100323 tests
mkid_final9e
sample rate = 340 MHz
2-carrier
PFB + 2^11 FFT + 2^6 FFT
FIR filter + decimation
final output ~100 Hz
tests as reported by Ran and Tom:
-------------
1a:
DAC directly to ADC
look at two bins, one with carrier, one without
carrier power:
f1 =
mean I = 180 bits
DAC side = full amplitude / 4
ADC side should be full amplitude * 0.7 / 4 / 2
(last factor of 2 is the division to avoid PFB overflow
= 179.2, so right on
mkid_final9e_carrier_23Mar_reproduce_onoff_r6_s1024_o0_byreal.eps.gz
no carrier bin shows expected -159 dBFS noise level
on-carrier bin is not gain-cleaned, so shows full 1/f gain noise. Hits white at about 10 Hz
1b:
same, except two carriers
f1 =
f2 =
carrier power:
mean I = 180 bits (f1), 173 bits (f2)
same idea as above, makes sense
mkid_final9e_carrier_23Mar_reproduce_r6_s1024_o0_byreal.eps.gz
after cleaning, i shows white noise at -156 dBFS down to 1Hz, q is a bit worse, white down to few Hz, slowly rising above there
---------------
2:
add simple IF system, parts hanging in midair
simple IF system =
DAC I/Q + IQ_upconvertor + IQ_downconverter + 20 dB amplifier + ADC I/Q
DAC full scale output = 2 dBm (rms power)
use DAC full scale amplitude / 4 -> subtract 12 dB -> -10 dBm
IQ upconvertor output gives -14 dBm to -16 dBm
-> case-by-case, we see IQ upconverter conversion loss is -5 to -6 dB
Simple IQ conversion process should give RF power equal to I or Q power (half of the sum of I and Q power is lost)
IQ downconverter has meausured -8 dB conversion loss (RF input to I or Q output), no measurement at output was done in this particular instance -> -22 to -24 dBm here
20 dB amplifier has NF = 3.85 dB -> 400-430K
expect -2 to -4 dBm at ADC input, measure -4 dBm
check voltage: -4 dBm = 400 uW rms -> 141 mV rms -> 200 mV amplitude, 400 mV pp. ADC full scale = 1.28 Vpp, so expect 320 bits amplitude. We see 350 bits, very good.
check noise levels:
* amplifier noise at ADC
TN = 400K
BW = 150 MHz due to LPF
rms noise power = 8.3e-13 W
ADC full scale power = (1.28 Vpp / 2.)^2 / 50 * (1/2)
= 4.1 mW = 6 dBm
so SNR = 4.1 mW / 8.3e-13 W = 4.9e9 -> 96 dB
ADC SNR expect = 64 dB. So amplifier should not be adding appreciable white noise.
The data are in files
2a: mkid_final9e_carrier_23Mar_withIFsystem_simple_on_off_r6_s1024_o0_byreal.eps.gz
2b: mkid_final9e_carrier_23Mar_withIFsystem_simple_bothon_r6_s1024_o0_byreal.eps.gz
2a tells us that in an off-carrier bin, we recover the expected -159 dBFS white noise level in I and Q, as expected the baseband amp adds no noise
2b: can clean out much of the gain noise, can get down to expected white noise ~5Hz and above in I, 10 Hz and above in Q
Conclusion: simple IF system adds less removable gain noise, but white noise is ok.
-----------
3: IF system to simulate dewar
DAC I/Q + IQ upconverter + 26 dB attenuator + 10 dB attenuator + 34 dB amplifier + 18 dB attenuator + 26 dB amplifier + IQ downconverter + 20 dB amplifier + 6 dB attenuator + ADC I/Q
DAC full scale output = 2 dBm (rms power)
use DAC full scale amplitude / 4 -> subtract 12 dB -> -10 dBm
IQ upconvertor output gives -14 dBm to -16 dBm as for case (2)
36 dB attenuation -> -50 dBm to -52 dBm
this is intended to be comparable to the carrier power that exits the dewar after the HEMT
adds a 300K noise temperature at this point
34 dB amplifier
-> has noise figure 1 dB -> 75 K noise temp
noise temperature = 375K x 34 dB = 940,000 K at output of amp
carrier power = -16 dBm to -18 dBm
18 dB attenuator
-> carrier power at output = -34 dBm to -36 dBm
noise temperature at output = 940,000K x -18 dB = 15,000 K
attenuator adds 300K of noise at this point, negligible
26 dB amplifier
-> carrier power at output = -8 dBm to -10 dBm
noise figure of amp = 3.5 dB -> 360K, negligible compared to 15,000K from prior stage, so neglect it
noise temperature at output = 15,000K x 26 dB = 6e6 K
measured carrier power at this point = -10 to -11 dBm, as expected
IQ downconverter: conversion loss of 8 dB
-> carrier power at output = -16 dBm to -18 dBm
noise temperature at output = 1e6 K
20 dB amplifier
-> carrier power at output = +4 dBm to +2 dBm
noise figure of amp = 3.85 dB -> 400-430 K, negligible compared to 1e6K at output of previous stage, so don't include it
noise temperature at output = 1e8 K
6 dB attenuator
-> carrier power at output = -2 dBm to -4 dBm
noise temperature at output = 2.4e7 K
measured carrier power = -4 to -5 dBm
check voltage: -4 dBm = 400 uW rms -> 141 mV rms -> 200 mV amplitude, 400 mV pp. ADC full scale = 1.28 Vpp, so expect 320 bits amplitude. We see 300 bits, very good.
check noise levels:
* all noises propagated to ADC
TN = 2.4e7 K
BW = 150 MHz due to LPF
rms noise power = 5e-8 W
ADC full scale power = 4.1 mW = 6 dBm from above
so SNR = 4.1 mW / 8.3e-13 W = 82000 -> 49 dB
ADC SNR expect = 64 dB. So we should see substantially higher white noise. In 0.1 Hz bins of final audio frequency FFT, expect SNR in FFT to be SNR in timestream x (340 MHz / 0.1 Hz) = 144 dBFS. This is in fact what we measure for the white noise in a bin without carrier in it, see
mkid_final9e_carrier_23Mar_withIFsystem_on_off_r6_s1024_o0_byreal.eps.gz
How cleanable? See
mkid_final9e_carrier_23Mar_withIFsystem_copy1_r6_s1024_o0_byreal.eps.gz
cleanable down to the higher white noise level.
This seems too good -- didn't think we had this much margin on HEMT noise. The max signal power in the readout spec is 30 pW/carrier with 144 resonators = 4.3 nW (-53 dBm). To make this match the full scale of the ADC, which is 4.1 mW, we would need a gain from the HEMT input of 60 dB, where the noise temperature is 2K. In the above setup, we have -50 dBm entering the lower noise LNA, with 375K noise temp, and -5 dBm entering the ADC, which corresponds to a gain of 45 dB. This gain does not include the HEMT gain, which is nominally ~30-40 dB, so we basically have much more gain relative to the input of the lower noise LNA than we would have in practice.
What's the right test? We want to use carriers that are at the high end of the power we might actually use with, say, 128 carriers, assuming clip avoidance works.
The DAC max output power is +2dBm -> 1.6 mW
The DAC output power per carrier is then 1.2e-5 W = -19 dBm
We typically want 10 pW per carrier at the HEMT, which is -80 dBm. So we need 60 dB attenuation from DAC output to HEMT input.
The total power over all carriers is -59 dBm.
The ADC max input power is +6 dBm. So we want about 67 dB of gain from the HEMT input to the ADC input.
The HEMT gain is about 38 dB, but there is cable loss of about 3 dB, so 35 dB gain.
So the power at the output of the HEMT is -24 dBm total, -45 dBm per carrier.
The room temperature low-noise LNA has 34 dB gain.
So power at output of low-noise LNA is -11 dBm per carrier, 10 dBm total.
IQ mixer wants 0 dBm input, so need to attenuate by 10 dB.
IQ mixer output will be -8dBm total -29 dBm per carrier.
ADC wants +6dBm input, so need to amplify by 20 dB and then attenuate by 6 dB. The single-carrier power at the ADC will be -15 dBm per carrier.
This corresponds to:
-15 dBm per carrier
40 uW per carrier mean power
112 mV pp -> 180 bits mean power
HEMT noise propagated to ADC.
total gain:
35 dB HEMT
34 dB LNA
-10 dB attenuator
-8 dB IQ downconverter
20 dB baseband amp
-6 dB attenuator
------
65 dB gain
noise temp: HEMT noise temp = 2K, so 6.3e6K at ADC
noise power = 6.3e6K in 150 MHz BW = 13 nW
ADC full scale = +6 dBm = 4.1 mW, so HEMT SNR at ADC = 55 dB.
For this test, we want 60 dB attenuation from DAC output to HEMT input (including IQ upconverter). Since IQ upconverter has -6 dB conversion loss, want 54 dB real attenuation. The dewar provides 43 dB, so we need to add 11 dB outside the dewar.
To mimic this without the HEMT, we want to present to the low-noise LNA the same power it would get in the above setup, which is -45 dBm per carrier.
In the mimicked setup, there is the following gain from the low-noise LNA input to the ADC:
34 dB LNA
-10 dB attenuation
-8 dB IQ mixer conversion loss
20 dB baseband amp gain
-6 dB attenuation
-----
30 dB gain
The LNA input will have a noise temperature of 375 K (from 300K attenuator + 75K amp noise), so 375,000K at ADC input.
The expected noise power is 375,000K x 150 MHz = 7.8e-10 W.
The full scale ADC power is +6 dBm = 4.1 mW.
So SNR will be 67 dBFS.
The desired attenuation on the input side is set by wanting the appropriate power level at the LNA input. That power level is -20 dBm total, -45 dBm per carrier. If we output from the DAC at the max possible levels of +2 dBm total, -19 dBm per carrier, we want 26 dB attenuation between the DAC and the LNA input, including IQ upconverter. Since the IQ upconverter has -6 dB conversion loss (one input to RF output), we need 20 dB real attenuation from IQ upconverter output to LNA input.
In Ran's prior tests with the dewar, he saw ~-125 dBFS white noise SNR in the 0.1 Hz audio-frequency FFT. Is this the expected value?
total gain:
35 dB HEMT
34 dB LNA
-18 dB attenuator
26 dB amplifier
-8 dB IQ downconverter
20 dB amplifier
-6 dB attenuator
------
83 dB gain
using HEMT TN = 3K, HEMT TN = 6e8K at ADC, in 150 MHz BW gives 1.2 uW power. The ADC full scale is +6 dBm = 4.1 mW, so we get a HEMT SNR of 35 dBFS. In the 0.1 Hz output spectrum, the ADC SNR of 64 dBFS corresponds to -159 dBFS, so we add (64 dBFS - 35 dBFS) = -130 dBFS. When we clean, we get an extra 3 dB, so expect -127 dBFS.