Electronics Weekly Electronics Design & Components Tech News
Texas instruments has revealed a buffer amplifier developed especially for oscilloscope inputs from dc to GHz.
BUF802 signal paths
“Oscilloscope manufacturers have got a choice: develop a discrete JFET input stage and risk limited performance, or develop their own asic,” TI general manager of linear amplifiers John Caldwell told Electronics Weekly. Developing a custom asic “is a massive undertaking where they need their own design team, or to pay someone else to design it. Which ever way, they have to commit to MOQs at a foundry. They said to us: ‘we really don’t want to do this'”.
TI’s aim was to create an integrated high-impedance low-capacitance buffer amplifier that could sit next to the input connector of a 3GHz 12bit oscilloscope and drive the subsequent programmable gain amplifier.
The result is BUF802, with a 100MHz – 3.1GHz large signal bandwidth at 1Vp-p (2GHz at 2Vp-p), 7kV/μs slew rate, >50GΩ input impedance and 2.4pF input capacitance. Input-related noise is noise is 2.3nV/√Hz – “better than any asic on the market”, claimed Caldwell.
Low noise is important because the buffer sits before the programmable gain amplifier, so its noise will be the minimum system noise displayed on the final instrument’s highest gain setting – and the ICs dynamic range must include all of the instrument’s gain settings.
Having that combination of bandwidth and high impedance, according to Caldwell, allows the scope manufacturer to use a simple single-pole switch in series with a load adjacent to the scope input for 50Ω-hiZ mode switching, rather than using series switches to swap separate 50Ω and high-impedance buffers into the signal path.
High input impedance and low noise comes from a carefully-crafted JFET at the input pin, in a circuit that Caldwell cannot reveal.
Again in a way that TI is keeping close to its chest, a pair of 100mA diode or diode-like clamping structures have been included within the capacitance budget to protect the input from end-user-abuse.
“100mA with 2.4pF on the front meets or beats all oscilloscope manufacturer’s requirements,” said Caldwell. “The team achieved it, and a really low-noise JFET with low capacitance, all in the in-house high-speed bipolar process we have.”
Like contemporary discrete scope front-ends, BUF802 is intended to be used with dc-blocking series capacitor between it and scope’s front-end connector, with a dc-accurate low-frequency amplifier in parallel with the capacitor to servo dc and low-frequency accuracy into the high-speed buffer’s output.
TI’s rendering of the traditional input that it intends to replace
However, said Caldwell, TI’s reading of existing discrete inputs (diagram left), is that the dc-blocking capacitor introduces ac positive feedback to the precision amplifier, with the attendant risk of a peak in the frequency response of the resulting compound amplifier – that could need tuning out individually per scope input.
BUF802 with ancillary input components
TI’s answer in the BUF802 (right) is to avoid potential positive feedback by give the buffer a separate input from the dc servo amplifier that adds dc-restoration after the JFET stage rather than at the JFET gate. Gain β will need to be trimmed per oscilloscope input, requiring a preset potentiometer, plus an optional preset capacitor trim in higher-performance applications (~ <0.03dB gain flatness).
If required, it can also operate in the more convention scope input circuit, or as a stand-alone buffer – both not using the auxiliary signal path.
The output stage is a classic Class-AB diamond buffer, said Caldwell – cascaded complementary emitter followers, that which can drive a 50Ω load in this case (below left). The central clamp circuit prevents the output going beyond pre-set limits, and reduces input over-load recovery time for the whole IC.
Buffer signal distortion was the “most difficult requirement oscilloscope manufacturers gave to us”, according to Caldwell. “We are not talking audio levels, but we are talking 12bit ADCs, and achieving low distortion at 1GHz or 2GHz is a significant challenge.” Especially, he added, as no distortion-correcting feedback is possible that this speed, the amplifier has to be open-loop.
However it was achieved, the result at 1GHz 1Vp-p is around-55dBc second harmonic distortion and -59dBc third harmonic.
This came at the expense of considerable current consumption: ~34mA from the (±4.5V to ±6.5V) power rails.
Artists impression: A high-frequency capacitor, a chunky 400V part in many scopes, is all that sits between the input connector and the BUF802
However, the scope maker can trade power consumption against distortion using a single programming resistor connected to a pin provided on the chip, which also allows the manufacturer to use a single front-end PCB for differently-priced instruments in its range.
Despite dissipating around 400mW, the package is a 3 x 3mm QFN, with a thermal pad.
End-to-end gain is unity – how much trimming, if any, is needed in production to get this, is another thing TI is not revealing, although the official unit price – $1.80 in quantities of 1,000 – suggests production is not too onerous, and could tempt manufactures of active probes as well as scope makers.
Update:
see this on-line TI article for more about avoiding peaking in a front-end buffer. Also see page 21 and subsequent pages of the BUF802 data sheet for α and β gain information, trimming, and a complete 1GHz scope front-end design.
There is a BUF802 evaluation board – a plain PCB that the user has to populate with components bought separately.
