hiss in preamp balanced output SSM2142

[jackinnj]

Meaghers, the SSM-2142 never was particularly famous for its noise performance. The spec' sheet quotes it as -109 dB re 10V rms. If we express this as dB's re 1V (o dBu) it is -89 dB.

The SSM2142 SNR is spec'd at -93.4dBu -- at least in the RevB datasheet. It still falls short of Thatcorp's 1646.

http://www.thatcorp.com/datashts/1600data.pdf

The 1646 is available from Mouser, $4.78 per. The SSM2142 is $4.22 at DK.

Before abandoning the SSM2142, tie a 100nF ceramic to each supply pin to ground, right at the V+, V- pins. It also helps to make your design as "symmetrical" as possible. Some of these chips can have a bit of offset -- that's why you'll see a 10uF cap between "Sense" and "Out" -- I've drawn in a jumper -- there's a nice WIMA MK2-SL 10uF cap which should fit in there (or you can use a 10uF NP Electrolytic).

EDIT: C5 -- optional -- you can use a 100pF cap which will shunt RFI to ground

 

[jackinnj]

I measured the noise using the AP Test procedure -- it's -89.3 dB. THD% is a pretty remarkable 0.00024% from 20Hz to 22kHz, and 0.00044% 20Hz to 80kHz.

The test setup isn't that exotic -- just a Tektronix PS5010 power supply, the SSM2142 with 100nF bypass, no electrolytics. I ran a piece of copper tape down the center of the perf-board to which the bypass caps are grounded. I used a shielded container from the line driver, unbalanced in, balanced out...turned out the fluorescent lights and soldering station. The shielded container was a Williams Sonoma Pfferneuse box which I prefer to the Danish Butter Cookies used at Linear Technologies.

The AP Test Procedure drives the amplifier to that point at which clipping starts to rise dramatically, then integrates across the audio spectrum. I don't have a CCIR-486 filter on this analyzer.

I think the problem with hiss relates exclusively to your power supply.

EDIT: A-Wtd, Quasi Peak the noise is -98.6dB

 

[PMA]

In case it was used as single-ended output, the -OUT MUST have been grounded (otherwise noise is horrible, as it just ADDs the unconnected -OUT as high impedance input to +OUT voltage). For balanced output, the noise is normally reasonable.

 

[meaghers]

WOW, thank you all so much for your replies. I am exteremely grateful for the help and attention you have all given me. Hopefully this information can be useful to someone down the line. I, however, have decided to leave my preamp unbalanced for now. I think I will try again when I have learned to fabricate my own pcb's. Does anyone have wisdom to give on this process? Where should I start?

 

[gain]

a quick down n dirty way i used to make my first pcb's (and still do for simple layouts) ...

take a blank pcb and clean it up real good with a piece of sandpaper or steel wool. then take a black waterproof sharpie and draw onto the pcb where you want the copper traces to be. make sure they are colored in nice and dark ... go over them a couple times to make sure. you can first draw the lines on with a pencil in case you make mistakes. after you have drawn on the traces, put the pcb in a small PLASTIC tray and pour in just enough etching solution (available at the 'shack) until the board just begins to float. the etching solution will dissolve away the copper NOT covered by the marker in about 10-15 min.

edit: be careful with the etching solution. it will stain anything it comes into contact with, including your skin.

 

[Keith Taylor]

Meaghers, After sending my earlier post I had a thought that could explain your poor noise performance. Could you check that you have the correct value for R9 (Rod Elliott's drawing) This is in series with the preamp output to improve the stability with capacitive loads. Perhaps, instead of 100R you have 100K or 10K etc. If this were so it would be bad news for noise, loss of high frequencies due to cable capacitance and not good for the SSM 2142 as it requires a low source impedance to drive it. As I said, just a thought.

 

[meaghers]

I'll check the value of R9 when I get back home on Friday. Thanks for the thought 🙂

 

[EWorkshop1708]

a quick down n dirty way i used to make my first pcb's (and still do for simple layouts) ...

take a blank pcb and clean it up real good with a piece of sandpaper or steel wool. then take a black waterproof sharpie and draw onto the pcb where you want the copper traces to be. make sure they are colored in nice and dark ... go over them a couple times to make sure. you can first draw the lines on with a pencil in case you make mistakes. after you have drawn on the traces, put the pcb in a small PLASTIC tray and pour in just enough etching solution (available at the 'shack) until the board just begins to float. the etching solution will dissolve away the copper NOT covered by the marker in about 10-15 min.

edit: be careful with the etching solution. it will stain anything it comes into contact with, including your skin.

That's awesome, I want to try that! 😱
If I get the Name Brand Sharpie, will there be better results?

 

[Minion]

I find the Best Way is just to print out your PCB design on some Photo paper or magazine paper useing a Lazer printer and then use an iron and iron the Design on to a Piece of Clean PCB material and then soak in water and rub the paper till there is nothing left but the Ink transfer....then soak the PCB in One part Muriotic acid and 2 parts hydrogen peroxide till the Copper disolves (about 5 min) and then rince it off with acetone to disolve the ink and drill your Holes and stuff the PCB with your components....
Another good way it to use the same above method but print your design on to Press N peel , it is a Bit quicker and more reliable useing it....

I have made Hunderds of boards this way , I have actually never had the need to use a Prefab board as this has allways worked extremely well for me....

Here is a Guide to the Press n Peel method ...

http://www.techniks.com/how_to.htm

here is a Guide to the Photopaper/Magazine Paper toner transfer method:

http://www.headwize.com/projects/garbz2_prj.php


Cheers

 

[PMA]

Here is the comparison of noise of balanced output chip operated as single ended. Green is for output from OUT+, OUT- left unconnected. Red is for OUT- grounded. You can see that one MUST ground the unused output in single-ended mode.

 

[jackinnj]

I got a couple of THAT1646 from Mouser -- here's how they compare (pretty darn well) to the SSM2142:

An externally hosted image should be here but it was not working when we last tested it.

 

[meaghers]

Meaghers, After sending my earlier post I had a thought that could explain your poor noise performance. Could you check that you have the correct value for R9 (Rod Elliott's drawing) This is in series with the preamp output to improve the stability with capacitive loads. Perhaps, instead of 100R you have 100K or 10K etc. If this were so it would be bad news for noise, loss of high frequencies due to cable capacitance and not good for the SSM 2142 as it requires a low source impedance to drive it. As I said, just a thought.

R9 is indeed 100R as indicated. Thanks for the thought 😀

 

[mediatechnology]

Hi - I just stumbled into this thread via a Google search and wanted to amplify - pun intended - the following statement as to why the SSM2142 might be noisy.

In case it was used as single-ended output, the -OUT MUST have been grounded (otherwise noise is horrible, as it just ADDs the unconnected -OUT as high impedance input to +OUT voltage). For balanced output, the noise is normally reasonable.

The SSM2142 and DRV134 use a cross-coupled output based on Thomas Hay's work at MCI while the THAT1646 uses OutSmarts based upon Chris Straum's work at ATI. When a single-ended load is connected to a CCOS or OutSmarts output, the opposing output must be grounded to prevent the high common mode gain of each part from being converted to differential noise.

The THAT1646 has an output noise of ~ -100 dBu while the SSM2142 (probably under slightly different test conditions) a figure around ~ -94 dBu. These are differential measurements. If one output is used to feed an SE load with the other output floating, the noise in each part will typically measure about 40 dB worse around -60 to -66 dBu.

To properly terminate these devices for an SE load, the opposing out must be grounded. Then the noise performance will be close to published specs.

Floating the opposing output is essentially the same as floating a line transformer's secondary. Unfortuanly there's one huge difference: With a transformer you'll immediately know you have a problem due to hum. With a active balanced output, there's just added noise. About 40 dB's worth...

I don't know if that's what happened to our OP, but my hunch is that the - output was floating. Grounding it will also add 6 dB of gain.

And yes Keith is right, the source impedance must also be low or the common mode gain gets converted to differential noise at the input.

For a THAT1646 (I don't know about the SSM/DRV) the inductance at pin 3, "ground" must be very low. The bypass caps should connect directly to pin 3 and it should be earthed through a very low "L" path.

 

[Bonsai]

Interesting discussion - I have been looking at using th e SSM2142 to convert a single ended output from a PGA2310 to balanced.

My thinking has been to provide 4 or 5 single ended inputs (aux 1 and 2, tuner, tape, recorder, phono etc) along with 2 balanced inputs with conversion to SE via a SSM2141. On the output of the PGA2310, I was going to use 1 x SSM2142 to convert the single ended output to balanced.

However, after reading this, I reckon a smarter way to go will be to use 1 x PGA2310 per channel and run these in fully balanced mode. for single ended inputs, I'll then wire some RCA sockets in parallel with the XLR's and convert the input to songle ended via a switch.

Any comments on this approach? BTW, I just read the review on the baby GamaT integrated that uses the PGA2310 - the reviewer was very impressed.

 

[mediatechnology]

Interesting discussion - I have been looking at using th e SSM2142 to convert a single ended output from a PGA2310 to balanced.

Thanks. I can't really comment on your alternative approach but you can still do the above.

The SSM2142 DRV134 and THAT1646 are all good performers as long as the opposing output is grounded when driving an SE load.

Another alternative is to disable the SSM/DRV's CCOS or THAT OutSmarts by simply not connecting the force+ and force- outputs to the sense+ and sense- inputs. This can be done if the load is known to be balanced and never a grounded SE connection.

You don't get the advantage of the 6 dB gain boost and it will also dump current if one leg is shorted to ground but the common mode noise gain is reduced. In a balanced situation the common-mode noise of any of these parts isn't an issue because it's in common mode. With a relatively poor differential amp following it (say 40 dB CMR) the difference in differential noise vs. common mode noise is moot.

It almost boils down to two sets of rules for these types of outputs. Use CCOS and, with SE loads, require grounding the opposing output. Or, don't use CCOS and never ground an output.