A lot of medium power and video transistors have low Rbb.
I am just building an Rbb meter so let´s see what i can find.
We used BC140 / BC160 in MC stages in the 80th. We selected them by listening over headphones.
What a problem can be with transitors not made for audio is that Hfe is sometimes quite low and not very linear with collector current. Hfe can also be very different in NPN and PNP devices, so is 1/F noise corner frequeny.
By the way, i have a working prototype of the BJT INA. To get that going i needed more time then i thought. A lot of circuit parameters had to be changed.
I am just building an Rbb meter so let´s see what i can find.
We used BC140 / BC160 in MC stages in the 80th. We selected them by listening over headphones.
What a problem can be with transitors not made for audio is that Hfe is sometimes quite low and not very linear with collector current. Hfe can also be very different in NPN and PNP devices, so is 1/F noise corner frequeny.
By the way, i have a working prototype of the BJT INA. To get that going i needed more time then i thought. A lot of circuit parameters had to be changed.
Personally, I like to buy quality INA's based on IC'. Twenty five years ago we first started with Scott Wurcer's AD524. While it was not super quiet, it was very quiet, and it have VERY good common mode rejection. This is most important. Today, you can buy even quieter INA's and use them.
Of course, the INA concept can be used for balanced in microphone preamplifier stages. In this case, all discrete designs are best. I have an 8 channel mixing board, returned to me, after years of use, using Toshiba 2SK146 diff pairs in a discrete op amp board, configured as an INA. I also made a similar configuration for Crystal Clear Records, decades ago.
Of course, the INA concept can be used for balanced in microphone preamplifier stages. In this case, all discrete designs are best. I have an 8 channel mixing board, returned to me, after years of use, using Toshiba 2SK146 diff pairs in a discrete op amp board, configured as an INA. I also made a similar configuration for Crystal Clear Records, decades ago.
So far, i did made some tests for a MC preamp with some well known
transistors.
BC550/560 are quiete good in respect of the age of these devices.
2SA970/2SC2240 are not better, or only marginaly.
The best results were those of the 2SA992/2SC1845 , on par with
the old 2SA872A/2SC1775A.
Using four in paralel rejected the preamp noise contribution to
a mere 0.32 nV/sqrt hz.....
Personnaly, i use 16 of them, as i did choose a symetrical differential
as topology, so i had to compensate for the increased noise of the
differential pairs compared to single ended by increasing the numbers.
Anyway, great work by there, we just wait for the final cut if
ever there is one, as you seems to be a die hard perfectionnist...
all the best in your tries..
It is important the amateurs and professionals alike to bring to the front what is really happening with transistors and noise. I hope this gets through.
First, it is important to note that there are 5 major sources of noise in a bipolar transistor.
The first two are related to shot noise. This is both input and output current shot noise. One gets manipulated mathematically and turned into En The other is converted to In. This is where we get the En and In figures on the spec sheet. En can be virtually as low as you could ever measure, BUT In will rise proportionally and ultimately dominate the noise. All you have to do is to change the output current in the bipolar transistor to adjust this.
The other three major noise sources are the departures from the ideal bipolar transistor. These are what makes the essential noise difference between transistors measurable.
These are Base Resistivity or Rbb', BETA (the higher the better, all else being equal). and 1/f noise (a number of mysterious causes).
First, it is important to note that there are 5 major sources of noise in a bipolar transistor.
The first two are related to shot noise. This is both input and output current shot noise. One gets manipulated mathematically and turned into En The other is converted to In. This is where we get the En and In figures on the spec sheet. En can be virtually as low as you could ever measure, BUT In will rise proportionally and ultimately dominate the noise. All you have to do is to change the output current in the bipolar transistor to adjust this.
The other three major noise sources are the departures from the ideal bipolar transistor. These are what makes the essential noise difference between transistors measurable.
These are Base Resistivity or Rbb', BETA (the higher the better, all else being equal). and 1/f noise (a number of mysterious causes).
Before i started the MPP thread i build a balanced phonostage with INA163 at the input.
Then comes the 75usec passive and the rest active in an OPA228. At the output i added a fast 100mA transimpedance OP as buffer. The stage is DC coupled with servo. At the output i added a foil safety cap. I used very good parts and even made an elaborated battery supply. I used this stage in the beginning as benchmark.
It does not sound bad at all but the problem is that the internal feedback resistors are 3kOhm. So for a gain of 38dB that is optimal to avoid noise in the next stage ( why this is so i will try to explain in a later post) you need an Rg of 80 Ohm. That brings the total noise to 2.2nV/qHz. Compare that with my 0.6nV/qHz at the same gain ( i am now so far that i can change the gain to virtually any value without added noise) and you see that my design is in a totally different ballpark.
I am not extremely sensitive to noise, because when the music is playing it is usually swapped when it is not totally incompetently designed but many more or less naive listeners crank their systems full and want noise as low as posible.
I know that a stage with extremely low noise is not a recipy for good sound "per se"
( look for example to some good sounding tube units that have a fair amount of noise) but it does not heart provided everything else is the same.
P.S. working on the BJT version i swapped the plus and minus supply today and have blown my only bord. I am quite frustrated. This BJT stgae is a pain the b.
Then comes the 75usec passive and the rest active in an OPA228. At the output i added a fast 100mA transimpedance OP as buffer. The stage is DC coupled with servo. At the output i added a foil safety cap. I used very good parts and even made an elaborated battery supply. I used this stage in the beginning as benchmark.
It does not sound bad at all but the problem is that the internal feedback resistors are 3kOhm. So for a gain of 38dB that is optimal to avoid noise in the next stage ( why this is so i will try to explain in a later post) you need an Rg of 80 Ohm. That brings the total noise to 2.2nV/qHz. Compare that with my 0.6nV/qHz at the same gain ( i am now so far that i can change the gain to virtually any value without added noise) and you see that my design is in a totally different ballpark.
I am not extremely sensitive to noise, because when the music is playing it is usually swapped when it is not totally incompetently designed but many more or less naive listeners crank their systems full and want noise as low as posible.
I know that a stage with extremely low noise is not a recipy for good sound "per se"
( look for example to some good sounding tube units that have a fair amount of noise) but it does not heart provided everything else is the same.
P.S. working on the BJT version i swapped the plus and minus supply today and have blown my only bord. I am quite frustrated. This BJT stgae is a pain the b.
Wondering, why use the phrase "dynamic headroom" for a preamp. This is usually used for power amps with unregulated supplies that are able to provide more short term peak power. Isn't this simply headroom?
Any unusual distortion mechanism would show up in a null test. Null testing has been used for many years as I'm sure you know. Don't care for his SIM name but the Baxandall configuration is also shown here: "Sound Impairment Monitor (SIM) - Is This The Answer?"
There is also the Hafler configuration for power amp testing.
I used both, the Baxandall and the Hafter test. You can do that digitally now. You can download the "Diff Maker" from Liberty Instruments for free.
Liberty Instruments, Inc. Home Page
Two amps i tested with the diff maker made a fantastic null but when i put them in my system they left a lot to be desired.
I can not explain at the moment why that is so.
When i whould try an explanation a lot of people whould think i am a subjectvist, and reading my theatrice is a waste of time. I did that mistake some years ago when i published my work about cables on the net. I was really surprised how seriously offended some people reacted only because i said that my measurements could explain SOME audible differences.
Still i allow my privat experience to shine through here sometimes.
This is the reason i called this thread MPP ( My Privat Phonosatge).
Liberty Instruments, Inc. Home Page
Two amps i tested with the diff maker made a fantastic null but when i put them in my system they left a lot to be desired.
I can not explain at the moment why that is so.
When i whould try an explanation a lot of people whould think i am a subjectvist, and reading my theatrice is a waste of time. I did that mistake some years ago when i published my work about cables on the net. I was really surprised how seriously offended some people reacted only because i said that my measurements could explain SOME audible differences.
Still i allow my privat experience to shine through here sometimes.
This is the reason i called this thread MPP ( My Privat Phonosatge).
Dear Joachim,
I can confirm your experience about the Diff-Maker test. I had it as a prototype version in my own software and found out that it was not corresponding to the results of a listening test. Good measurents did not necessarily mean good sound. But I'm used to this effect ;-) Anyway, it was very interesting and funny to play around with it. So I keep going for the built-in Belcher test and it's successor noise-distortion test in my analyzer that gives me more informations in relation to listening results. Of course it is necessary to understand the graphs, but that is more a thing of practice and experience. Have fun!
I can confirm your experience about the Diff-Maker test. I had it as a prototype version in my own software and found out that it was not corresponding to the results of a listening test. Good measurents did not necessarily mean good sound. But I'm used to this effect ;-) Anyway, it was very interesting and funny to play around with it. So I keep going for the built-in Belcher test and it's successor noise-distortion test in my analyzer that gives me more informations in relation to listening results. Of course it is necessary to understand the graphs, but that is more a thing of practice and experience. Have fun!
JurgenU, you are right. We have done the same thing, and found it wanting. We hear differences, yet a fairly straight forward differential test may show little or nothing.
I remember back in 1974 when we had our first FFT differential test (10bits) and we tried to do it with IC's vs discrete op amp designs. The differences were lost in the noise. It really impressed my associate, John Meyer, and he virtually became contemptuous of me, because his SPEAKERS showed BIG differences, and my IC's showed small, if even measurable differences. Years later, he went to use IC's almost completely, thinking that they were not audible, as they were not easily measurable. A big mistake, in my opinion.
I remember back in 1974 when we had our first FFT differential test (10bits) and we tried to do it with IC's vs discrete op amp designs. The differences were lost in the noise. It really impressed my associate, John Meyer, and he virtually became contemptuous of me, because his SPEAKERS showed BIG differences, and my IC's showed small, if even measurable differences. Years later, he went to use IC's almost completely, thinking that they were not audible, as they were not easily measurable. A big mistake, in my opinion.
Dear John,
Belcher test is the one suggested by R.A. Belcher. Here is a rough explanation (Belcher.pdf)
Measurement is shown in the pictures Belcher2.gif (with stimuli shown)
and Belcher2.gif (stimuli removed)
One of my favorites is the noise distortion measurement ( it is even more sensitive to distortions )
It is a kind of THD measurement using a band limited noise carrier as test signal. The level outside the noise carrier band is measured and related to the total level. Stimulus and analysis range are separated by a kind of brick wall filter with Kaiser window (highest side lobe damping).
The picture noisedistortion.gif shows the measurement.
I hope all data are legible, this board is a bit tricky, not sure if it works.
Belcher test is the one suggested by R.A. Belcher. Here is a rough explanation (Belcher.pdf)
Measurement is shown in the pictures Belcher2.gif (with stimuli shown)
and Belcher2.gif (stimuli removed)
One of my favorites is the noise distortion measurement ( it is even more sensitive to distortions )
It is a kind of THD measurement using a band limited noise carrier as test signal. The level outside the noise carrier band is measured and related to the total level. Stimulus and analysis range are separated by a kind of brick wall filter with Kaiser window (highest side lobe damping).
The picture noisedistortion.gif shows the measurement.
I hope all data are legible, this board is a bit tricky, not sure if it works.
Attachments
I use this test all the time and when you go back in this thread you will see that i posted some measurement. I think in the US it is known as spectral contamination.
Deane Jensen did a paper about that and there is some work from Rich about the optimum patter. Prof. Dr. Klippel uses something simimiliar that he calls "Sparse Spectrum".
I have a highly modified DAAS, a Klippel Distortion Analyser and i am the German importer of Praxis so i have a lot of experience with this test. It is also very telling in loudspeakers. I recently developped a new dome tweeter with SEAS. On the outside it looks very unspectacular and it is not even very expensive, but we where able to reduce spectral contamination by 6dB.
The sound got much better too. The resolution is now nearly as good as with a ribbon.
My reference is the big RAAL with amorfous core.
I whould really like to publish more measurement but my lab gets totaly new build and i will not be able to use it until mid april.
I also was helpfull to develop the distortion isolation Script in Praxis and that could be the solution for our quest. I just whould like to implement it in 24bit.
Deane Jensen did a paper about that and there is some work from Rich about the optimum patter. Prof. Dr. Klippel uses something simimiliar that he calls "Sparse Spectrum".
I have a highly modified DAAS, a Klippel Distortion Analyser and i am the German importer of Praxis so i have a lot of experience with this test. It is also very telling in loudspeakers. I recently developped a new dome tweeter with SEAS. On the outside it looks very unspectacular and it is not even very expensive, but we where able to reduce spectral contamination by 6dB.
The sound got much better too. The resolution is now nearly as good as with a ribbon.
My reference is the big RAAL with amorfous core.
I whould really like to publish more measurement but my lab gets totaly new build and i will not be able to use it until mid april.
I also was helpfull to develop the distortion isolation Script in Praxis and that could be the solution for our quest. I just whould like to implement it in 24bit.
Of course I know this, Joachim.
Explanation of the measurement was mainly for John, because he asked for it and I suppose he does not know this analyzer because it is nearly unknown in the U.S.
;-)
Yes, I'm aware of Diff Maker and I've used Bill Waslo's LAUD from Liberty Instruments for more than 10 years.
My point is that if anyone speculates on some sort of yet undefined form of distortion it will show up in a null test.
You mentioned something happening in the first 100 mS of a transient - 100 mS is a long time and it would surely show up in a null test. Did you mean uS ?
http://www.diyaudio.com/forums/analogue-source/154210-mpp-80.html#post2071505
I'm not saying that I agree that this is a problem in any good amplifier.
Last edited:
I'd call this high frequency headroom to indicate that the emphasis is on HF, but as I see it there is nothing dynamic about it. It does not change with level or time. OK, perhaps it is nit picking ....