Wouldn't that be the result of averaging a full bandwidth signal and the same signal through a 1st order LPF? I have to admit I didn't work fully through the math, but my intuition is that you end up with an average of the two signals.
Does the surface mounted version of the MXL990 have the same circuit as the through-hole version?
- Thread starter Resonite
- Start date
Help Support GroupDIY Audio Forum:
Wouldn't that be the result of averaging a full bandwidth signal and the same signal through a 1st order LPF? I have to admit I didn't work fully through the math, but my intuition is that you end up with an average of the two signals.
Fine, don't take my word for it - draw up the circuit in LTspice and see how not-3dB/oct that slope ends up as 
I myself have no real idea how one might achieve a shallower slope than a first-order (6dB/oct), with only passive components.
Noone said any signals get "averaged". The preamp resolves the DIFFERENCE between the "hot" input and the "cold" input, simple as that.
And on a side-note...
I myself have no real idea how one might achieve a shallower slope than a first-order (6dB/oct), with only passive components.
Noone said any signals get "averaged". The preamp resolves the DIFFERENCE between the "hot" input and the "cold" input, simple as that.
And on a side-note...
Update: I did end up soldering a 10nF cap parallel to R7, I also replaced C3 and C4 with 1uF film caps to shape the tone. I justed tested the modified microphone and it seems like a major improvement. The last thing I have to do is to change out the capsule.
The microphone did have an improved bass response and it does seem that I have achieved some sort of reduction in high frequencies.
The microphone did have an improved bass response and it does seem that I have achieved some sort of reduction in high frequencies.
I will probably do that later in Tina-TI; I installed that recently on Thor's recommendation to compare to LTSpice.
But for the moment let's take as a predicate that the phase splitter with capacitor on the drain works as described, and the cold side connection has a LPF behavior while the hot side does not.
Just as background to make sure everyone reading can follow along, the standard no filtered output to a diff amp; for easy math let's say the signal is normalized to an amplitude of 1 on each side at every frequency of interest, i.e. hot side signal amplitude is 1, cold side signal amplitude is -1 (same absolute value, but opposite polarity).
A difference amplifier takes the difference, i.e. subtraction, of the cold signal from the hot, so the output amplitude is:
1 - (-1) = 2
For completeness I show the case of LPF in cold side and hot side. For ease of math assume the filter response is -6dB at 2kHz, -12dB at 4kHz.
At low frequencies the amplitude will still be 1, so the output of the difference amplifier will be 1 - (-1) =2.
at 2kHz the output amplitude is 0.5 and -0.5, so the output of the difference amplifier will be 0.5 - (-0.5) =1
at 4kHz the output amplitude is 0.25 and -0.25, so the output of the difference amplifier will be 0.25 - (-0.25) = 0.5.
In dB that makes the 2kHz response 20*log(1/2) = -6dB
and at 4kHz 20*log(0.5/2) = -12dB
Now let us add a low pass filter in only the cold side, again for ease of math assume essentially flat at low frequencies, -6dB at 2kHz, and -12dB at 4kHz (6dB/octave 1st order filter). If the beginning amplitude is 1, then the -6dB amplitude will be 0.5 and the -12dB amplitude will be 0.25. The filter is in the return/cold side, so our convention is negative amplitude.
At low frequencies it will be the same as the unfiltered case, 1 - (-1) =2
at 2kHz it will be 1 - (-0.5) = 1.5
at 4kHz it will be 1 - (-0.25) = 1.25
So in dB at 2kHz, where the output of the cold side signal is -6dB, the output of the difference amplifier will be 20*log(1.5/2) = -2.5dB
At 4kHz where the output of the cold side signal is -12dB, the output of the difference amplifier is 20*log*1.25/2) = -4dB
The difference between linear and logarithmic shows up there, the output of the difference amplifier is not actually down 3dB at the 6dB point of the LPF, and it isn't quite 6dB down at the 12dB point of the LPF, but the slope is obviously reduced compared to the case
Hopefully that explanation is easy enough to follow.
The complete circuit diagram, of which OP took the impedance converter stage, can be found here: Nieuwe pagina 1
Btw, for those who don't know what DUS means: Diode Universal Silicon. AFAIK, this terminology was introduced by Elektor in the seventies, together with TUP and TUN. A DUS is typically a 1N4148.
On this page, you will also find some proposals for modifications. Note: by linking to this website, it does not mean that I endorse the statements made by the author, especially with respect to ceramic capacitors. As a.o. Krohn already pointed out, NP0/C0G caps are perfectly suited for audio applications. There are more statements that I disagree with, but I wanted to leave it at that for now.
Cheers, Jan
I had the MXL770/990 circuit readily available in LTspice, albeit with some minor changes because I used a different JFET for the simulation. So I can easily plot the Bode diagram of the circuit and visualize what happens if the value of C14 is enlarged, which btw has the same effect as putting a capacitor parallel to the Drain resistor R7. You can see the results in the attached screenshots. The first one shows the output voltage with both caps 1nF. The second one with C14 increased to 10nF. The bode plots show the differential output voltage. With asymmetrical caps, the gain of the negative output just drops off at a lower frequency. Above ~200kHz, the curve drops with -6dB/oct, as could be expected.
A disadvantage of this asymmetrical EQ is that the frequency-dependent output impedance of the T1 and T2 becomes different, hence making the CMRR worse with increasing frequency. For this reason, I would be more in favor of the EQ circuit as described on the Audioimprov site. However, the disadvantage of this circuit is that the noise level is increased by R7 and R8. Btw, by adding two or three more RC circuits parallel to R9/C14, you can make any LPF filter curve between 0 and -6dB/octave. I have played with that idea to change the FR of a flattish LDC into a dark sounding Ribbon mic with a gradual drop-off of -5dB between 400 Hz and 20 kHz.
Jan
A disadvantage of this asymmetrical EQ is that the frequency-dependent output impedance of the T1 and T2 becomes different, hence making the CMRR worse with increasing frequency. For this reason, I would be more in favor of the EQ circuit as described on the Audioimprov site. However, the disadvantage of this circuit is that the noise level is increased by R7 and R8. Btw, by adding two or three more RC circuits parallel to R9/C14, you can make any LPF filter curve between 0 and -6dB/octave. I have played with that idea to change the FR of a flattish LDC into a dark sounding Ribbon mic with a gradual drop-off of -5dB between 400 Hz and 20 kHz.
Jan
Switchable or not?
How about running some sims with a capacitor in parallel with, in your schematic, R7 & R7a? I've got a good hunch that shouldn't mess with the output impedance of the PNP pair in any meaningful way, while offering a low-pass effect (or a high-shelf attenuation, if you add a resistor in series with it).
Drop C3 & C4 to 22nF or 15nF - according to Digikey's RC calculator, that should net you a 48Hz or 70Hz cutoff frequency. There's a chance the PNP bases might skew that a bit.
Simulating the circuit should be reliable enough for so simple a parameter though.
Simulating the circuit should be reliable enough for so simple a parameter though.
In the MXL990, R7/R7a and R6/R6a are just single 2k2 resistors. My LTspice schematic is a universal schematic for the MXL770 and MXL990 and the 770 has the split source and drain resistors. In the 770, R7/R7a and R6/R6a resistors are used as a -10dB pad. It is useless as a pad btw, as it does not prevent overloading the JFET at high SPLs. And it has a switchable 150 Hz HPF (S2a, S2b, C16, C17 in attached schematic). If you want to change the cut-off frequency to ~50-70, C16 and C17 should be increased to 39...56nF, assuming you keep C3 and C4 on 220nF. If you just want to replace C3 and C4 and keep it non-switchable, use 22...33nF caps. The lower the capacitance, the higher the cut-off frequency. If you want a switchable HPF, maybe buy a used 770? They can be had for 50 to 70 Euros, sometimes even less.
Putting a cap across R7 has the very same effect as increasing the value of C14 because for high frequency AC signals and with C14<<C4, the Vplus power and the collector of T2 are connected to ground via the decoupling caps C5 and C6. So for high frequency signals, they are effectively shorted. To demonstrate that adding a cap across R7+R7a is similar to increasing C14, I added C8 to the schematic (9nF), which is now effectively parallel to C14 (1nF). So together they sum up to 10nF. The results can be shown in the screenshot below. The difference is less than an inaudible 0.1dB. I couldn't get the cursors on exactly the same frequency as in the previous simulation run with C14=10nF, but I guess you'll get the point.
Cheers, Jan
Putting a cap across R7 has the very same effect as increasing the value of C14 because for high frequency AC signals and with C14<<C4, the Vplus power and the collector of T2 are connected to ground via the decoupling caps C5 and C6. So for high frequency signals, they are effectively shorted. To demonstrate that adding a cap across R7+R7a is similar to increasing C14, I added C8 to the schematic (9nF), which is now effectively parallel to C14 (1nF). So together they sum up to 10nF. The results can be shown in the screenshot below. The difference is less than an inaudible 0.1dB. I couldn't get the cursors on exactly the same frequency as in the previous simulation run with C14=10nF, but I guess you'll get the point.
Cheers, Jan
Attachments
For a more accurate cut-off frequency calculation, you should bring the load of your mixer into the equation. At first sight, that may not seem very obvious, but the low input impedance of the mixer multiplied by the gain of the output transistors appears as impedance parallel to R3 and R4. So you cannot assume a 150k resistor in the RC cut-off frequency calculation. Again, some simulations to show the effect of the mixer load. First, with a 2k input impedance (=R8) and assuming C3 and C4 are 15nF. The cut-off frequency equals ~155Hz. This is quite close to the 150Hz HPF of the MXL770, which has 15nF in series with 220nF. So I guess MXL took the mixer input impedance of a typical mixer into account.
Now in the second picture, I increased R8 to 2 Meg. Lo and behold: the cut-off frequency went almost one octave down and is now 79Hz! Next to a better CMRR, this is another reason to why I prefer a CFP output over a single transistor as in the classic Schoeps circuit.
Jan
Attachments
R8 represents the input impedance of the preamp device.
AFAIK, 3k is a normal input impedance for a mic preamp. No need to change R1 and R2. And whatever value you take for R1 and R2, they should ALWAYS be of the same value for best CMRR (read: noise and hum reduction) and mijimal DC offset between pins 2 and 3 (required for transformer inputs).
I would say, just experiment with different capacitor values in the ranges mentioned in previous posts and see what works for you, rather than frantically trying to obtain a certain HPF frequency @ 3k mixer input impedance. The actual bass response of your mic also depends on proximity effect and even the sound source (point source or other).
Jan
I would say, just experiment with different capacitor values in the ranges mentioned in previous posts and see what works for you, rather than frantically trying to obtain a certain HPF frequency @ 3k mixer input impedance. The actual bass response of your mic also depends on proximity effect and even the sound source (point source or other).
Jan
Is there a reasonably straightforward mod for the 770 to make its pad switch function properly? Shortly before seeing this thread I bought one mainly for that pad switch, because I thought a pad was a pad. (I already had 990s lying around, but wanted something more flexible to put a b-stock flat 47 in. Actually it’s a Monoprice 600800, but that’s supposed to be the same mic w/ different silkscreening on the outside… I actually tested and bought it in a pawn shop before noticing the gold printing on the case didn’t say MXL.)
Last edited:
What's wrong with it now, as it is?
According to jp8 (in the 3rd & 4th sentences of the paragraph I quoted) it doesn’t prevent overloading the FET at high SPLs.
You could either shunt the capsule with a capacitor, or lower the capsule polarisation voltage. Took a quick glance on the PCBA and schematic and I think the easiest way is to shunt the capsule with a capacitor. Value would depend on the amount of attenuation required and capsule capacity. I would suggest to use an NP0/C0G type ceramic capacitor.
R6 and R7 (680R) should be changed to 2k2. Remove R6a an R7a (1k5) and replace by a jumper wire. Cut the upper three pins of the pad switch. Connect the left pin of the switch to ground. Connect the capacitor between the middle pin of the switch and the left insulated ptfe turret pin, which connects to the capsule. If my description is not clear, I can draw up the schematic and make some pictures and drawings if you like. Sometime next week, when I have more time.
Jan
R6 and R7 (680R) should be changed to 2k2. Remove R6a an R7a (1k5) and replace by a jumper wire. Cut the upper three pins of the pad switch. Connect the left pin of the switch to ground. Connect the capacitor between the middle pin of the switch and the left insulated ptfe turret pin, which connects to the capsule. If my description is not clear, I can draw up the schematic and make some pictures and drawings if you like. Sometime next week, when I have more time.
Jan
