Sponsored by GaGeReviewed by Maria OsipovaFeb 27 2025
Traditional ultrasonic imaging systems typically transmit broadband ultrasonic pulses across the entire sample volume under examination. Analysis of these pulses allows for the measurement of sample thickness, detection of reflective flaws, and identification of interfaces between different materials. Additionally, these pulses facilitate the study of fundamental material properties, such as elastic moduli.1
While offering the benefit of probing the entire sample volume, such imaging systems are generally constrained to lower ultrasonic frequencies due to the sample’s substantial ultrasonic attenuation, which increases rapidly with frequency. For instance, 10 MHz ultrasound travels only a few millimeters into soft biological tissue before experiencing severe attenuation.
However, despite undesirable increases in attenuation, the ultrasonic wavelength proportionately decreases as frequency increases, consequently enhancing spatial resolution of the system. As a result, higher-frequency ultrasonic systems inherently deliver superior spatial resolution.
Configuration
Higher-frequency ultrasonic systems are commonly referred to as acoustic microscopes, an example of which is shown in Figure 1. These microscopes typically operate with ultrasound frequencies ranging from 35 MHz to 1 GHz and beyond. Due to significant signal attenuation, these systems are often limited to probing only a few millimeters beneath the sample surface.
Furthermore, to optimize signal strength and spatial resolution, microscopy systems commonly utilize a focused transducer. The focused transducer is coupled to the sample using a droplet of ultrasonic couplant, often water, that is dragged along by the transducer as it is robotically positioned in a scan pattern across the sample surface.
Figure 1. Block Diagram of Scanning Acoustic Microscopy System. The red dot indicates the focal point of the transducer. Signal Out from the Pulser/Receiver is split in two by an RF splitter – one of whose outputs passes through a passive attenuator (ATTR) before entering the GaGe RazorMax digitizer. Image Credit: GaGe
The focal distance of the transducer defines a plane within the sample parallel to its surface that is scanned. This plane’s position can be adjusted slightly by moving the transducer perpendicular to the surface.
Transducers with varying focal distances can be used for larger adjustments. Contemporary microscopy systems now employ segmented transducers whose focal depth can be rapidly varied by altering the segment excitation pattern.
As in conventional ultrasonic imaging systems, microscopy systems excite a transducer by employing a broadband high-voltage excitation. Echoes reflected from scatterers within the sample are subsequently detected using the same transducer.
Figure 2 depicts typical ultrasonic waveforms, showing simple echoes from the sample's front wall and a scatterer at the transducer's focal point, as highlighted in Figure 1.
Scatterers are composed of interfaces between materials of different acoustic impedances that reflect ultrasound, providing contrast for images. The depth of a scatterer is calculated using the standard ultrasonic relation:
where v represents the acoustic propagation speed through the sample medium, and Δ t represents the elapsed time between the two echoes. The factor of 2 accounts for the roundtrip taken by the ultrasonic pulse. Additional details are provided by the amplitude ratio A2 / A1, which diminishes with attenuation by the media, and with scatterer depth, size, and contrast.
Figure 2. Two simulated ultrasonic signals were sampled at 1 GS/s, each of which contained two reflected echo pulses. The first echo is from the front sample surface and the second echo is from an embedded scatterer, which is shown in a magnified view within the green circle. Vertically separated for clarity, blue and red signals indicate unamplified and amplified signals, respectively, as described in the text. Conventional echo separation (Δt) and echo amplitudes (A1 and A2) are shown with dashed lines. Image Credit: GaGe
As mentioned earlier, acoustic microscopes can unveil internal interface structures within a sample by detecting echoes reflected from internal structures. Focused transducers may also capture additional ultrasonic echoes arising from ultrasonic mode conversion, surface waves, or dispersed, non-abrupt structures.
Although understanding the origin and nature of all detected echoes is ideal, users of acoustic microscopes may also operate phenomenologically, leveraging ultrasonic contrast to gain valuable information about a sample's interior without necessarily comprehending the precise underlying mechanisms.
A medical research client sought to develop an acoustic microscope capable of scanning biological tissue samples at ultrasonic frequencies up to 250 MHz, with a typical operating frequency of 70 MHz.
The researcher required a digitizer with high vertical resolution to detect small echoes and the capability to sample at up to 1 GigaSample per second (GS/s). The waveform lengths acquired by the digitizer must support depths within the sample of up to 6 millimeters.
The GaGe RazorMax CSE161G2 is a 2-channel digitizer capable of sampling at up to 1 GS/s, meeting the specified requirements. The RazorMax is a PC add-on card designed to fit into a PCI Express Gen3 x8 slot, which is available in many modern commercial PCs. Its 16-bit resolution enables an exceptionally low threshold of echo detectability, which can be further enhanced as outlined below.
The 6 mm depth requirement indicates that the sound will make a worst-case return trip of 12 mm, rounded up to 15 mm. Assuming an estimated propagation speed of 1500 m/second for tissue samples, the ultrasonic signal must be acquired for 10 microseconds (μs) following excitation.
Using RazorMax’s 1 GS/s sampling rate, these acquisitions will thus require the acquisition of (10 μs) / (1 GS/s) = 10,000 samples. The RazorMax can re-arm itself between successive waveform acquisition within 1 μs.
As a result, the RazorMax can trigger at rates of up to 90 kHz (= 1 / (10 μs + 1 μs) without missing triggers. The system will acquire waveforms at multiple positions scanned within a rectangular grid pattern. All system operations are managed by a Windows application developed in the C programming language.
This paper outlines three methods utilized in this application to enhance ultrasonic signal acquisition:
- Implementation of Dynamic Range Expansion (DRE) to maximize the detectability of small echoes.
- High-speed PCI Express Data Streaming to achieve maximum scanning speed.
- Application of GaGe’s Trigger Out signal for efficient and synchronous ultrasonic triggering.
Dynamic Range Expansion (DRE)
Ultrasonic signals frequently exhibit a high dynamic range, meaning they contain both high- and low-amplitude echoes within the same signal. To enhance the digitizer’s ability to detect smaller echoes, the user can employ Dynamic Range Expansion (DRE).2 Implementing DRE involves splitting the raw ultrasonic signal using an RF power splitter.
After being subjected to different levels of external amplification or attenuation, the splitter’s two outputs are then connected to two digitizer input channels, each subjected to different external amplifications or attenuations.
Figure 2 illustrates the simulated 2-channel acquisition of an ultrasonic signal. Shown in blue and red, respectively, the first splitter output signal is connected directly to the first digitizer channel, while the second signal first passes through a 20X (26 dB) amplifier before connecting to the second digitizer channel.
The signal acquisitions depicted in Figure 2 were simulated for 8-bit digitization to clearly demonstrate the effectiveness of DRE. The unamplified blue signal explicitly shows the acquisition of the first echo, as it vertically spans most of the digitizer’s 256 digital levels (28 = 256).
However, since it is 20 times smaller than the first echo, the second echo spans only 2 or 3 levels, as seen in the magnified view of the blue signal. It is evident that this second echo would have been missed if it had been less than one-256th of the first echo’s amplitude.
With its external amplification of 20X, the red signal in Figure 2 demonstrates that the first echo saturates the input and experiences severe clipping. Nevertheless, the magnified view reveals a much higher fidelity view of the second echo in red, as it now spans most of the 256 digital levels, compared to only two or three levels as stated previously.
DRE functions by using the signal for each echo from the channel that best suits it – namely, the unamplified channel for the larger first echo and the amplified channel for the smaller second echo.
DRE works effectively because high and low-amplitude echoes are separated in time, allowing for examination on separate channels at different times. DRE would not work, for instance, if the high and low amplitude components were separated in the frequency domain, as the resultant ubiquitous saturation of the signal would make it unusable.
In addition to requiring temporal separation of different amplitude signal components, DRE imposes two other conditions on the digitizer. First, the simultaneity of the two input channels must be precise. Any inter-channel time skew would distort the calculation of Δt. The RazorMax provides exceptional inter-channel alignment, with negligible channel skews under 30 picoseconds.
Second, the recovery time of the RazorMax from saturation must be quick enough to prevent signal distortion from persisting until the second echo. As shown in Figure 2, the RazorMax recovery time is clearly under 1 nanosecond, which is more than sufficient. Therefore, both the RazorMax and the user’s signal meet all the necessary criteria to use DRE.
In contrast to the 8-bit simulated case of Figure 2, where signal amplification was applied, Figure 1 demonstrates that the second signal passes through a signal attenuator. This compact, passive, in-line attenuator delivers 12 dB (4X) of signal attenuation with no degradation before the signal enters the 16-bit RazorMax’s optional ±240 milliVolts input range.
Although the enhancement provided by DRE in this actual case is not as dramatic as in the simulation of Figure 2, the 4X difference in signal amplitudes will enable DRE to improve the echo detectability threshold by up to 4X.
PCI Express Streaming Acquisition
The customer’s data throughput and data volume impose stringent demands on the digitizer.
As outlined earlier, typical waveforms contain 10,000 samples on two channels. The 70 MHz ultrasonic frequency results in an ultrasonic wavelength in water (v ≈ 1500 m/second) of (1500 m/second) / (70 MHz) ≈ 20 μm.
This wavelength corresponds to the approximate diameter of the region isonified by the transducer at its focal point. For consistency, this diameter should align with the spatial scanning increment.
Consequently, the user has decided to scan in a rectangular 2D grid with a 20 μm increment in each direction. With 2048 X 2048 grid points, the total grid area will be approximately 40 mm X 40 mm.
The total data volume will be:
- Total Data per Image = 2 Ch * 2 Bytes/Sample * 10000 Samples/Waveform
- * 2048 Waveforms/Line * 2048 Lines/Image
- ≈ 160 GigaBytes/Image
The large 8 GB memory onboard the RazorMax is insufficient to store all 160 GB of data for a single image. In traditional GaGe Memory Mode, this issue was addressed by halting RazorMax’s acquisition once its onboard memory was filled and all its data was downloaded. However, this download delay would substantially slow down the overall scan time.
The RazorMax allows its contents to be written to and read from simultaneously, thanks to its dual-port onboard memory. This capability facilitates RazorMax Streaming Mode, in which waveform data is streamed directly to the PCI Express (PCIe) bus via its onboard memory, functioning similarly to a FIFO to absorb small bus latencies.
Under optimal conditions, Streaming Mode acquisition does not require pausing for data downloads and can proceed uninterrupted throughout the image scan.
Although more powerful, Streaming Mode is more complex than Memory Mode and presents three limitations. First, the total data rate from the RazorMax cannot surpass the speed of the PCIe bus.
If exceeded, the RazorMax’s onboard memory will be filled with new waveform data at a rate faster than the PCIe bus can empty it, eventually resulting in memory overflow and data loss.
Similarly, the aggregate data rate cannot exceed the rate at which the Streaming Target can consume the data stream; otherwise, an overflow will occur. For instance, multiple parallel storage drives are frequently configured in a RAID array to support a high streaming rate.
Lastly, software applications running in Streaming Mode should not be written in inefficient high-level languages that hinder the instantaneous throughput of the streamed data. Ideally, the software should be written in C, a selection already made by the user.
If the acquired data stream were fully continuous without breaks, then the aggregate output data rate from the RazorMax in this application would be Data Rate = 2 Channel * 2 Bytes/Sample * 1000 MS/second = 4 GB/second.
The actual data rate will be lower than 4 GB/second because the acquisition will not be entirely continuous; instead, it will be broken up by dead spaces between triggered 10,000 sample waveforms.
The GaGe RazorMax indefinitely sustains data streaming rates up to 5.2 GB/second via its PCIe Gen3 x8 interface, ensuring sufficient speed for this application.
PC RAM serves as the PCIe Target for RazorMax’s data stream.
In fact, for all streaming acquisitions, RazorMax streams data to a pair of dual toggling PC RAM buffers, from which the data is moved to the Target. These buffers are allocated using special memory allocation routines provided by GaGe, as the RAM they consist of must be physical, contiguous, and limited in size (10 MB is typically used).
In this application, the user must allocate a 160 GB virtual, non-contiguous conventional RAM buffer. The toggling buffers are emptied into this 160 GB conventional buffer. Since the associated memory copy operations are quick, the RAM can easily keep up with the incoming data stream.
After acquisition, the software application retains all raw acquired waveform data in PC RAM.
From there, different sections of the data can be virtually examined, cropped, and used to generate map images of ultrasonic parameters extracted from the raw waveforms. It is possible to store only cropped data subsets or even reduced images instead of the enormous volume of waveform data.
Synchronous Triggering
Traditionally, ultrasonic systems configure a pulser/receiver to produce ultrasonic excitations at regular intervals. The pulser/receiver typically includes a Sync pulse output that coincides with ultrasonic excitation and can serve as an external trigger for the digitizer.
One limitation of this triggering method is that the Sync pulse is asynchronous or unrelated to the RazorMax’s 1 GHz sampling clock. The effects of this asynchronicity are demonstrated in Figure 3.
In Figure 3, Sync pulses from the pulser/receiver are represented with multiple black trigger edges.
Each of these edges results in the same indicated Trigger Point, regardless of whether the edge occurs immediately after the last Pre-Trigger Point or just before the Trigger Point. This uncertainty causes the corresponding repetitive signal acquisitions to dance left and right by exactly one sample.
Rather than being a flaw of the GaGe digitizer, this 1-sample jitter is a fundamental consequence of an asynchronous clock and trigger. Since the most critical ultrasonic parameters are typically the positions of echo pulses, asynchronous triggering inherently limits the accuracy of such echo position measurements to one sample.
Figure 3. Diagram indicating repetitive asynchronous triggers in black, whose edges wander in time with respect to the sample points. By contrast, the RazorMax’s synchronous Trigger Out pulse (shown in red) has an edge position that is fixed with respect to the sample points. Image Credit: GaGe
Like all GaGe digitizers, the RazorMax features a synchronous Trigger Out pulse, the timing of which is illustrated in Figure 3.
In contrast to the black asynchronous triggers shown in Figure 3, the Trigger Out pulse is synchronous with RazorMax’s sampling clock, ensuring that its rising edge consistently occurs at the same phase of the sampling clock.
As a result, when the RazorMax acquires the Trigger Out pulse, the pulse does not exhibit a 1-sample jitter. Instead, it will exhibit a significantly smaller jitter that is equal to the analog electrical jitter associated with RazorMax’s components, typically less than one-tenth of a sample.
By using the Trigger Out pulse to externally trigger the pulse/receiver, the ultrasonic excitation inherits the synchronicity of the Trigger Out pulse, thereby exhibiting minimal jitter.³
As shown in Figure 1, the Position Encoder is the original source of the trigger pulse, producing a pulse when the transducer reaches the next measurement position. This encoder functionality is sometimes referred to as Pulse on Position (PoP).
This set-up removes the need for slow and non-deterministic software communication between the Position Encoder, the PC, and the RazorMax, which would otherwise reduce scanning speed. Instead, the Position Encoder directly triggers the RazorMax in hardware, eliminating the need for any additional software-based synchronization.
The Position Encoder externally triggers the RazorMax, which in turn automatically generates the Trigger Out pulse. This pulse then synchronously triggers the ultrasonic excitation. By utilizing the triggering scheme described, the RazorMax minimizes signal jitter through the synchronous triggering of the ultrasonic pulser receiver.
Additionally, the quickest scanning time is achieved by triggering the RazorMax with a hardware trigger from the Position Encoder.
Conclusion
The GaGe RazorMax demonstrates the necessary digitizer characteristics for the application: specifically, a high 16-bit resolution for small echo detection, a high 1 GS/second sampling rate for the 70 MHz ultrasonic signal, a convenient PC form factor, and an affordable price.
Additionally, this article has illustrated that RazorMax can further enhance its exceptional echo detectability threshold through the use of DRE. The RazorMax’s capacity to stream data at speeds exceeding 5 GB/second enables 0 % dead time, allowing for continuous acquisition of ultrasonic waveforms throughout the scan.
Finally, synchronous triggering from the Position Encoder ensures the quickest possible ultrasonic triggering with minimal trigger jitter. The techniques outlined in this article can be applied to other fields, including laser microscopy and lidar.
Acknowledgments
Produced from materials originally authored by Andrew Dawson, Ph.D at GaGe (a Vitrek brand)
This information has been sourced, reviewed and adapted from materials provided by GaGe.
For more information on this source, please visit GaGe.