Products > Test Equipment
Upcoming Rigol DSG815/830
<< < (7/18) > >>
TurboTom:
For those of you who are interested in the DSG800 operational details...  :phew:
I don't guarantee for 100% accuracy but the diagram should be fairly correct.
TurboTom:
So here are a few remarks on the DSG815 (RF) hardware, looking at the block diagram I published yesterday:

It gets immediately obvious that Rigol's design is for the most part a simple "straight-through" approach. Since the main PLL oscillator covers a frequency range of 1:2 (1.82~3.64GHz), followed up by three by-two frequency dividers, the total frequency range is theoretically 1:16 (227.5MHz~3.64GHz) of which the range of 227.5MHz to 3.6GHz is used in the top-of-the-line model.

The lower range of 0~227.5MHz (okay, they start at 9kHz or the like, but for convenience I keep it "0"), the 910MHz signal that's available from the internal reference PLL divided by four, is mixed with the main PLL output divided by four (output range 910~682.5MHz), resulting in the required low frequency band as the difference. It's important to notice that now the control of the main PLL gets mirrored, i.e. all modulation schemes need to be mirrored as well (FM , PM and complex modulations will have to have the side bands reversed).

Another result of the straight-through approach is the necessity of a huge amount of filtering. In contrary to a full-range mixer approach with very high frequencies as input signals, say 5GHz fixed reference and 5~8.6GHz VFO to achieve a mixing product of 0~3.6GHz, where the sum and the mixer input signals are far away from the output signal range and the mixing product is fairly clean (if the mixer inputs were...), in this straight-through approach a lot of harmonics are generated due to the stacked dividers. To get this signal clean, a total of ten switchable low pass filters (plus a few that are doubled up) are required.

Since the filters can be arranged as distributed element configurations (for the upper end) and conventional passive filters, they are not really expensive. Most of the filter switching is accomplished by PIN diodes (HSMP-389B, now obsolete) which also isn't a driver for cost. The biggest advantage of this approach is that the highest frequency throughout the design is 3.64GHz, which may be just low enough to be able to use impedance-controlled FR4 as PCB material and not having to use more expensive and more difficult to process Rogers hybrid materials.

I didn't specifically address Rigol's modulation approaches (the two blocks that I identified as configurable band pass filters could also be amplitude modulators, utilizing HSMP-3832 PIN diodes, so bear with me...), but since the FPGA is controlling both the main PLL's reference clock and all the attenuators, it's well possible that these are the means for frequency, phase and amplitude modulation. The interface for complex modulation add-on circuitry is included in the signal path before the low-frequency band mixer. Still, this modulator will have to cover a very wide frequency range which is a considerable disadvantage of the straight-through approach.

There are two level detectors included in the design, one in front of the attenuator stack and one right at the output. The first one appears to be resposible for the level accuracy since it's heated and temperature stabilized. The second one is probably used as a protection monitor (in case RF is fed back into the output) and as a power monitor for higher output levels when the PA is used.

There is no reason that the entry level models of this instruments cannot be unlocked by software to cover the full frequency range unless on recent production models, part of the filters and switches aren't populated. But since these components really aren't ruining the bill, it would probably be less economical to run different production batches than having all the RF boards produced identically.

The DSG800 series is clearly built to a budget but I'ld still say that no real corners had been cut that would affect performance. The straight-through approach has also been utilized by other manufacturers, i.e. Dave's Marconi SG follows a similar route.

So much for that, if I'ld get my hands on a set of good photos of the "A" version torn down, I'll add some information on the I/Q modulator...
noreply:
TurboTom,

Thank You so much for your very insightful observations and explanations  :clap:

A few questions, from a novices point of view ;-

1. Given Rigol's 'straight-through' approach - are there any inherent performance (i.e. better phase noise, smaller harmonics, etc, etc) advantages over the full-range mixer approach?
2. Did Siglent adopt a similar approach in their SSG3000x range? or the full-range mixer approach?
3. Are there inherent performance benefits of the full-range mixer approach - if so, what are they?

Thanks

TurboTom:

--- Quote from: noreply on July 11, 2020, 11:44:01 pm ---TurboTom,

Thank You so much for your very insightful observations and explanations  :clap:


--- End quote ---

Thank you very much for your kind words  ;)


--- Quote ---
A few questions, from a novices point of view ;-

1. Given Rigol's 'straight-through' approach - are there any inherent performance (i.e. better phase noise, smaller harmonics, etc, etc) advantages over the full-range mixer approach?


--- End quote ---

Actually, there are: Provided the same "care" is taken designing an oscillator for the fundamental frequency or one for a higher frequency suitable for mixing, the phase noise would be better for the straight-through approach. Dividing the fundamental frequency to provide the lower bands would further improve the phase noise. Rigol provided a factor "N" in their frequency band table which actually resembles the reciprocal divisor by which the LO frequency is divided to provide the configured output frequency. If a mixing approach is taken to produce the desired frequency, the resulting phase noise will be the geometrical sum of the phase noise of the two frequencies to be mixed, whereas it scales with the "N" factor in case a straight-through approach is utilized. You've got to be aware that the lowest band of the DSG800 utilizes the VFO to be divided by four and then mixed with the 3.64GHz reference divided by four, which results in a worse phase noise than Band 2 which is generated form the VFO divided by eight (see Rigol's table).




--- Quote ---
2. Did Siglent adopt a similar approach in their SSG3000x range? or the full-range mixer approach?


--- End quote ---

Sorry, I cannot answer this question exactly since I don't know of a complete tear-down of one of Siglent's SSG3000's RF modules. But the specs of the SSG3000X make me believe that Siglent had a very thorough look at Rigol's design when they started work on their generator... Yet the second lowest frequency band is the mixing product of the VFO divided by two, so provided the phase noise of the VFO of these two specimen is similar, Siglent's approach will be inferior (for whatever reason, Siglent splits the "mixed" frequency range into two, first 9kHz~1Mhz with "N" = 0.25 and the second 1~250MHz with "N" = 0.5).




--- Quote ---
3. Are there inherent performance benefits of the full-range mixer approach - if so, what are they?

Thanks

--- End quote ---

There are a some benefits of the full-range mixer approach: Provided the mixing partners, i.e. local reference and VFO frequencies are sufficiently "clean", the amount of filtering / filter switching required to provide a sufficiently "clean" output signal is reduced considerably. It basically reverts to an LPF suitable to block the sum frequency and the feed-through of the individual mixing partners. Modulation may also be easier since everything can be accomplished at a single frequency (the internal reference). Albeit, this frequency is much higher than in case of the "straight-through" approach, so the semiconductors required may be more expensive.

Edit: Forgot to mention another substantial advantage of the full-range mixer approach: Since range switching isn't required, signal generators following this approach will be capable of seamless frequecy sweeps over large ranges without dropouts or "phase hickups" which unavoidably will result during range-switching of a straight-through SG. Same situation for (ultra) wide-band FM, if your carrier is right on the edge of a band.

I hope this answers your questions so far.

All the best,
Thomas
noreply:
Thank you once again for great explanation.

Yes - looking through the specifications for both the DGS800A series and the SSG3000X series - Rigol has slightly better phase noise figure quoted - from memory it was  -112 dBc/Hz typical to Siglent's  -110 dBc/Hz

Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod