I recently purchased two
LNA 10 low-noise oscilloscope preamplifiers to use in an application I am working on - measuring phase noise in some hobbyiest oscillators. In preparation for using them in this application, I did some simple testing, the results of which I thought I would share.
This is not a full or even a partial review of the LNA 10. Rather, it comprises some notes that I thought others might find useful, if they decide to purchase the device.
For the cost, $270, the LNA 10 is an impressive piece of equipment. It provides 10X, 100X and 1000X voltage gain with an implied bandwidth of 1Hz - 1MHz. I say this is the implied bandwidth because, rather surprisingly, the bandwidth is not actually specified in the documentation. However, it provides a selectable analog low-pass filter on the input that may be set from 1 Hz to 1 MHz, so the bandwidth should be within this range. Noise is less than 4 nV/sqrt(Hz) for offset frequencies > 100 Hz. For offset frequencies less than 100 Hz, its noise characteristics are:
Freq | Noise(dBm) |
10-100 Hz | 4.1 nV/sqrt(Hz) |
1-10 Hz | 6 nV/sqrt(Hz) |
10-2-1 Hz | 20 nV RMS |
10-4-10-2 Hz | 50 nV RMS |
Note that the last two noise values are expressed in terms of RMS noise, rather than spectral density noise.
Other features of the device are:
- Selectable DC/AC coupling on input
- One-sided or differential input options
- Input impedance : 500 Kohm
- Output impedance : 470 ohm
- Impedance between differential inputs - 1 Mohm
- CMRR on differential input : > 90 dB
- Gain accuracy : +/- 1%
- Offset adjustment, equivalent to +/- 1mV at input
In order to determine the suitability of the LNA 10 for my application, I ran some tests on its performance. I used a Rigol DG-1022 signal generator to produce a fixed frequency sine wave; a PicoScope 4262 in spectrum mode to analyze the results; and the LNA 10 to provide amplification. For the case of no gain, I connected the output of the DG-1022 directly to the PicoScope.
The goal of the first test was to determine the gain characteristics of the device at various frequencies and gain settings. The parameters of this test were:
Test ParametersFor all configurations - PicoScope: 16,384 bins; 30 segment averaging; Blackman-Harris Window; no-termination at PicoScope input; SG amplitude: 1 mV
P-P1 KHzLNA: 2 KHz LPF
PicoScope: 1 - 5 KHz span; 10 KS/s; 3.277s time gate
10 KHzLNA: 20 KHz LPF
PicoScope: 1 - 20 KHz span; 40 KS/s; 819ms time gate
100 KHzLNA: 200 KHz LPF
PicoScope: 1 - 200 KHz span; 400 KS/s; 81.92ms time gate
500 KHzLNA: 1 MHz LPF
PicoScope: 1 - 1 MHz span; 2 MS/s; 16.38ms time gate
Test ResultsFreq | No Gain (dBV) | 10X gain (dBV) | 100X gain (dBV) | 1000X gain (dBV) |
1 KHz | -57.41 | -38.41 | -18.3 | 1.509 |
10 KHz | -57.36 | -38.36 | -18.42 | 1.508 |
100 KHz | -57.42 | -38.85 | -19.02 | 0.978 |
500 KHz | -57.43 | -38.92 | -19.43 | 0.782 |
Freq | No Gain (dBm) | 10X gain (dBm) | 100X gain (dBm) | 1000X gain (dBm) |
10 KHz | -55.13 | -35.92 | -16.04 | 3.97 |
I setup the PicoScope to use log amplitude averaging for the first set of results (using units of dBV) and also ran a test on 10 KHz using log power averaging (using units of dBm). In both cases the change in values were roughly +20 dB for each increasing gain setting. I ran the dBm test, since in my application the measurements of interest are log power values.
Note that the LNA 10 yields fairly consistent gain characteristics for 10X gain. For 100X and 1000X gain, there is some roll-off in the gain at higher frequencies (100 KHz and 500 KHz).
One thing I ran across when playing around with the LNA 10 was an undesirable property that the noise inherent in the DG-1022 signal was not amplified the same amount as the coherent signal. Figures 1-4 are PicoScope plots for a 50 KHz signal amplified as follows: Figure 1 - no gain; Figure 2 - 10X gain; Figure 3 - 100X gain; and Figure 4 = 1000X gain.
Figure 1 - Signal with no amplification
Figure 2 - Signal with 10X amplification
Figure 3 - Signal with 100X amplification
Figure 4 - Signal with 1000X amplification
Notice that the noise floor in the 10X amplification case is about 50 dB greater than in the no gain case. For the 100X amplifcation case, the noise floor is about 20 dB greater than the 10X amplification case. For the 1000X amplification case, the noise floor is about 10 dB greater than in the 100X amplification case. This is a serious problem for my application unless a solution was available.
Fortunately a solution exists. I contacted AlphaLab and asked them about this problem. They replied as follows (I am quoting them with their permission):
To avoid high current surges (if there's a voltage difference between input and output grounds), the input grounds have a series 10Ω resistor. When using only the I+ (or I- alone) input, any ground noise is amplified. This can be substantial noise, because the Picoscope ground and signal gen ground form a loop. You can even see this if you set the knob to I+, and short the I+ input, then touch it with your finger, or touch the signal gen ground to the I+ ground. This excess noise is completely eliminated if you short the I- input, connect the signal to I+ , and set the knob to I+ - I-.
Best regards,
Bill Lee
AlphaLab, Inc.
I created a simple shorting device by cutting the BNC connector off of an RG-58 cable and soldering the signal wire to the grounding mesh. I then connected this device to the I- input of the LNA 10, connected the signal to the I+ input and used the I+ - I- input option. The result was much better. Figures 5-7 show the spectra produced by the PicoScope for the cases: Figure 5 - 10X gain; Figure 6 - 100X gain; and Figure 7 = 1000X gain.
Figure 5 - Signal with 10X amplification
Figure 6 - Signal with 100X amplification
Figure 7- Signal with 1000X amplification
In each case, the noise floor is amplified by approximately an additional 20 dB for each increased gain setting.
The next step for me is integrating the LNA 10 into the procedure for using a phase detector (in my case an HP11729C) to analyze phase noise in some hobbyist oscillators. (For those interested, see
this topic)