We have 3 different models of function generators/arbitrary waveform generators at work, and I figured it would be interesting to see the differences on the inside between all of them. They all do mostly the same things: DDS waveform generation, with one or two channels, and built-in plus user-specified arbitrary waveforms, topping out in the single-digit Mhz range (except for the BK 4063, with 80 Mhz max.).
First let's look at the "serious test equipment company" entry: the
Agilent 33210.
Here's the overview of the inside:
Not much on the bottom side of the main board except a good sprinkling of passives (mostly decoupling caps):
In the bottom-right is an Agilent-branded control chip (large BGA), likely with a processor and peripherals. Other chips nearby are likely RAM and flash.
Nothing exotic is visible for frequency sources, one of the crystals (probably the one near the FPGA, discussed next) serves as the frequency reference.
The FPGA at bottom-left probably does the actual waveform synthesis from parameters (frequency, symmetry/duty cycle, etc.), possibly with internal lookup tables for sinewaves but maybe fully-calculated on the fly. This drives a wide buffer (TI ...ACT16244, to its left) which probably interfaces it to the Analog Devices AD9744 DAC just above it. This is a member of the TxDAC family, specifically sold for waveform synthesis (SDRs, etc.); it has 14-bit resolution and outputs a differential current signal with externally-selectable full-scale (good for one level of amplitude control maybe?).
To the bottom-right of the FPGA is an ISSI IS61LV6416 64Kx16 10ns SRAM, almost definitely used by FPGA for waveform storage - at the very least this holds user arbitrary waveforms, and possibly also for generating its own pre-calculated lookup tables from the user parameters, to be able to just dump memory contents to the DAC when running instead of calculating everything on the fly. I don't know about the division of labor though - it's also possible that the main processor (instead of the FPGA) calculates every single point in the output waveform and just loads it up in this memory, while the FPGA does nothing but be a DMA-like memory-to-DAC communications bridge and the processor-to-memory comms interface. This would probably be the most flexible way to go, as the waveform generation happens in higher-level easier-to-modify firmware.
To the ADC's top-right is an AD1851 16-bit "audio DAC"...much slower but also precise. This probably controls either the offset voltage (if the offset is added in the analog signal chain, rather than having full-scale control digitally) or output gain scaling for the main DAC, if its reference current is adjusted directly.
The main DAC outputs go to 2x 69-ohm resistors for I-to-V conversion (likely tight-tolerance very stable thin-film), then into a relay which appears to select between two side-by-side chains of R/L/C, which are probably the low-pass "reconstruction" filters for smoothing. The two different filters might have different corner frequencies, if the DAC is set to different sampling rates for different frequency ranges.
Following upwards along the left edge of the board, more relays, ICs, and passives follow which are probably for the expected signal chain purposes: adding DC offset, changing gain ranges, turning the differential ADC signal into a single-ended output, etc. Relays have traces of flux around their leads, probably hand-soldered or soldered in a separate step as they have often have trouble with ultrasonic cleaning used post-reflow for the rest of the board.
Four unidentifiable (markings look like "ADG7139", but that part doesn't exist - these numbers were impossible to read in person too) MSOP-8s follow in a row - these may be output driver amps with exposed pads on the bottom for heatsinking. Finally, near the output connector (underneath that front panel overhang) is another relay and some large blue resistors, probably for the 50-ohm source impedance option. A 4-terminal piece of mystery magnetics (maybe a common-mode choke?) follows just before the (hidden) BNC output connector.
Power comes from a standard-looking AC-DC supply attached to the lid, and enters the main board into a connector at top-right.
The main board has a couple power supplies:
* A smaller one with 2 inductors (LTC 1940), maybe for the digital supplies.
* One with a couple external MOSFETs and a large transformer (but no serious isolation though) with at least 4 outputs, which probably supplies the positive and negative analog power rails. Two linear regulators with heatsinks are to its left, marked "+15V_ISO" and "-15V_ISO". An optocoupler seems to provide feedback.
The "ISO" is explained by the function generator output, which floats over a limited range (the case shows 42Vpk between output return and earth ground). This means that the whole output section, all the way from FPGA to output connector, have this isolated power supply, as opposed to the rest of the control and external comms (GPIB, USB, etc.) stuff which sits at earth ground for safety purposes. A few optocouplers sit between the main controller and FPGA, for the main controller to set parameters and download waveforms most likely.
Two linear regulators near the FPGA have large copper areas for heatsinking; an LM317 adjustable regulator (silkscreen: "+1...") and a smaller unidentifiable one (silkscreen: "+3"), which may provide clean power to the DAC...or just the isolated digital power from the +/- 15V isolated analog power.
Next up is the far-cheaper
BK Precision 4052. As a side note, the "&" in between the "B" and "K" in their name seems to come and go depending on the phase of the moon. It's absent here on the front panel.
The actual waveform generator circuitry is unremarkable and doesn't look very different from the Agilent: there's still the same recognizable outline of a processor, FPGA for waveform generation which feeds dual DAC chips, ranging relays, filters with lots of inductors, and op-amps for gain/offset. There's isolation here somewhere, as the output connectors are floating relative to earth ground but the USB connector is grounded; I can't tell exactly where it is though.
The most notable thing about it is the giant Siglent logo! Looks like it's just a re-branded Siglent product (although to be fair, the specs or some features of the design may be custom).
Overall, there doesn't seem to be any obvious "special sauce" that the Agilent has that this one doesn't, at least on the waveform-generation side of things. It's possible that the analog section gives significantly better performance on the Agilent, but otherwise, if I had to guess at the sources of the price difference, it would be exchange rate/engineering (having your whole company based in China, which especially helps for the engineering costs here, which unlike super-high-volume consumer electronics doesn't get amortized down to nothing over millions of units), a focusing on low cost through design and/or good-enough manufacturing (including maybe being willing to cut some mostly-inconsequential corners to get disproportionate drops), and cheaper parts sourcing maybe for things like the op-amps/LCD/connectors.
Also, the power supply is a custom many-output design integrated with the main board, rather than a separate off-the-shelf piece.
This right here shows the difference in philosophy between a high-end brand-recognition company like Agilent (now Keysight) and a cost-focused one like B(&)K Precision or Siglent. It doesn't make sense for Agilent to do a custom power supply for every product: it's a lot more time and money (not necessarily the design itself, but the validation, safety certification, EMI testing, etc.) for not really much benefit, as people aren't going to stop buying Agilent products if they're 10% more expensive - the company is much better-served by focusing on the part that they're good at, which is the measurement circuitry, and outsourcing the already-solved parts of the design to someone else by buying an off-the-shelf power supply (or maybe a standard design adapted to their requirements) from a 3rd-party supplier. That's why they have a standard-looking AC-DC supply followed by a few simple actually-custom DC-DCs for the exact voltages and isolation they need. With Siglent, on the other hand, they're trying to sell for as low of a price as they can with the given performance and features, so they'd rather cut out all unnecessary middlemen and have a well-integrated power supply design which is component-count optimized for exactly what they need, with a single stage and fewer parts overall, even if it takes much more effort to develop (and it will, with this 7-output thing, count the diodes!). I'm not making any judgements on either approach, as neither is strictly "better" or "worse" - it all depends on bigger goals and context.
Finally, here's the higher-frequency (80 Mhz) version of that last model: the
BK Precision 4063.
Like the other BK Precision, this is also a branded Siglent design. The digital control section is on a separate board here, maybe for shared use among their higher-end function generators, and it brings over some signals on a ribbon cable which go straight to a row of what looks like Analog Devices digital isolators for feeding the waveform-generation stuff on the main board.
The increased output bandwidth is visible here in the analog section. The layout is much tighter and components are more tightly clustered, as to be expected with higher-bandwidth signals for parasitics and EMI reasons. With these sorts of bandwidths you can't just leave an inch of extra trace in the middle of your filter, or there will be consequences for crosstalk (radiated signal to other channels), radiated EMI outside the instrument, and worst of all bandwidth flatness (resonances or low-pass filtering created by the parasitic inductances or capacitances).
The output drivers are also much beefier; in the 5 Mhz version they didn't stand out enough to be visible, but here each channel has 4 parallel amps (could be 8 actual amps total if those SOIC-8 packages have dual amplifiers in them) sharing the output drive through 8 output resistors.
The waveform generation part is also more interesting here too:
[
The DAC is another Analog Devices TxDAC, the AD9781, which has two separate outputs with 14-bit resolution, up to 500 Msps. Notice all the wavy differential traces leading to it: at the sort of digital data rate you need here, serial LVDS inputs are the way to go: clocking in some high-bitrate data streams over 14 parallel lines would mean having to route and synchronize those signals to each other with some seriously matched timing accuracy.
As for what feeds the DACs, there's a mystery chip under a heatsink and two Netsol S7N801831M 512Kx18 "NTSRAM synchronous pipelined burst" chips. The heatsinked mystery chip may be either something custom, or a fast FPGA for doing the communications from the memory to the DAC. These memories are interesting, since according to the datasheet they have a "burst mode", where you give it a burst command with a start address and length, and they do a fast sequential dump of a whole section of memory. I'd have to guess that this function generator takes advantage of that, with the high data rates necessary: the memory probably has the pre-calculated waveform points and the memories are just commanded to do repetitive bursts of that waveform data, which then gets sliced up and transmitted over the LVDS pairs to the DAC by the mystery-heatsink chip.
The power supply has the same number of outputs as the other BK, I think, but more diodes as some of them are paralleled. Interestingly, a lot of the diodes are directly fused.
Anyways, hope this was an interesting look inside. Apologies for the particularly bad photography on these.
Home
Help
Search
Login
Register
