sinara-hw / Zotino

ARTIQ - compatible 32 channel DAC card in EEM standard
10 stars 1 forks source link

Zotino v1.1 errata #22

Closed hartytp closed 5 years ago

hartytp commented 6 years ago
a-shafir commented 6 years ago

@hartytp I am not sure how it is possible to set the TEC without i2c with the constrains discussed above. It can be a CPU in the TEC regulator that will decide the target temperature but even in this case i2c interface to the host will help a lot.

Also in the view of low PPM what will be the target load of the board? There is 500r 1% serial to the output. So if the load will be even 1M 5% it is not so precise. Even the leakage of bav99 can't be ignored because it can be from uA to tens uA! bav199 has much better specs. According the opamp specs it has short circuit protection and stable witc C load. So I think the output filter (if need) can be optimized to minimize R (50r or so, simulation need most likely). For the best DC precision better to remove the RC at all.

For the BOM. There are 0.1% resistors in the opamp filter BUT the same circuit uses 5% capacitors. I have not simulated it but it looks odd.

gkasprow commented 6 years ago

The TEC should be connected to GND by capacitor and common mode choke.. Otherwise we will get ground bouncing caused by possible external GND loop which we cannot control right now.

hartytp commented 6 years ago

For the BOM. There are 0.1% resistors in the opamp filter BUT the same circuit uses 5% capacitors. I have not simulated it but it looks odd.

Why is that odd? 0.1% resistors are pretty standard at most assembly houses and are very cheap. 0.1% (or even 1%) capacitors are not.

For this circuit, 5% is fine (we don't need the pole frequency to be set more accurately than that). Remember that the load will often be 100pF or more due to cabling etc, which is non-negligible.

hartytp commented 6 years ago

@hartytp I am not sure how it is possible to set the TEC without i2c with the constrains discussed above. It can be a CPU in the TEC regulator that will decide the target temperature but even in this case i2c interface to the host will help a lot.

Current setup is fine. We have a microprocessor in the temperature regulator which can do all that (the control loop is all digital). a second I2C temperature sensor is not required.

According the opamp specs it has short circuit protection and stable witc C load. So I think the output filter (if need) can be optimized to minimize R (50r or so, simulation need most likely). For the best DC precision better to remove the RC at all.

No! That's not a good idea. Lowering that too much increases the dynamic current when the output value changes, leading to nasty self-heating effects that will make the stability worse.

Also in the view of low PPM what will be the target load of the board? There is 500r 1% serial to the output. So if the load will be even 1M 5% it is not so precise. Even the leakage of bav99 can't be ignored because it can be from uA to tens uA! bav199 has much better specs.

The load for this will generally be capacitive with >1MOhm of DC resistance. Remember that we don't care about the absolute accuracy of the voltage as much as its stability.

Those diodes were selected for very low leakage. Looking at the datasheet, even at 85C, the leakage is <1nA at for reverse bias between a few V and 10V (the range we're in) with very low temperature dependence. even 1nA*500R = 500nV. 1LSB=20V/2^16=300uV. So the leakage is completely negligible.

hartytp commented 6 years ago

The TEC should be connected to GND by capacitor and common mode choke.. Otherwise we will get ground bouncing caused by possible external GND loop which we cannot control right now.

I see, you want a CMC on the +12V and GND return line for the CMC. Seems sensible.

In any case, let's not worry about this for now as the +12V may not be required in the end if Thermostat ends up using PoE...

gkasprow commented 6 years ago

I want them for TEC pins to cut common mode switching noise.

hartytp commented 6 years ago

I see, you're worried about that coupling from the TEC to the ground plane?

FWIW, I think that the MaximIC peltier driver ICs are probably low enough noise that this isn't going to be a problem, but a cheap choke + capacitor is probably not a bad thing to have just in case.

a-shafir commented 6 years ago

Right, there was BAV199 already. Sorry.

The system CPU via the I2C needs to communicate with the TEC regulator CPU. Otherwise I see no way to have it working correctly (predictable) depending on the airflow and ambient air conditions.

I think it makes sense to put the note or label about the 500r output impedance.

I understand your concern about the output currents.

I see the cost of 0.1% (purchasing reel) is 23c each and 1% around 0.2c. Both digikey. 5% indeed cost the same as 1% for most of the good brands but 0.1% is a very different story IMHO, it is a different material and technology. For this cost you can put 1% capacitors easily.

For the TEC power filters it makes sense to put it on the load side (regulator extension PCB). In this case you can ensure it matches the module noises etc. Also if the customers will not need TEC in the mass production the cost will be saved.

gkasprow commented 6 years ago

Zotino 1.1 arrived. TEC heatsink fits well. 20180222_172114

hartytp commented 6 years ago

Looks really nice! Can you try powering it on. If it doesn't die after 24 hours, can you ship me one for further testing?

hartytp commented 6 years ago

Greg, just received this. Looks really lovely, well done! Particularly like the TEC mounting.

gkasprow commented 6 years ago

There is missing thermal paste below the sink and the TEC. Since we will probably dismantle the stuck up I did not place it to not make hands dirty :)

hartytp commented 6 years ago

Zotino V1.1 seems to work. Nothing feels hot, all else seems as it should be.

Will leave on over the weekend to see if it spontaneously dies again.

Will aim to upload an ARTIQ driver on Monday if anyone else wants to play with it.

Generally, I'm very happy with this PCB. Great job @gkasprow.

hartytp commented 6 years ago

Drum roll...

So, the channel I'm looking at is still putting out voltages correctly after a weekend of operation, and nothing feels particularly warm to touch (although, I've just come in from the "mini beast" and my fingers aren't fully defrosted yet, so sensitivity may not be optimal).

Provisionally, this all seems to work. Phew!

gkasprow commented 6 years ago

@hartytp On 4 remaining modules I installed TECs in a way that can be removed with use of a screwdriver. I also applied the thermal paste.

gkasprow commented 6 years ago

to not expose the I2C interface, I added two resistors that are DNP by default.

hartytp commented 6 years ago

Great! Thanks Greg!

gkasprow commented 6 years ago

Moved release 1,2 to new, dedicated repository https://github.com/sinara-foundation/Zotino