AMX RX

[ivconic]

Maybe I was wrong to post all this on this topic? I should have started a new topic with the name: "Poor sourcing components & soldering skills man AMX version"!

[ivconic]

How many "low power" modules will be there?
RX&ADC is solved by this (above).
MCU and RX will draw significant current, I suppose.
Audio output probably too.
What else? LCD... insignficant.
So, for the rest of the circuitry I also have "poor sourcing components & soldering skills man" solution too:
DC-DC StepUp regulator 4A with XL6009. Additional 7805/7905 regulators can be added too.
All the power suplies are supplied then from one set of batteries, one battery pack; made of 3 or 4 LiPo 18650 (or similar), tied in series using BMS.


- Input voltage: 3-32V
- Output voltage: 5-35V
- Max. current: 3A, up to 4A with heatsink
- Efficiency: 94 % (max) Normal efficiency at 85%
- Frequency: 400KHz
- Accuracy: ± 0,5%
- Working temp: -40° C do + 85° C
- Dimm: 43mm x 21mm x 14mm

Test refference:
- In 3V Out 12V 0.4A 4.8W - One 18650 LiPo
- In 5V Out 12V 0.8A 9.6W
- In 7.4V Out 12V 1.5A 18W - Two 18650 LiPo in series
- In 12V Out 15V 2A 30W - Three 18650 LiPo in series
- In 12V Out 16V 2A 32W
- In 12V Out 18V 1.6A 28.8W
- In 12V Out 19V 1.5A 28.5W
- In 12V Out 24V 1 A 24W

...
Let no one worry about possible noise, interference, cross-talking from this module to the rest of the device, because there won't be any.
The FelezJoo PI detector has been on the forum for a long time, it is "known" for its sensitivity even to a "sharp look"... everything bothers
it to work properly.
The smallest detail, the smallest tolerance... and that detector doesn't work properly at all... Who made it; he knows the suffering about it.
I have done it on several occasions and always powered it with this module. With the addition of a couple of "thick" elcos on the module itself.
There was no problem at all! The module did not introduce any noise to the operation of the detector. Everything worked as it should.

[ivconic]

Like I said; I "don't like" the power supply. Since all those "chips" are SMD and probably tough to solder on pcb.
Not to mention sourcing them from abroad.
So I have "better" solution, probably for my case only.
I can obtain indeed cheap and very small module:

DC-DC convertor step up 5V to +/-12V
+12V / 50mA
-12V / 30mA
Input voltage 2.8 - 5.5V
...
Futrher I can add 78L05, 79L05 and 1117-3v3 or LM337L regulators (along with bunch of caps all around) to get voltages properly.
All these as integral part of the RX&ADC module.
Sweet and easy.
If my calculatiuons are correct; ADC will consume 12mA at 500ksps, which leaves 38mA for the opamps at positive rail and 30mA at negative rail.
Question is are the 78/79L05 regulators "eats" some mA during the regulation? If yes; I bet it is insignificant value.

[ivconic]

I would say yes, C7 needs to be 10uF, low ESR cap. This provides current to the ADC's internal reference when the Conversion starts.
No experience with this MAX11158 but with the LT2380 we (at my work) had the reference droop during conversion. This was solved by putting a big low ESR cap (10uF) directly across the ADC's Ref to gnd pins.

Ok, clear.
Once the ADC is negotiated finally; I plan to do this in form of module. But anything I can put through-hole; I intend to put through-hole. Only what I can't; I will make it SMD.
Such big value for non polarized cap is easier to find in SMD variant. Through-hole one would occupy too much "square meters" on pcb!
Now I am only waiting clarifications on those resistors and final concensus about the ADC chip.

[waltr]

[QUOTE=ivconic;n410321]Say Carl; C7, is it really 10uF non polarized cap? Or it is just a typo?

I would say yes, C7 needs to be 10uF, low ESR cap. This provides current to the ADC's internal reference when the Conversion starts.
No experience with this MAX11158 but with the LT2380 we (at my work) had the reference droop during conversion. This was solved by putting a big low ESR cap (10uF) directly across the ADC's Ref to gnd pins.

[ivconic]

Carl I am bit confused about power supply pins at IC1-3 ???
You engaged decompensation pin 8. alright.
You haven't draw PS pins, so I assume those go to +/-5v, right?
Any decouplings close to PS pins at chips?
RX3&RX4 DNS??
R1-4 ??

I see decouplings now, sorry! Was lazy to scroll down!
I need only clarification on those "DNS" values as well as the reason of R1-4 existence?
Ok, R3&4 I would use as 1k. To equalize&limit the current only.
Value will mostly depend on coil(s). 1K is wild guess.
But on the other topic we discussed thermal noise, so those resistors will introduce certain noise on signal paths, therefore... ?

[ivconic]

Say Carl; C7, is it really 10uF non polarized cap? Or it is just a typo?

[ivconic]

Carl I am bit confused about power supply pins at IC1-3 ???
You engaged decompensation pin 8. alright.
You haven't draw PS pins, so I assume those go to +/-5v, right?
Any decouplings close to PS pins at chips?
RX3&RX4 DNS??
R1-4 ??

[ivconic]

...In case we have final concensus about MAX11158?

So, there is no final consensus still?!

[Willy Bayot]

After these noise recalculations, are you now able to make a final recommendation on ADC?

I guess the MAX11158 is now out of the list since it has a SNR of 94dB

• LTC23xx-yy?
• How many bits?
• What conversion rate?

[ivconic]

MAX11158 lowes price at supplier would be 33e. Ok, I can live with that.
Problem is it must be sourced from abroad.
Will see how to do it. In case we have final concensus about MAX11158?
The rest of RX schematic I like. Doable.
I don't like power supply schematic. But the bench option is alright.
Now you talking Carl! That's what I wanted the whole time; something to start with.

[Carl-NC]

If we go with an 18-24bit ADC we want to drive it fully differential NOT single ended with a ground common.

Your posts reminds me that a fully differential drive where each side is +/- 5V is a full scale of 20vpp, not 10vpp. So my analysis above needs correction:

Differential driver: "This is 555 codes per volt so a 20V full-scale input (now assumed) would need 11,111 codes, or 14 bits. Or for a 16 bit converter @ 20V we have about 5.9 LSBs of noise."

Single-ended (assuming differential ADC driver): This comes to 4.3 LSBs of noise @ 16b.

We can also calculate the AFE SNR. In the differential case we have 1.8mvpp of noise and a 20vpp full scale so (maximum) SNR is 80.9dB. For the single-ended case it's 83.6dB. We want the ADC SNR to be, say, 6dB better than this so I should probably stop fretting about whether the ADC is 100dB or 93dB.

[waltr]

Ivconic,
when using newer SM part one needs to learn how to do solder past masks and oven reflow. This is not hard to do and there is lots of info on the web to build a 'toaster oven' temerature controller.

Tinker,
If we go with an 18-24bit ADC we want to drive it fully differential NOT single ended with a ground common.
This typically requires either a single to differential op-amp or three op-amps with reference Voltages.
By adjusting reference Voltage and gains almost any input Voltage range can be converted to the ADC's full scale input.

[Carl-NC]

Here is an updated noise analysis, see this post and this post for the original. I'll assume current mode since I know it offers the lowest noise. The coil resistance is now 10Ω and the feedback remains 100Ω.



The feedback resistor (100Ω) noise is



The opamp (AD797) input-referred noise is 0.9nV/rtHz (note: in the previous analyses I used the wrong gain; it's 2*RFB/RL+1, not RFB/RL+1):



The opamp current noise is 2pA/rtHz and is even more in the mud than before so I will throw it out.

The total differential noise of the first stage is therefore



Here is a sim plot of the first stage noise:



Before applying the second stage gain, we can take this number and divide by the first stage gain to get the total input-referred noise. The first stage gain is 2*100/10 = 20 (because it's differential) so the input-referred noise is 1.4nV/rtHz. In an opamp gain stage, the input-referred noise fundamentally can't be any lower than the opamp's input-referred noise which is 0.9nV/rtHz. The result of 1.4nV/rtHz suggests that the differential topology has a noise that is sqrt(2) higher than a single-ended topology, which makes perfect sense because there are 2 opamps in the differential circuit. In short, we can reduce the noise by using a single-ended preamp.

Anyway, 28nV/rtHz is applied to differential-difference amps with an overall gain of 10. Previously I ignored the noise contribution of the second stage because I know it's in the mud due to the gain of the first stage. For example, the 1k input resistor adds 4nV/rtHz to 28nV/rtHz which is 28.3nV/rtHz. If we like, we can bump this up to 30nV/rtHz and avoid any more math. So the second stage output noise is 300nV/rtHz. NBW is still 1MHz so the total integrated noise is now 300uV rms or 1.8mvpp. This is 555 codes per volt so a 10V full-scale input (now assumed) would need 5555 codes, or 13 bits. Or for a 16 bit converter @ 10V we have about 11.8 LSBs of noise. You'll notice this is worse than the original analysis (4.1 LSBs with a 1nV/rtHz opamp) but the overall gain here is 200 instead of 50, so we're actually doing better. Here is a sim plot of the total preamp noise:



Let's take a quick look at a single-ended preamp. To maintain the same overall gain I'll bump up the feedback resistor to 200Ω.









Dividing by the gain of 20 gives us an input-referred noise of 1.03nV/rtHz, just above that of the opamp alone, which is expected. After the second stage the noise is about 220nV/rtHz or 1.32mvpp. This comes to 8.6 LSBs of noise @ 16b.

All this suggests that no matter what you do, you are fundamentally limited by the input-referred noise of the first stage opamp multiplied by the gain of the preamp. And that is true for an opamp-only solution. Then you might be tempted to reduce the gain to reduce the noise but this actually reduces the SNR. You want to run the maximum gain that doesn't overload the preamp and you want a fair amount of that gain to be in the first stage so that subsequent stages don't contribute any noise.

If you dive into a discrete transistor preamp then it's possible to reduce the noise further. I've played with common-base front-ends and gotten the input-referred noise down to 0.5nV/rtHz, which would roughly cut our single-ended noise by half, or 4 LSBs @ 16b. In a future version I might consider this but I don't want to complicate the design at this point.

[Carl-NC]

I'm moving further discussion of the RX AFE from the "Concept" thread to this ("RX") thread. Here is the AFE I've been working on:



It is based on Tony's original concept, see post #2 of this thread. I have modified it so it can be run in either voltage mode or current mode. The schematic values assume current mode. I have also added a second subtractor stage to maintain a fully differential signal all the way to the ADC. The ADC says MAX11158 just because that's where I left off during the ADC debate. I will probably change it.

Right now I have the AD797 for the first stage because it's low noise (0.9nV/rtHz) and also has low flicker noise. Note that pin 8 is a "compensation" pin. Other candidate opamps sometimes use pin 8 for Enable or Shutdown so I've included resistor options for tying the pin high or low. The second stage shows the MAX412 but that's just a placeholder; the actual opamp is TBD.

Note RX1 at the RX coil and RX6 on IC3b. These allow converting the AFE first stage to single-ended mode. The possible reason for this is to improve noise. The fully differential architecture has sqrt(2) times the noise of a single-ended version. I will do another noise analysis in this thread that updates the numbers to this circuit.

Here is the power supply for the AFE:



This is designed around a "true ground" principle instead of the usual rail-splitter fake ground. That is, the battery ground is the actual ground for the whole circuit. I have not decided what batteries this will run on, I'm leaning toward 7.2V Li-Ion but maybe it's 4-C cells. The Positive Boost stage is for use with a 6V battery and is probably not needed for the 7.2V option, therefore it has a bypass option. Both analog rails are powered by linear regs that can be adjusted in case we want something other than +/- 5V. The MAX1735, however, is limited to -5V so I need to look at other options. Finally, there is a "Bench Power Option" that uses grungy old LM317/337 regs for a quick-n-dirty way to get started without building all the other power circuitry.