Need LNA for low-frequency low-level preamp

[dnessett]

How about this 700 ohm noise equivalent amplifier:

http://www.emelectronics.co.uk/a29.html

Ok, not a lot of gain at 100kHz but enough to allow a much cheaper low noise amplifer for the rest of the gain. Not especially low noise at higher frequencies but no flicker noise and only around 9nVpp .1 to .33Hz - you'd be hard pushed to get much better from an off the shelf product for less than $200.

Not really what I am looking for. Minimum input levels appear to be 2 mV, whereas I need something that will amplify signals in the 10-100 uV range. Nevertheless, thanks for the info.

How about using different amplifiers for different frequency bands? You could use the even lower noise, 25ohm equivalent, A23 < $500. LF noise is only 1.3nV pp but unity gain bandwidth is only 10kHz. Use it for say < 100Hz and an LT1028 for 10Hz to 100kHz, (.85nV/rtHz).

If you want to go the route of paralleled opamps, bipolar amps have too much current noise but the JFET ADA4625-1 is quite impressive; 150nV pp 0.1 to 10Hz, 3.3nV/rt(Hz) @ 1kHz but only 4.9fA/rt(Hz) current noise so you can parallel as many as you can afford. Expensive at $7.8, @ 10off; the dual channel version is cheaper (per amp) but only seems to be available direct from Analog @ $5.99, MOQ 100.

If you need to make the measurement in one step, you could use both amps with a crossover set at some frequency where any gain and phase errors aren't important.

TiN has played with some A23s - take a look on his xdevs site.

[EDIT] You could parallel lots of A29 amps - current noise doesn't dominate below about 10k ohms source impedance. A bit pricey mind. 

All these ideas are interesting, but not suitable for my purpose. As I write elsewhere, I am not an experienced amp designer (and certainly not an experienced low-noise amp designer), so suggesting IC devices that might be included in a design isn't what I am looking for. But, hey, others might find this information useful, so kudos for forwarding it.

[splin]

Not really what I am looking for. Minimum input levels appear to be 2 mV, whereas I need something that will amplify signals in the 10-100 uV range. Nevertheless, thanks for the info.

You misread the datasheet:

Input Level

The A29 will operate with input levels up to plus or minus 2 mV.

It's designed for amplifying nanovolt signals.

All these ideas are interesting, but not suitable for my purpose. As I write elsewhere, I am not an experienced amp designer (and certainly not an experienced low-noise amp designer), so suggesting IC devices that might be included in a design isn't what I am looking for. But, hey, others might find this information useful, so kudos for forwarding it.

Fair enough; but can you split the measurments into two parts? Use an A23 for ultra low noise low frequency measurements, say up to 10Hz or even 100Hz, and a more conventional low noise amp for measurements in the band 10Hz to 100kHz?

When I say a more conventional amp it could be the simplest opamp inverting or non-inverting amplifier circuit - a couple of resistors for setting the gain and a few capacitors for decoupling and stability - well within the capability of many beginners (preferably one with access to a scope to check it doesn't oscillate!).

The difficult part of your problem is flicker noise (1/f noise) at low frequencies, hence the suggestion to use an off the shelf design.

[EDIT] There's a good thread here about testing the A10 and A23 amplifiers. Reply #25 includes the 2017 EM Electronics price list.

[dnessett]

You misread the datasheet:

Input Level

The A29 will operate with input levels up to plus or minus 2 mV.

It's designed for amplifying nanovolt signals.

I did indeed misread it. Thanks for pointing this out.

Fair enough; but can you split the measurments into two parts? Use an A23 for ultra low noise low frequency measurements, say up to 10Hz or even 100Hz, and a more conventional low noise amp for measurements in the band 10Hz to 100kHz?

When I say a more conventional amp it could be the simplest opamp inverting or non-inverting amplifier circuit - a couple of resistors for setting the gain and a few capacitors for decoupling and stability - well within the capability of many beginners (preferably one with access to a scope to check it doesn't oscillate!).

The difficult part of your problem is flicker noise (1/f noise) at low frequencies, hence the suggestion to use an off the shelf design.

It is a reasonable option that I will consider. However, I still would have to design and build the second amplifier, something that doesn't appeal to me at this point. But, I won't reject it out of hand.

[EDIT] There's a good thread here about testing the A10 and A23 amplifiers. Reply #25 includes the 2017 EM Electronics price list.

You seem to have left off the URL to the thread you reference.

[Vgkid]

Here you go
https://www.eevblog.com/forum/testgear/nanovoltmeters-performance/?action=dlattach;attach=453013

[maat]

I will chime in on amplifier designs. I you have a fairly low impedance source, you could try the amplifier used by Linear Tech/ADI to test some of their voltage regulators: https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf

It worked pretty well for me so far and I still do have about half a dozen PCBs left of that design if someone is interested. It is a fixed 10k gain amplifier with 10 Hz - 1 MHz Bandwidth. I usually use it in conjunction with an SR560 to add more gain and do some more filtering before passing the signal on.

Below 10 Hz you fare better using some zero drift design.

[maat]

These are some noise plots I made to benchmark the amplifier.  The bottom trace is with the input shorted and the upper one is with a 10 Ohm resistor across the input. The dashed line is the Johnson-noise of the resistor. The discontinuities result from my setup. Unfortunately my spectrum analyser starts at 7 kHz, so I sampled the data below 10 kHz with two other devices and calculated the noise density from that. I also used an SR560 to amplify and filter the data for each segment.

The result agrees pretty well with the Linear Tech App note.

[Kleinstein]

I will chime in on amplifier designs. I you have a fairly low impedance source, you could try the amplifier used by Linear Tech/ADI to test some of their voltage regulators: https://www.analog.com/media/en/technical-documentation/application-notes/an159fa.pdf

It worked pretty well for me so far and I still do have about half a dozen PCBs left of that design if someone is interested. It is a fixed 10k gain amplifier with 10 Hz - 1 MHz Bandwidth. I usually use it in conjunction with an SR560 to add more gain and do some more filtering before passing the signal on.

Below 10 Hz you fare better using some zero drift design.

That amplifier is BJT based and has high current noise. So it is not suitable for a 600 ohms source. The noise level is good at 10 Ohms, but expected to be poor for more than 100 Ohms.

[dnessett]

Thanks to everyone who comented on this topic. The contributions contained useful and germane information. Several commentors noted that oscillator phase noise increases significantly as offset frequency approaches 1 Hz and for this reason a good low noise preamp probably would suffice:

(1) Low noise oscillators use low flicker noise transistors for this very reason which suggests medium frequency amplifiers using bipolar transistors will always be sufficient for this application.

Oscillators tend to have more noise in the low frequency band - so some 1/f noise may be acceptable.

The AlphaLab Preamp should provide low noise and enough gain for either situation.

I decided to explore this line of reasoning using the AlphaLab Oscilloscope Preamplifier LNA10 as an example. I welcome comments on the following argument.

When using the HP11729C in Phase Detector mode, the lowest phase noise device should be the reference oscillator. Its phase noise should be less than either the DUT or the HP11729C itself. The device I am using as a reference oscillator is the Wenzel HF-ONYX-IV (part 501-11578-04), which has typical phase noise of -135 dBm at 10 Hz (which is less than that of the HP11729C and likely that of the DUT). Since oscillator specs generally quote phase noise as power relative to a 50 ohm load, this equates to 3.976e-8 V_RMS/sqrt(Hz) = 40 nV/sqrt(Hz). The input-referenced noise of the AlphaLab LNA10 is 6 nV/sqrt(Hz).

To obtain the gain I need requires cascading two AlphaLab LNA10s. Since noise adds like power, the noise resulting from two cascaded LNA10s is sqrt(2*(6^2)) = 8.5 nV/sqrt(Hz). So the noise contributed by the cascaded amplifiers is about 20% of the noise contributed by the reference oscillator.

It is desirable that the noise added by the preamp is no more that 10% that of the reference oscillator. Nevertheless, the reference oscillator phase noise should be much less than either the DUT or HP11729C phase noise, so the noise added by the preamp is likely less than 10% of the total system phase noise, which should suffice for my purposes.

From 10-100 Hz, the phase noise of the AlphaLab LNA10 is 4.1nV/sqrt(Hz), meaning in a cascaded configuration the added phase noise would be approximately 5.7nV/sqrt(Hz). This is about 15% of the noise contributed by the reference oscillator. This is still not at the 10% desirable level, but again probably sufficient for my purposes.

Given this analysis, I am leaning toward purchasing two AlphaLab LNA10s and using them as the preamp for the PicoScope 4262. If for some reason this approach doesn't work out, I may then take a closer look at the EM Electronics A29.

[Kleinstein]

Using 2 amplifiers in series does not increase the input noise very much. The noise of the 2 nd amplifier is attenuated by the gain of the 1 st. So with a gain of >=10 for the first stage, the second stage gives essentially no extra noise.

I don't think it even needs a second amplifier. The pico-scope should be well sensitive enough with only moderate gain (e.g. 10 x, maybe 50 x) up front. I kind of doubt the extra amplifier is that much lower noise than the scope.

Not only amplifiers have more noise at low frequency - also oscillator phase noise usually has an 1/f contribution. So one likely would not need the same noise level at very low frequencies. The LNA10 does not look that bad.

The 4.x nV  noise level is not that much higher than the noise of the 600 Ohms output impedance, which is sime 3.2 nV/SQRT(Hz). One number I am missing the the amplifiers data is the current noise or impedance for best noise figure. Chances are high the current noise is rather low, as a BJT based amplifier should have more like lower voltage noise. So the amplifier is likely JFET based.

[dnessett]

I don't think it even needs a second amplifier. The pico-scope should be well sensitive enough with only moderate gain (e.g. 10 x, maybe 50 x) up front. I kind of doubt the extra amplifier is that much lower noise than the scope.

The noise floor of the PicoScope 4262 is -92dBm. As a lower bound say the phase noise of the system is the phase noise of the reference oscillator at 10 KHz (it will be higher than this, but this is a worst case assumption). This is -165 dBc\Hz.

I have not yet posted the procedure for using the HP11729C in a Phase Detection configuration, but after calibration and corrections, the carrier suppressed signal is raised about 17 dB, so the effective target is -165 dBm + 17 dB = -148 dBm (The reference carrier input is set to 0 dBm during the set up procedure). To raise this level above -92 dBm requires about 56 dB gain. Since the LNA10 has a maximum gain of 30 dB, a single amp is insufficient. Cascading two together would give a maximum gain of 60 dB, which would just work.

Now, I will admit the calibration and corrections algorithm is a bit complicated. So, I'm not completely confident in the above analysis. Nevertheless, each LNA10 costs $275, which means two can be purchased for $550. This seems a reasonable investment and if only one is required, I can use the other for a different project.

Added 8-10-2019: I went to order the LNA10s and noticed I got the price wrong. Each costs $270, which means two cost $540.

[chuckb]

The LNA10 also has a 470 ohm output Z. So keep the Picoscope on High Z input.

One preamp has selectable 1000x gain this is 60dB.

[dnessett]

One preamp has selectable 1000x gain this is 60dB.

Power gain = 10 log( (V_in² / R_in) / (V_out² / R_out) ). R_in= 500 Kohms and R_out = 470 ohms. So, Power gain = 10 log( (V_in²) / V_out²) * (R_out /  R_in). R_out /  R_in ~= 10^-3, so Power gain ~= 10 log( (1000² * 10^-3)  )= 10 log(10³) = 30 dB.

On 8-10-2019 (forward reference) Kleinstein wrote:

In the calculation there is another mistake: the step from 500 K at the input to 470 Ohm at the output adds another 30 dB to the power gain. So looking at the power one would get 90 dB, not 30 dB.

Kleinstein is right. I created the derivation right before I went to bed and didn't double-check my work (always a perilous practice). The problem (in the derivation) is Power gain isn't 10 log( (V_in² / R_in) / (V_out² / R_out) ), it is 10 log( (V_out² / R_out) / (V_in² / R_in) ). This inverts the ratio of resistances, which therefore equals 10³  not 10^-3.

Also, V_in² / V_out², isn't 1000², it is 1000^-2. V_out² / V_in² = 1000².

The bottom line is power gain is 90 dB, as Kleinstein correctly points out.

[Kleinstein]

One preamp has selectable 1000x gain this is 60dB.

Power gain = 10 log( (V_in² / R_in) / (V_out² / R_out) ). R_in= 500 Kohms and R_out = 470 ohms. So, Power gain = 10 log( (V_in²) / V_out²) * (R_out /  R_in). R_out /  R_in ~= 10^-3, so Power gain ~= 10 log( (1000² * 10^-3)  )= 10 log(10³) = 30 dB.

The low frequency connections between the instruments are not using impedance matching. So the use of dBm is more confusing than helpful. The relevant  gain is only the voltage gain, not the power gain.

In the calculation there is another mistake: the step from 500 K at the input to 470 Ohm at the output adds another 30 dB to the power gain. So looking at the power one would get 90 dB, not 30 dB.

Anyway I have some doubt the HP11729C output noise is so low. Changes are there is quite some noise in the HP11729C, even if it is just the about 500 Ohms resistor used to make the output 600 Ohms impedance.

[dnessett]

In the calculation there is another mistake: the step from 500 K at the input to 470 Ohm at the output adds another 30 dB to the power gain. So looking at the power one would get 90 dB, not 30 dB.

I have corrected the error in the original post with an edit.

[dnessett]

The low frequency connections between the instruments are not using impedance matching. So the use of dBm is more confusing than helpful. The relevant  gain is only the voltage gain, not the power gain.

In the calculation there is another mistake: the step from 500 K at the input to 470 Ohm at the output adds another 30 dB to the power gain. So looking at the power one would get 90 dB, not 30 dB.

Anyway I have some doubt the HP11729C output noise is so low. Changes are there is quite some noise in the HP11729C, even if it is just the about 500 Ohms resistor used to make the output 600 Ohms impedance.

I address the point made in the middle paragraph in the original post with an edit.

In regards to the output noise of the HP11729C, at 10 Hz it is -115 dBc/Hz in the band 5 MHz-1280 MHz, which is much higher than the Wenzel Reference Oscillator. I used the figure for the Wenzel as a worst case scenario.

The first paragraph in the response raises two new issues.

1. The amplification by the LNA10 boosts the signal above the noise floor of the PicoScope 4262. However, its effect must be removed by a correction calculation on the data after the FFT is applied. Assuming I use the 1000X amplification setting of the LNA10, is the correction to subtract 90 dB from the values computed by the PicoScope? This seems correct to me, since the corrected PicoScope data will be in units of dBc/Hz, not V/sqrt(Hz). Comments?

2. Figure 1 shows the configuration of the test setup from the HP11729C to the PicoScope.



Figure 1 - Test Setup from HP11729C to PicoScope

My concern is reflections. The output impedance of the HP11729C is 600 ohms and the input impedance of the LNA10 is 470 ohms 500 Kohms. For the short lengths of RG-58 used to connect the devices reflections can probably be ignored. (Modified 8-11-2019.) The output of the HP11729C contains useful information from 1Hz-200KHz. The LNA10 has output filtering that can be set to low-pass 200 KHz. The 1/4 wavelengh of 200 KHz is 1230 feet. The RG-58 connecting the LNA10 to the PicoScope will be several orders of magnitude less than this. The input impedance of the PicoScope 4262 is 1 Mohm. So, I think the 600 ohm resistor at the Tee (shown in red) is unnecessary (as is the Tee) and should be eliminated to reduce noise (this is what has been suggested by several commentators). Agreed?

[Edits embedded in text. This has not been my weekend for clean arguments.]

[Gerhard_dk4xp]

The tee & 600 Ohm termination can be safely removed at 200KHz max. Your lab is
not so large as that this could matter.

The manual of my Agilent 89441A FFT analyzer recommends the Hann filter for noise
measurements; Blackman-Harris is more for sine signals with its steep transition bands.

I have finally completed my switchable delay line, but need a larger +24V brick
since I underestimated the current drawn by the coax relays.
There are 3 ranks in series: coarse, mid and fine. Each rank consists of 2 pcs. 1P6T
coax relays and the 6 semi rigid coax cables. (Nexans Quickform 3.5 mm o.d.)
The 1:6 relays reduced the number of contacts in series.

Making the cables was highly meditative, but those SMA connectors are surprisingly
easy to install on SemiRigid. I could check the actual delay with a TDR.

You need a lump of iron with a 3.5 mm bore hole abt. 5 mm deep. You insert the cable
as far as possible and circumcise it with a sharp knife. When you bend the cable somewhat,
the outer metal will easily break and can be removed. Then remove the Teflon flush with
the outer metal, solder on the center contact, shove it into the rest of the connector and
solder the outside.

Those 1p6T coax relays are normally quite cheap on ham flea markets; everybody
wants nothing but 1P2T. There are relays with built-in TTL drivers. Usually one of
of the drivers is defective, that's why they ended up on the flea market. 24/28V supply
and TTL inputs don't go together well. Since I need only the working coil, that could
be healed, but it does not look good.

I first wanted to use AirCell cable for its lower loss, but found out too late that these
special SMA connectors were too thick to fit onto the relays.   :-(
The delay line can bring anything from 4 to 200 MHz into phase quadrature, no matter
what amplifiers or filters may be in the path.
 

[chuckb]

You do not worry about impedance matching anything in the mixer output signal chain. Adding load resistors only attenuates your voltage signal. Your goal is to maximize the voltage supplied by the mixer and getting to the voltage measuring device (Picoscope).

When you calibrate the Mixer / delay line gain you will end up calibrating the Alpha Amplifier gain, and the PicoScope FFT gain all at the same time.

As you have pointed out, you are looking for dBc. The PicoScpoe displays in dBm. The correction factor for for dBc for your setup will be determined during your Mixer/Delay line calibration.

The mixer probably has a 50 ohm or less source impedance so the 500k input Z of the Alpha amp will reduce the signal less that 0.1%.
The 470 ohm output Z of the Alpha amplifier will likewise not be loaded (attenuated) by the 1Meg ohm input Z of the PicoScope.

The Alpha amp will need to be AC coupled to remove the DC offset voltage from the mixer.
You will probably also want to use the 100kHz low pass filter in the Alpha amp In addition to the scope low pass filter to help reduce unwanted 20 MHz frequency mixer products.


Mr Rubiola's web site teaches how to do all this. It seems like he has the most advanced techniques in the world.
http://rubiola.org/

[magic]

For just a 600 Ohms source a single BJT based OP like OPA209 would be more suitable.

They recently released a 2210 with slightly better noise and DC specs.

[Gerhard_dk4xp]

It irritates me, that the 11729C is always drawn with an amplifier after the low pass.

If there is one, noise behaviour of the external preamp should be not an issue.

If there is none, the effective source would be just the ring mixer and the real source
resistance would be smaller. The prescribed 600 Ohm load would only be there
to keep the high discrimination sensitivity and to keep the filter happy.
But the investment into a better voltage noise spec than for 600 Ohm would still pay.

[dnessett]

For those interested, I have posted some notes on my experience with the LNA 10 here.