anyone ever use AI to analyze a circuit?

[ners]

I wanted to learn more about my Orange Thunderverb 200 and found this preamp schematic, then ran it through Gemini and was surprised at the level of detail....but realizing I probably don't know enough to figure out how accurate the AI analysis actually is.

http://www.theallens.theallens.com/Thunderverb_main_preamp.pdf

I don't know how to share a link to the response, but it generated some interactive elements that let you click through each stage and it explains what it does.

This PDF (ORA-CD209, Issue 1.1, dated 23 November 2006) is the factory schematic for the Orange Thunderverb Main Preamp PCB (Sheet 1/2: Preamp 1 & 2, Loop, Reverb & Phase Splitter), designed by Adrian Emsley and drawn by Paul Stevens at Orange Musical Electronic Co. Ltd.

1. Document Metadata & Valve Complement​

Parameter	Factory Specification
Drawing Number	ORA-CD209 (Sheet 1 of 2, Issue 1.1)
Designer / Drafter	Adrian Emsley (Circuit Design) / Paul Stevens (Schematic/PCB)
Preamp Gain Valves	V10, V9, V8 — 3 × ECC83 (12AX7)
FX Loop Buffer Valve	V7 — 1 × ECC81 (12AT7)
Reverb Driver / Recovery	V6 — 1 × ECC81 (12AT7)
Phase Inverter Valve	V5 — 1 × ECC83 (12AX7)
Switching Relays	RL1, RL2, RL4 (Omron G5V-1 single-pole) & RL3 (Omron G6A dual-pole)

2. Stage-by-Stage Circuit Analysis​

A. Input & Cascaded Gain Stages (​

Both channels share the initial input stage before branching into separate gain/voicing paths switched by relays RL1-A and RL4-A (G5V-1

• Shared First Stage (V10-A Input jack SK1 feeds parallel/series 68K grid stoppers (R43, R45) and a 220n input capacitor (C19) into pin 2 of V10-A. The cathode is biased at 1.5K (R46) bypassed by 22µF (C20/C22) for full-bandwidth gain, with a 100K plate load.
• Tight Interstage Coupling (1nF → 2.2nF → 4.7nF Exiting V10-A (pin 1), the signal splits through notably small 1n0 (1nF) coupling capacitors (C24 and C25). Just like in high-gain British topologies, rolling off sub-bass right after the first stage prevents low-end blocking distortion and keeps the subsequent stages tight. Later stages step up to 2n2 (A2 off V8-B) and 4n7 (C26 into V9-A).
• Dual-Gang Gain Controls (RV1, RV4 Rather than controlling a single grid, Orange uses dual-gang logarithmic potentiometers (RV1-B with a 56pF bright cap C29 and 470K grid resistor; RV4-A / RV4-B A1M0 dual-gang 1M pots) to adjust drive simultaneously across multiple cascaded triodes (V10-B, V9-A, V9-B, V8-A, V8-B).
• High-Gain Biasing & Filtering: V10-B uses a 1.5K cathode resistor bypassed by 10µF (C21) and a 100pF plate-to-grid/snubber capacitor (C30) to tame ultrasonic parasitic oscillation. V9-B runs a warmer 820R cathode resistor (R49) bypassed by 100µF (C33) with 100K 1W plate load resistors (R39, R40, R41).

B. Tone Stacks: Channel A (3-Band TMB) vs. Channel B ("Shape" Filter)​

One of the defining characteristics on this schematic is how the two channels are voiced differently:

1. Channel A — Low-Slope 3-Band EQ (RV5, RV6, RV7
   • Uses a classic passive Treble/Middle/Bass network (C36 = 560pF treble cap, C31/C32 = 22n bass/mid caps, RV5 = A500K, RV6 = B25K linear Middle, RV7 = B250K).
   • The 22K Slope Resistor (R61 Where a traditional Marshall uses a 33K (or 56K) slope resistor, Adrian Emsley drops R61 down to 22K. Lowering the slope resistance shifts the midrange dip downward in frequency and lets a massive amount of low-mid thickness pass through, giving Channel A its signature chewy, stoner/doom-friendly midrange weight.
2. Channel B — Dual-Gang "Shape" Control (RV2-B / SH
   • Built around a B500KDG (500K Linear Dual-Gang) potentiometer (RV2-B labeled SH), paired with a 56pF high-pass cap (C35), 22n coupling/filter caps (C34), and a 15K tail/shunt resistor (R58).
   • Turning the single Shape knob clockwise simultaneously deepens the mid-scoop while boosting bass and treble—sweeping continuously from a fat, mid-forward British bark (counterclockwise) to a scooped modern metal contour (clockwise).

C. Tube-Buffered FX Loop (​

Both utility sections use higher-current, lower-mu ECC81 (12AT7) dual triodes instead of 12AX7s:

• FX Loop (V7-A & V7-B
  • Drives LOOP_OUT (Send) and recovers LOOP1 (Return) via 220n coupling caps (C1, C4, C5), 82K plate/load resistors (R5, R26), 56K (R17), and a cathode network using 680R (R27), 15K (R29), and 100R (R30).
• Spring Reverb (V6-A & V6-B
  • V6-A (Driver Biased with a 1.5K cathode resistor (R21) and 220K grid network (R22), driving the reverb transformer primary (TX11 Red / TX12 Blue) with tank connections on TX13/TX14 (Black / Green).
  • V6-B (Recovery Amplifies the returning reverb signal (REVERB2, 1.5K cathode, 22n coupling C14), with footswitchable muting/bypass handled by relay RL2-A (G5V-1).

D. Phase Splitter & Post-PI "Attenuator" (​

• Long-Tailed Pair Phase Inverter (V5-A & V5-B
  • Signal enters V5-A via a 47n coupling cap (C2) with 1M grid leaks (R19, R25), a 1.5K cathode bias resistor, and a 33K tail resistor (R13).
  • Negative feedback (-FB) from the output transformer secondary is injected into the phase inverter via a 6.8K resistor (6K8) and 100n cap.
  • Plate loads feed the balanced power amp drive lines PS1 and PS2 through 100n coupling capacitors (C6, C7, C8).
• The "Attenuator" Secret (RL3-A & RL3-B — Omron G6A
  • Notice the dual-pole G6A relay (RL3-A / RL3-B) wired directly across the phase inverter outputs PS1 and PS2. Despite being labeled "Attenuator" on the front panel, the Thunderverb's attenuator is not a speaker-level power soak—it is a footswitchable Post-Phase-Inverter Master Volume (PPIMV) that attenuates the drive signal exiting V5-A/V5-B right before the power tubes. This allows you to push the entire preamp and phase splitter (V5) into clipping at bedroom or stage volumes, then kick the relay out via footswitch for a full-volume solo boost.

Interactive Signal Flow & Stage Explorer​

Use the interactive diagram below to trace the signal path across ORA-CD209, toggle between Channel A (3-Band TMB) and Channel B (Shape), and inspect the exact component values at each valve stage:

Design Takeaway: Two details make the Thunderverb preamp (ORA-CD209) stand out from a standard British high-gain circuit: the 1nF (C24/C25) coupling caps immediately after V10-A (which strip sub-bass mud before the dual-gang gain cascade), followed by the 22K tone stack slope resistor (R61) on Channel A (or the B500KDG Shape filter on Channel B), which restores massive low-mid girth after distortion is generated.

Probably gonna get some hate on this lol...

 

[BlueParfait]

I have, with ChatGPT 6, on my Ceriatone Yeti just weeks ago, actually

Granted it's a layout, not a schematic, but it took around 5-10 minutes and was able to answer questions I later verified as true

It also did make a few mistakes a few times, but I caught them. Usually by misreading and making logical leaps in effect/marketing language rather than getting something fundamentally wrong

I cannot speak for Gemini's accuracy in analyzing circuits. I've only tested it (I have 3.1 Pro and Argon. It comes with the Google Drive space I wanted. Can't buy the space separately) with just text that I wrote myself and even then it made a lot of mistakes.

 

[ners]

I have, with ChatGPT 6, on my Ceriatone Yeti just weeks ago, actually

Granted it's a layout, not a schematic, but it took around 5-10 minutes and was able to answer questions I later verified as true

It also did make a few mistakes a few times, but I caught them. Usually by misreading and making logical leaps in effect/marketing language rather than getting something fundamentally wrong

I cannot speak for Gemini's accuracy in analyzing circuits. I've only tested it (I have 3.1 Pro and Argon. It comes with the Google Drive space I wanted. Can't buy the space separately) with just text that I wrote myself and even then it made a lot of mistakes.

I did this with Argon High since I got it for free with my phone.

3.1 Pro got stale like 6 months ago, I used it to analyze some Wizard schematics and while it was fine it didn't really provide the same level of insight as Argon did here.

I wish I could figure out how to share a link because the interactive UI it built is really nice to click through the gain stages instead of reading a massive wall of text.

 

[Bad Brain]

I just used chat GPT to troubleshoot a blown fuse in my ceriatone 2203. Found it helpful and knowledgeable but as noted it does make mistakes.

I also had it compare differences between the clone diagram and Marshall's schematic. Found a few and made some changes, found subsequent improvement.

 

[ners]

I just used chat GPT to troubleshoot a blown fuse in my ceriatone 2203. Found it helpful and knowledgeable but as noted it does make mistakes.

I did this as well for my Mesa Mark III++, was helpful to figure out which tube was bad (without having proper testing gear).

Definitely useful for generic troubleshooting steps, which is probably my weakest area of knowledge since I lack real world experience working on amps/etc.