There are really two different types of "low noise" transistors optimized for different circuit environments. Most BJTs that I have seen marketed as low noise devices are optimized for low noise with medium to high source impedance environments. One of the better known of these types is the 2N5089. These devices have very high current gain, but often at the expense of higher base resistance. They are NOT particularly low noise in low-impedance circuits. On the other hand, the 2N4401 is known to be a very good low-noise performer in low source impedance environments, like a moving coil cartridge pre-preamp. These devices have very low base resistance, but not great current gain.
Cheers,
Bob
Early BJTs had huge excess hiss, static, popcorn, and grunge.
Improved processing (say 1970; thanks Andy) made a real difference. As said, the 2N5089 types boasted "low noise". To make the most money per nano-acre of processed Silicon, they were very small devices. The "low noise" was very true for sources 5k-20k. This covers many tape-heads, phono, and microphone transformers. We could basically stop selecting-out the least-hissers and use "any" transistor of that type.
Silicon cookers have never (maybe recently) grasped the LOW-z audio noise market. The required large area offends their bean-counters. Why would SMALL-signal designers want large (expensive) transistors? There have been a few types specifically masked and marketed for low-Z noise; sadly most did not sell in quantity so were dropped.
As Bob says, there ARE low-rb high-hFE parts. "Switches" are optimized for large current with low drive power. So lowish rb and high-ish hFE. A really big base (low rb) can degrade hFE (more relative leakage around the edges) but switch-makers do as good as they can. If made in dirty Silicon they would still hiss. But dirty Silicon also reduces yield and field-reliability. Does terrible things to dense logic (they all use the same bar of Silicon). So modern Silicon is usually squeaky-clean. Parts like 2N4401 are historically Low Noise (in low-Z work) even though sold as "Switch".
Many of the old Marketing buzzwords just never go away. There are still catalogs which list the old-old-old '741 opamp as "improved". Yeah, better (maybe) than the even older '709.
Improved processing (say 1970; thanks Andy) made a real difference. As said, the 2N5089 types boasted "low noise". To make the most money per nano-acre of processed Silicon, they were very small devices. The "low noise" was very true for sources 5k-20k. This covers many tape-heads, phono, and microphone transformers. We could basically stop selecting-out the least-hissers and use "any" transistor of that type.
Silicon cookers have never (maybe recently) grasped the LOW-z audio noise market. The required large area offends their bean-counters. Why would SMALL-signal designers want large (expensive) transistors? There have been a few types specifically masked and marketed for low-Z noise; sadly most did not sell in quantity so were dropped.
As Bob says, there ARE low-rb high-hFE parts. "Switches" are optimized for large current with low drive power. So lowish rb and high-ish hFE. A really big base (low rb) can degrade hFE (more relative leakage around the edges) but switch-makers do as good as they can. If made in dirty Silicon they would still hiss. But dirty Silicon also reduces yield and field-reliability. Does terrible things to dense logic (they all use the same bar of Silicon). So modern Silicon is usually squeaky-clean. Parts like 2N4401 are historically Low Noise (in low-Z work) even though sold as "Switch".
Many of the old Marketing buzzwords just never go away. There are still catalogs which list the old-old-old '741 opamp as "improved". Yeah, better (maybe) than the even older '709.
So has anyone compiled a list of suitable transistors.
I found this small list. Not sure of its accuracy.
2N4401 - GP
SS9014 - LN
MPSA18 - LN/HG
MPSA05 - GP
MPSA06 – GP
MPSA42 – GP
PN100 - GP
KSC815 - HG
KSC1008 - GP
KSD1616 - GP
ZTX1051a - LN/HG
ZTX690b - HG
Legend:
LN = Low Noise
HG = High Gain
GP = General Purpose
I found this small list. Not sure of its accuracy.
2N4401 - GP
SS9014 - LN
MPSA18 - LN/HG
MPSA05 - GP
MPSA06 – GP
MPSA42 – GP
PN100 - GP
KSC815 - HG
KSC1008 - GP
KSD1616 - GP
ZTX1051a - LN/HG
ZTX690b - HG
Legend:
LN = Low Noise
HG = High Gain
GP = General Purpose
It might be a better idea to create a list of candidates and then decide for yourself which are suitable / unsuitable to a particular task. One way to get a large number of possible candidates, is to look at the 3rd edition of Art Of Electronics, by Horowitz and Hill. Table 8.1a mentions more than 50 different BJTs, and Table 8.2 has more than 40 different JFETs.
It also might be a good idea to scan through Bob Cordell's power amplifier book {the subject of this Forum thread} and notice which transistor types he uses. They might be good candidates, for you to decide whether you think they are suitable or unsuitable.
And you can also visit Bob's website "Cordell Audio". He gives away a free library of transistor model parameters for SPICE simulation. Maybe the transistor types that Bob has bothered to prepare for simulations, might be good candidates for you to decide whether you think they are suitable or unsuitable.
It also might be a good idea to scan through Bob Cordell's power amplifier book {the subject of this Forum thread} and notice which transistor types he uses. They might be good candidates, for you to decide whether you think they are suitable or unsuitable.
And you can also visit Bob's website "Cordell Audio". He gives away a free library of transistor model parameters for SPICE simulation. Maybe the transistor types that Bob has bothered to prepare for simulations, might be good candidates for you to decide whether you think they are suitable or unsuitable.
Greg Erskine i believe have such compilations on his webpage and a thread here...Transistor data with sorting
Thanks mark. I do have Bob's files and am familiar with running simulations with Ltspice. I will take on board your advice. Thank you.
I guess, I was just wondering if anyone else had some they recommend.
I have seen that you have in the past posted some plots of the gain characteristics of certain transistors.
My original question was would a higher gain transistors like a KSC1845 be a better choice in this location.
So I really just wanted some clarity on that. I will look into the suggestions made by yourself and others.
Thanks
I guess, I was just wondering if anyone else had some they recommend.
I have seen that you have in the past posted some plots of the gain characteristics of certain transistors.
My original question was would a higher gain transistors like a KSC1845 be a better choice in this location.
So I really just wanted some clarity on that. I will look into the suggestions made by yourself and others.
Thanks
Bob answered by saying,
In the application of an power amplifier IPS, one assumes that the source R from the pre-amp, is rather low, <1KOhm, so a KSC1845 is not necessarily the best choice and a 2N5551 may very well be the better choice from a noise perspective. I guess one way to improve its noise performance is to use a 2n4401 and cascode it for higher voltage gain stages, but is it necessary?
In the book, p114, Bob's measurements on the BC-1 amplifier example (using a 2n5401) is 9nV/rtHz, saying it is respectable. Input series R4 and FB shunt R6 each 1K are the largest noise contributors.
Unfortunately, in real world -music- there's no high speed square wave signals artificially created for the
designer's pleassure.
Is the relatively slow musical signal and it's journey into the amplifier
who has to move ours solid state devices , charge and discharge them with
the precision required .
That is the dificult task .
BR
Last edited:
You are correct. We always must remember that measurement results in most cases are just looking for symptoms of what might be shortcomings that will affect the sound.
Amplifiers that had inadequate slew rate in the past often did not sound so good. Then there is the question of what is adequate slew rate. We don't listen to full-power 20 kHz sinewaves either, but the rate of change of such a signal gives us a starting point to think about what is adequate in the worst case. Let's not forget that the anti-aliasing filters in a CD recording chain will often put fairly hard limits on achievable relative slew rate on CD recordings (maybe less-so in some other H-Res formats).
If we are talking about really hi-end audio, we never even want to get close to an amplifier getting to slew rate limiting. As discussed previously in this thread, we want some margin against slew rate limiting. A factor of 5 is not unreasonable, and not difficult to obtain with a skilled designer.
At the same time, having a high slew rate does not necessarily guarantee that the amplifier will accurately respond to high frequency material with very low distortion. An example of this is an amplifier that achieves high slew rate by using a class B or class AB front-end arrangement. In my view, every stage in an amplifier, with the exception of the output stage, should always be operating in class A. Again, this is not hard to do.
Cheers,
Bob
Thanks Mark, I'll check it out tonight.
Can you do me a favor is Imax the same as Itail ?
Oh. I see my mistake Itail does equal Imax but re' is not caculated on the quiescent current but at Itail or Imax.
So re' is really re' (max), that's were I went wrong. But this was a great learning experience so its all good.
Now I'm really confused because I just read this at the end of chapter 2.
the value for Itail is 2mA and its saying that only 1mA will be available to charge or discharge the capacitor from the bias point which is 1mA.
Could someone please clear this up for me. It seams now that I was wrong again. I(tail) does not equal I(max) instead I(max) = I(tail) - I(quiescent)
Sorry to ask what to most of you is probably a basic thing. But I must be missing something. If you turn to page 162 it says The LTP of Figure 6.13 etc... When you look at figure 6.13 it has a IPS gm of 2ms. and a I(max) of 1mA. So I assume that I(tail) = 2mA
Now if you turn to page 238 you will see a schematic which seams to match the Re and re' values. It has a I(tail) of 2mA but it has a IPS gm of 4ms.
Bob would you mind clearing this up for me a its really got me stuck.
Now I'm really confused because I just read this at the end of chapter 2.
the value for Itail is 2mA and its saying that only 1mA will be available to charge or discharge the capacitor from the bias point which is 1mA.
Could someone please clear this up for me. It seams now that I was wrong again. I(tail) does not equal I(max) instead I(max) = I(tail) - I(quiescent)
Sorry to ask what to most of you is probably a basic thing. But I must be missing something. If you turn to page 162 it says The LTP of Figure 6.13 etc... When you look at figure 6.13 it has a IPS gm of 2ms. and a I(max) of 1mA. So I assume that I(tail) = 2mA
Now if you turn to page 238 you will see a schematic which seams to match the Re and re' values. It has a I(tail) of 2mA but it has a IPS gm of 4ms.
Bob would you mind clearing this up for me a its really got me stuck.
Imax, the magnitude of current available to charge or discharge the Miller capacitor, depends on the type of load circuit that is used to load the IPS LTP. In circuits where the load is a simple resistor, the net maximum current available to charge the capacitor is Itail - Iquiescent of the LTP, since the load resistor is conducting a nominal amount of current that is 1/2 I tail. This means that Imax will be 1/2 I tail.
The situation is different if the LTP is loaded with a current mirror. In this case it is easy to see that Imax will equal Itail because all of the tail current can be routed to the Miller capacitor when either of the LTP transistors is fully on and the other is off.
Cheers,
Bob
Thanks Bob, now it makes sense why it appeared that the value of Imax was not consistent.
So just to ensure that I have got this right now.
In the case where the LTP uses a current mirror where Itail is 1mA the re' of each input differential pair transistor in their quiescent state will be Vt / Ic => 26/0.5 => 52 and Vt / Ic => 26/1 => 26 in the condition where Imax equals Itail. Which is the point at which the LTP it slew rate limited by Imax in this case 1mA.
Last edited:
OT Alert!
Apologies if it’s been mentioned already in this thread, but Bob will be giving a seminar with Demian Martin, on measuring amps and loudspeakers on the first day of Burning Amp, Saturday Nov. 9 in San Francisco.
It’s a 2 day event this year. Saturday is the Seminar and the Preamp Building Camp
Sunday will be like the traditional Burning amp. With lots of presentations and components to look at and admire.
Info: burningamp.org
Apologies if it’s been mentioned already in this thread, but Bob will be giving a seminar with Demian Martin, on measuring amps and loudspeakers on the first day of Burning Amp, Saturday Nov. 9 in San Francisco.
It’s a 2 day event this year. Saturday is the Seminar and the Preamp Building Camp
Sunday will be like the traditional Burning amp. With lots of presentations and components to look at and admire.
Info: burningamp.org
Last edited:
