reducing hiss in mic preamp

[dai h.]

I'm trying to see if I can reduce the hiss on one of the XLR input channels on my old Tascam 488mkII. Since it seemed to help in another mixer section on a separate cassette Portastudio, I tried changing the two opamps in the channel to NJM5532Ss (from the original NJM4565LD), as well as changing the alu electrolytics to Rubycon Black Gates (which I scrounged up out of my junk pile). Compared to the next channel, it's slightly better, slightly more hi-fi but not substantially different and I'd like to make it even quieter. The input area seems to look like this:

http://sound.westhost.com/project66.htm

(and also another circuit for a balanced input in a book on hand.)

The westhost article states, "metal film resistors in the collectors and emitters of the input pairs [should be used] for least noise", and in fact right near the XLR input there are some metal films used. On what appears to be the next section in the circuit after that, there are some appearing to be pairs near two transistors (which I assume is part of the balanced circuit segment) but these are small (1/6W?) carbon film. I'm going to try changing to metal films, and the question (finally, lol...) is how much does their relative tolerance matter? I can obtain up to 0.1% tolerance metal films but these are significantly more expensive (around 11 times more than a regular metal film--around a dollar ea.). I'm looking through my parts and I'll use what I have on hand but will need to purchase what I don't have. Thanks.

oh, please see pic for a visual.

[Rob Mercure]

Hey Dai,

While this really isn't my area of best expertise as best I know the resistor tolerance has little to do with its noise characteristics other than more expensive parts generally performing better overall. Since you don't seem to have the circuit diagram you might just be shooting in the dark with the resistors - if replacing them doesn't help I'd look to the transistors you've identified.

But you might want to buzz Enzo as his SS expertise seems to combine the best of theory and real life practice.

Rob

[Steve Conner]

Most mic pres use two discrete transistors in the front end, because op-amps are too noisy. These transistors largely determine the noise performance. If Tascam cut corners to get the price down, they may just have used ordinary cooking-grade transistors in this very demanding role.

The first obvious upgrade would be to change them for special low-noise types, or if you can't get any low-noise ones, you could try 5 regular transistors in parallel. Rod Elliot recommended 2N4403s: these are fairly chunky power transistors with a large junction, which helps the noise performance the same way as paralleling small devices does.

You could also try increasing the bias current on the transistors: Tascam may have skimped on that to extend the life when it was running off batteries.

The resistor tolerances don't really matter too much: you would use metal films purely for their lower excess noise. The gain control pot is part of the circuit and can contribute noise too.

If the noise is even worse than the noise of a 4-track cassette, it must be REALLY bad. :-o

[dai h.]

thanks for the responses guys, I guess my assumption that the resistors would have an effect was incorrect. This is actually a 4-track (sort of, actually an 8 track cassette portastudio), and the hiss is not really that bad. I suppose I'm being a bit greedy/nitpicky but if possible I want something not really noticable even when cranking the HIGH EQ knob. The next (and only other) XLR input ch. (which is 99.9% unmodded except for two 223 films for power supply bypass in place of the stock 223 ceramics--some goofy thing I tried since I stumbled into a bag of 223 1% polypropylene surplus for cheap) in comparison doesn't sound much different really. I tried to compare the sound as best as I could speaking into the chs. with a cheap mic and playing with the EQs, and what I noticed is that the (mostly) unmodded ch.--when you crank the HIGH EQ all the way up--the character of the hiss is harsher sounding than the modded ch. I also had to make an XLR cable conincidentally, and instead of wiring it normally I attached a separate wire for the shield for the end going to the channel input and hooked it to the bottom of the Tascam where there is a small bottom plate which seemed to be more of a chassis ground (has the power Xfrmr. connected to it plus some other grounds from the power supply and metal frame of the cassette player connected to it) which I have a ground wire going to the ground on an AC outlet since it seemed more "correct" in terms of grounding. Seemed to make a subtle difference but nothing really noticable changed.

You could also try increasing the bias current on the transistors: Tascam may have skimped on that to extend the life when it was running off batteries.

This portastudio is AC powered with no battery operation option, so perhaps that doesn't apply? (Not sure since I'm not familiar enough with the circuit though.)

The gain control pot is part of the circuit and can contribute noise too.

I tried changing the 1000uF 6.3V alu electrolytic (which appears to be connected to the gain pot and 100ohm) to a Black Gate(had one unused) since in the article I was looking at with the mic pre cct. stated "tantalum was preferable (but probably not obtainable or easy to obtain)" but as stated I'm not really getting anything dramatic in noise lvl. difference.

Also, one thing I noticed from talking/comparing the channels was that the modded one has slightly less low freq. response. I think this is due to me changing two 10uFs right at the input to 3.9uF films (had 10uF films but it seemed to be too much and maybe causing distortion on peaks?). I guess they need to be a bit higher. Not very noticable though.

Lastly I tried playing some Queen (music) as a source for the two XLR input chs., and the EQs (HIGH, LOW, and MID with variable freq., lvl. but no "Q") don't sound that bad but if they could be subjectively nicer sounding I'd prefer that.

[Amp Kat]

I don't think it has anything to do with ground or power supply problems. It's a gain issue and noise and gain are directly proportional. It may be possible your overdriving the input or a huge mismatch inpedance wise. Try a direct box and see if it quiets down a hair. Dynamic or Condensor ? A input transformer can be a huge difference in quality of noise as it amplifies and bucks canceling hum. Check out the jensen website for more details as almost everything about mic-pres is there http://jensen-transformers.com

[Enzo]

This cassette recorder is pretty basic, and I don't think other than reasonable levels that noise was a primary consideration in the design. You can [probaly get some incremental improvements piece by piece. Maybe a couple fancy B-B op amps.

Try grounding off the siignal path with a large cap - 0.1 or something - stage by stage to see what contribution each provides in the hiss department.

Can you configure an external mic pre and run in as a line level signal?

[dai h.]

thanks for the additional responses Amp Kat and Enzo,

I guess I should've communicated better, but I think what I'm trying to do is more increase performance than solve a "bad hiss problem", since it's not really so bad as what one would characterize as "horrible". I'd like to try some of the Burr-Brown OPA2604, but they might not fit since the ones in there now are SIP (did find some NJM5532LD googling which is the lower noise ver. of the NJM5532D in an 8-pin SIP package and only a buck ea. too to might try some of those). The external mic pre amp idea--maybe I can use this backwards, i.e. output the 488mkII mic pre out of the insert jack to something else and maybe be able to judge how much the circuitry after--the fader and summing amp?, headphone monitor--is adding hiss. The transistors toward the input are 2SC1844 which are apparently audio freq. low noise transistors, FWIW. Later tonight I'm going to just go ahead and change some more resistors to metal films I've got on hand. Oh yeah, the mic was a cheap dynamic and I'm trying to spend as little as possible since I think it'd make more sense to buy an audio interface with good mic preamps like an E-MU, etc. so I probably won't go as far as buy an input transformer('though I'm sure Jensens are nice). My labor doesn't matter, since it is free though, lol.

[Amp Kat]

If you go to low noise you'll start getting a sound that reflects glass breaking. The BB 4134 (I think) is a single op and probably the best headphone amp but it's obsolete for some time now but that would help if you can find one. With the dynamic I'd still try the direct box. The problem with reducing the noise is once you do then it won't be loud enough to get the meter reflection that you need for a hot signal to the recorder. Certainly not as hot as +4dbu but maybe 0 or line level as Enzo pointed out a seperate preamp coming in maybe better than what you have. They also have some pretty cheap boxes out there for less than $100 that would work. Presonus,Art, Behringer and even Rat shack but I'd test them first as you never know what your going to get or if they work right.

[dai h.]

well, dug through my parts for appropriate value metal films and changed out some of the Rs in the one XLR input ch. plus the amp around the fader, the summing amp and around the monitor out (hdphone out). A lot of 10k, 100ks later in the signal chain and 5.6k, 15k, 10k, 100k in the one channel. Had to double up 200ks to get 100k, and stand up some (since 1/4W and not the 1/6W used as stock Rs), put one on the back of the PCB, but there was enough space for the 1/4W and doubled 200k 1/6Ws to fit. Also two cheap looking mylars (2n2 and 3n3) around the EQ to more upmarket-looking ERO polypropylene I had on hand(had to change one of the 100uF 16V Black Gates back to an original regular alu. electro. due to space issues to fit the 3n3).

Played back a gtr. signal from the output of a spk. emulator recorded earlier, and subjectively it sounds worse , however, this is actually more accurate to what I heard through the 'phones when I recorded it. (Something I noticed before was that the recorded gtr. would actually sound a little noticeably better upon playback. I'm guessing this has something to do with the limited freq. response of the tape recorder--15.6kHz according to a review of the unit in an old iss. of Recording mag.--plus some "grunge" from the signal chain.) The hiss seems slightly less.

Did a bit of googling on metal film Rs in mic pres and tolerance, and it seems 1% is often used, so appears 1% tol. is adequate except for the Rs feeding the phantom power. For some reason (which I'm not aware of yet) a really tight tolerance seems to be desirable here, and the ones right by the XLR input may be such ( around 13k paralleled for around 6.8k). Therefore, it looks like there are no metal films stock except for those. And, re: the 2SC1844s, according to the NEC site(maker of the Trs), these are apparently an obsolete type (ended prod. in '97) and are "an ultra-low noise transistor which in relative to conventional noise specs (en=1.2nV/Hz) is 1/2 in comparison (en=0.6nV/Hz) and since (furthermore) has a high hfe (400 typical), it is possible to build an amp having superior S/N specs such as a hi-fi MC cartridge EQ amp, or utilize 2SC1844s for the playback amp input stage in a tape deck." So a "low-noise Tr. + no metal films for the collector and emitter Rs" together may mean I can eke out a bit better performance by changing those (didn't have the right values so will have to go buy them).

Also, didn't have any 47k Rs as well so (I'm guessing) since they are a relatively high value with probably many used for summing signals, if I change all those I'll get less additional hiss.

[f.leghorn]

Low noise metal films

A very good low noise metal film at 40 cents each is the PRP (Precision Resistive Components) type PR9372, 1/2W 50ppm. Designed specifically for low noise audio applications. www.sonicraft.com .

[dai h.]

thanks. Well a small update. I bought the MFs and changed out most of the ones around the input. A couple of 4.7ks which look like loads on the base(s), a couple of 2.2ks and 4.7ks for the emitter and collectors (didn't work out which was which yet but I followed them to V+ and V- for the opamps), a 20 ohm in series with the gain trimmer, a bunch of 47ks later in the signal chain plus some odds and ends--couple of 82ks in the monitor section, 18k and 47ks on one master fader section, etc. Compared to the other XLR input ch. which is mostly unmodded, I don't really detect much of a difference in the sound or the hiss. It's still reasonably quiet but I don't really feel I've gotten closer to my ideal of "near silence even at extreme gain/EQ settings" though. The only ones in the mic pre input section I didn't change were two 10ohms in series with the signal (pins 2,3) which appear to make up an (RF?) filter along with two 2200pF ceramics to ground.

I also remembered having some cheap Telecom transformers with a primary of 600 ohms/ sec. 1200 ohms C.T. ("STFT-9004" below). Bought a couple a couple of months ago since they were real cheap surplus stuff (around US2.50??). Would they work? (I'm sure they wouldn't compare to a Jensen, etc. but just wonder if I could just try one for kicks).

http://www.njrc.jp/product/pdf/densi/STFT.pdf