John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.

iko

Ex-Moderator
Joined 2008
Ikoflexer, it is the wrong piece of equipment for the job. Just use an AC voltmeter with a 10K rolloff filter added. Look for Total noise up to 10KHz. Is it 7uV or 70 uV?

Why is it wrong? I'm curious.

The hp3581a is a selective voltmeter as well; it has a 300Hz wide filter that can be moved up to 50kHz. OK, it doesn't give me the noise up to 10kHz all at once, but it'll give me a good idea of what's going on, no?
 
Why is it wrong? I'm curious.

The hp3581a is a selective voltmeter as well; it has a 300Hz wide filter that can be moved up to 50kHz. OK, it doesn't give me the noise up to 10kHz all at once, but it'll give me a good idea of what's going on, no?

There is a app note that says noise figures are meaningless for audio design. Nat Semi or Analog Devices.

What are you trying to achieve ?
 
Last edited:
Use a low noise voltage reference (such as discussed in the last few pages of this thread) in a voltage regulator to obtain a low noise psu.

a low noise psu ?? what's that exactly ?

it's job is to reject ac ripple! How can it have low noise if it's PSRR is 70db and it's self generated noise is -130db.

Never did get that one!

PS. I think the definition used by the switch mode designers (PSRR+noise) is better.
 
Last edited:
Please everyone, don't play on the thread. If someone doesn't know what we are talking about, then educate yourself with Google, using key words. I don't teach basic electronics here.
We are NOT talking about noise figure, here. If you don't know the definition of noise figure, please don't pose an opinion about it. Yes, and by the way, I own 2 HP 3581 analyzers, and used one on my bench for years, you can TRY to measure the broadband (10-10,000Hz) noise with the HP 3581, but it will probably lead you to confusing results, since you don't see or can't independently compute the answer to the measurement in the first place. How will you know how to add your measurements to compare to National's? If you don't know, you should trust the opinions of people who do. What is in it for us, to mislead you?
 
Last edited:
3581A can go down to 3Hz BW.

Fanuc is right. It's hard for anybody to understand why there's such a desperate need for a pathologically low noise power supply. I've mentioned myself, that's only because the Vendetta input stage has virtually no PSRR. To me it's wrong design decision, but then I cannot debate your "sounds good" class of arguments. I've had enough subjectivism in the cable thread.

I'm asking again, can I post the Vendetta schematic so that anybody could understand what you are talking about?
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
How much more psrr your cascode adds in your stage 2 phono with the 4 parallel 2SKs and common source resistor, on top of what just a load resistor could do for that input stage, and then again how much more your stage 3 with the folded cascode, syn08? Also how much it cuts on input capacitance by cascoding with so low degeneration help by the necessarily diminutive Rs for keeping their noise contribution negligible? Does the topping BJT contribute its residual noise directly to the shorted input noise of the stage and also the Vref noise for its base bias, or those two factors they should be non linearly added? Thanks.
 
For the record, all I have to do is to find a 'perfect' current source, with about 40ma Iq, to replace a pair of resistors that provide current to the first and second stage of the complementary folded cascode input stage. Right now, distortion is OK, noise is OK, but I could save a few cents on the power supply buffer, by making it noisy.
 
How much more psrr your cascode adds in your stage 2 phono with the 4 parallel 2SKs and common source resistor, on top of what just a load resistor could do for that input stage, and then again how much more your stage 3 with the folded cascode, syn08? Also how much it cuts on input capacitance by cascoding with so low degeneration help by the necessarily diminutive Rs for keeping their noise contribution negligible? Does the topping BJT contribute its residual noise directly to the shorted input noise of the stage and also the Vref noise for its base bias, or those two factors they should be non linearly added? Thanks.

About 40dB in all cases. Input capacitance has no bearing with the PSRR or degeneration, it's the price you have to pay for paralleling JFETs. Make no mistake, there's no way around and all solutions (with many small of few very large low noise Toshiba JFETs) give about the same results.

No, the cascode has virtually no contribution to the noise. Noise always adds squarely, if the noise sources are uncorrelated. Take a look at Pass Pearl input stage (XONO is about the same).
 
Last edited:
Well, Syn08, what happens if you put the bw control on 3 Hz? Rough ride? You know, bouncy, bouncy, bouncy? Even with a perfect eyeball to average the random noise, this shows that you don't even know how to use the test equipment efficiently.

Insult reported to the moderators.

And now, on the technical stuff. At 300Hz BW, a 50nV/rtHz source will render some 850nV. OTOH, if you are looking in the specs, the 3581A itself has -130dBV noise in the same bandwidth, that is some 350nV. That's to much to have a precise measurement, and substracting the instrument noise at each frequency is very inconvenient.

At 3Hz BW, the instrument has -150dBV, which is already good enough to be neglected. The only caveat is that the operator has to wait until the filter step response settles. I have measured noise with 0.01Hz bandwidth, that's indeed a pain.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
About 40dB in all cases. Input capacitance has no bearing with the PSRR or degeneration, it's the price you have to pay for paralleling JFETs. Make no mistake, there's no way around and all solutions (with many small of few very large low noise Toshiba JFETs) give about the same results.

No, the cascode has virtually no contribution to the noise. Noise always adds squarely, if the noise sources are uncorrelated. Take a look at Pass Pearl input stage (XONO is about the same).

OK. Good, adds like squares, helps. But capacitance has to do with thd and I am hopping that the cascode shields the elsewise huge Miller to a point. Strong source degeneration would do more but we can't afford high resistor there in this application. I am not about to use loop feedback so I have to keep residual thd acceptable, I skip the Rfb noise on the other hand. I am also not fond of high resistor vref multiplying jfet nvsq by gm, strikes me like a vicious circle, so I think of using a strong idss high gm BL jfet feeding 4 leds for the cascode bjt base bias at about 7v on top of 4 2sks //// with just a common sub 10 Ohm Rsource. Will it be worthwhile to slap an electrolytic across the leds, or add RC time constant? I am saving to purchace a 0.25mV 14 ohm source LowMC cart you see and I wanna build a quiter and higher gain DIY phono version for it.
 
OK. Good, adds like squares, helps. But capacitance has to do with thd and I am hoping that the cascode shields the elsewise huge Miller to a point. Strong source degeneration would do more but we can't afford high resistor there in this application. I am not about to use loop feedback so I have to keep residual thd acceptable, I skip the Rfb noise on the other hand. I am also not fond of high resistor vref multiplying jfet nvsq by gm, strikes me like a vicious circle, so I think of using a strong idss high gm BL jfet feeding 4 leds for the cascode bjt base bias at about 7v on top of 4 2sks //// with just a common sub 10 Ohm Rsource. Will it be worthwhile to slap an electrolytic across the leds, or add RC time constant? I am saving to purchace a 0.25mV 14 ohm source LowMC cart you see and I wanna build a quiter and higher gain DIY phono version for it.

You should be very careful with electrolytics. They have their own noise, as a response to normal ambient vibrations. The cascode base circuit can safely decoupled with a large, good quality, electrolytic but a local regulator (capacitor multiplier topology) may not be ok, if the input stage doesn't already have enough PSRR.

Feedback doesn't help reducing the nonlinear input capacitance (Ciss) distortions, Ciss is outside the loop. But it's not that bad, HPS3.1 measures better than 0.01% at a pathological overdrive of 32Vpp output. See my web site for the exact results and distortion spectrum. Your vinyl pops will be distorted 0.01%, that's not bad :)

Good luck with your cartridge adventure!
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Its the very low signal level that helps us, else without loop feedback the whole phono circuit it would jump to a whopping 2ndH. Anyway, I do have my doubts about the electrolytic that is why I ask, so what do you reckon about less than 200 Ohm worth of Leds DCR driven by 10mA idss 1nV 22ms BL for biasing the base? Leave it alone and skip the cap's distortion and esr too? Because it has to be about 1000uF for low corner. If I feed a 100uF with an R at the bias junction instead, the R part is gonna add its Johnson?

P.S. Let me show you what I am about to make. (19mA Ibias @ 46V).
 

Attachments

  • 1stLN.gif
    1stLN.gif
    3.4 KB · Views: 272
diyAudio Chief Moderator
Joined 2002
Paid Member
Good luck with your cartridge adventure!

I see that you use a VPI HW19 with a unipivot(?), what cart you use and you made 0.4nv phono? Also what kind of speakers? Me I have 95dB spl sens. Can not hear anything with 1 2SK and 0.5mV carts for 56dB phono gain, but with 60-62dB and half the cart signal gen, I think I am pushing it. Hence the 1st stage mod I am thinking to make and we discuss.
 
Its the very low signal level that helps us, else without loop feedback the whole phono circuit it would jump to a whopping 2ndH. Anyway, I do have my doubts about the electrolytic that is why I ask, so what do you reckon about less than 200 Ohm worth of Leds DCR driven by 10mA idss 1nV 22ms BL for biasing the base? Leave it alone and skip the cap's distortion and esr too? Because it has to be about 1000uF for low corner. If I feed a 100uF with an R at the bias junction instead, the R part is gonna add its Johnson?

P.S. Let me show you what I am about to make. (19mA Ibias @ 46V).

I'm not a big fan of LEDs as low noise references. Some models can be pretty silent (in particular IR LEDs), but there's no guarantee, there are huge variations from manufacturer to manufacturer and lot by lot. Definitely avoid anything blue or violet, those are implicitely very noisy.

Anyway, you still need to decouple the reference. Don't worry, that 1000uF cap won't add to the noise, it is shunted by the reference low impedance. That application is a very good candidate for LM329 instead of LEDs, that's what I'm using in HPS3.1 with very good results. You can also add a small resistor to help filtering, the impedance seen towards the cascode base is still much larger, so the cap noise is still shunted.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
John when he picks on you, you protest, now you pick on him, he protests, you pick on him again, will we make a quiet diy phono here, or we will wear the cop hat? I am between Bell labs and NASA here, and instead of helping me not to through good money on a low MC and listen to hiss, you behave like kids. What is this? Half a million people maybe reading globally across a year! Grow up!
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Good idea, Salas, but too much Gm. Why does this matter? (Hint: the current source amplifies its own self noise).

You are considered about gm*DCR Leds I think. On the other hand low gm jfet has more nv, and will not let the leds reach vf bcs low gm is low idss, upping their number and internal impedance. Catch22. Cap across leds is best of both worlds or rc filtered low noise vref like syn's LM or 431 at 1mA me thinks then.
 
Status
Not open for further replies.