Try LM317 as it is slightly better on hiss compared with the LM7812. Roughly speaking the LM317 could be said to be a LM781.25. The gain is set like an op amp. Usually 240R recomended for the upper arm. 2K0 and 240 R. gives 1.25 {( 2000/240 ) +1 } = 11.67 V. A capacitor of between 10 uF and 47 uF can be used to loose most of the hiss if across the 2K resistor. Although there are better devices LM317 is not at all bad. You can use a simple RC stage by finding the right values. That sometimes beat everything.
I tried simplifying the Quad 33 M1 just to give 3180/318 uS. I used the C1 section. I did the 75 uS with a buffer and passive EQ. The big surprise was no real difference. Although the Quad RIAA is arrived at by tweaking the gain and EQ to offset the lack of HF loop gain it is not a bad compromise. My surprise is the Quad is not quite as open as the simplicity suggests. It's active 75 uS is not it's worst problem. An op amp that out performs it is MC33078. NE 5532 slightly less good whilst being better than the Quad.
I tried simplifying the Quad 33 M1 just to give 3180/318 uS. I used the C1 section. I did the 75 uS with a buffer and passive EQ. The big surprise was no real difference. Although the Quad RIAA is arrived at by tweaking the gain and EQ to offset the lack of HF loop gain it is not a bad compromise. My surprise is the Quad is not quite as open as the simplicity suggests. It's active 75 uS is not it's worst problem. An op amp that out performs it is MC33078. NE 5532 slightly less good whilst being better than the Quad.
I don't see any hints about noise performance in the Data Sheet. Not surprising, since the devices are intended to drive motors, SMPS switch transistors, etc.
Certainly plenty of gain to play with! And they're plenty fast enough, despite un-spectacular collector capacitances.
The voltage rating may be a problem unless you run the devices in cascode.
In the 1970's there was said to be a Philips TV transistor that was ideal. Philips made them in small batches for repairs and sold them at a loss. In the end Philips insisted the chassis number of the TV given when ordering. I didn't take much notice except to note small power transistors should work well. I don't know the rbb- of BD139/140. It should be reasonable. Lentek used devices like these, they removed the numbers. T0126 types.
I recently looked at the question of statistical noise cancellation. It seems to me the paralell resistance of the combined devices comes to much the same numbers as statistical averaging. However the noise spectrum to my ears of combined devices is less blue. This better supports the statistical averaging. It also says to me lower noise due to multiple devices is worthwhile.
I recently looked at the question of statistical noise cancellation. It seems to me the paralell resistance of the combined devices comes to much the same numbers as statistical averaging. However the noise spectrum to my ears of combined devices is less blue. This better supports the statistical averaging. It also says to me lower noise due to multiple devices is worthwhile.
This thread prompted me to make some (rather informal) tests on a large variety of transistors, including traditional low-noise types (the ones intended for that application), medium power in TO126, TO5, Japanese, Europeans, US, linear, switching, high-voltage, etc.
The low noise ones were certainly not the best, BC338's were somewhat better, but the clear winner (for my selection anyway) was the ZTX690, with the 2N5321 a close second.
I made the measurements at Ic=1mA and Vce=10V
The low noise ones were certainly not the best, BC338's were somewhat better, but the clear winner (for my selection anyway) was the ZTX690, with the 2N5321 a close second.
I made the measurements at Ic=1mA and Vce=10V
I know it's against the science. That's how I hear it. Could be when hiss is very low it's a function of ears and not devices. Fair enough as it is ears we need to please. I don't think anyone has ever asked that question in this way. If I add two noise sources at the level where microphones typically work I don't just hear a reduction in noise. I hear much less high frequency noise. Being very happy it worked at all I never looked at it on an analyser. I just have a hunch there is a greater effect than just simple rbb-. If I'm wrong no big deal.
I hear the higher frequecies as louder. A resistor seems to have a whiter spectrum. Seems is the word as I never tested the reality of it. I don't say it is an excact 3 dB per octave ( 3.0103 ? ). It sounds bass filtered. Add three near identical noise sources and the reduction seems greater than the maths suggests. Listening carefully the HF seems to have dropped more. Most reading seems to suggest noise is always white. I have my doubts. As this is at the level when we stop hearing hiss it's almost certainly how my ears hear. We are alas far behind our tools.
Not true, resistance at zero DC current yes, in a semiconductor device as only a partial contributor you have to look at everything.
Thanks, I actually have a few LM317 to try but wasnt sure of the best way to use it here. Fantastic information!
/Martin
Measuring en and rbb'
I scoured datasheets, and selected and purchased hundreds of low-noise BJT transistor candidates. Paul made a low-noise single-stage transistor amplifier, see AoE III, Figure 8.92, in which we could place our D.U.T. candidates, and he took spectral plots of their noise at different operating currents. I entered the masses of data into Excel, and modeled the e_n, i_n and r_bb' noise components to determine the minimum-possible noise from r_bb' Johnson noise, but at Ic currents low enough not to create base-current i_n * r_bb' noise. Then I made a spreadsheet of about 60 useful transistor candidates, see Table 8.1 on page 501, for y'all to feast on. As a bonus, you also get measured values of the Early voltage V_A, and plots of beta vs current, Figure 8.39 for all the transistors.
We also went through the same exercise for JFETs, and created tables and graphs from our measurements.
This is information I would have died for in years past, that's what motivated me, and now I can go forward on designs using all this grand data. DIY experimenters can too, if they get a copy of our book.
I scoured datasheets, and selected and purchased hundreds of low-noise BJT transistor candidates. Paul made a low-noise single-stage transistor amplifier, see AoE III, Figure 8.92, in which we could place our D.U.T. candidates, and he took spectral plots of their noise at different operating currents. I entered the masses of data into Excel, and modeled the e_n, i_n and r_bb' noise components to determine the minimum-possible noise from r_bb' Johnson noise, but at Ic currents low enough not to create base-current i_n * r_bb' noise. Then I made a spreadsheet of about 60 useful transistor candidates, see Table 8.1 on page 501, for y'all to feast on. As a bonus, you also get measured values of the Early voltage V_A, and plots of beta vs current, Figure 8.39 for all the transistors.
We also went through the same exercise for JFETs, and created tables and graphs from our measurements.
This is information I would have died for in years past, that's what motivated me, and now I can go forward on designs using all this grand data. DIY experimenters can too, if they get a copy of our book.
Thank you. Actually, I need to point out one thing: I'm not a PhD anything. After a degree in Physics, I dropped out of grad school in Chemical Physics and went back and got an EE degree; that's where I found my real passion. I'm a practicing circuit-design engineer, a pretty good one I'd like to think, but just an engineer who loves to show others his design tricks. I'm the engineer (with all the stuff that needs to be taught) in the Horowitz & Hill team, and Paul is the teacher.
We're working hard on the x-Chapters book, our editor wants it this summer, but I want to put more stuff in first, we'll see.
No one of these, the classic came with a transistor/JFET head (well the classic one had "real" meters).
Quan-Tech 5173 - In Stock, We Buy Sell Repair, Price Quote
Ahh, I see. Now the modern way is to use a highly-linear 24-bit ADC and do an FFT to see the noise and distortion vs frequency. You can aid calibration by testing with low-distortion oscillators, etc., but if noise and harmonics aren't showing, they aren't there. If they do show, maybe they still aren't there, that's one reason for low-distortion oscillator testing.