Visual walkthrough — TVC dynamics — gimbal servo bandwidth, time delay

3.5.45 · D2 · Physics › Guidance, Navigation & Control (GNC) › TVC dynamics — gimbal servo bandwidth, time delay

Hum ek sawaal ka jawaab de rahe hain: "Jab computer nozzle se koi angle maangta hai, toh actually kya nikalta hai — aur kitni der baad?"

---

Step 1 — Physical cheez: ek nozzle pivot par

KYA. Socho ek rocket engine nozzle jo ek hinge par swivel kar sakta hai (ek gimbal). Hum swivel angle ko $\delta$ (Greek letter "delta") kehte hain. $\delta =0$ matlab nozzle seedha rocket ke axis ke neeche point kar raha hai; positive $\delta$ matlab thoda tilt hai.

KYUN yahan se shuru karein. Kisi bhi maths se pehle, humein jaanna hai kaun sa physical object hum model kar rahe hain. Baad ka har symbol is ek hinged nozzle ki property hogi.

TASVEER. Figure dekho: burnt-orange nozzle ek pivot se latka hua hai. Angle $\delta$ dashed centre line se tilt hai. Iske upar chaar forces kaam karte hain — unse hum aage milenge.

---

Step 2 — Hinge par chaar forces → rotation ke liye Newton ka law

KYA. Jo bhi cheez spin karti hai woh rotation wale $F=ma$ ko follow karti hai. Rotation ke liye yeh padhta hai: "torque = inertia × angular acceleration." Hum nozzle par kaam karne wale har torque ki list banate hain.

KYUN yeh tool. Humein ek equation of motion chahiye — ek rule jo predict kare ki $\delta$ waqt ke saath kaise badlega. Newton ka second law (rotational form) exactly woh tool hai jo "yeh forces kaam karte hain" ko "yeh aise move karta hai" mein badalta hai. Koi doosra tool forces se motion nahi deta.

TASVEER. Figure mein har coloured arrow ek torque hai:

• Actuator torque $M_a$ (teal) — motor nozzle ko jahan chahiye wahan push karta hai.
• Damping $c\dot{\delta }$ (plum) — friction/oil resistance. $\dot{\delta }$ matlab "kitni tezi se $\delta$ badal raha hai" (rate, upar ek dot = per second). Tezi se motion → zyada drag, toh yeh $\dot{\delta }$ ke proportional hai. Yeh hamesha motion ko oppose karta hai.
• Stiffness $k\delta$ (orange) — structure se spring-jaisi restoring pull jo $\delta$ ko zero par wapas laana chahti hai. Zyada tilt → zyada pull, toh $\delta$ ke proportional.
• Inertia $J\ddot{\delta }$ — nozzle ki apni spin change karne ki reluctance. $\ddot{\delta }$ (do dots) = angular acceleration, "kitni tezi se rate khud badal raha hai."

Yeh ek mass–spring–damper hai, wahi maths jaisi ek weight spring par bounce karti hai — exactly isi liye servo ek second-order system nikalta hai.

---

Step 3 — Motor ko bataya jaata hai: "command ko chase karo"

KYA. Motor randomly push nahi karta. Yeh proportionally push karta hai ki angle kitna galat hai. Command ${\delta }_c$ define karo = computer ne jo angle maanga. Error hai ${\delta }_c-\delta$. Motor torque hai $M_a=K({\delta }_c-\delta )$, jahan $K$ ek gain hai (error ke har radian par kitna hard push karta hai).

KYUN. Yeh sabse simple possible feedback rule hai: "jitna zyada off ho, utna harder target ki taraf push karo." Ise proportional inner loop kehte hain. Isi wajah se nozzle commanded angle par kabhi jaata hai.

TASVEER. Figure mein do dials hain — commanded (teal) aur actual (orange) — aur unke beech ek red gap. Push arrow gap ke saath barta hai aur zero ho jaata hai jab dono align ho jaate hain.

$M_a=K({\delta }_c-\delta )$ ko Step 2 mein substitute karo aur $\delta$ terms left mein gather karo:

$J\ddot{\delta }+c\dot{\delta }+k\delta =K({\delta }_c-\delta )$ $J\ddot{\delta }+c\dot{\delta }+\underset{\text{total stiffness}}{\underbrace{(k+K)}}\delta =K{\delta }_c$

---

Step 4 — "Frequency lens" par switch karo (Laplace)

KYA. Hum "time derivatives" ko ek single algebra symbol $s$ se replace karte hain. Rule: time mein derivative lena = is nayi lens mein $s$ se multiply karna. Toh $\dot{\delta }\to s\delta$ aur $\ddot{\delta }\to s^2\delta$ (agar sab rest se shuru ho).

KYUN yeh tool aur koi nahi. Differential equations directly solve karna mushkil hai. Laplace transform calculus (derivatives) ko ordinary algebra ($s$ se multiply karna) mein badal deta hai. Yeh sawaal ka jawaab deta hai "agar main input ko frequency $\omega$ par wiggle karun, kya nikalta hai?" — jo exactly woh bandwidth sawaal hai jis ki hum parwah karte hain. $s$ ko socho ek knob ki tarah "kitni tezi se wiggle kar rahe hain."

TASVEER. Figure swap dikhata hai: left par wiggly time-domain equation right par ek clean polynomial-in-$s$ ban jaati hai, jaise chashma pehanna jo blur ko formula mein snap kar deta hai.

Rule apply karo aur $J$ se divide karo:

$(\underset{\ddot{\delta }}{\underbrace{s^2}}+\underset{\text{damping/inertia}}{\underbrace{\frac{c}{J}}}s+\underset{\text{stiffness/inertia}}{\underbrace{\frac{k+K}{J}}})\delta (s)=\frac{K}{J}{\delta }_c(s)$

"Output over input" mein rearrange karo — transfer function (servo ka input-to-output rule):

$\frac{\delta (s)}{{\delta }_c(s)}=\frac{K/J}{s^2+\frac{c}{J}s+\frac{k+K}{J}}$

---

Step 5 — Do numbers naam do jo sab kuch control karte hain

KYA. Messy fractions $\frac{k+K}{J}$ aur $\frac{c}{J}$ mein do meaningful numbers chhupe hain. Hum universal second-order template $s^2+2\zeta {\omega }_{n}s+{\omega }_n^2$ se match karke unhe naam dete hain.

KYUN. Physics mein har second-order system — springs, circuits, servos — ek hi shape share karta hai. ${\omega }_n$ (speed) aur $\zeta$ (settling style) ko naam dekar hum sab ke baare mein same tarike se reason kar sakte hain bina re-derive kiye. Yeh vocabulary ka badlaav hai, naya physics nahi.

TASVEER. Figure terms ko colour se match karta hai: constant term ${\omega }_n^2$ banta hai, middle coefficient $2\zeta {\omega }_n$ banta hai.

Yeh assume karte hue ki motor strong hai ($K\gg k$) steady gain $\approx 1$ hai (maango $5°$, eventually $5°$ milega), aur clean servo lag bachta hai:

$G_{\text{servo}}(s)=\frac{{\omega }_n^2}{s^2+2\zeta {\omega }_{n}s+{\omega }_n^2}$

---

Step 6 — Bandwidth: sabse tezi wiggle jo servo follow kar sakta hai

KYA. Servo ko frequency $\omega$ par sine wiggle feed karo. Slow $\omega$ par nozzle faithfully follow karta hai (output size = input size). Tezi karo aur eventually nozzle keep up nahi kar paata — uska output shrink hota hai. Bandwidth ${\omega }_B$ wahan hai jahan output $1/\sqrt{2}\approx 0.707$ ho jaata hai input ka (ise "$-3$ dB" kehte hain).

KYUN $1/\sqrt{2}$? Yeh standard "half-power" line hai: $1/\sqrt{2}$ amplitude par power (amplitude squared) exactly half hoti hai. Yeh "faithfully following" aur "visibly lagging" ke beech agreed border hai. (Yahi soch Bode plots padhne mein use hoti hai.)

TASVEER. Magnitude curve $|G(j\omega )|$ $1$ par flat shuru hoti hai, kabhi kabhi bump up karti hai (agar under-damped ho toh resonate karti hai), phir roll off karti hai. $0.707$ par dashed line ${\omega }_B$ par cross karti hai.

Output $1/\sqrt{2}$ kab hai yeh dhundne ke liye, $s=j\omega$ set karte hain (frequency $\omega$ par pure wiggle; $j$ phase track karta hai). Kyunki $s^2=(j\omega {)}^2=-{\omega }^2$:

$|G{|}^2=\frac{{\omega }_n^4}{({\omega }_n^2-{\omega }^2{)}^2+(2\zeta {\omega }_n\omega {)}^2}=\frac{1}{2}$

$x=(\omega /{\omega }_n{)}^2$ lo ise quadratic mein tidy karne ke liye:

$(1-x{)}^2+4{\zeta }^{2}x=2\Rightarrow x^2-2x(1-2{\zeta }^2)-1=0$

$x$ ke liye solve karo (sirf positive root, kyunki $x$ square hai aur positive hona chahiye):

---

Step 7 — Edge cases: $\zeta$ ke extremes par kya hota hai

KYA. Ek formula jis par tum trust karo use apne extreme inputs survive karne chahiye. Hum damping ke teen regimes test karte hain.

KYUN. Parent ne warn kiya tha "${\omega }_B$ aur ${\omega }_n$ ko confuse mat karo." Reason poori tarah is baat mein hai ki $\zeta$ bandwidth formula ko kaise bend karta hai. Hum sab cases dekh lete hain taaki koi scenario surprise na kare.

TASVEER. Ek plot par teen magnitude curves:

• $\zeta =0.5$ (under-damped): $1-2{\zeta }^2=0.5>0$. Curve girne se pehle bump up karti hai (resonance) → ${\omega }_B\approx 1.27{\omega }_n$, ${\omega }_n$ se bada.
• $\zeta =0.707$ (critical sweet spot): sabse flat possible curve, ${\omega }_B={\omega }_n$ exactly.
• $\zeta =1$ (over-damped, sluggish): $1-2{\zeta }^2=-1<0$, curve jaldi droop karti hai → ${\omega }_B\approx 0.64{\omega }_n$, ${\omega }_n$ se chhota.

Recall Bandwidth cases par quick check

$\zeta =0.5$ ke liye, kya ${\omega }_B$, ${\omega }_n$ se bada hai ya chhota? ::: Bada — lagbhag $1.27{\omega }_n$, kyunki under-damped resonance roll-off ko baad mein push karta hai. $\zeta =1$ ke liye, bada ya chhota? ::: Chhota — lagbhag $0.64{\omega }_n$; heavy damping jaldi roll off kar deta hai.

---

Step 8 — Delay: same shape, late aayi, aur phase chura leti hai

KYA. Nozzle response sirf lagged nahi hai, yeh ek delay $\tau$ ke waqt mein shifted bhi hai (sensor sampling, computation, valve transport). Frequency lens mein ek pure delay factor $e^{-s\tau }$ hai. Iska size $1$ hai (amplitude ke baare mein kuch nahi badalta), lekin yeh phase ko $-\omega \tau$ rotate karta hai.

KYUN yeh dangerous hai. Ek control loop unstable ho jaata hai jab uska total phase lag crossover frequency ${\omega }_c$ par $-180°$ reach karta hai (jahan loop gain $1$ pass karta hai). Delay free mein pure phase lag add karta hai — yeh silently tumhara phase margin spend karta hai magnitude plot ko touch kiye bina jahan tum notice karte.

TASVEER. Do identical sine waves: input, aur output $\tau$ se right mein shifted. Shift seconds mein chhoti lagti hai lekin high frequency par ek cycle ka bada fraction cover karti hai → large phase angle.

Kyunki $e^{-s\tau }$ transcendental hai (polynomial nahi), root-locus tools ise directly nahi chaba sakte, toh hum ise ek rational stand-in se replace karte hain — first-order Padé:

$e^{-s\tau }\approx \frac{1-\frac{\tau }{2}s}{1+\frac{\tau }{2}s}$

Numerator mein minus sign ek right-half-plane zero create karta hai $s=2/\tau$ par — yeh mathematical fingerprint hai "output pehle galat direction mein jerk karta hai," jo delay ka asli destabilising sting hai.

---

Ek-tasveer summary

Sab kuch ek canvas par: physical hinge (Step 1–3) ek second-order transfer function ban jaata hai (Step 4–5), jiski magnitude curve bandwidth ${\omega }_B$ deti hai (Step 6–7), jabki delay $e^{-s\tau }$ saath mein quietly phase rotate karta rehta hai (Step 8). Full servo jo computer actually baat karta hai woh product hai $e^{-s\tau }{G}_{\text{servo}}(s)$.

$\frac{\delta (s)}{{\delta }_c(s)}=\underset{\text{late, phase chura leta hai}}{\underbrace{e^{-s\tau }}}\underset{\text{itni hi tezi se move kar sakta hai}}{\underbrace{\frac{{\omega }_n^2}{s^2+2\zeta {\omega }_{n}s+{\omega }_n^2}}}$

Yeh seedha autopilot loop mein jaata hai, aur iska bandwidth pehle bending mode ke neeche rehna chahiye.

Recall Feynman retelling — poora walkthrough simple words mein

Hum ek nozzle se shuru karte hain jo hinge par latka hai. Teen cheezein usse move hone se rokti hain — uski apni heaviness (inertia), oily friction (damping), aur ek springy structure jo use wapas kheenchta hai (stiffness) — aur ek motor use push karta hai. Motor simple-minded hai: jitna zyada nozzle wahan se door hai jahan humne maanga, utna harder push karta hai. Woh likhhna ek bouncing-spring equation deta hai, aur har bouncing-spring equation ke do personality numbers hote hain: kitni tezi se swing kar sakta hai (${\omega }_n$) aur kya yeh overshoot karta hai ya crawl karta hai ($\zeta$). Agar hum command ko jaldi jaldi wiggle karein, kisi speed par nozzle aur keep up nahi kar paata — woh speed bandwidth hai. Under-damped servos girane se pehle ek chhoti resonant bump se bhi show off karte hain; heavily damped wale jaldi thak jaate hain. Aakhir mein, ek reaction delay hoti hai: nozzle sahi cheez karta hai, bas ek beat baad. Woh lateness motion ko shrink nahi karti, lekin timing ko rotate kar deti hai — aur ek feedback loop mein, kaafi rotation help ko wobble mein badal deta hai. Toh golden rules khud se nikhal aate hain: servo ko fast enough banao, delay tiny rakho, aur kabhi bhi bandwidth ko rocket ke bendy parts mein crash mat karne do.