Booster v1.1 testing

[hartytp]

Booster v1.1 here and looking really nice. The mechanics feel vastly improved from the initial version.

One minor niggle is that it would be slightly nicer if the screws on the FP were CS rather than cap head. Otherwise, it looks great.

Also, the lid was a bit tricky to remove because it fits so snugly (had to stick a screwdriver through the fan holes to push it out). However, once I did get it off I was rewarded by seeing how beautiful it all looks inside. I'm really impressed.

[hartytp]

Okay, cool. Cross-talk is better than 100dBc up to 400MHz. Nice! I'll post proper data on that later.

So, that more or less concludes all the measurements of analog performance I need to make. Performance is as good/better than I would have expected/hoped for by every measure! Well done to everyone involved.

Just need to check the interlock + power measurements and I can start using this in the lab.

[hartytp]

Gain for all channels:

• channel to channel variation in gain is about +-0.5dB so, within spec
• between 50MHz 450MHz the gain variation is just over +-1dB so, again within spec

[hartytp]

Cross-talk measured with 0dBm applied to channel 0. Cross-talk is the ratio of the power measured at output of channel n (input terminated) to the power out of channel 0.

[hartytp]

Same, but with channel 4 as the aggressor.

Wow! That's not much cross-talk, nice job (NB this is very close to the noise floor of the VNA, which wasn't carefully calibrated for this measurement)

[hartytp]

NB output power for that measurement was approx 1W (-10dBm + 40dB) so, while not a P1dB measurement, it's a good indication of the cross-talk at high output power...

[hartytp]

Some quick stress testing:

• turn Booster on/off a couple of dozen times with various gaps between on/off. done with rf applied to all channels (0dBm/all outputs 50R terminated) and with no RF applied.

[hartytp]

Quick test of output power interlock using RF at 200MHz. Interlock set to +37.0dBm. Actually trips at +35.7dBm. This is consistent with https://github.com/sinara-hw/Booster/issues/27#issue-341533805 where the power measures about +1.3dB to high on the power detector. So, that seems to work within spec! Good!

[wizath]

HW interlock is set at 36 dBm

[hartytp]

So, do you mean that the output power interlock is currently implemented in software, and that the hardware interlock (via the comparator) has a fixed threshold of 36dBm? I would much prefer it calling the INT:POW command also adjusted the hardware interlock (that way, the interlock can respond very fast to protect sensitive loads)

[hartytp]

Spent some time trying to break Booster by power cycling and turning it in/out of standby mode with and without input power (at varying levels) applied. So far, it all seems to be very robust and we don't seem to have managed to break anything. Great!

NB we always kept the outputs 50R terminated, as the reverse power interlock hasn't been implemented yet

[hartytp]

Looking for issues, such as PSU issues, when all channels are fully loaded. Running all outputs at 4W, 200MHz and looking at RF on a spectrum analyser.

[hartytp]

Same but very narrow span. Peak is 35dBm. See some spurs at -100dBc ish level. Will check whether they are on the synth or not.

Edit: spurs at around 200kHz and 600kHz offsets. SMPS?

[hartytp]

Looked at large signal cross-talk by driving ch 3 at 4W, 200MHz. All other channels at 201MHz, same RF power. The spur (combined cross-talk) is -92dBc. So, cross-talk with all channels loaded at max power is roughly as expected from the previous measurement. Good!

[hartytp]

Here is the synth on the spectrum analyzer.Carrier is -5dBm. Noise floor is a bit worse for this measurement, but at least the major spurs we saw were due to the synth, so I currently have no evidence of any spurs on Booster's output. We'll check this more carefully at some point in the future with a phase noise meter if we find the time.

Edit: oops, forgot to save data. Anyway it was basically the same as the above plot

[hartytp]

A different kind of cross-talk check:

• looking for slow thermal cross-talk or things like ground plane/power line sag when other channels are switched on
• ran channel 3 at 4W, 200MHz and let warm up while monitoring output power via a V3500A power meter
• once that had stabilized, we applied the same RF power to all the other channels and continued logging
• I don't see anything I can measure on a seconds timescale. On a timescale of about 1 hour, I see a 0.05dB decrease in the output power of channel 3. This is consistent with a change in temperature of a couple of degrees due to the other channels heating up a bit.

So, that looks good.

[hartytp]

Edit: fixed previous post -- the change in output power of channel 3 when all other channels are run at max power is 0.05dB, not 0.5dB (and that drift takes something like an hour). i.e. there is almost no thermal cross-talk to speak of!

[hartytp]

Also checked for fast glitches pm channel 3 caused by turning all other channels on, by looking with a 423B diode detector (see transient measurements I posted previously for the setup). Wasn't able to see anything.

So, the channel-channel isolation really is awesome!

[Jochenwolf3]

Further to post https://github.com/sinara-hw/Booster/issues/56#issuecomment-418664006, here is the power drift of a single channel driven at 200 MHz. Initially, only that channel is being driven to output 4W. After 40 min, all other channels are being driven to output 4W, too. After an hour, the resulting drift is just 0.05 dBm, so pretty good!

[Jochenwolf3]

Measuring the continuous mains power draw, it is:

• All channels in standby, fans off, no RF input: 30(10) W
• All channels enabled, no fans, no RF: 70(10) W
• All channels enabled, fans on, no RF: 70(10) W
• All channels enabled, fans on, 4W of RF: 70(10) W

For reference, our standard MiniCircuits amplifier is a ZHL-03-5WF+, which takes 2.8 A at 24V (67W) per channel, i.e. about the same as the whole 8 channels combined!

[hartytp]

NB for the 2W output powers that we typically operate at, we should be able to reduce the power dissipation by reducing the bias current to 80mA or so (although, this needs testing) if anyone can be bothered to try that out.

[hartytp]

For operation at the full 5W RF power, that's a wall-RF efficiency of 57% (!). Even at 2W, we're at 22%, which is not bad.

The MCL amps actually end up drawing more power than the data sheet values once AC to DC converter losses, etc are taken into account, so Booster is much better...

[hartytp]

Remaining tests on my to do list:

• check for switch glitches when the interlock is reset
• look at output RF with a phase noise meter (Booster performs well enough to use with Raman lasers etc, so this is a good measurement to have)
• generally check that firmware works
• confirm interlock functionality/accuracy over frequency/power/temperature
• confirm power measurement accuracy over frequency/power/temperature

[hartytp]

NB the power measurements are actually VA measurements, need to check whether a correction factor is needed to get the true power consumption from that. In any case, they are definitely roughly correct!

[gkasprow]

The DC/DC converters have power factor corrector circuit so it's very close to 1.

[hartytp]

In any case, the number I quote above has to be an underestimate of the efficiency since the power dissipated has to be lower than VA. So, the efficiency is extremely good!

It's kind of crazy to me that we can use Kasli + Sampler + Urukul + Booster to build a 16 channel AOM driver + feedback that dissipates less power than a couple of old fashioned lightbulbs!

[gkasprow]

Well, take into account that I used resonant converters that have 95% efficiency, unlike some popular flyback ones where one gets 75% for low cost models.

[hartytp]

Yes, that was a really good choice. Not sure if you've read all the data I've taken, but I've been extremely impressed with how well this all works.

[gkasprow]

200 and 600kHz are SMPS. Can you change relative position of mains and measurement cables and see if this changes level of the spurs?

[hartytp]

@gkasprow I measured approximately the same spurs when I connected my synth to the spectrum analyzer, so it seems that the problematic SMPS is inside the synth. I'll have another look at this with our R&S phase noise meter, which has a much lower noise floor.

[hartytp]

In any case, 100dBc is fine for these spurs IMHO (if you have good enough T&M equipment you always find something).

[gkasprow]

It's quite possible that the SMPS is leaking via the mains entry, then PE wire, then analyzer mains and in this way creates the ground loop. Isolation transformer won't probably help much due to its capacitance. One can check it by winding the SMA cable around big toroidal RF core. EMI cable clips can also help.

[hartytp]

@gkasprow I still see these spurs when I completely remove Booster and only look at my synth on the SA. So, I do not believe that they are related to Booster. I'll look at this again with a phase noise meter at some point (I might also try passing the RF output of Booster through an 30dB attenuator + isolation transformer to check for grounding issues).

Related: are you using a power entry socket with RF filtering (IIRC you are, but I wanted to double check)?

[gkasprow]

Yes, the power entry has RF filter in it.

[hartytp]

Thanks for confirming

[hartytp]

Another measurement of Booster's step response. This time, we toggle Urukul between two ampliutudes

Using a fast scope + 30dB attenuator. Peak to peak overshoot is about 1.7dB

[hartytp]

If we use the Urukul switch instead of toggling amplitudes then we see this:

The overshoot is quite a bit lower in this case, presumably because the rise time of the switch helps to smooth out high-frequency behavior

[hartytp]

@gkasprow @wizath two other things that popped up during Booster testing that we didn't report yet:

1. One one amplifier, one module had some issue, which meant that the red channel LED came on. However, at some point this went away. We haven't been able to reproduce it, so maybe it was a mechanical issue.
2. Yesterday we were setting up the new Booster that arrived. We measured the gain on one channel to be very unstable and about 35dB, so 10dB lower than normal. After a while that issue went away and we haven't been able to reproduce it. I remember having this same issue with another channel on a different booster some time ago. Not sure what the problem was, maybe some issue with channel initializing that was fixed by power cycling the unit?

Anyway, we can't reproduce these issues, so can't investigate further, but I thought it was worth mentioning them for posterity. @wizath have you had similar experiences on any of the channels you've used?