It is true that power amps don't often have much of a noise problem with the input stage, BUT if you used the same topology for a line preamp, or even better yet, a phono preamp, then fet noise becomes very important.
However, this is how it usually works:
Fet noise in Toshiba jfets is usually almost always what it appears in its 'typical noise' graphs. This will be between 0.6 and 2 nV/ rt Hz. at 1 KHz or so. The biggest difference between jfets of a particular type is LOW frequency noise. This can be very different with different batches of parts, but it usually is fairly consistant in one batch, and in general. Sometimes, even Toshiba screws up, and goes out of spec., but not often.
Measuring individual fets requires that you run a specific current, and also know exactly what transconductance the fet has at this current. This is not easy for an amateur. It is easiest to use the Toshiba spec sheets. They are accurate. However, some American manufacturers can lie to you, and give you a noisy device, even when the spec sheet looks good. Beware! This is because they changed how they processed the fet over the decades, and never updated (downgraded) their noise specs. I got caught twice with useless parts, because of this.
However, this is how it usually works:
Fet noise in Toshiba jfets is usually almost always what it appears in its 'typical noise' graphs. This will be between 0.6 and 2 nV/ rt Hz. at 1 KHz or so. The biggest difference between jfets of a particular type is LOW frequency noise. This can be very different with different batches of parts, but it usually is fairly consistant in one batch, and in general. Sometimes, even Toshiba screws up, and goes out of spec., but not often.
Measuring individual fets requires that you run a specific current, and also know exactly what transconductance the fet has at this current. This is not easy for an amateur. It is easiest to use the Toshiba spec sheets. They are accurate. However, some American manufacturers can lie to you, and give you a noisy device, even when the spec sheet looks good. Beware! This is because they changed how they processed the fet over the decades, and never updated (downgraded) their noise specs. I got caught twice with useless parts, because of this.
Hi, John
One of them was J110, as long as I know it... You pointed them a long time ago with semi-words when trying to help an Aussie guy to repair a JC-2 (also with Elso Kwak's help...)....
These parts were Siliconix (the former company)... Then it was bought by Harris Semiconductors... And now it Siliconix again !!!... As an evidence, the production process should have changed with time with such proprietors changes...
Second part could be a dual high Idss claimed to be a close challenger to the 2SK147... Was a Crystalonics C413 if I recal fine, a part that was later new labelled in 2N6550...
Won any prize for pointing them out ? LOL !
Jbaudiophile
One of them was J110, as long as I know it... You pointed them a long time ago with semi-words when trying to help an Aussie guy to repair a JC-2 (also with Elso Kwak's help...)....
These parts were Siliconix (the former company)... Then it was bought by Harris Semiconductors... And now it Siliconix again !!!... As an evidence, the production process should have changed with time with such proprietors changes...
Second part could be a dual high Idss claimed to be a close challenger to the 2SK147... Was a Crystalonics C413 if I recal fine, a part that was later new labelled in 2N6550...
Won any prize for pointing them out ? LOL !
Jbaudiophile
Hi, JustCallMeDad,
When building the Vendettas, John used to measure NOT the noise guaranteed from datasheets... But their ACTUAL NOISE SPECTRUM... As you may know, such noises add vectorially when correlated... But John's trick was NOT to let them have any correlation... That's why I targeted John's own works first... He just uses academic theory with double thinking applied to...
Only some guys were in this league 25 years ago : Stan Curtis (now in charge of Mission), John Linsley Hood, Jean Hiraga and Masao Noro (Stax)... They are all in my best engineers contest hit-parade...
Simply hope that I didn't spoiled out a former John's secret... In fact, I don't think so, because it was in the 2N4401/4403 secret time... Just not so secret with time... (just see Toshiba with low Rbb' 2SB737/2SD756...).
But hope this helps !
Jbaudiophile
When building the Vendettas, John used to measure NOT the noise guaranteed from datasheets... But their ACTUAL NOISE SPECTRUM... As you may know, such noises add vectorially when correlated... But John's trick was NOT to let them have any correlation... That's why I targeted John's own works first... He just uses academic theory with double thinking applied to...
Only some guys were in this league 25 years ago : Stan Curtis (now in charge of Mission), John Linsley Hood, Jean Hiraga and Masao Noro (Stax)... They are all in my best engineers contest hit-parade...
Simply hope that I didn't spoiled out a former John's secret... In fact, I don't think so, because it was in the 2N4401/4403 secret time... Just not so secret with time... (just see Toshiba with low Rbb' 2SB737/2SD756...).
But hope this helps !
Jbaudiophile
JB..., it is not that complicated. I parallel large fets to get lower noise. A Vendetta has 4 large Toshiba fets on the input. I have a noise analyser, so I can pick out the lowest noise units and put them at the input.
Signal is correlated, noise is almost always uncorrelated, so it tends to 'interfere' with itself and not increase as fast as the signal does, when you parallel fets, (or bipolar transistors).
Signal is correlated, noise is almost always uncorrelated, so it tends to 'interfere' with itself and not increase as fast as the signal does, when you parallel fets, (or bipolar transistors).
Well... John,
I was just elarging to JustCallMeDad some pepers I had here and that were just that difficult to obtain here in Europe... (you lived in Swissland for a time... So you can hear what I am saying, eh ?).
Was in June 88 Stereophile issue, page 106... In a special note, J.Gordon Holt put "John Curl hand selects FETS for the SCP-2 with a transistor tester which not only measures noise but classifies that noise according to frequency. Thus he can select both on the basis of low noise and noise with a particular spectral density"...
In read something confirming that in another journal (perhaps UK one... Must check !), that the target was as little spectral noise correlation as possible... I went back to maths books, and made sure that noise adds vectorially...
Simply hope that when enlarging data to other guys about your own work, I just not confuses more !!!
Biut anyway, much thanks for clarifying that point !!! Interesting !Jbaudiophile
I was just elarging to JustCallMeDad some pepers I had here and that were just that difficult to obtain here in Europe... (you lived in Swissland for a time... So you can hear what I am saying, eh ?).
Was in June 88 Stereophile issue, page 106... In a special note, J.Gordon Holt put "John Curl hand selects FETS for the SCP-2 with a transistor tester which not only measures noise but classifies that noise according to frequency. Thus he can select both on the basis of low noise and noise with a particular spectral density"...
In read something confirming that in another journal (perhaps UK one... Must check !), that the target was as little spectral noise correlation as possible... I went back to maths books, and made sure that noise adds vectorially...
Simply hope that when enlarging data to other guys about your own work, I just not confuses more !!!
Biut anyway, much thanks for clarifying that point !!! Interesting !Jbaudiophile
Jba -
Siliconix never was part of the RCA-Harris -GE-Intersil square dance. They were acquired by TEMIC/ Telefunken, which was then acquired by Vishay. As far as I know, they are still using the same fab for their devices (I drive by it all the time). They are one of the few US customers that still bothers to maintain detailed data sheets on their JFETs, though they stopped releasing their product family data sheets a long time ago.
Fairchild has acquired the small signal line of several companies (National, for example). Their data sheets are so cursory that they more or less grudgingly aadmit that the part in question is a JFET and has a breakdown voltage. If you want to use the part for anything else but its intended purpose, you're on your own. I have no idea where they fab their small signal devices.
Siliconix never was part of the RCA-Harris -GE-Intersil square dance. They were acquired by TEMIC/ Telefunken, which was then acquired by Vishay. As far as I know, they are still using the same fab for their devices (I drive by it all the time). They are one of the few US customers that still bothers to maintain detailed data sheets on their JFETs, though they stopped releasing their product family data sheets a long time ago.
Fairchild has acquired the small signal line of several companies (National, for example). Their data sheets are so cursory that they more or less grudgingly aadmit that the part in question is a JFET and has a breakdown voltage. If you want to use the part for anything else but its intended purpose, you're on your own. I have no idea where they fab their small signal devices.
Something like this would be appropriate, for noise measures?
An externally hosted image should be here but it was not working when we last tested it.
?
I would use an AD797. The noise generated by a single JFET will be masked if the opamp is too noisy. You could even dispense with the Source resistor, as it is adding at least some noise.
Although the test will not give absolute measurements, if you check enough devices, you will get some idea of where the bell curve is. That will give you an idea of what the noise variation is for a given population of devices. At that point you can begin choosing the less noisy parts, even if you don't know what their absolute noise levels are.
Grey
Although the test will not give absolute measurements, if you check enough devices, you will get some idea of where the bell curve is. That will give you an idea of what the noise variation is for a given population of devices. At that point you can begin choosing the less noisy parts, even if you don't know what their absolute noise levels are.
Grey
you're forgetting current noise; ad797, lt1028 are best for much lower Z than the 1 K feedback R here
opa227/op27/lt1007 would be better with the 1K feedback shown, ad743/745 is competitive too
AZ op amps might be good for 1/f noise region, perceptual weighting makes this noise unimportant in flat amps, for riaa it is critical
opa227/op27/lt1007 would be better with the 1K feedback shown, ad743/745 is competitive too
AZ op amps might be good for 1/f noise region, perceptual weighting makes this noise unimportant in flat amps, for riaa it is critical
John, I have a few questions concerning the PCB's used in the Blowtorch. My apologies in advance if these have been answered already in this thread.
Looking at the Line Stage boards on the right side I wonder if the JFets and Mosfets are mounted underneath the boards and on the cabinet for better heat dissipation. Assuming that these boards are double-sided are there components, be they SMD or discrete, mounted on the bottom side of the boards?
Switching to the left and looking at the stacked boards I assume that these too are double-sided, right? Are there also SMD or discrete components soldered underneath each of the boards?
Thanks in advance for clearing up things.
Looking at the Line Stage boards on the right side I wonder if the JFets and Mosfets are mounted underneath the boards and on the cabinet for better heat dissipation. Assuming that these boards are double-sided are there components, be they SMD or discrete, mounted on the bottom side of the boards?
Switching to the left and looking at the stacked boards I assume that these too are double-sided, right? Are there also SMD or discrete components soldered underneath each of the boards?
Thanks in advance for clearing up things.
All discrete, except for minidip IC's. Teflon board double sided, heavy ground plained, 6 to220 case mosfets on bottom of line stages. All on same side of Vendetta gold plated boards on left.
I discovered that 2n4401-3 transistors were extra quiet , back in 1968, and I used these parts in the Levinson JC-1 pre-preamp in 1973. However, I did get a footnote in about them in 'Wireless World' in 1971-72 sometime, connected to another article on noise. They are good parts, but Rohm and several other parts are better. It all has to do with base resistivity, or Rbb'.
I discovered that 2n4401-3 transistors were extra quiet , back in 1968, and I used these parts in the Levinson JC-1 pre-preamp in 1973. However, I did get a footnote in about them in 'Wireless World' in 1971-72 sometime, connected to another article on noise. They are good parts, but Rohm and several other parts are better. It all has to do with base resistivity, or Rbb'.
Thanks jba for all the valuable info on noise considerations (low rbb’, selected fets etc.).
Thanks,Grollins and jcx, I looked at the various op-amp datasheets, for R values in the range 500R /1.5K, as you pointed out LT1007/1037 is the best choice and the AD 743 for higher R values. For lower values<500R the LT1028 is the best candidate but the gain could be on the lower side for low transconductance devices.
Thanks,Grollins and jcx, I looked at the various op-amp datasheets, for R values in the range 500R /1.5K, as you pointed out LT1007/1037 is the best choice and the AD 743 for higher R values. For lower values<500R the LT1028 is the best candidate but the gain could be on the lower side for low transconductance devices.
According to the photo we can notice the SY circuit description at the beginning of the thread, on the other hand, like points out Courage, the place between the input fets look particularly empty ? Fets biasing and gain resistors must be under the board? As well as the 1K drain loading resistors?
John, thanks for explaining.
Still puzzled a little.
I assume that on both boards, line stage and Vendetta's the Ground plane is on the side we see in the picture (component side), right?
Is there a specific reason why you have chosen to have the boards upside down, so to speak, because the picture of the inside of the Blowtorch was taken with the bottom panel removed.
Is that correct?
Thanks again.
Still puzzled a little.
I assume that on both boards, line stage and Vendetta's the Ground plane is on the side we see in the picture (component side), right?
Is there a specific reason why you have chosen to have the boards upside down, so to speak, because the picture of the inside of the Blowtorch was taken with the bottom panel removed.
Is that correct?
Thanks again.
- Status
- Not open for further replies.