optimum results. Usually, some devices will current-hog and basicially turn off the others. Kind of a waste, isn't it?
Now, someone here provided a link that SHOWED that 2SK170's and 2SJ74's were still for sale at a reasonable price in Europe. Then why the whining that you can't get 'exotic' parts? You haven't even tried.
john curl said:
Wrong again. That's not going to happen, unless you feed the JFETs from a current source, which doesn't make much sense anyway. Otherwise, paralleling unmatched JFETs is (from a noise perspective) exactly equivalent to a single device carrying all the current. If N/P devices have to be used (like in your Vendetta) all you have to match is the sum of all Idss, and that's for keeping the offset within reasonable limits,
You can avoid any kind of matching if you are using a servo as in my headamp. This allows using unmatched devices straight out of the tube. I swapped randomly enough JFETs (tenths of both types) to confirm that the noise doesn't change a iota, while the servo keeps the output within under 1mV offset.
Again, there is a good case for matching for distortion cancellation in complementary differential pairs, which is not the case here.
john curl said:
Can't happen within a color grade. You can pickout at random from a bag of 2SK170-GR's and the current hogging is small potatoes. I will go further and bet any group of 8 picked at random can be divided into two groups of four that match quite well even for distortion. Bit of work swapping around, but saves money.
Get a good look that avatar will go soon. (AD524 paper, 1982)
- Status
- Not open for further replies.