Question on input noise current

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Fred,

What you say is most certainly right, and John Curl has said
the same thing too. However, we were more specifically
discussing op amps, or at least I were. AFAIK there is still no
FET-input op amp that can rival the noise figure of the bipolar
ones for low source impedances. However, the question how
the bipolar ones perform noise-wise for ill-matched impedances
seems to be an unknown in the equation.
 
Re: Maybe not....

Fred Dieckmann said:
[B
My favorite low noise jfet has a voltage noise of less than 0.8nV per square root hertz, a little better than the MAT-02 bipolar. [/B]

And the '797 has an en of 0.9nV per rt Hz, but that's the noise density for the whole circuit, not just the input stage transistors. I don't know for sure, but I'll bet there are discrete bipolars out there these days with even lower noise voltage densities; higher intrinsic transconductance has its merits.

That's not to say that you can't get really good performance out of FETs- my current phono stage is darn quiet (not as much as the '797 stage, but quiet enough!) and uses the cheap NTE458 as the bottom of a cascode.
 
I think you must state what typical source you have in mind because only MC-pickups, strain gauges microphones and thermo elements need VERY low noise amps. If you are after noise meaurements and the output is an amplifer, the source impedance will be low => no problem with choice of BJT, JFET input.

If you want this amp to be a preamp for oscilloscope or noise meter, don't overdo it. A single LT1028/1115/AD797 feed with batteries will do the jobb.

EDIT: Correct there is no JFET opamp in the world which can compete with a BJT opamp for LOW IMPEDANCE signal sources. If you want JFET's you have to buiild a discrete amp.
 
peranders said:
I think you must state what typical source you have in mind because only MC-pickups, strain gauges microphones and thermo elements need VERY low noise amps. If you are after noise meaurements and the output is an amplifer, the source impedance will be low => no problem with choice of BJT, JFET input.

With the possible exception of MC pickups, I had not planned
to measure any of those types of devices. There are quite many
others which might require a low-noise amp. Voltage references,
eg. diodes, is one such example. I leave the rest of them to
your imagination. In most cases the source will most likely be of
low impedance. However, I think there might be cases where
one wishes to measure also higher-impedance sources, but if
those cases are important enough to consider is something I
haven't made up my mind about.


If you want this amp to be a preamp for oscilloscope or noise meter, don't overdo it. A single LT1028/1115/AD797 feed with batteries will do the jobb.

I can't believe what I hear, or see rather. The master of
million-component designs tells me not to overdo it!!! ;)

You do have a point, though, I happen to have a tendency to
overdo things. As for supply, I definitely will use batteries. I
see no point in other solutions for this type of device. Whether
just a single low-noise op amp would do is also a question I
have asked myself. Not having fully convinced myself either way,
it seems the instrumentation-amp variant would have its
advantages in some cases. Feedback is welcome on this issue,
though.
 
diyAudio Retiree
Joined 2002
Ignore the man behind the curtain

"And the '797 has an en of 0.9nV per rt Hz, but that's the noise density for the whole circuit, not just the input stage transistors."

No, that is input referred noise and the major source of this noise is the input diff pair. Also of interest is that the AD797 requires a 100 ohm series resistor for stability for non inverting circuits that bumps the noise voltage up to about 2 nV per square root hertz. This is about twice the voltage noise of a 2SK389.

Strain gauges can easily be a high enough resistance (1 K ohm and more) that the current noise of a AD797 would make it unsuitable for a strain gauge amplifier. Some of the high output cartridges have coil resistances exuding 200 ohm making the AD797 noisier the using a good discrete jfet design.

It is not that hard to build a hybrid circuit with 3 or 4 low noise jfets and a decent but not ultra low noise op amp. I have seen app notes for such hybrid op amp circuits. they don't have be that complicated.

Do think maybe John Curl has actually learned this stuff better than the rest of us in 30 years of designing low noise audio circuits?

Could be..........
 
peranders said:
Christer,

1 LT1028 etc
2 9 V batteries
2 electrolythics 100-470 uF
2 100 nF polyester, ceramic
resistors for 100 or 1000 gain

Maybe input protection, zeners...

That's all.

Note also that this will get you broadband noise, not weighted.

Yes, I know it can be that simple. It doesn't make a true
diff. amp though, which might be useful sometimes. I will
probably add some selectable LP and HP filters. Probably
also input protection, although I am afraid that might add noise,
so it should probably be possible to switch it off.

BTW, I couldn't get that link to work either.
 
Re: Ignore the man behind the curtain

Fred Dieckmann said:


No, that is input referred noise and the major source of this noise is the input diff pair. Also of interest is that the AD797 requires a 100 ohm series resistor for stability for non inverting circuits that bumps the noise voltage up to about 2 nV per square root hertz. This is about twice the voltage noise of a 2SK389.

Strain gauges can easily be a high enough resistance (1 K ohm and more) that the current noise of a AD797 would make it unsuitable for a strain gauge amplifier. Some of the high output cartridges have coil resistances exuding 200 ohm making the AD797 noisier the using a good discrete jfet design.

It is not that hard to build a hybrid circuit with 3 or 4 low noise jfets and a decent but not ultra low noise op amp. I have seen app notes for such hybrid op amp circuits. they don't have be that complicated.

Input referred, yes, but it's still the aggregate noise of the entire circuit. And there's the issue of PS noise rejection, where the '797 really shines and is difficult to match with discretes. When you look at the box of gain as a whole, you DO have to consider all noise sources, not just input noise.

It's possible to do better with discrete, but damned hard and certainly more expensive. And for a typical MC or MM application, which really is the only place that ultra-low noise behavior is important for home audio, you can still do better with bipolars than FETs for the same number of paralleled devices. And there's always transformers, which at these levels are enormously quieter than even the best active devices.

No argument that at high source Z, FETs are a better choice. With strain gauges, there are other attributes which become more important, most notably CMRR. But other than a few exotic no-longer-made phono cartridges, no-one uses strain gauges in audio.
 
peranders said:
Think again Christer, mostly you don't a diff amp and when you I think the common mode range will be too small. If you only use battery powered equipment you don't need any diff amp, only when you want to connect to an oscilloscope, powered from mains.

And what makes you think that both my scope and my PC
are battery powered? Neither of them is. However, as I don't
have any safety ground in the wall outlets, this is not necessarily
a problem. The scope/PC will be floating wrt to the DUT. I also
agree that the common-mode range is not large enough to be
very useful. However, if one wants to use the amp to measure
one non-ground-referenced voltage and simulataneously
wants to use the other scope/PC channel to measure som
other voltage, then there might be a problem. AC-copuling
solves the problem of low common-mode range. However,
perhaps it might work as well to use a non-differential amp
and AC-couple both inputs anyway, despite one of them actually
being ground. I have been thinking about that, but not quite
figured out if it will work as well.
 
diyAudio Retiree
Joined 2002
tag team?

I notice SomeonE has been very quiet lately and wondered if perhaps you were carrying his load and arguing for two.

"Input referred, yes, but it's still the aggregate noise of the entire circuit."

No..... it's really not. With a typical high open loop gain op amp circuit, the noise the noise currents are in parallel with the both input terminals and the voltage noise source in series with one. loop feedback will suppress other noise contributions due to loop feedback. The noise contributions are those outside the correction of the negative feedback. the input diff pair will dominate the noise contribution to the circuit.

http://www.analog.com/UploadedFiles/Application_Notes/5480117281535838576388017880AN358.pdf

Power supply rejection and a number of other of other noise sources are important and can be greater than the input pair noise but that is not what this discussion is about. There are many High output MC cartridges that will be nosier with the AD797 that a good discrete jfet circuit. Even for low impedance moving coil cartridges, the noise of the quietest jfets is lower than the AD797 op amp or the MAT-02 BJT transistor. I believe that the noise reduction for devices in parallel is the same for jfets and BJTs and is reduced by the square root of the number in parallel. BTW be careful about base current from BJTs going though your cartridge. It is believed by many to effect the hysteresis of the magnetics.

Don't take my word about jfets, ask yourself what would JC (John Curl) do?
 
Christer, I strongly suggest that you fix a power outlet with protective earth, beacuse some parts in your PC will float on 115 VAC through the mains filter. This can be not healthy for connected gear and you can also get higher emission than necessary.

Also, measuring uV (at any potential) via an amp and a PC seems to be a hopeless task. This is sensitive things. You have emission sources everywhere, PC, mobile telephone, PC monitor etc.
 
diyAudio Senior Member
Joined 2002
Hi,

I believe that the noise reduction for devices in parallel is the same for jfets and BJTs and is reduced by the square root of the number in parallel. BTW be careful about base current from BJTs going though your cartridge. It is believed by many to effect the hysteresis of the magnetics.

Good one, Fred...not too many people actually are aware of that, the latter part I mean.
I think your analysis is correct on both counts...not that you'd need my confirmation of course.

Cheers,;)
 
peranders said:
Christer, I strongly suggest that you fix a power outlet with protective earth, beacuse some parts in your PC will float on 115 VAC through the mains filter. This can be not healthy for connected gear and you can also get higher emission than necessary.

Well, you know as well as me that its quite expensive to fix
that. Otherwise I would have done so long ago. The PC ground
certainly floats at about half the mains voltage, which is a PITA
when connecting the soundcard to the stereo.


Also, measuring uV (at any potential) via an amp and a PC seems to be a hopeless task. This is sensitive things. You have emission sources everywhere, PC, mobile telephone, PC monitor etc.

Yet, a lot of people seem to use the soundcards for audio
measurements, including noise measurements, with seemingly
reasonable results. The signals I get through my soundcard
also seem to be remarcably clean. i agree that seems like a
miracle, considering it is located inside the PC, but real
measurements support this miracle. Still, the purpose of the
amp would be to amplify the signal, of course, so one need
not trust the LSB's.
 
Poor debating tactics, Fred. Besides which, you know that my arguments are strictly my own. Gods and idols just aren't my shtick.

Input referenced noise versus input pair noise: Straw man. I wouldn't argue, and didn't, that input referenced noise isn't dominated by the input pair. But it's not the only noise source. Input referenced noise DOES include the rest of the circuit- if you claim that the Analog sheet you reference says otherwise, please cite a page and line, or just quote it. As far as I understand noise specs for opamps, it's something measured directly at the output, then referenced to an equivalent noise source at the input.

Moving Coil Source Z: I've measured a few of them (I'm not speaking of high-output, just "normal" MCs), going back to the days when I was designing head amps. The Zs are pretty darn low, at least on product from Dynavector, Technics, Ortofon, Linn, Denon, and Supex (showing my age there!). And when paralleled with the typical 100 ohm (or thereabouts) load... Can you cite some source Z numbers and demonstrate with a calculation the superior noise performance of any FET compared with something like a '797? You may well be right, I just need to be led by the hand through that one, since it's contrary to my experience. And I don't pretend to be aware of every FET on the market. Electronics is a hobby for me ;)

Is there any actual evidence (not anecdotes) of base currents of the order of magnitude shown by the input of the '797 causing a problem with the magnetic system of an MC cartridge?
 
idol speculation Sy? ..........sigh....

Forgive my bit of fun with Mr. Curl's iinitials. I really can't think of any preamp designer who would be more reputable on the subject and the specifications seem to bear him out when reads the data sheets and a couple of articles on noise in op amps ((discrete or IC)


High output MC coil and load resistance.

http://www.dynavector.co.jp/english/pdf/20x_emanual.pdf

I will leave it to someone like Mr. Carr for expertise very small DC currents in cartridges. Perhaps he would comment on his take as a world famous cartridge designer. I would value his opinion as about as credible source as they come on the subject. My understanding is the it is a no no. Even very small DC flux effects magnetic materials. I designed SPDIF and AES pulse transformers and will tell you that DC currents a thousand times smaller that the signal current screwed up the sound to a degree that I put polypropylene coupling caps in front of all my pulse transformers. I did not say that the bias current the AD797 is necessarily a problem, but parallel bipolars without careful base current cancellation seems a very risky practice. I would also make sure your circuit powers up in such away as not to put transient currents through the cartridge during power up.

Some Bipolar designs
http://users.ece.gatech.edu/~mleach/headamp/

Mr. Pass uses four paralleled 2SK170 for the Pearl and Aleph Ono designs and quotes 1 nV noise for the Pearl preamp. Perhaps he would comment on the subject of jfet vs. BJT for moving coil preamps.


Fred

PS I do take back one speculation.... actually I think it might be SE standing in for SY rather than the converse.
 
I can't be SE, I'm much taller.

Heh, heh, heh. :devily:

The 1 nV noise spec, that's input referenced to what gain and over what bandwidth? How does that number compare to a similar stage with the same gain, the same number of input devices, but using ultra-low-noise bipolars in the input stage?

I'd be interested in Jonathan's input on permissable bias currents in MCs. Microamps? Picoamps? Femtoamps?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.