Abstract:

I measured transfer functions of all filters that can be switched from DGS in PLL CMS (S2415893).
Though Common BOOST 2/3 filter seem to have slightly lower zeros, it is probably acceptable.

Detail:

To replace the noisy PLL circuit (klog18014), I tested spare PLL CMS at test bench.

What we did:

1. Measure TF from IN1 to SLOW_OUT without turning on the filters
The last low-pass filter on the slow path should be modified, and it should have 500kHz cut-off frequency.
Figure 1 shows the result.
Though it has an additional phase delay at high frequency but should be acceptable because UGF of PLL loop is about 30kHz.

2. Measure TF from IN1 to SLOW_OUT with turning on filters one by one
Figure 2-6 show the TF when turning on COMMON COMPENSATION (p:40, z:4k, g:100), COMMON GENERIC (g:1), SLOW BOOST(p:4, z:400, g:100), SLOW COMPENSATION (p:4, g:100), and SLOW GENERIC (g:1)
All TFs seem fine.

3. Close SLOW loop with Moku:Go
Since SLOW boost filter has integration filter (p:0, z:10), it is difficult to measure the TF of SLOW BOOST 1/2/3 without closing the loop.
So, I made simple low-pass filter (fig7) with Moku:Go and close the CMS loop.
Cablings are just CMS SLOW_OUT -> Moku:Go IN1 and Moku:Go OUT1 -> CMS IN1, and CMS setting is as shown in fig 8.
Figure 9 shows the OLTF of closed CMS loop, which has about 30kHz UGF (ssame as real PLL UGF).

4. Measured OLTF with BOOST 1/2/3 and estimate TF of each filter
Since SLOW BOOST filters cannot turned on one by one, I measured OLTF with BOOST1 (fig10), BOOST 1+2 (fig11), BOOST 1+2+3 (fig12).
From these TF measurement, I calculated the TF of each boost filter (fig13:BOOST1 (p:0, z:10, g:1), fig14:BOOST2 (p:20, z=2k, g:100), fig15:BOOST3 (p:20, z=2k, g:100)).
Due to the high gain at low frequency, it is difficult to identify pole frequency of BOOST2/3.
In addition, zero frequency of BOOST 2/3 seem to slightly lower than the theory but probably it is acceptanble.

So, spare PLL CMS seems to work well in terms of the TF of each filter.
All measured data and plot are uploaded in JGWdoc (S2415893).

Images attached to this report

30942_1725262195_CLOSED-SLOW-10kLP_COMM-BOOST1-on_1Hz-1MHz.png

30942_1725262203_CLOSED-SLOW-10kLP_COMM-BOOST1-2-on_1Hz-1MHz.png

30942_1725262210_CLOSED-SLOW-10kLP_COMM-BOOST1-2-3-on_1Hz-1MHz.png

Comments to this report:

takahiro.yamamoto - 19:20 Tuesday 03 September 2024 (30961)

I tried to estimate a precise zero/pole of the 2nd and the 3rd stages of common boost filters.
For simplifying the fitting, I computed the ratio of two measured TFs which are with common compensation + 1st boost + 2nd boost and with common compensation + 1st boost.
This ratio should correspond to the transfer function of only the 2nd boost filter.

According to the fitting results (see Fig.1), zero/pole is 1.57kHz/16.7Hz.
Design value with R74:8.06k, R78:82 and C119:1000n are 1.96kHz/19.7Hz.
These differences seems to be slightly large in order to explain by the accuracy of registers and a capacitor.

The 3rd boost filter which is same design with the 2nd one was also fitted with the same manner as shown in Fig.2.
Estimated zero/pole is 1.61kHz/17.0Hz.

Though these differences doesn't seem to be serious for the PLL lock, it's better to check current PLLY CMS after bringing it back to Mozumi.

Images attached to this comment

30961_1725358830_Screenshot from 2024-09-03 18-41-16.png

30961_1725358833_Screenshot from 2024-09-03 18-41-31.png

takafumi.ushiba - 12:26 Saturday 07 September 2024 (31024)

I measured the BOOST1/2/3 TF of PLL CMS, which was used during O4a and replaced yesterday.
Figure 1, 2, and 3 show the TFs o BOOST 1, 2, and 3, respectively.
Old and new in the legend mean the data of S1809505 and S2415893, respectively

For all BOOST filters, old and new CMS measured data matched well.
This implies that there might be systematic shift of zero-pole frequency at BOOST 2/3 filters.

Images attached to this comment