Reports 1-1 of 1 Clear search Modify search
DGS (General)
takahiro.yamamoto - 2:57 Saturday 14 May 2022 (20715) Print this report
ADC glitch problem on many suspensions


- This work is related to klog#20658, klog#20667, and klog#20682.
- Low frequency noise was able to be removed by using a dedicated power supply unit for AA.
- GND of DC18V seems to be dirty.
- Next step is to find that a cause exists on the power supply unit (upstream of 18V power strip) or on the vacuum tank side (downstream of 18V power strip).


Because I wasn't able to reproduce the noise problem on the test bench with the combination of ADC and AA which seemed to be bad ones,
I did some trials in the mine again.

At first, I checked changes in ADC noise level by connecting GND of DC18V and DC24V.
Before this work, there was ~500ohm between GND of DC18V and DC24V. Voltage between two GNDs was ~70mV.
Both GNDs (at the center pin of the power strip) were connected to the digital rack.
Though I measured and compared ADC noise before and after this change, there was no improvement as shown in Fig.1 ~ Fig.4.
(Reference plots represent spectra before connecting GNDs.)

As the next, a dedicated power supply unit for AA chassis were installed.
AA chassis was unmounted from digital rack at this time.
So AA chassis should be completely separated from the original power supply unit for DC18V.
In this case, noise level was drastically improved as shown in Fig.5 ~ Fig.8.
This noise level is roughly same one in unplugging the SCSI cable between AA chassis and ADC.

And also, I tried to connect a common GND of the dedicated power supply unit to the GND pin of the power strip.
In this case, ADC noise got louder even if AA chassis was driven by the dedicated power supply unit (See Fig.9 ~ Fig.12).

According to these tests, It seems to be better to not connect AA chassis to current GND of DC18V.
I'm not sure GND was contaminated at the power supply unit or by connecting a vacuum tank via many signal cables yet.
Because I had no enough time to check this fact, I'll do that in next week.

Images attached to this report
Comments to this report:
takahiro.yamamoto - 16:42 Saturday 21 May 2022 (20815) Print this report
[Kamiizumi, Tomura, YamaT]


- We continued to survey the ADC noise issue on ITMX.
- We was able to confirm that there is no change in the glitch situation by the difference of KEPCO 30A power supply unit and a dedicated one.
- Presence of the 3rd AA (S2111159) seems to cause ADC glitches on the 1st and the 2nd ADCs though the 3rd ADC seems to be most quiet one.
- All measurements are stored in /users/DGS/measurements/ADC/K1IX1/2022/0520/*.xml
- After this work, I found too large glitches on MN_PS_V1.
- Now MN_DAMP_{R,P} cannot be engaged and Ushiba-kun implemented OLDAMP as temporal solution.
- We will check MN_PS issue on next Monday. (Hopefully Satellite box or CoilDriver was damaged.)


As I reported in klog#20715, I found ADC noise can be removed by using a dedicated power supply unit for AA chassis on ITMX.
On the other hand, a dedicated power supply unit didn't work well on ETMY as reported in klog#20762.

The difference between surveys on ITMX and on ETMY was the number of AA chassis which was connected to the dedicated power supply unit.
Only the 1st AA (AA0) was connected on ITMX and all three AAs (AA0, 1, 2) were connected on ETMY.

Difference by power supply unit
In addition to the above uncertainty, a question that which KEPCO 30A power supply or a vacuum chamber side is cause of this problem.
In order to find the answer of them, we checked the ADC noise with various situation listed as follows.
  1. Only AA0 was connected to KEPCO 30A power supply
  2. All three AAs were connected to KEPCO 30A power supply
  3. Only AA0 was connected to the dedicated power supply
  4. All three AAs were connected to the dedicated power supply

At this time, all circuits except AA chassis were disconnected from power supply.
Only in the case 1. and 3., ADC noise became quiet and glitches presence in the case 2. and 4.
This means that there is no difference between KEPCO 30A power supply and the dedicated one.
And also, some of (or combination of) AAs are noise sources.

Combination of used AAs
As the next, we checked the combination of AAs which was turned ON as follows.
  1. Only AA0 was connected to the KEPCO power supply
  2. AA0 and AA1 were connected to the KEPCO power supply
  3. AA0 and AA2 were connected to the KEPCO power supply
  4. AA2 were connected to the KEPCO power supply

In the case 1. and 2., ADC noise became quiet and in the case 3. and 4., ADC glitches can be seen on all ADCs connected to AA.
(The case that all AAs were connected to the KEPCO power supply was already confirmed in the check of power supply difference.)

AA2 itself seemed to be healthy because channels on AA2 were most quiet in all AAs on ITMX.
But when AA2 was connected to same power supply with AA0 and/or AA1, glitches didn't occur only on ADC2, but also on ADC0 and/or ADC1.

Swap test of AA and ADC combination
Because of this results, we doubted the AA2 and/or ADC2 as the cause of ADC glitches.
Finally, we measured the ADC noise by swapping the combination of the connection between AA and ADC.
In this test, ADC0 and ADC2 were connected to AA2 and AA0, respectively.

A result is the ADC0-AA2 had many glitches and ADC2-AA0 didn't have.
So we concluded the cause of this issue was AA2, not ADC2.

We plan to replace AA2 as spare one and current AA2 will be tested at Mozumi in next week.

Accidental troubles on MN_PS_V1
After this work, we reverted all situation and tried to revert ITMX as ALIGNED.
But MN_PS_V1 has too loud glitches which cause the DAC saturation in DAMPED state.

This channel is assigned on CH4 of ADC1. Because we mainly touched ADC0 and ADC2 today, we cannot any idea about the reason why such a change in the situation...
Because we didn't enough time to investigate this problem, we plan to do detailed checks in next week.
Hopefully, coil driver and/or satellite box may be damaged at turning ON/OFF power switch of each circuit (not a damage on photo sensor in the vacuum...)

As the temporary solution, Ushiba-kun implemented the MN_OLDAMP filters. So ITMX can be damped and aligned now by using MNOL instead of PS.
takahiro.yamamoto - 15:15 Saturday 18 June 2022 (21185) Print this report
[Kamiizumi, Tomura, YamaT]


- We tried to change a jumper pin configuration of AA. But there is no effect on ADC noise situation.
- ITMX cannot go ALIGNED just after recovering DGS racks. So we may need to wait until temperature of circuits became stable.
- ITMX is now in LOCK_ACQUISITION though coil output for GAS is around 30000ct.


Because Kamiizumi-san found a difference on jumper pin configuration from LIGO's one,
we measured ADC noise again with the configuration same as LIGO.

As reported in , we had already tried to change a jumper pin configuration on the filter board.
In this time, configuration on the interface board (
D1909971) was changed.
'P10' is opened as a default for AA chassis made by Hitz. On the other hand, jumper pin is inserted as a default for AA chassis which are LIGO's ones and a new product lots in KAGRA.

Basically its effect may not be so large when we use ADC as differential read mode.
But it's better to short here for RTN pin on ADC board itself.

We applied this change for S2113203 and measured ADC noise again. Unfortunately, we cannot see any improvement on ADC noise situation.
After the measurement, we left jumper pins in S2113203 and we plan to insert it all other AA chassis in the future.

Just after finishing our works, ITMX cannot go ALIGNED state because GAS output saturated.
After turning ON all circuits, readout value on LVDT GAS is change o(100um) in 1hrs shown in Fig.1 (We turned ON circuits around t=-1h10m).
Drift occurs in the direction of the set point value. So after waiting for a while, DC loop can be engaged.

Because IXV room is too cold, circuit temperature is drastically changed if we turned OFF circuits during almost one day.
I guess this is not only the drift of GAS position itself, but also the stability of the temperature of circuits.
Anyway, we need to wait at least a few hours for requesting ALIGNED after such kind of work.
Images attached to this comment
Search Help