ETS and increasing the Bandwidth of the ADALM1000

In two previous blogs we discussed how to use equivalent time sampling (ETS) techniques in the software to display higher frequency signals with the ALM1000 hardware. In this blog we will talk about some ways to increase the usable bandwidth of the analog inputs.

The stuff inside the blue dashed box in the schematic represents the signal path between the pin and the ADC. C5 is the lumped representation of all the stray capacitance on the pin from the analog switches and the protection diodes. The ADA4661 opamp used as the input buffer has a 1.8 V/uS slew rate and 2.1 MHz small signal (at  20 mV p-p) closed loop bandwidth. The performance of this opamp will ultimately set the bandwidth. To limit the bandwidth and reduce noise, R20, R62 and C42 form a single pole low pass anti-alias filter with a -3 dB bandwidth of about 150 KHz.

The simplest way to compensate for the on board low pass filter is to insert a high pass filter with the zero at the same frequency as the low pass pole. Normal high pass RC filters do not pass DC so R1 is added to pass DC. This structure is often called a shelving filter because of the shape of its frequency response. It has a DC gain of 1/5 and a high frequency gain of 1 when the impedance of C1 is much smaller than R1 (C1 shorts out R1).

Passive high pass filter schematic

This equalization of the frequency response can be effective as we see in the following scope screen shot. Channel A (green trace) is uncompensated and channel B (orange trace) includes the high pass filter.

Passive high pass filter on CH B, ETS 25 KHz square wave

A significant down side to this simple passive filter is that it can be a substantial load on the driving signal. It looks like 1.5K at DC and 300 ohms at high frequencies. We can add a unity gain buffer stage to drive the filter as shown in the next schematic. The OP27 was chosen from the ADALP2000 parts kit for this because it is the fastest unity gain stable amplifier in the kit. It is not however a rail to rail amplifier. The OP484 quad might be an option but it has only a 4 MHz GBW vs. 8 MHz for the OP27.

Buffered high pass filter schematic

The addition of the buffer can have very similar results as we see in the next scope screen shot. Channel A (green trace) is uncompensated and channel B (orange trace) includes the buffered high pass filter.

Buffered high pass filter on CH B, ETS 25 KHz square wave

Another down side of the filter used in the first two examples is that it attenuates the DC signal (by 1/5). The gain is scaled up in the software to calibrate out this attenuation. But this also increases the apparent noise referred to the input by the same amount (5X). Another shelving filter structure is shown in the next schematic. It has a gain of one at DC and a high frequency gain of 5. We can use the even faster OP37 amplifier in this case because of the AC gain of 5.

However, this filter has a problem as well. Because of the gain at higher frequencies, the input amplitude will need to be reduced so that the output does not clip. The OP37 is also not a rail to rail amplifier so the clipping level is even smaller.

High Pass filter with unity gain at DC simulation schematic

In addition to its use as a teaching tool, the ALM1000 is an open hardware development board and we encourage advanced users to dive into the schematic and change the hardware. This is a good example of where more skilled users might modify their boards. If you have access to a surface mount rework soldering station and have some skill using it you could of course remove capacitors C42 (for channel A) and C26 (for channel B) and increase the bandwidth that way. This next scope screen shot shows the effect of removing the channel A low pass filter capacitor. Now we see a much faster rise and fall time basically limited by the 1.8 V/uSec slew-rate of the ADA4661.

Equivalent time sampling 25 KHz square wave, with (CB) and without (CA) filter

The low pass filter was included in the design for a reason and that was to reduce the noise. These next two scope shots compare the noise on the two channels with and without the capacitor. The first is with trace averaging off.

Just Channel A 10nF low pass filter cap removed, trace average off.

The increase in the noise on channel A is noticeable but still less than 0.5 mV p-p. This next one is with trace averaging turned on.

Just Channel A 10nF low pass filter cap removed, trace average on.

Trace averaging helps but there is still more noise in channel A. Because of the bandwidth and slew-rate limit of the opamp we don't need to remove all the capacitance and could add back somewhere between 1nF and 2nF. The next two screen shots of the noise is with 1.8nF added to channel A with trace averaging turned off and on.

Noise with 1.8 nF low pass filter cap on channel A vs 10 nF on channel B

With trace averaging on

Now the noise level is back to nearly what it is with the 10 nF cap.

So in conclusion we have shown that by using a few external components or modifying the board we can increase the usable bandwidth of the ALM1000 to where it is set by the performance of the ADA4661 opamp.

As always I welcome comments and suggestions from the user community out there.

Doug