Due to a relatively low signal to noise (S/N) ratio of TRV900E (see measurements), especially in manual gain mode, I have studied the signal path from mike input to A/D conversion in TRV900.
First, the mike signal goes through "Stereo zoom microphone amplifier for camcorders" - BA7780 (both for manual and automatic gain control). According to data sheet it amplifies the signal by factor 25 (28dB) for external mike and by factor 8 to 17 (18.5dB to 24.5dB) for internal mike [the drawback is that the frequency response of the signal becomes quite limited]. Then the signal follows to the aforementioned "electronic volume" chip M5222 (this chip attenuates signal according to selected mike gain in MIC LEVEL control). The signal is then amplified by factor 48 (33.6dB) by the low-noise operational amplifier 4572. After that both signals (auto and manual gain controlled) go to "Audio in-out selector" AN2902FHQ which is some kind of audio switch with built-in AGC control (just for mike signal in auto-gain mode). The signal is then transferred into digital form by Burr-Brown (Texas instruments :-) A/D converter PCM3006T.
I have made some experiments and found out that the main reason for relatively large noise signal is probably "electronic volume" chip M5222. Therefore, I have connected the mike input signal from the input jack directly (through 1k resistors just for the safety reason) to the input of the operational amplifier 4572, as shown in these pictures (whole picture, resistors on mike in jack, soldering on 4572). Of course, before doing that, contacts No. 3 and 5 of the chip 4572 should be carefully!!! unsoldered (just in case it was done by my skilled co-worker) and slightly lifted up (there should be no contact with the PCB) by some sharper and narrower tool. The used cables should be very thin and soft.
By the way: how to open the camcorder? The guidelines are given on John's site!
And
the results? Bad news first: Since M5222 is not used, the gain in manual audio
mode is now fixed :-(to
about 50 = 34dB - so similar to choosing 5 for mic gain in MIC LEVEL control).
Good news: The noise becomes quite low. When no signal is attached to mike input
(11k resistors are connected between signal and GND), the average RMS becomes
-76.9dB for left and -78.5dB for right channel (noise levels were even lower -
about -82dB
The frequency characteristic (span) of the modified mic input is also much better than before (comparable to line in characteristic).
Frequency
[kHz] |
Gain |
Gain
[dB] |
0.01 |
0.44 |
-7.1 |
0.02 |
0.77 |
-2.3 |
0.03 |
0.87 |
-1.2 |
0.05 |
0.95 |
-0.5 |
0.1 |
1.0 |
0 |
1.0 |
1.0 |
0 |
10 |
1.0 |
0 |
20 |
1.0 |
0 |
I
have tested the modified TRV900E with external microphone AKG
C391. First I tested my voice (mike
has been placed in my hands 20cm away from my head), then I played two songs on
my Hi-Fi, which
"frequency characteristic" is not perfect (CD player is about 15 years
old SONY, amplifier is NAD 312 and loudspeakers are BW301). Mike has been placed about 45cm from a
loudspeaker. First was a song
"Playtime" from "Pacific Blue
II (NorthSound)", the second was "Those
Good Old Dreams" from Carpenters. The original songs
(ripped from CDs) are here: Playtime, Those
Good Old Dreams (mono versions (only left channel) in order to make fair
comparison with microphone). Hear more tests...
I must tell that I am *very* satisfied with obtained results.
Caution! Note that the modification given above is only for educational purposes. I won't be responsible for any demage. The modification is on your own risk! Note that the operational amplifier (4572) might be demaged by static electricity (not very probable, since it is bipolar op-amp) or by larger input signals; soldering might demage operational amplifier, the contacts might be broken, etc.
Some schemes... Block diagram, mike input scheme, audio out scheme.
I have built some kind of elastic suspension ("shockmount") for TRV900 and built my own mike preamp and attenuator (for capsule AKG CK91) in order to get "decent" sound even when the mike is "mounted" on TRV900. Read here.