Teensy Convolution SDR (Software Defined Radio)

Tisho, congratulations on the great progress in the development of the project. Your idea of an open source version of the Malachite DSP is one step closer.


1. Do you plan to release PCB drawings in Kicad format or at least GERBER files in the future?

2. If you plan the receiver as portable, it might be worthwhile in the next version to implement an impedance matching circuit with 2SK544 in the receiver for connecting a telescopic antenna. I can confirm that with this circuit the received signals have better SNR than if the telescopic antenna connects directly to 50R input PE4259.

View attachment Optional_PCB.pdf

Regards
 
Hi Radek,

Here are the complete Altium projects as well as the Gerber files:
https://www.dropbox.com/sh/3ogyd11vek3po2t/AAD1Ae3N5vt__J_iJlIMK1Z-a?dl=0
The production of 5 pieces of the main board was around 40 Euro with delivery (a bit pricy but it is 4 layer board), and the production of the BPF board was 15 Euro (as well 4 layer but smaller dimensions)

For the impedance matching I am completely agreeing with you. I plan although a bit different approach to it. I think our day op-amps can perform pretty good up to 25-30Mhz.
View attachment LT1818_AMP.pdf
This I build on small board and a piece of wire hanging out of the balcony (something like Mini Whip antenna but less power consuming) and I can say I am very satisfied, but I am opened to new ideas :)

Cheers
 
Last edited:
Hello all together,

I see that the Msi001 topic is slowly getting hot :) Here I will share with with you my progress up to now.

I already built the a prototype merging the Franks software and pieces of the gmtiis (I corrected a fragments for the setting frequency function and the gain setting of the Msi001, in my eyes now it is much better) with the nice RF front end of the Malachite DSP project. I added additionally on extra board a digitally switchable 8 channel band-bass filter.
As you can see from the attached circuit diagrams the idea is to have autonomous SDR (open source version of the Malachite DSP).
I hope you will like it and will get more people in to it to make it even better :)

Circuit diagram of the main board
View attachment 20690

Circuit diagram of the optional band-pass filter
View attachment 20691

Sorry for the quality of the pictures, needed better light
View attachment 20694

View attachment 20693

The inside out :)
View attachment 20692

Hi Tisho,
Nice work in the receiver. I have a question.
The data sheet for the PE4259 says that the bandwidth is 10 MHz - 3.0 GHz and the PE42641 states 100 MHz - 3.0 GHz. Have you tried these switches at lower frequencies? I know data sheets don't always tell the whole story.
Bill
 
Hi Bill,
I also was a bit concerned about it in the beginning but they work just fine.
On a piece of wire (around 2 meters) have no problem receiving down to 80m band, with the active antenna (impedance matching circuit with LT1818 amp from previous post) I had no problem receiving the DCF77 (77.5kHz)
 
Thanks Tisho. Those are nice little switches if they work well at low frequencies. The manufacturer designed them for GSM and WCDMA so they likely work fine but maybe they didn't bother to test them outside the intended range. I am a little worried about the footnote in the PE4259 data sheet (footnote 1, Table 1). It says "Device linearity will begin to degrade below 10 MHz". I may buy a couple and test them with my little homemade VNA - it could be a non-issue.
 
Thanks Tisho. Those are nice little switches if they work well at low frequencies. The manufacturer designed them for GSM and WCDMA so they likely work fine but maybe they didn't bother to test them outside the intended range. I am a little worried about the footnote in the PE4259 data sheet (footnote 1, Table 1). It says "Device linearity will begin to degrade below 10 MHz". I may buy a couple and test them with my little homemade VNA - it could be a non-issue.

That would be great, I was actually thinking to take one of those for the exact same reason:
https://www.ebay.de/itm/50KHz-900MH...274004737981?_trksid=p2385738.m4383.l4275.c10

But I am not sure if man can trust them, have to investigate a bit more, now I see that there are some reviews in youtube.
 
I have a nanoVNA. It works well for an inexpensive VNA and you can’t beat the price. But I also built a VNA from an article in the Jan 2020 issue of QEX that I prefer. Either one will work well for testing the switch. And they really help with filters and antennas.
 
Tisho, thank you for publishing PCB files.

Today I ordered Teensy 4.0, display, SGTL5000 and other things ... In the meantime, I will probably modify the PCB to use PCM1808 and PCM5102. According to Frank B, they should be better than SGTL5000.

Regarding the circuit for impedance matching - I think the RX9CIM is designing a variant of the circuit with 2SK544 due to the correct function at a relatively low supply voltage. Works LT1818 impedance matching circuit with a supply voltage of 3.3V correctly ?

Regarding PE4259 and PE42641 - according to UT4UBK "they have a generator inside to generate key management voltage. And he gives a hump in the region of 1.5 MHz." Probably for this reason, the manufacturer has increased the lower limit of frequency switches. The older revision of the datasheet states DC-3GHz. However, I use them in Malachite RX and they work well.

WH7U - only be careful when buying these switches with Aliexpress - defective non-functional pieces have been reported at some retailers. I can confirm that.
 
If you decide to get a nanoVNA, make sure you don’t get one of the cheap clones. Go to nanovna dot com and get that one. I don’t know why people are making cheap clones of something already so cheap but it is a big problem according to the nanovna forums. I will post a picture of mine which is the original authentic version. It works great. I recommend it. I only prefer the vna I built because I built it. This one does what it is supposed to and does it well. There is a bit of a learning curve.
 
IMG_4730.jpeg
Here is a picture of my nanoVNA. This is the authentic version. The two 50 Ohm terminators that are not in the box I purchased separately (the ones that come with the nanoVNA are not very accurate). I recommend getting a couple of 50 Ohm terminators that are 1% or 2% if you want to do accurate measurements. It's a great little tool. I used VNAs at work over the years and they were about the price of a luxury car - not something I ever though I would own. With this little unit, I feel like I can do reasonable RF work at home. It is not as accurate as the commercial units but it is perfectly fine for building radios up to UHF frequencies if you learn its limitations. The important thing to know is that it is accurate at 50 Ohms impedance, but don't trust it to be exact if you are measuring something at 2 Ohms or 20K Ohms. There are ways around this. But if you are tuning a filter or something, it is excellent. It is a game changer for those of us who tinker at home. The Jan 2020 QEX magazine has a review on the nanoVNA - this is what inspired me to buy one.
 
WH7U - only be careful when buying these switches with Aliexpress - defective non-functional pieces have been reported at some retailers. I can confirm that.[/QUOTE]

Thanks for the warning radek. I usually buy chips at Mouser here in the US. I've had bad luck with chip clones over the years.
 
Tisho, thank you for publishing PCB files.

Today I ordered Teensy 4.0, display, SGTL5000 and other things ... In the meantime, I will probably modify the PCB to use PCM1808 and PCM5102. According to Frank B, they should be better than SGTL5000.

Regarding the circuit for impedance matching - I think the RX9CIM is designing a variant of the circuit with 2SK544 due to the correct function at a relatively low supply voltage. Works LT1818 impedance matching circuit with a supply voltage of 3.3V correctly ?

Regarding PE4259 and PE42641 - according to UT4UBK "they have a generator inside to generate key management voltage. And he gives a hump in the region of 1.5 MHz." Probably for this reason, the manufacturer has increased the lower limit of frequency switches. The older revision of the datasheet states DC-3GHz. However, I use them in Malachite RX and they work well.

WH7U - only be careful when buying these switches with Aliexpress - defective non-functional pieces have been reported at some retailers. I can confirm that.

1. Do you know why exactly the PCM1808 and PCM5102 are better in comparison to SGTL5000 (to be honest I didn't compare one to one the datasheets yet).
2. I can confirm that the LT1818 works normal with voltage over 3V.

Bill thank you for the advice I will try to get the right VNA :)
If it is not trouble could you share the article desribing your home made VNA, I just got curious about it now, but couldn't find it available online
 
1. Do you know why exactly the PCM1808 and PCM5102 are better in comparison to SGTL5000 (to be honest I didn't compare one to one the datasheets yet).
2. I can confirm that the LT1818 works normal with voltage over 3V.

Bill thank you for the advice I will try to get the right VNA :)
If it is not trouble could you share the article desribing your home made VNA, I just got curious about it now, but couldn't find it available online

Tisho, regarding the audio codec - this is a recommendation from Frank B (post # 331) without further technical details. However, when quickly previewing and comparing parameters in catalog sheets, both ADC and DAC have better parameters in THD + N and SNR. I could not find the dynamic range parameter in the SGTL5000 datasheet.
 
Bill thank you for the advice I will try to get the right VNA :)
If it is not trouble could you share the article desribing your home made VNA, I just got curious about it now, but couldn't find it available online

QEX requires a subscription. But all the software and PCB files etc. are available here:
http://www.arrl.org/qexfiles
Look for the January/February article by James Koehler. I am almost positive that the ARRL made the article available publicly, but I can't find the link.

This is the magazine - The Jan/Feb issue contains a review of the nanoVNA and also has the article on building your own Vector Impedance Meter which is basically the same thing. I will look around and see if I can find a way to get you the article without getting into legal trouble.
http://www.arrl.org/files/file/QEX_Next_Issue/Jan-Feb2020/Jan Feb TOFC.pdf

There is some discussion about the two options in the nanoVNA users group:
https://groups.io/g/nanovna-users/topic/69424021#9156
 
Thank you guys for the info :)
Now I really got interested about the potential improvement of my design. Here is small comparison of the parameters between SGTL5000 and PCM1808, PCM5102 option:
ADC-DAC comp.png
It really looks like that the PCM1808 and PCM5102 are superior in the performance.

There is one thing bothering me, as usual different semiconductor producers make their datasheets differently and when you take a look to the PCM1808 documentation it is nice and detailed explained the test condition for the THD:
PCM1808.png
And SGTL5000 just give one value.

Then I question myself is it possible that they are compared not in exact same conditions, and more important will it be there noticeable difference between both options in the real design :)
To hear the opinion of people who compared both options will be very useful.
 
Last edited:
Thank you guys for the info :)
Now I really got interested about the potential improvement of my design. Here is small comparison of the parameters between SGTL5000 and PCM1808, PCM5102 option:
View attachment 20736
It really looks like that the PCM1808 and PCM5102 are superior in the performance.

There is one thing bothering me, as usual different semiconductor producers make their datasheets differently and when you take a look to the PCM1808 documentation it is nice and detailed explained the test condition for the THD:
View attachment 20737
And SGTL5000 just give one value.

Then I question myself is it possible that they are compared not in exact same conditions, and more important will it be there noticeable difference between both options in the real design :)

Hi Tisho,
I always choose TI parts if I have the option. My only reason for this is that their data sheets are very good and almost always have the necessary info for a designer. If they don't tell you how they tested something, it isn't really much use, but TI does tell you most of the time. I have plenty of experience with parts that don't meet the data sheet specs - sometimes by a large margin. I have never had that trouble with TI and I have used hundreds of their ICs over the years. I trust TI data sheets so I don't mind paying more for their parts if I need to. What good is a part if you can't trust the data sheet? All that being said, the specs on either IC are really good. You are never going to hear the difference between 85 dB and 93 dB THD in your headphones. (I will probably have a bunch of audiophiles yelling at me for this but it is true). Either DAC will give excellent performance.
The SNR and Dynamic range of the ADC is very important though. The dynamic range of the receiver is determined here to a large extent. But in order to achieve 99dB SNR (or dynamic range) you need to do some special things. 100 dB is an amplitude difference of 100,000. What it means in the receiver is this. If you have a strong signal that is being digitized by the ADC but you are trying to hear a weak signal near the strong signal, you need the dynamic range to handle this. The ADC needs to digitize the strong signal without distortion (which splatters it all over the weak signal).
Say you design your receiver so that it can handle a maximum signal of 10 mV at the antenna without distortion. If there is a strong signal of 10mV at your antenna and you have a receiver with 100 dB dynamic range, you can hear a signal of 10mV/100,000 = 0.1 uV. This is about right for a good commercial quality communications receiver. And if your SNR is also 100 dB, your 0.1 uV signal will not be buried in the noise either (the noise will also be 0.1 uV) in this case.
I hope you don't mind me getting on my soap box again, but I hope to save you disappointing results in your receiver.
Where people generally go wrong here is to assume they will get 100 dB SNR because that is the IC spec. This is not true. If you have 1 mV of power supply noise (which is a very quiet power supply) on the analog power supply line of your PCM1808, you will have a large fraction of that 1 mV imposed on your signal by the chip itself. Say the PCM1808 PCR (power supply rejection ratio) of the PCM1808 is 20 dB (TI didn't tell us the PCR - it is rarely specified except for linear regulators). But guessing it is 20 dB, the noise you are adding to your signal will be 1 mV * -20 dB = 100 uV. The SNR of your radio will be your maximum signal / noise = 10 mV / 100 uV = 40 dB. This is a poor quality receiver. It means that the weakest signal that you will receive is 100 uV and that signal will also be equal in level to the noise in your receiver.
This is why TI gives you two power supply pins on the PCM1808. You will never get less than 1 mV of noise on a digital circuit so don't share that pin. So they allow you to have a power supply for the digital circuit and a separate very quiet power supply for the ADC. What TI is saying when they say 99 dB of SNR is: if you have a design with no external noise, the chip is capable of 99 dB SNR. This is nearly impossible to achieve in real life, but you can do it. Use a separate high PCR linear regulator for the analog pin on the PCM1808 and you can get really outstanding SNR and dynamic range with the PCM1808.
 
I need to fix my example above. I forgot to include the gain of the audio amps. But the principle is still correct.
 
Hay Bill, thank you so much for the detailed explanation.
I already see few points where the improvement can be done (the analog supply 1,8V of the SGTL5000, the SGTL5000 itself and so on). Just to do it a bit more quantitative, or in other words to make easier to compare, correct me if I am wrong but I think the Noise Figure (NF) is a good indicator to compare receivers. Coming days will try to measure the NF of the current design.
Here is the algorithm I plan to use, and if you see some flaws please correct me :)
1. Will calibrate the dBm meter of the receiver let say at 80m band (3.7Mhz) with signal generator and attenuator
2. At the same gain settings, with terminated 50ohm input will measure the salve noise Psn of the receiver in dBm
3 Then NF = Psn - (-174dBm/Hz + 10 * log10(BW)) where:
NF is the noise figure. Psn is the measured total output noise power. -174dBm/Hz is the noise density of 290°K ambient noise. BW is the bandwidth of the frequency range of interest.

Once having the NF can start modifying and see if there is improvement or not, very curious what it is possible with my design :)
 
Tisho, thank you for publishing PCB files.

Today I ordered Teensy 4.0, display, SGTL5000 and other things ... In the meantime, I will probably modify the PCB to use PCM1808 and PCM5102. According to Frank B, they should be better than SGTL5000.
...

Instead of the PCM1808, take a look at the PCM1804 or PCM4202.

Regards

Karlchen
 
The two 50 Ohm terminators that are not in the box I purchased separately (the ones that come with the nanoVNA are not very accurate). I recommend getting a couple of 50 Ohm terminators that are 1% or 2% if you want to do accurate measurements. It's a great little tool.

I'm a new NanoVNA-F owner. Can you recommend a good source for quality terminators?
Tnx!
Todd
 
I'm a new NanoVNA-F owner. Can you recommend a good source for quality terminators?
Tnx!
Todd

I bought my terminators from this place:
https://www.w5swl.com
And this is the model:
https://www.w5swl.com/Mini-Circuits-SMA-Male-14-Watt-Termination-Load-DC-2GHz_p_6170.html
Measured with my old Fluke 79 III meter, which I admit has not been calibrated in years, I measure 49.9 Ohms on both of them. These are mini-circuits terminators which are very good quality. This model has been discontinued so that might be why they are only $3 USD each. The replacements are more like $12 USD.
You want to try and stick with a brand name and avoid the clones which are notorious for containing a 52 Ohm 5% 1/4W resistor in a housing that looks like a real terminator. They will cause your calibrations to be way off at higher frequencies.
 
Hay Bill, thank you so much for the detailed explanation.
I already see few points where the improvement can be done (the analog supply 1,8V of the SGTL5000, the SGTL5000 itself and so on). Just to do it a bit more quantitative, or in other words to make easier to compare, correct me if I am wrong but I think the Noise Figure (NF) is a good indicator to compare receivers. Coming days will try to measure the NF of the current design.
Here is the algorithm I plan to use, and if you see some flaws please correct me :)
1. Will calibrate the dBm meter of the receiver let say at 80m band (3.7Mhz) with signal generator and attenuator
2. At the same gain settings, with terminated 50ohm input will measure the salve noise Psn of the receiver in dBm
3 Then NF = Psn - (-174dBm/Hz + 10 * log10(BW)) where:
NF is the noise figure. Psn is the measured total output noise power. -174dBm/Hz is the noise density of 290°K ambient noise. BW is the bandwidth of the frequency range of interest.

Once having the NF can start modifying and see if there is improvement or not, very curious what it is possible with my design :)

Hi Tisho,
Your plan to measure noise figure sounds excellent. It is a great way to quantify your improvements as you refine your receiver. The tricky part of NF measurements is the bandwidth. Noise figure is the commonly used performance metric for UHF and microwave receivers because they have a fixed bandwidth. It is perfectly valid but less commonly used for HF receivers because the receive bandwidth is not consistent. It seems like every receiver has a different bandwidth for SSB and there are all sorts of CW bandwidths to choose from, not to mention FM etc. If you select a CW bandwidth of 300 Hz it will give a very different result than the same receiver in WFM mode.
Personally, I would do the NF measurement with a SSB filter in place for HF and use an FM filter for VHF or UHF. If you specify the filter you use, others can compare their results with yours. I'm excited to hear what you get. NF tells the whole story as far as how weak a signal a receiver can receive and is a good way to compare everyones ideas and approaches.
Do you own a book called "Experimental Methods in RF Design" by Wes Hayward, etc. ? It is a bit out of date, but it has a lot of good info on measuring receiver performance without costly test equipment.
 
Hi guys,
before going to the noise measurements few words about the test setup I used to calibrate the receiver, as a source I used a Stanford Research DS345 signal generator and a 40dB attenuator, and each band was calibrated at S9 (-73dBm or 50µV rms). Generator was set to -33dBm and the 40dB attenuator was added on the output.

I kept fixed the receiver settings for each band:
- Premplifier on the main board "ON"
- Premplifier on the BPF board "OFF"
- Msi001 attenuation 40 dB
- RF gain (menu) 0 dB
- Filter bandwidth 5400 Hz


Here are my noise measurements
Noise_Measurement.png

Then I made a bit of research to see where actually I am with this data, in comparison to the other receivers, and I found very useful data in the SDRPlay forum:
https://www.sdrplay.com/community/viewtopic.php?f=5&t=3685&sid=f0528dc0462c23a3151ed3888bce46c3
Here is a similar measurement (smaller Bandwidth was set for the measurement 1800kHz) done with the RSPduo.
Revised RX Noise table.jpg
*Pay attention to the 50 ohm inputs

In the beginning couldn't believe that the NF is so close to the SDRplay even on some points slightly better, so I checked the calibration again and repeated all measurements but everything looks correct. :)

Bill I definitely will check for this book, I guess can be useful :)
 
Those are impressive measurements Tisho. If you want to double check that you are getting good numbers, you can add more attenuation and see if the NF changes (it should not). But with more attenuation you also have to be careful that your signal is not leaking from the signal generator to the receiver.
 
Back
Top