Incorrect. There is a 1.5k emitter resistor and ~ 22k load.
The VAS thus only offers 15V/V gain.
Lets look at both the scenarios you propose, that is, if the cap is either 330p or 330u. If it is 330u, the 1.5k resistor is bypassed by a capacitor, in this case the transconductance of the VAS is gm=0.5mA/26mV = 19.23 mA/V and the Gain is Av = -gmRL = 19.23mA/V * 22K = 423 V/V, but this is not entirely true, I estimated an Early voltage of 50V from the 2SA970 datasheet, which would mean that the gain is more like 19.23mA/V*22k||100k= 347 V/V, if the cap is 330pF, the gain is a bit lower than 22k/1.5k = 15 V/V, lets say 13.
Actually like 225pF. So slew rate with miller loop closed looks like 4.5V/uS and the low VAS Gain means modest input skewing may not saturate the the VAS.
Miller effect on the capacitor is given by C(1-A), where A is the gain, but, since the gain is inverting and its value is 347 as I mentioned, then the cap is 15p(1+347) = 5.22nF, if the emitter cap is 330p, then the Miller cap is 15p(1+13) = 210 pF. It is important to note that this is a
small signal effect. SR, which refers to a large signal effect depends on that cap absolute value and the maximum current it can be provided to it, which in this case is the tail of the diff pair, which is 1 mA. The current through a capacitor can be expressed as i = C dv/dt, hence dv/dt = i/c = 1mA/15pf = 66 V/usec, the actual figure will be lower than this, though. If, as you mention, Tascam had low TIM, then this unusually high value of SR would make sense.
These are likely 2SK170 or 2SK369 or equivalent duals. So the Transconductance of the differtial Amp is probably around 15mA/V.
The 2SK170 has an Idss between 2.6 and 20 mA, let's say 10mA since this is one of the values of the plots reported in the data sheet, it also has a pinch-off voltage of around -0.5 V, its intrinsic transconductance is gm0 = 2IDSS/|Vp| = 40mA/V. The schematic shows a VGS = -0.3V, hence the transconductance of each transistor is gm0(1-VGS/Vp) = 16 mA/V, so yes, your 15 mA/V is a good guess.
Assuming 2SA970 for the VAS (another standard "Japan audio part"), the load for the input stage is likely around 400 * 1.5k or 600k, leading incidentally to a similar 80dB open loop gain estimation with around 18kHz -3dB Point.
Regarding the value you estimated for the load resistor: 1.5k*400 (I am assuming you are using beta=400). If the 1.5k resistor is bypassed by the 330u cap, the load is more likely 400*(26mV/0.5mA)=20.8k AKA rpi in the standard pi model, if it is 330p then it is 600k as you say. Then the gain of the first stage is 16mA/V*20.8k = 333 V/V, this is the DC gain of course, however, as I will show next this value is actually 16mA/V*20.8k||400k = 316 V/V, the 400K resistor (which I obtained from simulation) being the 'Early' resistor of the 2SK170, and that is assuming the BJTs have a beta = 400, considering a beta = 200, which is the lowest specified by the datasheet then the gain is 162. If however, the cap is 330p, the resistance is indeed 600K, however, being paralleled by the 400k resistor gives a dc gain of gm*600k||400k = 3840. All of these unusually high values of gain for the diff input pair makes sense, since the transistors are not degenerated by any resistor.
The total DC open-loop gain with 330uF would then be 316*347= 109.6 kV/V = 101 dB or 162*347 = 56.2 kV/V = 95 dB. If the cap is 330p as you say, then the gain is roughly 3840*13 = 49.9 kV/V = 94dB. Of course, this is considering that they indeed used the parts we are assuming they used.
I plotted the 2SK170 IV curves on a simulator, and estimated a lambda = 0.005/V (output resistance parameter). With the 0.5 mA of drain current, this would yield roughly an Early resistance of 400K.
Regarding the -3dB Open loop Bandwidth. We see that in the case of the 330u cap, the OLBW is essentially 1/(2pi*400k||20.8k*5.22n) = 1.5 KHz, and with 330p it is 1/(2pi*400k||600k*210p) = 3.16 kHz. Now, lets look at the unity gain frequency, in the case of the 330u cap, assuming a gain of 101 dB this happens at 168 MHz, with the 330p cap and a gain of 94dB, it happens at 158 MHz. There is, however, another 15p cap in parallel with the 12k resistor, which would limit the frequency response to 884 kHz in closed loop.
Bottom line: in either case (whether the cap is 330u or 330p) the results are similar, I don't like that the input diff pair stage has no degeneration, if indeed the cap is 330u, it would mean that there is practically zero local feedback being used. On the other hand, if the cap is 330p, there is local negative feedback on the VAS due to that 1.5k resistor. Secondly, the lack of a Darlington or cascoded VAS is also not something I am very fond of. If I were to modify this amp, I would start by adding an emitter follower between the diff amp and the PNP common emitter.
