Re: [softrock40] Increasing the usable dynamic range of the RXTX 40/30/20
Putting the attenuator after the QSD means the the QSD is not the source of the intermod. The op amps are overloading. 600mV peak to peak will not overload the QSD. Rather than adding an attenuator and modifying the filters for 50 ohms, you could just change the gain setting resistors for the op amps.
Dave - WB6DHW
On 2/24/2013 6:08 AM, gerhard972 wrote:
I have been following the recent discussions on SoftRock dynamic range and intermodulation with great interest. Here is my personal experience.
During the evenings I received a lot of intermodulation products on my RXTX 40/30/20 in the 20m band.
Several AM modulated carriers spaced at exactly 5kHz did not shift up or down during tuning. My first impression was that the SI570 didn't change frequency. But it obviously are intermodulation products from broadcast stations. I uploaded a HDSDR screen capture to the PA1GVE album to show this behavior.
Tuning across the full range revealed that broadcast stations in the 49m band (6MHz) were the cause of the IM products. Levels of up to S9+60dB were measured. 600mVpp was measured on RX ANT (Q11.d) and 400mV on the antenna input. The audio output was less than 100mVpp.
Placing attenuators just before and just after the QSD mixer showed that the QSD mixer was the source of the IM products.
Since the RXTX showed 25dB of noise floor rise on 40m when connecting to the antenna and only 10dB-12dB on 20m, the obvious solution to eliminate the IM is to change the filter response such that 6 and 7MHz are attenuated 18 to 20dB and 14MHz 5-8dB at maximum.
A transfer measurement from the antenna connector to the primary of the QSD transformer T5 showed that in the original circuit 14MHz is 4dB weaker than 7MHz, contrary to what is needed.
This response is probably not as it was intended. Maybe because 2 filters were placed in series without proper terminations. The first filter (L3, L2, C26, C25, C24) is not terminated into 50 Ohm and the second filter (C27, L4, C39, T5) is not fed from a 50 Ohm source. Inserting a 3dB pad between both filters already removed the dip at 17MHz and the peak at 18MHz.
Some tweaking resulted in the following changes: C27 1nF (was 330p), 8dB pad (120, 56, 120 Ohm) between Q10/Q11 and L4, 100pF in series with L4, C39 330pF (was 470p), C25 390pF (was 470pF). The change of C25 also reduces the insertion loss on 20m to below 0.5dB.
Compared to the original circuit 6MHz is reduced by 18dB, 7MHz is reduced by 15.5dB, 10MHz is reduced by 11dB and 14MHz is reduced by 5dB only.
The measured receive filter responses can be seen in the PA1GVE album.
The result is that all IM products are gone, even if the RXTX is set to 14MHz and the antenna tuner is set to 6MHz. All three bands still show a 5-10dB noise floor increase when the antenna is connected and properly tuned.
Similar considerations might be valid for the other (bands) versions of the RXTX. These could also benefit from the receiver attenuator and the decoupling of the two filters.
My RX II Ensemble has much less IM problems. On 14MHz the preselector bandwidth is from 8MHz to 16MHz. So 6MHz is sufficiently attenuated (18dB). Both preselector filters below 8MHz have attenuators (close to 15dB).
If the RX II Ensemble is set 14MHz and the antenna tuner is set to 6MHz then sometimes broadcast stations can be heard on the spectrum center line (LO freq=tune freq), independent from RX II LO frequency. But it is hardly visible in the spectrum.
73, Gerhard PA1GVE
- I had an interesting day back in the lab after 10 days on the road. I won't run through the grueling details but I now have a 20 dB audio attenuator cable with 3.5 mm ends and I also built a hybrid RF combiner that allows me to combine or split two 50 ohm sources with 3 dB loss in each one. Without any trimming i got about 25 dB isolation between the ports up to 30 MHz. It will be a handy tool.
My attenuator cable idea works for utilizing the unused noise floor in either a 16 bit or a 24 bit card. That is the good news. The bad news is there seems to be little point. On the high end, as Alan suggested, I found that by minimizing the PC input gain I was within a couple dB of the clip points on the op amps anyway. So, even though I was able to "raise the roof" on the sound card, I only gained 2-3 dB before the op amps clipped. This interested me so I ran through the 4 RXTX that I have here in Beijing to check at what level the op amps went into clipping:
Radio Band OP Amp Saturate
dBm (WSPR Center Freq)
RXTX 160 160 -14
RXTX 80/40 80 -15
RXTX 30 -11
30/20/17 20 -10
RXTX 15 -19
15/12/10 12 -18
The answer is "anywhere from -10 to -19 dBm". So even if I could get the sound card range shifted up I was still limited by the op amps.
The other problem with "raising the floor" is it is ugly down there! With my attenuator cord in place I was able to move the Realtek and the iMic card floors up to match the antenna noise. With the 24 bit X-Fi I was able to get it up within 10 dB of the noise. But with the iMic and especially with the X-Fi I found that the noise, spikes, and spurs were so prevalent in the bottom 10-30 dB of the card that I really did not want to see them. In real life the antenna noise hides a multiple of these digital sins. The X-Fi useable noise floor, once you get above the noise and spikes, is not a significant improvement over that of iMic and not as good as the internal 16 bit Realtek.
So, again in real life, I am bounded on the top end by the op amps, let's say -17 dBm on 40 meters, and by the noise floor on the bottom end, normally about -110 dBm. That gives me 93 dB of useable dynamic range which is perfectly accommodated by a 16 bit card.
I do confess to having an E-MU 0204 on the way by courier from the US. It should be here next week. It will be interesting to see if the lower 20 dB will be useable and I am justifying the purchase on the basis that it will be a lab tool rather than operational.
Closing the book on this one for now ..
Warren Allgyer - W8TOD