Let us look at this again, I copied this from the schematic for the Cobra 2000GTL instead, which shows (missing in the above 148 schematic) the AMC control pot (VR12 500 Ohms) and D64 from Base to Ground on TR26, which was in EVERY Cobra 148GTL (and Grant XL) I had worked on up until the year 2000, when I moved to Florida to do something else for my main source of income (steady job, steady pay). This pot and diode are missing in every MB8719 Cobra 140/142, Washington and Madison also, and I replaced the preset 1.5k resistor (R130) in this circuit to give adjustment to the ones that were missing.
C109 has it's Positive terminal going to the main +8v source that feeds the Mic Input and AMC/ALC circuit. It has it's Negative terminal going through R128 (15k) and R129 (33k) to the same +8v source, or a total series resistance of 48k. R129 also biases the Base of TR25, a PNP 2SC733. Under no modulation conditions, both sides of C109 are at +8v, or possibly 7.3v, or 8v minus the Emitter/Base diode drop of 0.7v. R127 (1 Meg) has little or no effect on this circuit, other than giving it a very slight signal path to reference Ground. C109 does not charge or discharge through R127, it Charges Negatively (pulled down from +8v) from either the AMC Sense or RF Sence in SSB modes, and discharges back up to +8v through R128 and R129. When the Base of a PNP transistor is pulled Negatively, it causes the transistor to conduct, thereby pulling the Collector more positively, and in so doing, it pulls the Base of TR24 up, causing it to clamp on the Microphone Audio Input.
In the AM Mode, the Emitter of TR26 is connected to the feed for the Final and Driver through R166 (3.3k) then the Mode switch. The AM position of the Mode Switch goes to the AM Modulator. TR26 is a Common Base Amplifier, or sometimes called an Emitter Follower. Which means the Base and Collector are "following" whatever voltages the Emitter "sees". It ONLY reacts to Negative pulling voltages. The Collector of TR26 "pulls" the Base feed Resistor R128 of TR25 and C109 stores this negative pulling voltage, thereby Sampling and Holding the Negative Voltage swing. The harder you pull the AM Regulator circuit on it's Negative Peaks to Ground, the harder and longer it clamps down on the Input Audio being fed to the Microphone Amplifier Circuit. Also, the Mode Switch has a +8v constant supply that when in the AM mode, supplies +8v to R159 (8.2k) and Diode D42, to the Emitter of the RF Peak sensing transistor TR34. This effectively "turns off" this transistor, unless VR11 is overadjusted to the Ground voltage side of the control, thereby causing this circuit to "clamp" the AMC/ALC circuit above.
When in the SSB modes, the Final and Driver are hard clamped to +12v through the Mode Switch, and no longer on the AM regulator. This effectively turns off TR26, because it has no negative going voltages to sense. The Mode switch also releases the +8v AM voltage to the Emitter of TR34 which effectively turned it off in AM mode, so it now functions.
When I was experimenting with changing C109 to lower uF values, like from 10uF @16v to 4.7uF or 3.3uF, down to even 1uF, it didn't hold long enough and distorted the audio waveform on the Oscilloscope. So one day I was looking at the schematics for the PC122 (I have one, still), and noticed they had a resistor in parallel with the Sample and Hold cap in that radio, so I tried different values on Cobra 148, came up with 22k, which in parallel with R128 and R129 (total series resistance of 48k), causes the overall parallel resistance across C109 to be about 15k (15085). This causes C109 to still charge up (down?) on instantaneous peaks, but release a whole lot faster, or at an almost Syllabic Rate.