Op-Amp = 2Sk170

Status
Not open for further replies.
Jump to Latest
Oneminde

Oneminde

Member
Joined 2014
#1
Did I grab your attention... if so, good. I have a serious question to ask the community.

I know the Toshiba 2Sk170 is THE reference JFET to use, but I want to shed some light on that. This transistor has nV/√Hz (Voltage Noise) of roughly 0.82 nV/√Hz from 10Hz which is outstanding.

Linear Integrated Systems made the LSK170 which is meant to replace the 2SK170... however, its not as close as one might think. In the data sheet they speck the voltage noise as 1.9nV/√Hz 1kHz which would indicate that it is closer to more generic JFETs out there.

The On Semiconductor 2SK3557 does not show its nV/√Hz 1kHz but rather as a NF figure as 1dB.

The NXP BF862 report 0.8nV/√Hz 1kHz - on par with Toshiba or even slightly better.

The Linear Technology Op-amp LT1028 has 0.85nV/√Hz Typ at 1kHz and 1.0nV/√Hz Typ at 10Hz.

Lets put this into perspective.

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


The op-amp LT1028 is not part of the test but should be fairly close to BF862 which by the way is obsolete. Looking at the curves, the 2Sk3558 is on par with LSK170.

Cost:
Toshiba 2SK170: roughly $10-20 if you are lucky to find a real one.
Linear Systems LSK170: roughly $3.50-4
On Semiconductor 2SK3557: roughly $0.4-0.5
Linear Technology LT1028; roughly $9.
NXP BF862 ... obsolete.


1. Why are people lusting for the LSK170 when 2SK3557 ?

2. Why JFET when LT1028 ?

Article source; Measurements Rate SMT Low-Voltage n-JFETs Under Consistent Conditions | Electronic Design



 
#2
You haven't looked at the equivalent input noise current: a few fA/sqrt(Hz) for a JFET and 3.4 pA/sqrt(Hz) for an LT1028 (the datasheet says 1 pA/sqrt(Hz), but that is only true with equal impedances driving both inputs, not in a normal application). For source impedances exceeding about 250 ohm the noise current of an LT1028 will have more impact than its noise voltage. By the way, an AD797 has about the same noise voltage and less noise current than an LT1028.
 
#3
A unity gain buffer amplifier may be constructed by applying a full series negative feedback (Fig. 2) to an op-amp simply by connecting its output to its inverting input, and connecting the signal source to the non-inverting input (Fig. 3). Unity gain here implies a voltage gain of one (i.e. 0 dB), but significant current gain is expected. In this configuration, the entire output voltage (β = 1 in Fig. 2) is fed back into the inverting input. The difference between the non-inverting input voltage and the inverting input voltage is amplified by the op-amp. This connection forces the op-amp to adjust its output voltage simply equal to the input voltage (Vout follows Vin so the circuit is named op-amp voltage follower).
The importance of this circuit does not come from any change in voltage, but from the input and output impedances of the op-amp. The input impedance of the op-amp is very high (1 MΩ to 10 TΩ), meaning that the input of the op-amp does not load down the source and draws only minimal current from it. Because the output impedance of the op-amp is very low, it drives the load as if it were a perfect voltage source. Both the connections to and from the buffer are therefore bridging connections, which reduce power consumption in the source, distortion from overloading, crosstalk and other electromagnetic interference.
Click to expand...
Source; https://en.wikipedia.org/wiki/Buffer_amplifier#Op-amp_implementation




 
Last edited:
#4
"This connection forces the op-amp to adjust its output voltage simply equal to the input voltage (Vout follows Vin so the circuit is named op-amp voltage follower)."

Hence, there is no noise voltage either! Perfect noise-free amplifier made by just quoting Wikipedia out of context!
 
#5
Perhaps... however, the thread is more than the op-amp. Its also about the choice of JFETs, like the 2Sk3557. I am sure that in the right circuit, an op-amp in real life will perform rather nicely. Because in real life, the architecture and choice of accompanying components are everything. The FET or op-amp is just one component and matched to a poorly designed circuit or used with low quality component, the S/N and THD will increase compared to the theoretical performance.


 
#7
In any case, I'm sure there will be something about input bias currents of op-amps on Wikipedia or somewhere else on the Internet, and it will be explained that those are not eliminated by negative feedback. The same holds for random variations of the input bias current, also known as equivalent input noise current.

Because of its equivalent input noise current, an LT1028 will never perform nicely in an amplifier that has to work with high source impedances and to have a low noise. Typical examples would be moving-magnet phono amplifiers and the amplifiers built into condenser microphones. JFET op-amps can work well in those applications, although most are outperformed by discrete JFETs.

For most low-noise highish-source-impedance applications, the JFET with the lowest noise voltage density will work best. At extremely high source impedances, like inside a condenser microphone, you also have to take the noise current of the JFET into account. That consists of shot noise of the input leakage current and of a term that comes from the noise voltage and input capacitance.

By the way, the IF3601 and IF3602 are missing from your overview. They are quite expensive and their huge input capacitance makes them less suitable for microphones, but they have a really low equivalent input noise voltage density.
 
#8
2SK170BL Vds max 40V, Idss 6~12mA, max dissipation 400mW
2SK3557-6 Vds max 15V, Idss 10~20mA, max dissipation 200mW
2SK209GR Vds max 50V, Idss 2.6~6.5mA, max dissipation 150mW

2x 2SK209GR in parallel come closest to a single 2SK170BL, IMHO.
Costs about 1 USD.
https://www.diyaudio.com/forums/pas...source-follower-applications.html#post5583280

2SK3557 has too low Vds rating for many applications, as well as too high Idss.
If you accept the higher noise, it is an acceptable replacement for BF862.


Cheers,
Patrick
 
#10
I never said anything about using any discrete devices at max rated dissipation.
I personally always use a safety factor of 1.5~2, or more.
But the same applies also to SOT23 as well as TO92.

For a 2SK209GR at say 4mA, you can theroetically use Vds up to 37V.
Quite save at 18V, for example.

And I sometimes use custom heatsinks for both packages.


Cheers,
Patrick
 
#11
Operating at high VDS can increases noise, though:
higher junction temperature -> more thermal noise generated in the channel -> higher equivalent input noise voltage
higher VDS -> higher gate leakage due to impact ionization current -> more equivalent input noise current
 
#13
I was being a bit sarcastic. But it occurs to me that the 2sk170 must then have a rather low thermal resistance from the die to the outside surface of the package, to have a 400mW limit with infinite heatsink. So the die must be very large?
 
Status
Not open for further replies.
Top Bottom