SSM 2220 micro amplifier and discrete semiconductor alternatives

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

miquel

Active member
Joined
Nov 6, 2009
Messages
39
Location
Barcelona
Hola

Excuse me my wrong English.

I am interested in experimenting with the microphone preamplifier schematic that appears in SSM 2220 datasheet, specifically the three SSM paralleled (see datasheet) but with the AOP habitual realimentation configuration (see image attach).

datasheet ssm2220 example:
http://mihalcz.ingyenweb.hu/riaa/MCasyElek.pdf

In Elektor 12/98 appears a version for MC cartridge preamplifier, that is my inspiration, (see link). I think it’s easy to convert a microphone preamplifier:

http://mihalcz.ingyenweb.hu/riaa/MCasyElek.pdf

As alternative I would change the OP27 with OPA 627 or another best known audio OPA.

In Pordigy search, don’t appears information for this circuit (or at least, I do not find).

I have some questions and doubts and do the following comments:

1.- Has someone experience with this circuit?

2.- The Elektor circuit give great importance to the R9 (realimentation resistance) thermal noise effect, (in that circuit shows only 5,62 Ohms…!).

5,62 Ohms = 4,51 * 10-8 Vrms thermal noise = -144,7dBu noise

3.- But with this little resistance appears the problem to obtain a correct AOP realimentation loop (at least 600 Ohms in total), see R11 trick in Elektor project.

4.- But in  SSM2220 datasheet appears this resistance with large value (100 Ohms).

100 Ohms = 1,91 * 10-7 Vrms thermal noise = -132 dBu noise.

Obviously there isn’t a low value resistor loop AOP realimentation problem.

5.- In really, I think if you have a standard 200 Ohms microphone calculation noise (= -129dBu noise), is there no importance if you install a 5,6 Ohms or 100 Ohms R9 resistance, and then, datasheets circuit are OK.

6.- MAT03 isn’t in the usual market in Spain (Farnell, RS, etc.) (neither the venerable and ubiquitous LM394 not applicable because is NPN)

7.- Farnell (the best shop in Spain) has the SSM2220, but at the moment I think with a little stock, and expensive (It is like to eliminate it…?). But there are another PNP modern matching transistors SMD as:

BCM 856
BCM 857
DMMT 3906

And… Very very cheap (less than 50 cents…)

8.- Has someone experience with this semiconductors to substitute SSM 2220?.

9.- Why in that semiconductors datasheets appears only the (strange for my) NF (Noise Figure) in dB, and not the habitual Noise Voltage Density (NVD) in nV/sqroot Hz.

10.- How can I convert the NF in NVD (if that is possible).

Can sombebody help my?...

Very thanks from Spain

Very thanks

Miquel
 

Attachments

  • SSM2220 2.jpg
    SSM2220 2.jpg
    77.5 KB · Views: 54
I will try to answer best I can, your links didn't open and your questions are a little unclear.

Using multiple bipolar transistors in parallel reduces the noise voltage but increases the noise current so you need to do the math before using a preamp optimized for moving coil phono carts for microphones. MC carts are probably 1/10th the source impedance of microphones, so a good tradeoff to accept increased noise current for lower noise voltage.

------

You should be able to substitute any unity gain opamps in that circuit, and perhaps others with attention to their compensation needs.

------

I have not experimented with that exact circuit, but many similar to that over the years.

--------

Since I can't see the Elector reference I will guess about your questions.

Yes for low noise the feedback network also need to use low value resistors. For a mic preamp you can have difficulty with drive current, There are some modern opamps that will drive relatively low value loads that might drop in, or you could add discrete transistor buffers to a general purpose opamp.

-----

Yes, mostly, noise of feedback network is added to source impedance noise... 205 is less than 300, but in the grand scheme not huge depending on your other noise sources. It is not unusual to see 10-20 ohm max gain resistors.

------
You can flip the circuit around to use NPNs, or even use separate transistors, but with higher DC offsets.


I prefer constant NF curves for optimizing noise in design.  NF is with respect to a resistor value, so you can calculate the nV/rt HZ for the resistor value then add the NF,  i.e. a 1 dB NF @ 100 ohms, is 1 dB more noise than 100 ohm resistor's nV/rt Hz.

Have fun...

JR
 
That circuit is not directly useful for BALANCED microphones because it is unbalanced input.

You could put a transformer in front to convert balanced to unbalanced.

If you use a transformer, now you can also convert impedance from very low 150 ohms to some higher value which may be more suitable for low-noise amplification.

> Why in that semiconductors datasheets appears only the ...NF (Noise Figure)

If you can chose ANY source impedance, you pick the impedance which gives the best NF.

> habitual Noise Voltage Density

In transistors, total noise is BOTH voltage noise and Current noise.

Noise Figure combines both noise values, but only at the specified source impedance.

> DMMT 3906 Very very cheap

This is a dual 2N3906, one of the most popular small transistors on the planet.

2N3906 data

Page 4 shows noise curves. Noise may rise below 1KHz, but is fairly flat 1KHz-20KHz, which is what matters in audio.

The top-left chart shows best NF is for 2K source with 100uA of current in the transistor.

The top-right chart shows that for 100uA, the NF at 1KHz is low from 1K to 4K source impedance.

Therefore with a transformer, we would select a 150:2K ratio and 100uA current.

But good transformers are expensive. For transformerless operation from 150 ohm microphones, the top-right chart shows that the best NF is 3.5dB (and 1mA current).

3.5dB NF is good but not very-good.

The 2N3906 is a "small" transistor. When (no transformer) source impedance is as low as 150 ohms, we need a bigger device.

The SSM2220 plan you posted shows 4mA in Q2 or 2mA in each half of Q1. This is typical for best noise from 150 ohm sources.

SSM 2220, MAT02 MAT03 LM394 are the "good" devices for transformerless microphone amps. These have larger junctions with high gain, and are tightly specified for noise and match. They are expensive, demand is small, they can be hard to get.

Many inexpensive but good microphone amps use larger "switch" transistors. They are not specified for low noise, they are not matched, but experience says that in the right conditions, 99 out of 100 will be very low noise.

To handle Balanced inputs correctly you usually want more than one op-amp.

THAT Corp, TI, and others put low-noise transistors and all the op-amps and connections on a single chip. PGA2500, THAT 1510/1512, and others. You will probably have to get these from Farnell or other large international distributors.

> DMMT 3906 Very very cheap

Just for thinking: if one '3906 at 100uA has good NF for 2K source, then thirteen '3906 in parallel would be good for 150 ohm source. Total current would be 1.3mA, similar to best 150 ohm conditions for the devices rated for low 150 noise. The cost could be near 5 dollars, very affordable; but thirteen very-small 6-leg SMT devices is too many for me to solder (your eyes and hands may be better).

Finally: low-low-low-noise is NOT the biggest feature of a mike amp. I have made many recordings, small classical, even children's choir, with electronic NF above 5db. In most rooms, the room-noise exceeds the mike noise.

Also modern affordable Condenser mikes have high output levels which makes the "preamp" noise insignificant. (The real preamp is inside the microphone; our box does not have to be super-good as with dynamic or ribbon mikes.)
 
Very thanks for the answers, you are a lecturers....

The link is not OK, I try to attach an image with the Elektor schematic but I can't... I don't understand why.

Now I need to read carefully the answers. I send a new post this week

Muchas gracias

Miquel

 
> In Elektor 12/98 appears a version for MC cartridge preamplifier, that is my inspiration, (see link).
 

Attachments

  • MCasyElek.pdf
    307.6 KB · Views: 26
Very thanks for your attentio and for the link, you are the best in the world...

My idea it was to use an input transformer (I like iron).

Oh my God...! the DMMT3906 (odd name) is the same than two matched 2N 3906. I am not a brain luminary.

I have seen the SSM2220, LM394, and 2N3926 noise figure graphs, and I am thinking about intensity noise conception and source impedance.

I see the better noise situation for 2N 3906 is Ic 0,1 mA and Rs 2 KOhms (Noise figure 1,2 dB aprox.), at this point, rising intensity the noise increase.

The better noise situation for SSM 2220 is at Ic 1mA and Rs 1KOhms (Noise figure 0,7 dB). But that is in the upper limit of the graph (1 mA). It doesn’t show the noise for more intensity but it seems with more intensity it can handle less Rg at optimal noise.

The 1KOhm Rg bends for the LM394 noise are similar to SSM 2220, but are much better for higher Rg.

I have also seen the family datasheets THAT 1510, SSM 2019 and INA217 (THAT, Analog and Texas War…). INA specificity the optimal performance is at Rg 200 Ohms. SSM shows a formula (pg 6 datasheet), where the noise (total input) is every time directly proportional to the Rs, for that….. low Rg, low noise…!

Something else

There is a urban legend that the LM390 is not a discrete semiconductor, in really, is a array with a lot of discrete semiconductors paralleled, to work in better place for optimum noise.

1.- ¿Is for that than the LM390 works better at 1Kohm source, and the 2N3906 kid “neds” 2 Kohm?.

The SSM 2019 family works fine at very low impedance, the INA advice 200 Ohms and SSM 2029 says: “when less better” (formula dixit).

2.- ¿Is the front end pair configuration SSM 2019 family similar to discrete semiconductors array?. (The datasheets don’t specificity).

3.- ¿Is also the SSM 2220 front end a configuration transistor array?

Excuse me my wrong English, I can’t express “ideas”.

Miquel
 
> There is a urban legend that the LM390 is not a discrete semiconductor, in really, is a array with a lot of discrete semiconductors paralleled, to work in better place for optimum noise.

ALL discrete transistors are made on integrated circuit processes. The difference is how many transistors and interconnections are in the final product.

Not "legend". Clearly stated in one of the LM394 introductory papers.

Not "noise", but "balance".

My brother and I share a field for a garden. We want exactly-equal land so we can have equally good crops. We do not have time to map the wet, dry, rich, poor spots in the soil. There is room to plow 200 rows. Plow, then I use rows 1, 3, 5, 7..., he uses rows 2, 4, 6....  Most problems will average out equally between me and my brother. Even if one row is "dead" (flooded, all-rock, etc), the un-balance is only 1%.

Yes, that would be an awkward garden. But in integrated circuit production, it is easy. With one or two transistors per device, one bad spot gives significant un-balance. With 200 transistors connected even/odd, bad spots cause much less unbalance.

But DC balance is not the most important thing in an _Audio_ amplifier. We do not hear DC. DC bothers loudspeakers. We will always block the DC.

> use an input transformer (I like iron).

Then transform-up to several-K impedance, use any small high-gain transistor (2N5089 is a very good choice), or nearly any bipolar-input op-amp (NE5532 is still an excellent choice).
 

Latest posts

Back
Top