Micpre of Graham
in [Design]
|
From: cledus on 21 Mar 2007 11:29 Jim Thompson wrote: > On Tue, 20 Mar 2007 23:08:46 GMT, cledus <cledus(a)noemail.net> wrote: > >> Ban wrote: >>> The schematic >>> http://rapidshare.com/files/21272377/mic_amp_2.jpg >>> looks pretty simple, but it still needs work. I found a couple of gotchas: >>> 1. When you switch on the phantom power the Vbe of the transistors gets >>> reversed momentarily(+17V instead of -0.7V), degrading beta and Vos. This >>> will slowly destroy the input devices. This happens always in normal >>> operation with or without a mike. >>> 2. The power supply rejection is very poor(-20dB) especially at higher >>> frequencies. Here current sources might improve the situation. A lot of >>> additional filtering is also needed. >>> 3. When saturating the opamps will return to normal operation in a staggered >>> way, creating spikes in the O/P signal. >>> 4. The offset voltage varies with the gain, making it sensitive to >>> variations in gain setting. >>> I have attached a link to a commercial product, just to show that the art of >>> making a good preamp is not *that* simple. >>> http://rapidshare.com/files/21831341/mic_pre_02.png >>> >> Because I had the day off and was fascinated by Graham's "improved mic >> preamp", I threw it into my Spice simulator and tinkered around a bit. >> Here are some of the results >> >> My simulator does not have built-in models for the transistor and op-amp >> that he used. So I substituted the venerable 2N2907 and LM833 parts to >> see what happened. If you can believe the simulator, the noise >> performance is impressive at around 2.5 nV/rt-Hz referred to the input. > > I got ~960nV/rt-Hz I substituted 50 ohms for the 18k input resistors (R3 & R5). I forgot to point this out. This may, in part, account for the differences in our results. Not sure what makes up the rest. Perhaps the noise model for the 2N2907's? > >> However, the distortion leaves a bit to be desired. At a gain of ~30 >> and driving with +/-100mV pk-to-pk, the third harmonic is about -60 dBc >> at 100 Hz (~.1% THD). At 1kHz it gets better at ~-90 dBc for both the >> 2nd and third harmonics. If I try substituting the LT1028 model in my >> simulator for the op amp, the circuit goes unstable. The circuit may >> depend on a slower op amp to keep it stable. > > I suspect you made a mis-entry. You may be right. I don't trust the model I have for the LT1028. I entered a different high bandwidth op amp that seems a bit more trustworthy (MAX4106). It appears to be stable with this component. In fact the peak in freq response moves from 1 MHz using LM833 to greater than 10 MHz with the MAX4106. > >> For fun, I attempted to simulate a plain-jane LT1028 inverter based on >> the built-in model for my Spice simulator. I'm not sure that I can >> trust the model. I could get no where close to the noise performance >> claimed in the data sheets. And the noise was orders of magnitude worse >> than Graham's circuit. Like the data sheet recommends, I used 1.8k >> feedback and 60 ohms input resistors. The noise shows around 1.75 >> microVolts/rt-hz referred to the input as opposed to less than 1 nV! >> Maybe I am doing something wrong, or maybe the model is not trustworthy. >> But the distortion looks very impressive and is similar to the data >> sheet. At +/-100 mV and gain of around 30, harmonics were all >> suppressed well below 100 dBc for input freqs of 10 Hz, 100 Hz, 1 kHz >> and 20 kHz. Because of the noise discrepancy, I don't know how reliable >> these results are. But they do seem to follow the data sheet >> extrapolation at these input levels and no external load. >> >> Anybody know where I can get a reliable Spice model for the LT1028? >> >> -c >> > > From LTC ?:-) When I feel ambitious, I will dowload their freebie SPICE app and give it a try. I have used CircuitMaker2K for awhile now and need to get over the inertia to learn something new. > > ...Jim Thompson TNX, -c
From: Ban on 21 Mar 2007 11:26 Eeyore wrote: > Phalluson wrote: > >> "Eesyore" >>> Ban wrote: >>> >>>> 1. When you switch on the phantom power the Vbe of the transistors >>>> gets reversed momentarily(+17V instead of -0.7V) >>> >>> Can you explain how you think that happens ? >> >> ** Of course, you know the colossal fool cannot. >> >> If the +48 volt supply is somehow snapped on, then about + 15.7 volts >> momentarily appears on each base. >> >> Not a reverse Vbe situation at all, as the emitters are supplied >> from +17. > > Exactly. > > I was hoping that Ban would see his error himself. > Well, the thing is there is a divider formed by the 4k7 from 17V and the 6k8 from the opamp output. In the first moment the opamps have 0V, when the transient arrives, one opamp goes to the pos rail and forces the other to the negative. Now there are 4k7 || 5k(pot) with the 470u cap voltage, which can be as high as 4V, depending on the history. That makes 1.5V across the first junction and around 10V across the other one. -- ciao Ban Apricale, Italy
From: Jim Thompson on 21 Mar 2007 11:52 On Wed, 21 Mar 2007 10:29:59 -0500, cledus <cledus(a)noemail.net> wrote: >Jim Thompson wrote: [snip] >> >> I got ~960nV/rt-Hz > >I substituted 50 ohms for the 18k input resistors (R3 & R5). I forgot to >point this out. This may, in part, account for the differences in our >results. Not sure what makes up the rest. Perhaps the noise model for >the 2N2907's? The BJT model IS the noise model. [snip] ...Jim Thompson -- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | | http://www.analog-innovations.com | 1962 | America: Land of the Free, Because of the Brave
From: Eeyore on 21 Mar 2007 13:28 Ban wrote: > Eeyore wrote: > > Phalluson wrote: > > > >> "Eesyore" > >>> Ban wrote: > >>> > >>>> 1. When you switch on the phantom power the Vbe of the transistors > >>>> gets reversed momentarily(+17V instead of -0.7V) > >>> > >>> Can you explain how you think that happens ? > >> > >> ** Of course, you know the colossal fool cannot. > >> > >> If the +48 volt supply is somehow snapped on, then about + 15.7 volts > >> momentarily appears on each base. > >> > >> Not a reverse Vbe situation at all, as the emitters are supplied > >> from +17. > > > > Exactly. > > > > I was hoping that Ban would see his error himself. > > Well, the thing is there is a divider formed by the 4k7 from 17V and the 6k8 > from the opamp output. Yes, that's equivalent to 2k8 from 10.23V. > In the first moment the opamps have 0V, when the > transient arrives, one opamp goes to the pos rail and forces the other to > the negative. What makes you think that happens ? The 'transient' appears in common mode on *both* inputs. Of course there won't be a transient with a properly ramped 48V supply of course. > Now there are 4k7 || 5k(pot) with the 470u cap voltage, which > can be as high as 4V, depending on the history. That makes 1.5V across the > first junction and around 10V across the other one. You've missed something very basic. Read Phalluson's post and see if the number he quotes rings any bells for you. You *are* wong about this I have to tell you. Graham
From: Kevin Aylward on 21 Mar 2007 15:44
Ban wrote: > Phil Allison wrote: >> "Ban" >>> >>> I have attached a link to a commercial product, just to show that >>> the art of making a good preamp is not *that* simple. >>> http://rapidshare.com/files/21831341/mic_pre_02.png >>> >> >> >> ** Good pre-amp ? >> >> With a pair of 2.2 uF film caps ( = 1.1 uF ) in series with the mic >> input ? >> >> Not too good for the noise figure at or below 1kHz. >> > Why do you think a reactive element increases the noise? Input shot noise droped accross the capacitor. Say Ic=2ma, hfe=200, In=sqrt(2.Ic/hfe.q) = 2pa/rthz. At 1k, Xc = 160 ohms, giving 0.3nv/rthz. -- Kevin Aylward ka(a)anasoftEXTRACT.co.uk www.anasoft.co.uk SuperSpice |