_whitelogger has joined #tinyqma
<whitequark> it's here
<whitequark> so, my issue was, most stability plots in literature are done for a/q on axes
<whitequark> which is relevant if you want to keep frequency constant
whitequark changed the topic of #tinyqma to: design of an open hardware DDS-based QMA with a low-voltage dc/rf stage :: http://irclog.whitequark.org/tinyqma
<whitequark> but that doesn't really work for a fixed rf voltage, changing dc voltage and rf frequency
<whitequark> so the foremost priority in the design would be to work out stability conditions for these two parameters
<whitequark> bofh__: ^
<bofh__> yeah, your a needs to be totally different for starters
<bofh__> hm.
<whitequark> I failed diffequations eight times
<whitequark> which should give you an idea of how well I understand the derivation
<whitequark> actually, wouldn't it be the same stability plot? except instead of a straight line, we would trace some weird curve on it
<whitequark> bofh__: ^
<whitequark> so what i gather
<whitequark> is that i need a plot for µ
<whitequark> hm, no, not quite
<whitequark> so, actually, my q is fixed
<whitequark> oh, no, it's not, I'm varying the frequency
<whitequark> right. so. normally, you would keep everything constant except U and V
<whitequark> the varying Q/A ratio would change both q and a in the same proportion
<whitequark> which gives us the straight line on the plot
<whitequark> with the slope given with U/V
<whitequark> in my case, everything is kept constant except ω and V
<whitequark> the varying Q/A ratio *still* traces a straight line on the plot
<whitequark> except now the slope is given by uhhhhhh
<whitequark> still U/V, of course
<whitequark> ok, I got it
<whitequark> I keep U/V constant at all
<whitequark> and since my V is constant, U is also constant
<whitequark> thus what I have is the ratio of x/ω, where x=Q/A
<whitequark> or rather x/ω²
<whitequark> by varying ω², I change the range of ion masses which goes into that tiny peak slice on the stability plot
<whitequark> bofh__: does that make sense?
<whitequark> so let's say... 3mm rods, V=100V, m=0..2000
<whitequark> Q/A=0..2000 rather
<whitequark> at Q/A=1, omega=sqrt(100/(14.4434*1*(0.3)**2))=8.7
<whitequark> at Q/A=2000, omega=sqrt(100/(14.4434*2000*(0.3)**2))=0.19
<whitequark> or more precisely, 0.196123
<whitequark> at Q/A=1999, omega=0.196172
<whitequark> in this case, U would be 16 volts
<whitequark> 16.7843 for infinite resolution, a bit less for... finite
<whitequark> the smallest precision ground rods I see available are 2mm
<whitequark> which actually makes the job easier
<whitequark> er, above it's not omega. it's f in MHz
<whitequark> so for 2mm rods, V=100V, Q/A=1,2,1999,2000, f=13.1, 9.3, 0.294258, 0.294184
<bofh__> yes this makes sense
<bofh__> (sorry, just got back, laptop died so I used the interruption to get breakfast)
<whitequark> an additional advantage of using such tiny rods is that the capacitance is really small
<whitequark> maybe dozens of pF
<whitequark> which means your rf power is in milliwatts
<bofh__> er
<bofh__> x / <unicodefuck>^2
<bofh__> is what I'm seeing
<whitequark> <unicodefuck>?
<whitequark> it's supposed to be an omega
<bofh__> that's what I thought
<bofh__> it's showing up as unicode replacement char
<whitequark> ah
<whitequark> $7 per 1.8m
<bofh__> heck even at ~1nF the thought that I might be able to directly feed the output of the hackrf into this is sot of hilariously silly (max output power ~150mW)
<whitequark> 1nF at 13MHz is quite a bit
<whitequark> that's... 65W
<whitequark> also hackrf doesn't give you 100V
<bofh__> yeah it also isn't close to frequency-stable
<whitequark> so what I'd use here is a linear amp
<whitequark> directly from DDS output
<bofh__> exactly what I'd use as well
<whitequark> 100V... might be a stretch, but not by much
<bofh__> I can chain the output of a nice solid-state amp into a 3CX800A7-based linamp
<bofh__> that gets me 1kW @ ~200V if need be
<whitequark> holy fuck that's an enormous one
<bofh__> and that's just from crap I have lying around
<bofh__> $7 / 1.8m is, again
<bofh__> about an order of magnitude smaller than what I guessed it'd be
<whitequark> actually, I'm not even sure what linear amp to use for this
<whitequark> needs a tiny gain and large supply voltage
<whitequark> actually, why even bother with supplying it directly with 100V
<whitequark> 1:10 trafo is trivial
<bofh__> ...good point
<whitequark> air core
<whitequark> power it with 12V
<bofh__> (also if you need high voltage @ near unity gain, I believe I know the tube for that as well)
<whitequark> nah, I want a tiny PCB that can be mounted mechanically onto the flange
<whitequark> with the entirety of electronics
<whitequark> 100V is well within what can be just mounted on a PCB
<whitequark> take advantage of this!
<bofh__> yeah, 100V I'd not bother
<whitequark> heck even 600
<bofh__> 600V...meh. 1kV is when solid-state starts to really start being a pain.
<whitequark> have you heard of a cascode ladder?
<bofh__> built a few :P
<bofh__> (yes)
<bofh__> also wow
<bofh__> for Q/A ~ 2000 f is sub-MHz
<bofh__> with 2mm rods
<whitequark> yeah
<whitequark> this is actually fairly convenient that the relationship is with 1/omega^2
<bofh__> I am just surprised how small all these numbers are. Totally different from what I'm familiar with when I think QMA.
<whitequark> instead of just omega^2
<whitequark> because the resolution of DDS rises with m/z ratio
<bofh__> yeah
<whitequark> unlike regular QMA, where the resolution of the supply usually falls as you venture into HV
<whitequark> this is even better of an idea than what I initially thought
<bofh__> yeah wow
<whitequark> the annoying part here is keeping U/V ratio, as usual
<whitequark> the characteristics of the 1:10 air core trafo are not good
<whitequark> or rather, they are not precisely known offhand
<whitequark> in principle, it's possible to sample the rectified RF output
<whitequark> which introduces loss on the diode, etc
<whitequark> a better idea I think is just calibrate the instrument on a gas with known atomic weight
<whitequark> we know r0 to +-75 micron, omega to sub-ppm, and U to at least 1%
<whitequark> so we tune at, say, argon
<whitequark> singly ionized
<whitequark> or to hell with argon, just water, it's not like you can get rid of it even if you really want
<whitequark> and then you raise U until you just get out of range
<whitequark> that will be the vertex of the stability plot.
<whitequark> yeah, this sounds better, especially since you sidestep any confounding variables
<whitequark> such as crappiness of your ADC, losses in your diodes, parasitic inductance or capacitance
<whitequark> a second problem is high leakage inductance of an air core transformer.
<whitequark> so you'd need to also compensate for that to keep a constant U/V ratio among all the working frequencies
<whitequark> 0.06 Hz resolution. that's... four orders of magnitude better than what we need
<whitequark> hrm, it's up to 8 MHz
<whitequark> so we'll start from m/z=3. no big deal
<bofh__> just about any of the DDS chips AD makes that go up into the MHz range *should* work. TI makes similarly absurdly overkill-specced ADCs so that's not going to be an issue.
<bofh__> I'm slightly worried about the air-core transformer though.
<whitequark> bofh__: how in particular?
<bofh__> you mentioned leakage inductance, in my limited experience with them it's been very high
<bofh__> like you can compensate as you mentioned above, but I'm just wondering if there's a better part to use here
<whitequark> I haven't found RF transformers
<whitequark> 8MHz is ... way past what you'd use a ferrite for
<whitequark> there are premade RF baluns
<whitequark> but not 1:10
<whitequark> I could use some microstrip insanity at 8MHz, probably, barely
<whitequark> but not at 200kHz
<bofh__> I feel like I've seen 1:9 premade RF baluns but I can't find them right now so okay yeah air-core is our best bet here