Ditching the Step Attenuator

11 views
Skip to first unread message

NZ0I

unread,
Jan 16, 2017, 4:27:07 PM1/16/17
to Receiver Development Platform
Hi Jerry,

I think you said you'll be traveling for a while, so the request below might have to wait a while. But I wanted to write this before I forget.

I recently realized that we do not yet have a way to attenuate the received signal at the SA605 level. Attenuating at the IF stage is going to be important for realizing all the attenuation that is needed for an ARDF receiver. And, if we want to have smooth attenuation steps, we will need the IF attenuation to also happen gradually, in small steps. But if we accomplish small attenuation steps in the SA605, we will no longer need the step attenuator in the front end for smoothing out the large gain adjustments - we can use the IF attenuation action for that. So I suggest that we make the step attenuator disappear. Poof.

But we will need to add support for deafening the SA605 in small steps.

I think we can attenuate the SA605 in small steps in a straightforward manner: using an I2C digital pot to apply a bias to one or more of the clock input pins (pin 4 or 10). We know that biasing both those pins with about +1V will make the SA605 very deaf. That's why we had such lousy sensitivity before you installed the DC-blocking capacitors. So all we should need to do is apply an external bias in small steps. I can calibrate the steps to 1dB increments using a manual step attenuator that I have. The calibration profile can be stored in EEPROM.

Here is an experiment that would be very helpful for determining which pins to externally bias, and how much attenuation we should expect:

1. Measure the sensitivity with no external bias applied. (~10uV we already know)
2. Measure the voltage at pins 4 and 10. That will tell us the bias that the SA605 is applying internally.
3. Tie only pin 4 to ground through a ~100-ohm resistor. Then measure the sensitivity, and the resulting bias on pin 4 when tied to ground through ~100 ohms.
4. Tie only pin 10 to ground using a 100-ohm resistor. Then measure the sensitivity, and the bias on pin 10.

If we can achieve at least 50dB of attenuation in step 3 or step 4, that is probably all we'll need. If both pins fall short of that mark, then you might run one final experiment, in which you tie both pins 4 and 10 to ground through (separate) 100-ohm resistors, and measure the sensitivity with both pins tied low.

Gerald Boyd

unread,
Jan 16, 2017, 10:23:46 PM1/16/17
to NZ0I, Receiver Development Platform
Yes I can try this when I get back. 

Found resistors at fry's tonight after work for the gain resistor test. 

Looks like I will need to order the 34.8k and 14k for the attenuator test from digikey. They skipped those values at both fry's and the electronic parts distributor in Albuquerque.

I need some surface mount tweezers and I should find some other items I need to get the order past the $25

Sent from my iPhone
--
You received this message because you are subscribed to the Google Groups "Receiver Development Platform" group.
To unsubscribe from this group and stop receiving emails from it, send an email to receiver-development...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/receiver-development-platform/a0076704-3782-4d38-960e-446ad950e704%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Charles Scharlau

unread,
Jan 17, 2017, 9:26:24 AM1/17/17
to Gerald Boyd, Receiver Development Platform
The attenuator resistor values are not critical. If you can get parts within 20% of specified values it should be plenty good enough. I've specified 1% resistors just because they're the same price as 5% parts.

Thank you for all your help with testing, reviewing, and all the great ideas.

To unsubscribe from this group and stop receiving emails from it, send an email to receiver-development-platform+unsub...@googlegroups.com.

Gerald Boyd

unread,
Jan 21, 2017, 2:40:01 PM1/21/17
to Charles Scharlau, Receiver Development Platform
Back in town today.

Was able to find parts within 1k or r207,208 (13 k in place of 14k) and within 1.8 k of r205 and 206 (33k in place of 34.8k).  So I have att parts and gain killing resistor test parts. 

Catching up from being out of town today and also have family visitors. Will get some testing done this weekend. 

Will Also look over the other emails regarding the input matching  simulation and it's suggested set of experiments and the other email regarding an experiment for reducing sensitivity( applying bias to one of the clock inputs). 

In town this week so I should be able to look at all the items.
Will do in this order starting tonight.
1.Put Attenuators on both clock inputs.
2.Gain killing resistor test. 30k,27k and 24 k.
3.bias to clock input test ( measure bias then put 100 ohm resistors to ground and retest sensitivity per email)
4.Matching circuit test per recommendations in simulation email.

Jerry



Sent from my iPad

Charles Scharlau

unread,
Jan 21, 2017, 3:04:23 PM1/21/17
to Gerald Boyd, Receiver Development Platform
Thank you, Jerry.

I've been working on laying out the Rev B receiver board. It is looking like it might be beneficial to move away from using our Rev-A pin configuration: those two rows of 14 pins interfacing the Motherboard and IF Board. 

So far, I've kept the Rev-A pin configuration unchanged, so that it would be possible to plug the IF board you built into the Rev-B Digital Interface board. There are some "extra" pins that I've added (and you would lack the front-end components), but the original rows of 14 pins would still be plug-in compatible. But now it is looking like the Rev-B layout could be improved by changing those two rows of 14 pins as well, and moving them farther apart, making the Rev-A IF Board incompatible.

Please let me know if you have a strong opinion about that, since it will mean having to build some kind of adapter mechanism to make the Rev-A IF Board interface with the Rev-B Digital Interface Module. Or you would have to build up a Rev-B Receiver Board (at least populate the 2m components) to work with the other Rev-B hardware. I'm hoping you will want to do that (build and continue experimenting with the Rev-B Receiver Board) but I don't want to presume.

Gerald Boyd

unread,
Jan 23, 2017, 3:31:21 PM1/23/17
to Charles Scharlau, Receiver Development Platform
Charles and Patrick stay tuned for report later today. Been experimenting last night and today.
Rift now I am at 1 microvolts. RSSI working and no oscillations. Major change today when playing with reducing gain killing resistor.  Checking system stay tuned.

Charles on the subject below if changing the IF board pinout makes for a better outcome let's do that. Also if making a new if Board makes sense let's do that. We are in development and I have no problem with assembling a new if board if it helps makes a better receiver platform. 

If we make a new IF board  I could turn the old board into a standalone RF sniffer and still get use out of it.

Jerry

Sent from my iPad
To unsubscribe from this group and stop receiving emails from it, send an email to receiver-development...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/receiver-development-platform/CAK-GiwYhaztjk_mN9onsf%2Bo2i4QO_Pcg0OKNu%2B_uU3TcWqrJsw%40mail.gmail.com.

Gerald Boyd

unread,
Jan 23, 2017, 6:26:57 PM1/23/17
to Charles Scharlau, Receiver Development Platform
Test data placed on drive. File name is 1-22-23 testnotes.docx. Note not year 2023 January 22and 23.

Interesting day. Please look it over. We can also see where the limiter kicks in.
Jerry

Sent from my iPad

Charles Scharlau

unread,
Jan 23, 2017, 9:53:24 PM1/23/17
to Gerald Boyd, Receiver Development Platform
Those are interesting results. Thank you very much for all your investigative work. A few thoughts below.

Regarding the "finger fix" for sensitivity: your finger would have been adding resistance, capacitance and inductance in unknown quantities, and to various points on the board. It might be very difficult to determine an equivalent circuit. But the fact that sensitivity improved with the removal of the 51-ohm series attenuator resistors might be a clue as to what was happening. Perhaps the finger load was coupling more clock signal across the attenuator and into the SA605. 

The expected impedance from the Si5351 is vague in the documents I've found, but it is supposed to be able to drive something between 50 and 85 ohms. But the signal amplitude might drop under those loads. The input impedance into the clock inputs of the SA605 are also vague. They likely vary with frequency just like the RF input pin does. One application note indicates that the input impedance might be around 10k, so that's what I assumed when calculating the resistor networks. All that is to say that experimentation is almost certainly needed: after all, we aren't exactly following any of the circuits provided by the manufacturer, and we don't have SA605 models (even poor ones) with which to test our designs.

You have established that it should be possible to achieve ~1uV sensitivity with the SA605 alone. That is within ~14dB of our goal of 0.2 uV. So our front end should have more than enough gain (provided that it all works!) to get us to our goal. And since we can totally remove all the gain from the front end, and attenuate the signal by more than 10dB by turning off the pre-amp and gain block, we shouldn't have to worry about there being too much gain in the front end.

The extended range of the RSSI is quite interesting. In whoopee mode, the user will get the full benefit of the extended RSSI dynamic range. AN1996 provides a circuit that is supposed to extend the RSSI range even further. But perhaps by the time the RSSI voltage is maxing out, the RF detector circuit we added will be detecting the RF signal directly and will provide an alternative RSSI source that works to much higher signal levels.

Based on your testing I will remove the option to externally bias pin 10, and leave only the option to bias pin 4. Based on your numbers, I think you are seeing at least 40dB of attenuation available by that method. That is good, but maybe not as much attenuation as we'd like. Also, I don't like its effect on the RSSI voltage.

There are still a couple of other options for deafening the SA605: biasing pin 1 (and/or 2); or coming up with the equivalent of a variable R212 gain killing resistor. Both of those options might have the benefit of not up-biasing the RSSI. I might look into adding pads that will allow us to experiment with those two options - especially since it looks like biasing pin 4 alone might not provide all the attenuation we want.

Thanks again for all your help!












--
You received this message because you are subscribed to the Google Groups "Receiver Development Platform" group.
To unsubscribe from this group and stop receiving emails from it, send an email to receiver-development-platform+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Receiver Development Platform" group.
To unsubscribe from this group and stop receiving emails from it, send an email to receiver-development-platform+unsub...@googlegroups.com.

NZ0I

unread,
Jan 24, 2017, 10:04:14 AM1/24/17
to Receiver Development Platform
I've revised the IF design to provide for the possibility of biasing pins 1 and 2 to provide adjustable attenuation.

I like the concept of having a processor-adjustable gain killing resistor to replace R212. A digital pot will only work up to ~2MHz or so, and would present inductance and capacitance values that might vary depending on the setting. Ideally, what we would like is a variable RF attenuator with a characteristic impedance of ~1500 ohms. But such a thing does not seem to exist. We might construct such an adjustable Pi-network of JFETs, but would need to adjust the JFET bias of the JFETs independently, and the design would involve a good deal a modeling/experimenting to get it to work properly. It would also involve adding a lot of new components.

So, back to the variable RF attenuator:

Above is a very small, simple, DC-GHz, low-power device, and not too expensive. The problems: it requires -3V, and it has 50-ohm impedance. There might be alternatives that work with positive voltages, or we can certainly add a small negative supply. Regarding the characteristic impedance, I wonder if we could simply isolate the ground pins using a 1500-ohm resistor to ground?

I'll look around for similar parts having better specs, but I think this type of device might get the job done. 
Reply all
Reply to author
Forward
0 new messages