
SEM
Serhio4522/SEM_hitachi-
Digitization of a Hitachi scanning electron microscope (SEM) — hardware design, FPGA firmware, and analog frontend schematics.

artificial waveform generator / function generator for DHO800 and DHO900 series (AWG/AFG)
AWG is an open-source printed circuit board by MatthiasElectronic, from the MatthiasElectronic/AWG_DHO8-900 repository on GitHub, published under the CERN-OHL-W-2.0 licence. It is a 4-layer board of 58 × 77.4 mm with 208 parts on 145 nets, routed with 1,331 tracks and 216 vias.
It is built around the AD9744ACPZRL7, NE5532DR-MS and OPA2673IRGVR, with the LM393DR2GH, BC846B and MMBT3906 among its other parts. Its parts include 90 resistors, 60 capacitors, 14 inductors, 11 diodes and 8 transistors.
This is a daughter board for the Rigol DHO800 or DHO900 series oscilloscopes to add the missing hardware for the function generator.
State: finished and working
Download all files as zip archive here
Discussion thread: https://www.eevblog.com/forum/testgear/awg-function-generator-for-dho800-900-series-oscilloscopes/
Hands-On video by Thomas from "hochohmig": https://www.youtube.com/watch?v=W03NUFdP6oc
V1: Original version, not recommended due to an offset bug and smaller mechanical misalignments.
V2: Second version, not recommended due to a bug with OPA2673 sometimes getting hot. The fix with V3 is to connected Pins A0 and A1 to GND instead of -6.5V.
V3: Recommended version.
More details are in the history page in the schematic.
All signals are created with this AWG board, mounted on a DHO804 with a DHO924S vendor.bin.
Measured on V1 PCB with the modifications of the V2 schematic.
The switching edges are clean. trise & tfall are 15.5ns. This is a picture of maximum amplitude (+/-5V), but rise and fall times were measured identical for smaller voltages.
The -3dB point is slightly above 25MHz. At the maximum selectable frequency of 25MHz, a peak-to-peak voltage of 719mV is measured, which is just above the 707mV of the -3dB point.
Distortion is over the most parts independent from the amplitude, just at very small amplitudes this is different.
Here, a measurement of a 1kHz sine signal at 10V and 100mV peak-to-peak voltage is provided.
At 10Vpp, the dominant HD3 is at ca. -62dB, so THD should be below -60dB.
Noise measurement setup:
A sine signal at very low frequency is applied, the trigger is set to the crossover and the horizontal range is small such that the signal appears as DC.
→ This way, AC RMS can be measured and equals the RMS noise at that signal amplitude. However, ultra low frequency noise is not visible.
Noise at 2Vpp signal amplitude:
Noise RMS = 334µV, Signal RMS = 707mV => SNR = 67dB
Noise at 10Vpp signal amplitude:
Noise RMS = 5.2mV, Signal RMS = 3.54V => SNR = 57dB
Noise source
The measured noise levels are dominated by ripples/spikes from radiated EMI of the switching requlators on the mainboard with unshielded inductors.
A 1.3MHz ripple is always present.
A 740kHz ripple is only visible for >2Vpp, when the x10 gain stage (relay K4) is active..
In the following FFT of a 10Vpp 1MHz sine signal, both ripples appear as peaks:
Luckily, this is coupled into the circuit before the attenuators, so at smaller signal amplitudes also the EMI ripple is attenuated.
In addition, there are random (burst mode) spikes clearling coming from the PSU brick. They can even be measured with the AWG switched off.
To clarify: The noise/ripple is small enough, it is not visible in time domain unless having a very small amplitude.
My prototype has -3mV DC offset, coming directly from the motherboard on the target "DC offset" pin. This is the most noticable offset error, because this one is fully visible at the smallest signal amplitudes.
Fixed this for my board, but this is completely related to tolerance. Additional offset errors are given in dependency of the AC gain and range selection, also this is amplified by 10 when the output signal exceeds 1Vp.
My oscilloscope mainboard has a gain error of 3% between highest and lowest gain reference value. This is the most dominant part of the gain error, so if there is a need to fix this tolerance, this has to be done specific for each oscilloscope.
Change vendor.bin
To add this AWG to the oscilloscope, a modified vendor.bin (DHO914S or DHO924S) to enable the AWG software functionalities is required.
https://www.eevblog.com/forum/testgear/hacking-the-rigol-dho800900-scope/msg5344076/#msg5344076
Offset Removal
After modifying the vendor.bin, a DC offset in the analog measurement channels is there and might even not be removable by recalibration.
A firmware update AFTER changing the vendor.bin, followed by a recalibration helps. I assume that the FPGA fimware is changed during a firmware update in dependency of the vendor.bin.
Hardware version resistors
Update: There is no need to change the resistors for version detection - both DHO800 and DHO900 series oscilloscopes work without changing resistors.
Add missing components to the mainboard
The DHO800 series oscilloscopes are missing some parts on the mainboard.
Luckily, all remaining components (resistors, reference voltages, digital interface, …) are already assembled.
It is NOT necessary to add any additional RAM to the DHO800 series.
Total cost for 1 board should be ca. 50€.
All components (apart from the connectors) are soldered on the top layer. The PCB is a four layer board.
This allowed economic assembly with JLCPCB (not sponsored), which is very cheap when using the preferred components. So the whole design is optimized around these preferred components to keep cost low. However, there are some components that are extended (and this costs ca. 3€ each extra). So most of them were ordered on LCSC, because this order was anyways necessary for the connectors and the OP-Amps on the main board.
All parts in this design files have a LCSC number, which makes ordering at LCSC or JLCPCB very simple.
Just be sure that you
When keeping extended components low (I only chose the SMD inductor), then this can be quite cheap. I ended up with 34€ on JLCPCB for 5 partly assembled boards and 41€ on LCSC for components for 2 boards, including tax and shipment. This adds up to 75€. It should be ca. 53€ if I only ordered components for 1 board. (05/2025)
CLK = Sample rate = 156MHz
The digital data is ALWAYS completely using up the headroom of the DAC.
Adjustments on the gain are only done by the analog gain input voltage, not digitally.
With Amplitude = Vpp set to 2V, all attenuation/amplification stages K1-K4 are disabled and the analog gain input is at its maximum value.
Still unclear is the "Protection" output signal, also it seems strange that this has to be pulled down to negative voltage.
The 5 relays are: /sqrt(10), /10, /10, *10, On/Off
The topolology is
The DAC itself is the same as on the original AWG board, but in a much cheaper QFN package. The allowed DAC output current was significantly overdriven in the original design, thus I reduced this current.
The OP amps are completely exchanged in favor for the price. The original 1GHz part just seemed to be overkill for a 50MHz signal generator. I chose the 230MHz OPA2673 instead. Sadly, there is nothing cheap in between that has high voltage and high bandwith. But the slew rate is sufficient.
Power rails are reduced to only +6.5V and -6.5V, which is just sufficient for the maximum +5V / -5V signal output voltage and at the border of the allowed OP amp operating conditions. These are both switchers, ripple is kept low by having forced PWM and filtering directly at the OP amp inputs. There is no need for LDOs as most LDOs cannot remove the 1.1MHz ripple. The 5V relays are rated for this voltage, so this is great as well. Although the original board had +15V/-12V, the same maximum output signal is achieved. Having only 6.5V means that the DC offset stage had to be redone, therefore all PI attenuators and OP amp gains had to be changed. As an easy addon, I also added matched impedances on the PCB.
The DAC output filter is completly redone. The original filter required very specific and uncommon component values, which are nowhere availlable, so this had to be completely changed.
In the original design there was a 9th oder elliptic filter, which sounds great for HF suppression, but has a terrible step response. As I wanted solid edges for rectangular signals, I opted for a filter with a much better phase. While also optimizing for minimum component variation, this turned out to be a bit „creative“ and a 5th order lowpass filter (somewhere in between butterworth and bessel). Simulation in Spice looked promising, but the measured edges (see above, trise/fall = 15.5ns) are slightly slower than simulated (10ns).
As can be seen below, the step response is much better while not exceeding the OP ampl slew rate. Looking at the attenuation over frequency, obviously the elliptic filter is better. My filter has a slight attenuation in the upper passband and a not so perfect attenuation at the sample frequency (156MHz).
Attenuation in the passband: This is noticable for frequencies above 10MHz, but can be corrected for sine signals by manually adjusting the amplitude.
Aliasing: Aliasing can only occur above the Nyquist frequency (78MHz). For signals with lower frequencies (e. g. <1MHz), where fsig << fsample, this is not important as the sinc-behavior in the frequency domain of the sample and hold circuit does sufficently suppress any aliasing (sinc(1-1MHz/156MHz)=-44dB), in addition to the lowpass attenuation. For larger frequencies, this will be worse and probably noticable in the FFT or a spectrum viewer.
Conclusion:
→ a worse aliasing attenuation, relevant only at high frequencies
→ a better behavior in the step and impulse response
...which I am not going to solve, but feel free to contribute.
Be warned that on the left hand side from the relays, the plastic case gets very close to the PCB, so even large MLCCs or SMD inductors can be too large and the relays definitly are.
Thanks to
Everybody contributing to the Hacking the Rigol DHO800/900 Scope thread
Pu6k1n [EEVBlog] for the schematic of the reverse-engineered original AWG, the pinning helped a lot
AndyBig [EEVBlog] for all information about DHO800 to DHO900 vendor.bin modification and reporting a bug with V2
Mechatrommer [EEVBlog] for inspiration from his own AWG attempt and lots of tiny information
Thomas from hochohmig.de for providing calculation tools, testing the board and creating a demonstration video - check out his channel
Contributers to KiCad and the library plugins
This is published with CERN-OHL-W-2.0 license, which is accessible in license.txt.
No liability is provided for this not working or causing any form of damage or harm. This is not a product, but a development board. Use at your own risk.
This project is influenced by the original AWG from Rigol, so only recommended for personal use. Although not checked, selling this AWG could possibly violate patents or rights of Rigol.
| Qty | Value | Footprint | References |
|---|---|---|---|
| 21 | 100nF | PCM_JLCPCB:C_0603 | C201, C202, C203, C209, C210, C305, C310, C312, C313, C315, C321, C323, C324, C325, C326, C328 +5 |
| 14 | 10uF | PCM_JLCPCB:C_0805 | C204, C205, C206, C211, C212, C213, C304, C309, C311, C314, C322, C327, C402, C403 |
| 12 | 100nH | PCM_JLCPCB:L_0603 | L301, L302, L303, L304, L305, L306, L307, L308, L309, L310, L311, L401 |
| 11 | 3.3kΩ | PCM_JLCPCB:R_0603 | R203, R208, R209, R408, R433, R434, R437, R438, R439, R502, R507 |
| 11 | BAV99_C7420336 | PCM_JLCPCB:SOT-23-3_L2.9-W1.3-P1.90-LS2.4-BR | D301, D302, D401, D402, D403, D404, D405, D406, D407, D408, D501 |
| 10 | 22kΩ | PCM_JLCPCB:R_0603 | R202, R205, R206, R210, R212, R435, R436, R440, R441, R442 |
| 8 | 100Ω | PCM_JLCPCB:R_0603 | R307, R310, R317, R319, R422, R423, R424, R425 |
| 8 | 1uF | PCM_JLCPCB:C_0603 | C406, C409, C410, C411, C412, C413, C502, C503 |
| 8 | 68Ω | PCM_JLCPCB:R_0603 | R204, R211, R318, R443, R444, R445, R446, R447 |
| 6 | 47uF | PCM_JLCPCB:C_1206 | C207, C208, C214, C215, C216, C217 |
| 6 | BC846B | PCM_JLCPCB:Q_SOT-23 | Q202, Q401, Q402, Q403, Q404, Q405 |
| 5 | ~ | AWG_Footprint:RELAY-SMD_G6K-2F-X-XX | K1, K2, K3, K4, K5 |
| 4 | 100pF | PCM_JLCPCB:C_0603 | C317, C318, C320, C329 |
| 4 | 10Ω | PCM_JLCPCB:R_0603 | R414, R415, R416, R417 |
| 4 | 180Ω | PCM_JLCPCB:R_0603 | R418, R421, R432, R460 |
| 4 | 390Ω | PCM_JLCPCB:R_0603 | R303, R314, R427, R428 |
| 4 | MountingHole_Pad | AWG_Footprint:MountingHole_4.0mm_SmallPad | H101, H102, H103, H104 |
| 3 | 100kΩ | PCM_JLCPCB:R_0603 | R403, R505, R508 |
| 3 | 15Ω | PCM_JLCPCB:R_0603 | R402, R405, R409 |
| 3 | 2.7kΩ | PCM_JLCPCB:R_0603 | R306, R312, R316 |
| 3 | 4.7kΩ | PCM_JLCPCB:R_0603 | R401, R419, R420 |
| 3 | 47pF | PCM_JLCPCB:C_0603 | C407, C408, C504 |
| 3 | 56Ω | PCM_JLCPCB:R_0603 | R304, R430, R431 |
| 3 | ~ | PCM_JLCPCB:FB_0805 | FB201, FB202, FB203 |
| 3 | ~ | PCM_JLCPCB:Hole_Tooling_JLCPCB | H105, H106, H107 |
| 2 | 0Ω | PCM_JLCPCB:R_0603 | R102, R103 |
| 2 | 100Ω | PCM_JLCPCB:R_1206 | R410, R413 |
| 2 | 10pF | PCM_JLCPCB:C_0603 | C301, C302 |
| 2 | 1kΩ | PCM_JLCPCB:R_0603 | R101, R504 |
| 2 | 22uH | Inductor_SMD:L_APV_ANR5045 | L201, L202 |
| 2 | 24kΩ | PCM_JLCPCB:R_0603 | R201, R207 |
| 2 | 3.9kΩ | PCM_JLCPCB:R_0603 | R501, R506 |
| 2 | 4.7Ω | PCM_JLCPCB:R_0603 | R301, R406 |
| 2 | 49.9Ω | PCM_JLCPCB:R_0603 | R305, R311 |
| 2 | 5.6kΩ | PCM_JLCPCB:R_0603 | R302, R315 |
| 2 | 560Ω | PCM_JLCPCB:R_0603 | R308, R313 |
| 2 | 68pF | PCM_JLCPCB:C_0603 | C303, C306 |
| 2 | LMR54406FDBVR | Package_TO_SOT_SMD:SOT-23-6 | U201, U202 |
| 2 | MMBT3906 | PCM_JLCPCB:Q_SOT-23 | Q201, Q301 |
| 1 | 1.8kΩ | PCM_JLCPCB:R_0603 | R458 |
| 1 | 120Ω | PCM_JLCPCB:R_1206 | R457 |
| 1 | 270Ω | PCM_JLCPCB:R_0805 | R404 |
| 1 | 4.3kΩ | PCM_JLCPCB:R_0603 | R407 |
| 1 | 47Ω | PCM_JLCPCB:R_0603 | R426 |
| 1 | 5.1kΩ | PCM_JLCPCB:R_0603 | R503 |
| 1 | 6.8kΩ | PCM_JLCPCB:R_0603 | R429 |
| 1 | 82Ω | PCM_JLCPCB:R_0603 | R455 |
| 1 | AD9744ACPZRL7 | AWG_Footprint:LFCSP-32_L5.0-W5.0-P0.50-BL-EP3.1 | U302 |
| 1 | BNC_Amphenol_ICC 031-5329-51RFX | AWG_Footprint:BNC_Amphenol_ICC 031-5329-51RFX | J101 |
| 1 | Conn_02x05_1.27mm_SMD_Header_M | Connector_PinHeader_1.27mm:PinHeader_2x05_P1.27mm_Vertical_SMD | X2 |
| 1 | Conn_02x20_1.27mm_SMD_Header_M | Connector_PinHeader_1.27mm:PinHeader_2x20_P1.27mm_Vertical_SMD | X1 |
| 1 | LM393DR2G | PCM_JLCPCB:SOIC-8_3.9x4.9mm_P1.27mm | U501 |
| 1 | NE5532DR | PCM_JLCPCB:SOIC-8_L4.9-W3.9-P1.27-LS6.0-BL | U303 |
| 1 | NetTie_2 | NetTie:NetTie-2_SMD_Pad0.5mm | NT401 |
| 1 | OPA2673IRGVR | Package_DFN_QFN:Texas_RSA_VQFN-16-1EP_4x4mm_P0.65mm_EP2.7x2.7mm | U301 |

Serhio4522/SEM_hitachi-
Digitization of a Hitachi scanning electron microscope (SEM) — hardware design, FPGA firmware, and analog frontend schematics.



cjheath/jlcbasic
JLC Basic parts library for Kicad fetched and converted from EasyEDA

energietransitie/twomes-opentherm-monitor-hardware
Open hardware design files for the Twomes OpenTherm Monitor device

Eight4aWish/eurorack_electronics
Analogue designs and breadboard layouts: the Dual Pingable LPG, and the board behind Boy.

nietaki/robal
Debugging expansion for the LABOR modular prototyping board by Moritz Klein and Erica Shnths

norbertsawe/custom_synth_controller
An open-source, high-performance synthesizer controller built around the Raspberry Pi RP2040 microcontroller. Designed for low-latency MIDI processing, versatile hardware control surfaces, and custom electronic…
It measures 58 × 77.4 mm and has 4 copper layers, 1.61 mm thick.
208 parts, built around the AD9744ACPZRL7, NE5532DR-MS, OPA2673IRGVR and LM393DR2GH. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.
From the MatthiasElectronic/AWG_DHO8-900 repository on GitHub, at the commit this page was made from: the board is V2/KiCad_project/AWG.kicad_pcb and its schematic V1/KiCad_project/AWG.kicad_sch. PCBKit opens the board in the browser, with nothing to install.
Its repository publishes it under the CERN-OHL-W-2.0 licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).