Notes for users of Tigertronics hardware.
There are some noise problems reported with certain Tigertronics units. This is reported by John KF5SAB:
I thought I would post a note about my experience using a SingaLink USB with FreeDV in hopes that others might find it useful. This isn't exactly specific to FreeDV though, so please anyone else who has any tips for improving the SignaLink USB is more than welcome to chime in. I am by no means an expert on this topic, and am grateful to learn the experience of others. If my PSK31 QSOs or this mailing list are anything to go by, this device a tremendously popular piece of hardware, and I suspect a lot of people will want to use it for FreeDV. I sure did. Despite it's popularity, the SignaLink USB has some pretty grievous and long-standing hardware issues that have not been that well publicized, and worse (according to others) have apparently been somewhat swept under the rug by TigerTronics. That said, the hardware is still a nice little package and generally very functional for what 99% of people probably use it for.
It is my opinion that the hardware problems affect 'wider' band modes such as the FDM modem used in FreeDV more severely than they affect narrow band modes such as PSK31. Ultimately though it comes down to how the filters in your transceiver interact with the performance of your audio interface on both transmit and receive. In my own experience, my station is marginal enough that I was able to pick out the SignaLink USB itself as my weak point. I have been listening to FreeDV/Codec2 on the air for about a month now, but was not ever able to be received until after I made modifications to improve my interface hardware. Since FreeDV currently operates without any error correction, a few dB can mean the difference between a readable signal, and total noise. The audio interface to your radio is completely in your control, so there is no reason to let it trash your signals!
But no use complaining, right? We have already thrown down our cash for SignaLinks; we like using the little knobs instead of messing with windows volume sliders; and we are perfectly capable hams. So let's get to fixing them!
The two best resources that I found when I first started investigating this problem are:
These documents go into some deep detail about various issues and document a few fixes for each. I have tested and modified a few of these interfaces since I started searching. All have exhibited them to some degree, and all have seen considerable improvements since modification. If you decide to do anything based on my recommendation, please follow the direction in these guides.
The #1 problem to fix (and one of the easiest) is the USB power noise feeding into the op-amps 2.5V bias. The method I chose is to modify the resistive divider supplying the 2.5V to take its supply from the 3.7V LDO regulator providing the PIC with its Analog reference. This is detailed most clearly in the PDF linked above. All SignaLink USB's that I have tested exhibit low frequency noise peaking about 30-40dB above the noise floor with the harmonics stretching up to 5-600hz. When you consider that 10dB is an entire order of magnitude, that is really disappointing performance. I am thankful to David Rowe for making a change to the waterfall display in FreeDV to ignore <200Hz signals when setting the gain (color scaling) because of this hardware.
The #2 problem to fix (if you need to) is the frequency response. TigerTronics uses junky (this is the technical term for inconsistent tolerance) transformers, so you might have perfectly flat response, or your response at 2khz might be 20dB down from 500Hz. Mine have all been mostly OK, but a FreeDV signal will benefit from a nice flat response. Salvaging 1:1 transformers from things like old modems seems to be the recommended approach, and if I had to guess, I'd suggest that 56k modems probably have nicer transformers than old 14.4 modems, but YMMV.
There are of course other things you can fix or improve - adding more filtering/decopuling caps, improving shielding etc., but these two improvements are the main ones. All of the SignaLinks I have adjusted have had tremendous improvement but all still exhibit some type of remaining noise. One has a small spur that drifts around between 500-800Hz. Another has a fairly large spur at 1khz and its harmonics. I have not determined a good way to eliminate these problems yet, but at least they are only 2-3dB problems, not 40dB!
So do yourself a favor and check and repair these 2 things on your SignaLink if you are able. You may have perfect success without doing a thing, or you may not. I sure operated a ton of PSK and JT65 just fine and noticed nothing wrong.
To accurately measure the noise floor of your device, first disconnect all the audio input/output from the back of the signalink. Install Spectran V2 from http://www.weaksignals.com/ or another suitable FFT program that will show unscaled values in dB. Set the input volume gain in windows to 0dB in the levels tab of the input device properties (you can right click on the number and change the readout from "percentage" to "decibels") Yes, 0dB has the slider almost all the way down - only about 3% scale. Fire up Spectran, set the sample rate to 8khz and take a look at the spectrum. It will probably appear similar to one of the screenshots shown in the documents I linked earlier, and you will probably see a bunch of low frequency noise. Take a screenshot so you can compare before/after. You can use Spectran to analyze frequency response if you have a suitable signal generator also, or just feed white noise from your radio into the SignaLink and observe the response curve (this will be affected by your radio's SSB roofing filter as well, so you might compare between a couple of sound cards if you can)