Sunday, December 16, 2018

Low Noise Measurement Preamp


Recently I needed a low noise Preamp to measure voltage noise density of some low frequency amplifier circuits with my custom built FFT Analyzer [1]. This is surprisingly easy to do now with just a few IC’s from Linear Technology.

The basic specifications of what I needed are,
  • DC Coupled
  • Noise at 1 Hz   <=  10 nV/rt-Hz total in the measurement system.
  • Noise at 1 kHz <= 2.5 nV/rt-Hz total in the measurement system.
  • Frequency Range DC-100 kHz.
  • 1k Ohm input resistance.
  • Adjustable gain so that a wide variety of conditions can be measured.
  • Power from +/- 5 Volts.
  • Output Voltage at least +/- 2.5 Volts to drive FFT Analyzer Input
Some back of the envelope calculations suggested that I needed an overall gain of about 5000 for the Preamp to overcome about 90% of the FFT Analyzers input noise. With this in mind and to keep things simple, I picked a 2 stage preamp design consisting of a low noise OPAMP followed by a programmable gain amplifier.

The lowest voltage noise OPAMP for some decades now has been the LT1028 [1]. The low 0.9 nV/rt-Hz midband noise combined with the very low noise 1/f corner frequency was perfect for my needs. To expand the dynamic range of the preamplifier I added a post amplifier using the LTC6910-1 Programmable Gain Amplifier. The LTC6910-1 provides programmable gains of 1,2,5,10,20,50 and 100.

Setting the LT1028's gain to 52.1 V/V combined with the LTC6910-1's maximum gain of 100 combines for a maximum preamp gain of 5210 V/V [3].

Figure 1 – Using a low gain of 52 V/V on the LT1028 sufficiently overcomes the downstream noise and still allows for wide dynamic measurement range on even the highest gain settings. The overall gain can be set from 52.1 to 5210 V/V by programming the LTC6910. 

When measuring the noise floor of anything, the preamp noise is only part of the problem. It is equally important to have extremely well filtered power supplies and control lines to prevent,

#1 Spurious signals coupled in from the power supplies and showing up in the noise floor. As any switching noise that makes it way to the preamp will show up as a spurious peak in the measurement.

#2 Excellent damping and low impedance to prevent power supply feedback coupling and oscillation due to the large gains involved in the preamp. This is also important from one preamp channel to another as this design was ultimately a 2 channel design.

These two requirements are met with the application of the classic, simple and lowest of the low noise regulators, the: “Beta Multiplier” circuit. The circuit in Figure 2 is mounted close to the preamps and is in addition to the per-channel linear regulators on the main board. The combination of the two, then serves to help rid the noise floor of all but the most stubborn spurious signals.


Figure 2 – In any noise floor measuring system the preamp power supply is just as important as the preamplifier design. This adaptation of the classic and very low noise “Beta Multiplier” provides > 40 dB of isolation and very low noise at 100 Hz. The addition of the inductor provides continued rolloff well beyond the 100 kHz measurement bandwidth. High quality Tantalum’s must be used for the 220 uF capacitors.

The one downside of the Beta Multiplier is that it adds about 0.6 Volts of voltage drop. Hence the raw +6 and -5.5 Volt inputs end up being around +5.4 and -4.9 volts at the preamps voltage terminals in this application.

Since the ultimate usage of the preamp was in a dual channel design. To prevent coupling from one preamp channel to another, the circuit of Figure 2 was duplicated for each input channel.

To further prevent noise from coupling into the preamps noise floor, the LTC6910-1 control lines were heavily filtered and isolated from the microprocessors control bus. Control bus isolation was provided by using a MAX7317 I/O Expander in between the analyzer control bus and the preamp channels. This combined with the heavy filtering on the control lines at each LTC6910-1 provided the required isolation.


The preamp was attached to the FFT analyzer and the combined system noise floor was measured by shorting the input of the preamp.


Figure 3 – The overall system noise performance of a LT1028 and OPA140 version of the preamp with shorted inputs. The FFT Analyzers ADC and ADC Driver amplifier also add noise to the overall result.

The LT1028 input is suitable for measuring a large class of circuits since the useful measurement nodes tend to be low Impedance (i.e. regulator or amplifier outputs). However there are instances where the impedance is higher than the optimum source impedance of the LT1028 (which is quite low at around 160 Ohms @ 1 Hz). In these cases a higher impedance preamp is desirable. By simply substituting a JFET OPA140 for the LT1028, circuits up to the Megohm range can be measured. For comparison, the noise performance of both the LT1028 and OPA140 versions of the preamp are plotted together in figure 3.


The base Analyzer that this preamp was designed for has a mezinnine connector for the Analog input and utput circuits, this allows the overall instrument configuration to change with very little work. Just whip up a new Analog Input Board, add jst a few lines of embedded C code to the instrument application to 'teach' it how to run the new hardware, plug in the board. QED!


Figure 4 - The completed Low Noise PreAmp as implemented. The input used BNC connectors and both channels can clearly be seen. Plenty of room left on the board for future expansion.


Bonus Figure: Since Linear Technology does not publish low frequency curves of the LTC6910 Programmable Gain Amplifier. I have measured a typical device and present it here for a Gain of 100.


[1] Hageman, Steven C., “A Modern DSP Based Lock-In Amplifier”, 2018

[2] Linear Technology has a redesigned LT1028, the LT6018 but it doesn’t have significantly lower noise and unlike the LT1028 the noise isn’t 100% tested on every device.

[3] Clever readers will note that the LTC6910-1 does not have a minimum guaranteed Gain-Bandwidth product (GBW) that meets the required 100 kHz bandwidth at a gain of 100. I was only building a few of these preamps and the ones I had do meet the typical specification specification of 11 MHZ GBW, so I went with the typical specifications which met my requirements.

Article By: Steve Hageman

We design custom: Analog, RF and Embedded systems for a wide variety of industrial and commercial clients. Please feel free to contact us if we can help on your next project.

Note: This Blog does not use cookies (other than the edible ones).

No comments:

Post a Comment