UADC4 Universal Analog to Digital Converter for Zero IF Software Defined Radios

Transcription

1 UADC4 Universal Analog to Digital Converter for Zero IF Software Defined Radios INTRODUCTION Zero IF receivers or Direct Conversion Receivers (like SoftRocks, IQ mixers, WSE converters and IQ+ receivers) are the most popular SDR implementations used by Ham Radio operators. The introduction of PC audio sound cards created an extensive platform for SDR experimentation. As a result, Ham Radio operators have enjoyed significant benefits for more than 25 years. But as usual, performance is a strong trade-off negotiation; whereby the new PC audio cards cannot improve performance in our SDR RX systems, and the reason is very simple: PC audio cards are just PC Audio cards ; they are not ADC systems designed for Software Defined Radios. Even if software developers are proclaiming software is everything, that is an incorrect assumption because room still exists for HW improvements. Software cannot fix what HW cannot deliver! The adoption of PC audio Cards as an ADC system is a typical example of the inventiveness and creativity of Ham Radio operators over the last 20 to 30 years, but it is not the ultimate solution for Analog to Digital Conversion in a Ham radio SDR

2 system for Weak Signal Communication like EME. It s just a cheap readily available solution running at its limits and has been for many years! The UADC4 is not a PC audio card and cannot compete with professional audio cards as a sound card; similarly, sound cards do not offer ultimate performance for your Zero IF SDR. The UADC4 is specially designed for SDR applications and in conjunction with your SDR receiver it will create a robust platform with outstanding performance. The limiting factor will no longer be your PC audio card used as an ADC; but will be your Local Oscillator or your mixer. In this area many options exist to create a high performance SDR receiver with greatly superior performance than traditional SDR receivers based on simple PC Soundcards. This document is an explanation of what I did while in search of better performance for my IQ+ SDR radio, resulting in the creation of a dedicated ADC with some special characteristics you will not find in any commercial PC soundcard. Are ZERO IF RECEIVERS old fashioned technology? Many people think YES, they are! But with respect, I think they are wrong! The market is inundated with DDC receivers (Digital Down Conversion Receivers). One of the most often postulated argument for their use is: DDCs SAMPLE AT THE ANTENNA. THE REST OF THE JOB IS A MATTER OF SOFTWARE. That statement looks convincing but when viewed from the Ham Radio operator s perspective it is not realistic, because to a ham, both money and performance counts. Certainly, DDC Receivers have extraordinary advantages in areas where Zero IF Receivers cannot compete. For example, the use of pure software NCOs (Numerically Controlled Oscillators) fully implemented in a FPGA by software, dramatically reduces the tedious problem of phase noise and reduces reciprocal mixing problems. However how many of these radios fulfil this well-known promise? The answer is very few and they are too expensive for a Ham budget. Some of the limitations of DDC receivers are related to sample capacity, the dynamic range they can actually deliver and the excessive bandwidth over which they sample. Most of the commercially available chips today can sample at no more than 16 bits other than a few that can sample at 18 bits but only at frequencies below 50 MHz. Devices capable of real DDC conversion at over 100 MHz are rare and completely out of reach for a Ham budget; they are mainly designed for military purposes. Those with very high frequency response have less than 12-bit resolution.

3 Just reviewing some theory: In a digital system the maximum theoretic Dynamic Range is defined by the following formula: DR = (6.02 x N) - 3 Where N is the number of bits your ADC can sample, for example: A 14 bit = (6.02 x 14) -3 = 81.28dB A 16 bit = (6.02 x 16) -3 = 93.32dB A 18 bit = (6.02 x 18) -3 = dB A 20 bit = (6.02 x 20) -3 = dB A 24 bit = (6.02 x 24) -3 = dB But those figures are theoretical and in practice the reality is somewhat different. A rule of thumb for a 24-bit system is to decrease theoretical value by 20 db leaving a respectable 121 db Dynamic Range. This is true for one strong signal in your pass band; as soon as you have more than one, the SUM of those signals cannot exceed the saturation point of your ADC, each signal decreases the effective Dynamic Range of your system. But for most DDC receivers the basic dynamic range is below that of the above example. The ADCs of most DDC receivers are typically sampling at 16 bits or even less. Now compare and see how much effective Dynamic Range you lose just for sampling at the antenna with a large BW? Remember each individual signal SUMS and eats dynamic range, so a wide bandwidth will have more signals. As soon you are forced to use analogue down converters with mixers and Local oscillators to place the higher frequency within the operational BW of your DDC, the promise of sampling at the antenna just vanishes, introducing all the behaviour of traditional receivers with mixers. We are in a transitional period where hybrid systems (DDC + Zero IF Receivers) are gaining popularity due to the absence of affordable DDC chips that can operate at over 100MHz. Zero IF Receivers may be old, but are still alive, and they will remain useful for many years. Pure 24-bit sampling on frequencies above 100 MHz is only possible with DDC chips for military and commercial applications, because the cost per chip can exceed 5,000 USD, and they are not available in small quantities. But even with the poor dynamic range they can deliver they are able to sample many Megahertz in a fraction of a seconds. Is that really what you want for Weak Signal Communications like EME, Satellite or even a shortwave receiver?

4 An economic and affordable solution is to use Direct Conversion Receivers with a properly designed ADC and NOT with a PC sound card acting as an ADC. DDC receivers and Zero IF receivers were the subject of many comparisons in recent years: Leif SM5BSZ did a professional comparison, testing 3 well-known DDC receivers (Perseus, SDR-14 and SDR-IP) which showed good dynamic range. A similar comparison was done by Loftur TF3LJ/VE2LJX on a Softrock 6.3/Mobo/SDR- Widget Lite with a 24-bit audio card. Both comparisons showed very similar performance and both systems demonstrated better performance than a traditional Super Heterodyne Receiver. ABOUT DYNAMIC RANGE When I spoke about Dynamic Range in recent years some people told me. Dynamic Range is not a problem for me. and my answer was always good for you being a privileged EME operator. Dynamic Range is not the only value you analyse in a RX system but it s one of the most important, especially in today s congested world, where more and more Ham Radio operators locate antennas in cities and densely populated areas with tremendous man-made noise sources near/around the antenna. Just a few operators have the privilege to run EME antennas located in remote areas where a high Dynamic Range does not add any value to the system due to the absence of strong signals. Just to have an idea, here in South Africa, the SKA (Square Kilometre Array) radio telescope in the Karoo Valley exists It is equipped with hundreds of 12m offset dishes (when finished), running from 1GHz to 5GHz. The government, to protect this world-wide project, created an exclusion zone of several square kilometres around the antenna site where every single town, farm, industry, power line cable and train railways were removed and relocated, away from the exclusion zone. Even airplanes when they approach the site are forced to disable transponders and many electronic devices such as weather radars, etc. You cannot visit the site carrying any electronic device like 2-way radios, cameras, laptops or mobile phones they are strictly forbidden why?? Well the DDC converters on those antennas sample at very high speed and at high frequency, producing terabytes of data per day, BUT with an ADC sampling at just 8bits. Which means very poor Dynamic Range! But they don t need more due to the fact they run the installation in a protected area. Is your antenna located in a similar environment? Dynamic Range is very important and becomes more important every day!!

5 LIMITING FACTORS OF PC SOUND CARDS FOR ZERO IF RECEIVERS As I mentioned earlier, PC sound cards are just PC sound Cards and are not designed as Analog to Digital Converters for Zero IF SDR radios; they are designed for home or professional audio applications like voice and music in a very limited band width. PC Audio Cards, even if they can sample up to 192kHz, are designed to operate in a restricted bandwidth. The specifications only hold true for a few tens of khz, from 20 Hz to 20kHz to be precise. The reason for that is the human brain and ear response. Nothing above 22kHz can be heard by human beings. So, it doesn`t make sense to extend high tech specifications to a spectrum impossible to hear by the consumer. Nevertheless, the audio industry is pushing sampling rates up to 192kHz with the aim of reducing noise. This is one of the most polarizing topics within the audiophile arena. In the late 80 s clocking was one of the biggest problems for the primitive PC audio cards; with high jitter those cards introduced many errors in the ADC process. That was the time of 44.1kHz sampling; those cards also suffered from poor anti-alias filters used to remove inaudible supersonic frequencies to avoid artefacts which masked, harmed or even destroyed the audible spectrum. Even the move to 48kHz was not a solution until better filters were available, and which point supersonic intermodulation distortion was no longer a problem. A small increment of the sampling rate from 44.1 to 48kHz, in conjunction with better clocks and filters resulted in better audio cards. The extensive use of super high frequency equalizers, superfast compressors and electronic synthesizers (The 80s were the boom time for Electronic Music) justified the introduction of oversampling. Sample rates moved to 88.2kHz, when most PC audio cards started using oversampling for DSP applications and later were extended to 96kHz. BUT always the limiting factor of the human brain and ear resulted in a final BW of just 20kHz. Today 192kHz sampling rates in PC audio cards is getting more common but still under development. They have the same 20kHz BW limitation. Why would the industry develop chips with better flat response above 20kHz if the limiting factor is the human ear and brain? Waste of money and for sure SDR applications are never considered. Today many small audiophile houses are producing bespoke converters (AKA audio cards) with just 44.1 to 48kHz sampling with outstanding performance, much better than the most respected PC audio card brands that can sample at 192kHz. I will leave that discussion to the audio professionals, but it looks like having a human condition as a limiting factor, expressed in just 20 khz of BW, makes it pointless to produce high quality converters for higher sampling rates. If you have access to a 192kHz sampling audio card you will realize that the frequency response is terrible;

6 most of those audio cards maintain a flat spectrum for just 20kHz (the human limit) and after that the noise rolls up 20 db at the corners. It appears that we are confronted with the common idea that more is better, e.g. more pixels will result in better videos, more sampling speed produces better audio and your computer clock running at more Megahertz produces faster computers. Intel, after sales numbers decreased, abandoned the widespread idea that more Megahertz leads to faster CPU s, instead of that, Intel moved to moderate clock speeds and increased the numbers of cores per CPU, which creates really fast computers. But at that time the CPU industry was governed by the idea of More Megahertz more sales. Similarly, there exists a misconception about DDC receivers: Sampling at the antenna with a large BW, because more bandwidth is better. Completely wrong from the perspective of EME and Weak Signal Communications. LIMITING FACTORS OF EME SYSTEMS We can create a list of all subsystems in a typical EME station. The interaction of all of them will determine the final performance of the system and any improvement of one single subsystem will not necessary produce better overall performance. The limiting factor will change as soon as we start improving the performance of some subsystems: performance is like a moving target. Here a simplified list of common subsystems of an EME station, starting from the antenna to the operator: - Antenna (including a tracking system) - Pre-amplifier - RX cables and relays - RF Power amplifier (and TX coax lines) - Receiver (Transverter, superheterodyne, SDR, PC Sound Card) - Computer (with the associated software) - Operator experience Now think: If with that system we are unable to hear our own echoes, an easy approach will be to increase the RF power 3 to 6 db, or we double the size of the antenna, or we optimize the feed for better gain and/or lower noise, or we replace a lousy preamp with 0.7dB NF with a good one with just 0.2dB NF, or just replace a single piece of RG-8 in the RX system before changing the preamp for a better one. Or we change all of them. Many areas, different approaches, but now think if you just replace the radio; instead of and old RX receiver you decide to install a new Icom IC7300 with all the latest technology in his excellent receiver, BUT you keep all RX subsystems intact. Your antenna system has two sets of antennas, one for Vertical and one for Horizontal and you switch the RX from V to H manually. The result is unlikely to be any better than the previous system, and you already spent good money for the new radio for NIL performance increment. Instead of that, for less than 100 USD, you can install two Softrock receivers and setup Linrad in your PC, then you will defeat faraday

7 rotation by running an adaptive polarization system. The performance increment will be tremendous for just 100 USD with very little effort when compared with the expensive IC7300 which will add very little or nothing to your system. Optimization doesn t need to be expensive in terms of money; all you need is to be clever enough to identify the weakest points on your system and spend money on improving those areas. Just a drop in replacement is not always a good tactic, especially for systems that already have medium to good performance. If you already have an optimized system and you want to optimise every single subsystem, then you face a job of titanic proportions. This is especially true of the receiver subsystem. If your system is already well optimized, you run a good antenna array, for example 4 long yagis for 144 MHz, a good preamp, good quality cables and so on. Additionally, you are use an RTL or FDC dongle to feed MAP65 and constantly notice a large amount of interference in your pass band, sometimes in specific directions, then most probably you are running short of Dynamic Range (the USB dongles have moderate Dynamic Range and they overload very easy). After a proper assessment when you definitively identify your problem stems from a lack of Dynamic Range, you will need a better SDR with 20 or 30dB more Dynamic Range to solve the problem. WHY PC SOUND CARDS ARE NOT SUITABLE FOR ZERO IF RECEIVERS FOR ADC CONVERSION? Well, I have said they are but with limited performance. In the initial part of this document I explained the limitations and the reasons for them. These views were inspired by many documents and public information, but for better understanding I will invite you to read what Dan Larvy wrote about high sampling rates and audio cards. He is one of the most respected designers of audio converters in the world and a die-hard opponent of ultra-high sample rates. His observations are limited to the audio spectrum, but we can easily understand that the same theory governs the world of ultra-high speed DDC Receivers. The Nyquist Theorem is valid independent of the frequency and cannot be ignored. (link here) In summary Dan Larvy wrote:. Nyquist pointed out that the sampling rate needs only to exceed twice the signal bandwidth. What is the audio bandwidth? Research shows that musical instruments may produce energy above 20 KHz, but there is little sound energy at above 40 KHz. Most microphones do not pick up sound at much over 20 KHz. Human hearing rarely exceeds 20 KHz, and certainly does not reach 40 KHz. The above suggests that [even] 88.2 or 96 KHz would be overkill. In fact all the objections regarding audio sampling at 44.1 KHz are long gone by increasing sampling to about 60 KHz.

8 Now, how we can pretend PC sound cards are suitable for SDR receivers when they are optimised for use in a 20kHz bandwidth? The excess sampling capabilities in PC sound cards have the first 20kHz of response well designed but with the rest of the BW behaves like a noisy add-on with several limitations. We used and are still using PC sound cards for ADC conversion because they were the only affordable solution for decades, those expensive Audio cards still dominate nowadays, and they are a serious limiting factor in Ham radio SDR systems. For that reason, I decided to look for a proper ADC to do the job and keep audio cards to provide just PC sound, Skype or for listening to music. I have already retired my PC sound cards from the Radio station.

9 UADC4 DESIGN I started the process 2 years ago. As I had no experience at all in ADC design, in the beginning I asked myself where to start? Definitely any attempt to design my own ADC had to follow the basic rules of AD conversion, and learning those rules was a tedious job with extensive reading and testing. Leif Asbrink SM5BSZ played a key role by teaching me the many aspects of AD conversion. After understanding the principles of AD conversion, I was looking for the right ADC device. Then I was confronted with the reality of how difficult the selection could be. Another important consideration was how to stream the digital data into the computer without having to invest too much effort in extensive programming. The use of FPGA chips was rejected due the complexity and cost. Due to the low frequency involved, I was forced to examine Audio ADC chips with all pros and cons already previously explained. I narrowed my choice down to the following 4 manufacturers: - Analog Devices - Texas Instruments - Linear Technologies - AsahiKASEI I choose AsahiKASEI who has proven experience in ADC chips with outstanding (audio) performance. The selected chip, the core of the UADC4, is the AK5574, a 4- channel differential 32bits ADC capable of a sample rate of up to 768kHz and up to 121 db Dynamic Range in single mode and up to 127 db in 4 to 1 configuration. I decided to use the AK5574 based on its flat response above 20kHz (where most of the audio chips have a terrible increase in the noise floor). Compared with other ADC s, the AK5574 has a flat response above 40kHz to 96kHz, after that the noise

10 floor increases rapidly as it approaches 192kHz. Above 192kHz the chip suffers the same flaws as other brands; the noise floor is no longer flat at all. The use of a preamp in front of your SDR receiver and Linrad calibration will take care of any noise floor slope, but I was looking for a device with better intrinsic characteristics. The AK5574 can select the sampling speed without any special programming, the chip locks its sampling speed according to the software request you use to open the data stream. Linrad opens with a range of sampling speeds up to 192 khz with no problem and MAP65 opens data stream at 96 khz with no programming intervention. The same is true of HDSDR. The output data format supports 24/32-bit MSB, justified, I²S or TDM in PCM mode; in DSD mode it supports DSD native 64, 128 and 256. Additionally, the AK5574 has internal anti-alias filters; compared to other chips the filters have a good response. You can configure the PCM roll-off filter response with a single jumper between sharp and slow and you can configure the PCM digital filter delay between short and normal. In the best configuration at fs=96khz the passband is selected at 48.8kHz with -82dB Stopband KHz.

11 THE BUFFER STAGE by THE MANUFACTURER Once the ADC chip was selected I tried to understand why almost all ADC chips available for audio use suffer from an asymmetrical noise floor. I did measurements on many audio cards with different ADCs and I discovered peculiar buffer stages in the input circuit. Independent of the internal configuration of those chips, where apparently, they concentrate the perfect response in just 20kHz (others, like the AK5574, extend excellent performance up to 96KHz) the buffer stage has a lot of influence on the final performance. Referring to the manufacturer s documentation, this is the buffer stage proposed for the AK5574: This buffer stage is definitely not suitable for our purposes, it is fine for audio and music but has many pitfalls for an SDR application. - The NJM5534 is an opamp with an input noise voltage of 3.3nV/ Hz; this can be improved with a better opamp. - The AK5574 requires a differential input but this buffer is not really a differential amplifier. I did hundreds of hours of simulations in LTSpice and the output amplitude, in this configuration, has an asymmetrical amplitude at around 80 mv when measured at pins AINn+ and AINn-. To properly feed the ADC, both signals must have the same amplitude, no differences at all are acceptable. - The first opamp causes a delay in the signal, as you can see, when JP1 and JP2 are in place, the output of the first opamp goes to the input of the AIN+ opamp, but the AINn- opamp receives the signal without any active device, direct from the input source. Even if the first opamp has his source resistor and feedback resistor in a ratio of 1:1, there is a delay in the signal, which is totally unacceptable for an SDR application. - The values proposed for the source and feedback resistors are too high. The higher the values, the higher the noise caused by the resistor (thermal noise), but for audio applications it appears that the noise is acceptable, but not for SDR.

12 - Last and not least, this pseudo single end input to differential buffer has no limiting control in the input. It is probably expected that the designer will add some kind of potentiometer or digital resistor at the input to control the amplitude either from the front panel or via software. However, the opamps supply rail is +/-15VDC, which means that the output of the opamps can swing the entire rail to rail voltage, thus the risk of overloading and destroying the ADC is very high. For sure you can add, as I said, some kind of amplitude control at the input but this will add more noise to this already noisy buffer. THE UADC4 BUFFER STAGE (Sorry that the circuit diagram is not provided because there are clones and illegal copies appearing on ebay. These are mainly from Asia, the latest scandal is with the mchf SDR radio, have a look at Google.) To correct all the problems in the reference design buffer a new buffer suitable for our SDR application was designed: - The NJM5534 opamp was discarded and two new devices from Linear Technologies were selected. These devices have a lower input noise voltage of just 1.0nV/ Hz - The first device now works as a limiter to prevent ADC destruction due to excessive input amplitude. - For the second opamp a real differential amplifier was selected to correct the asymmetrical output seen in the previous reference design. A low pass filter was also incorporated to reduce frequencies above 100kHz - The values of the feedback and source resistors were selected to minimise noise contribution. - The resistors are high quality parts with tolerances expressed in ppm, equivalent to less than 0.2%. - A simple jumper matrix is used on the input. This can accommodate input values from 0-14 Vpp in 5 steps and completely avoids any noisy potentiometer or digital resistor. - To set the correct input gain, the user will hard wire the input using jumpers and replace the lid. For example, with the jumper in position 1, the maximum input will be 3.5 Vpp producing 2.8 Vpp at the output. If more level is applied the first opamp will saturate, the output will be distorted but the amplitude to the ADC will never exceed the 2.81 Vpp, protecting the ADC from any damage. THE CLOCK One of the most common problems with PC Soundcards is poor clock quality. Normally the clock is generated by cheap oscillators with high phase noise and poor

13 stability. This may be acceptable for music but for SDR applications low PN is very important to keep reciprocal mixing as low as possible. Any device with a phase noise greater than is not good enough. The UADC incorporates a very precise and low jitter clock. The internal clock in the UADC4 is the GXO-7506, it s an Ultra-low jitter Oscillator. It is used by some professional audio cards but it is uncommon in consumer PC audio cards. At 1KHz separation the GXO-7506 has an extraordinary performance with just - 155dBc/Hz. At 100KHz separation the phase noise is only -174dBc/Hz. The RMS phase jitter (from 12KHz to 5.0 MHz) is in the range of 0.04ps (pico seconds). The ADC As previously stated because we are dealing with base band signals in the range of 1Hz to 192kHz a commercial ADC used for audio applications was the best choice. The selection was rigorous and several different candidates were tested. The AK5574 from AsahiKASEI meets the design criteria, particularly with respect to the flat response beyond the traditional 20kHz umbra. To recap, up to 96kHz the noise floor response is very flat with only a small increase between 96 and 192KHz. If you want to read detailed information about the AK5574 here is a link to the datasheet: Internal power supplies The UADC needs several different voltages to work properly, the primary input voltage is in the range 11 to 14.5VDC but internally different voltages are used for the different stages. Linear regulators were rejected as contrary to popular belief they produce noise at different frequencies. The voltage regulators are switching types and implementation is based upon tried and tested ADC reference designs from Linear Technologies. The use of switching regulators for this application is not easy but when properly implemented they deliver good level of current with very low noise. ADC systems are particularly sensitive to ripple and switching oscillator noise which normally enters the system via the DC path. This dictates the use of high quality capacitors with extremely low ESR and carefully selected inductors for the switching oscillators. Additionally, every single DC path is extremely well decoupled to its respective ground plane to avoid artefacts entering via the DC path. For optimal DC decoupling it is imperative that the analogue and digital ground planes are segregated. This segregation means that separate grounds on the PCB

14 are mandatory. Adequate performance is only possible with a 4-layer PCB and not 2 layers as used by most of the consumer audio cards. The PCB layout If the success of an electronic design was just a matter of picking and placing the right parts it would be very easy to produce high quality devices. For high performance ADC applications, no other part has more impact on the performance than the PCB layout. Firstly, we need to follow the layout recommendations of the ADC manufacturer and secondly, we must avoid interference from the many signals that are present on the PCB. The UADC PCB is a 4-layer PCB that utilises micro vias to pass signals between layers. By far the most critical aspect of the PCB is the definition of the Analog Ground and the Digital Ground, they must be isolated BUT, in the end, connected at just one single point to avoid ground loops. An optimised 4-layer PCB is not easy to realise and demands a lot of testing. Four PCB iterations were required before optimal performance was obtained. One of the most difficult problems to resolve is the magnetic coupling between traces. Poor isolation causes signals to be induced on to circuits where they shouldn t be, reducing the channel separation and dramatically increasing the reciprocal mixing in the SDR. This is one of the most common problems with PC audio cards and why they don t reach a high level of performance. These induced currents are very sensitive and difficult to reduce, especially in the input buffer. Just moving a capacitor few millimetres in one direction can represent an improvement or degradation of performance by several db. Basic PCB design rules were followed but, in the end, optimisation is a tedious trial and error task which obviously has an impact on time and costs.

15 This is proto1 PCB, was rejected due to the presence of magnetic loops in the input stage The XMOS USB interface It is one thing to convert base band signals to digital data but an entirely different thing to present that data in a format easy to manage by computers and software. We evaluated several different types of interface, like Ethernet, SPDIF (optical) and USB. We found that the USB interface was the most convenient solution. Ethernet was not considered due to the inherent complexity as well as potential timing problems. Timing errors are particularly problematical for adaptive polarization receivers. An optical interface was rejected due to cost and low popularity. After all, only a moderate bandwidth is required. The USB port is an excellent interface for the required data rate and in order to speed-up our design phase the DXIO32ch produced by DIYINHK which is based on the XMOS XU was used. With 2000MIPS in dual issue mode it can work up to 384KHz with excellent performance and ZERO impact on the signal quality. Simply request USB.2.0 and the driver can deliver up to 4 analogue channels at up to 384KHz, 6 channels up to 384KHz in I2C mode and up to 8 analogue channels with appropriate firmware. Maybe overkill but at an extraordinarily competitive price with top performance. We just plugged the XMOS board into our prototypes and the digital stream from the ADC was available to the PC without investing the time and money required for a proprietary design. The board is delivered with an ASIO driver, which is an excellent way to manage low frequency data with extremely low latency.

16 UADC4 Design process, Performance and production During the initial phase of this project I began with the assumption that PC Audio cards are the limiting factor for ZERO IF receivers. We can do better. The numbers don t lie, a 15 to 20dB performance increase was expected and the initial tests clearly demonstrated and confirmed my hypothesis. The UADC4 was born as a decentralized design, with work done in Canada, Sweden and Switzerland with central coordination in South Africa. The Canadian part was operated by George Boudreau, a well renowned Electronic designer with a lot of experience in SDR radios. He produced the famous Widget SDR radio and several DAC's like ODAC as well as USB interfaces for audio. In Sweden Leif Asbrink SM5BSZ gave the benefit of his 50 year experience to the UADC4 project. Leif is a highly reputable and rigorous engineer who designed the famous SDR WSE units many years ago. He is THE worldwide leader in SDR HW performance tests and also developed the excellent SDR software LINRAD. Last but not least I am responsible for conceptualization, the prototype production and performance tests together with Leif, using experience accumulated in the last 10 years. We are all very proud of results obtained with the UADC4. I want to emphasise that Leif has no commercial links at all with the UADC4 or LinkRF, he is just interested in high performance hardware. Any endorsement from him is because the UADC4 meets his expectations. You can obtain more details directly from him. He will speak freely and independently about UADC4 performance, if you are interested contact him directly or follow the tests on his YouTube channel.

17 At the end of the document you will find a table with all the data collected by Leif SM5BSZ where he compares many PC Soundcards with a Softrock SDR used as a ZERO IF mixer in front. The UADC4 is located in 1 st place in the table with outstanding results and the only reason performance was not even better is because now the limiting factor is the ZERO IF mixer. A better mixer will produce much better results, as you can see the performance reached by the WSE units in one of Leif videos. The WSE units were designed 20 years ago and Leif has never previously been able to realise the level of performance he does when using UADC4 as dedicated ADC. Here some statements made by Leif after the final UADC4 prototype was tested: Mainstream development is towards direct sampling in the VHF range and many regard units like the Softrock Ensemble more like a toy than a real radio. Major improvements are possible however. Alex, HB9DRI, is working on a 4 channel ADC for usage with direct conversion SDR. The noise floor should be something like 15 db lower than the best soundcard in the table below when used with a Softrock having the Si570 replaced by a good crystal oscillator. The WSE RX2500 was designed in year 2001 with components of that era. The mixer uses 74HC4052 with about 70 ohm switch resistance as compared to 4 ohms in the FST3253 used in the Softrock Ensemble. Yet, mixer performance is dramatically better as you can see here: uadc4wse Noise is 8 db lower and harmonics -70dB as compared to -45 in my most modified Softrock. What I see is reciprocal mixing noise floor@10khz = -156 dbc/hz. That is something like 10 db better than the Perseus. It uses the UADC4 much better than the Softrock Ensemble, but improvements are still possible. The very sharp anti-alias filtering provides some noise and a too high gain. By running the UADC4 at 192 khz one could move the anti-aliasing into the digital domain and get a significantly better performance. A low Q anti-alias cutting at 96 khz would not have to add gain or noise. Further, the UADC4 provides some filtering above 96 khz. Popular trends today say direct sampling is SDR for the future. The UADC4 demonstrates that direct conversion today can be better. The direct conversion architecture of Softrock would need a lower noise LO, a better mixer and good front end filtering to become better than the best direct sampling radios. They also need good RF filtering.....end of Leif SM5MSZ quote We can see a dramatic improvement in noise floor with the UADC4. If we compare the combination of a WSE and Delta44 the best noise floor obtained was -145dBc/Hz at 1KHz separation. The WSE-UADC4 combination attains an outstanding noise-floor of dBc/Hz at 1KHZ, this is an improvement of 17.8dB in the noise-floor, a tremendous performance increase by just replacing the Delta 44 audio card with the UADC4 In terms of harmonics with the receiver close to saturation the WSE-Delta44 combination was -45dBc and the WSE-UADC4 combination was -70dBc up to the 4 th harmonic and better than -85dBc for the 5 th harmonic and above. Leif SM5BSZ did an extensive test using a SoftRock Ensemble with the UADC4 and the result was also impressive, you can directly review the comparison here:

18 The SoftRock Ensemble with the UADC4 is in 1 st place compared to the most popular soundcards and as Leif explains during his YouTube channel test the performance can be even better if the mixer is improved. Another good example of the excellent performance was obtained when I interfaced the UADC4 with the IQ+ revb. An unmodified IQ+ when used with the UADC4 will not result in major improvements. The IQ+ must be modified because it was designed to work with sound cards with a much higher noise floor. In order to exploit the excellent performance available with the UADC4 the analogue gain of the IQ+ must be reduced by around 20dB, this moves the saturation point from -26dBm to - 9dBm approximately. This means that if the IQ+ is modified and interfaced with the UADC4 it will tolerate a 17dB stronger signal than the normal IQ+ with sound card. Regarding reciprocal mixing we observe no contribution from the UADC4, that moves this weak point to the local oscillator used by your mixer. As already started the performance of Zero IF SDR radios is a moving target. For years the limiting factor was the soundcard, and this prevented better IQ mixers from appearing on our bench but now with a better ADC replacing the lousy soundcards the limiting factor is the IQ mixer and the Local Oscillator. Areas where it is possible to make big improvements.

19 Interfacing the UADC4 with your IQ mixer Three different IQ mixers in front of the UADC4 were tested and in all 3 cases much better performance was obtained than when those mixers were used with the best sound card. WSE converters: These units were designed 20 years ago, and have some limitations, nevertheless by adding a 10dB attenuator at the input of the WSE RX2500 the noise floor improves and overall performance improves by more than 10dB. This performance was never previously attained, even with the expensive EMU1616 soundcard, one of the best. For more details about the test please refer to Leif s YouTube Channel. Ensemble SoftRock: Leif has also published a video where you can see how a SoftRock performs with the UADC4. He proposes some modifications to reduce the analogue gain to achieve better performance with the UADC4. IQ+ rev A and B: As previously explained IQ+ RX boards need to be modified for best performance with the UADC4, the modifications are simple and straight forward. Only the removal of 3 resistors and the replacement with 4 resistors is required to reduce the analogue gain by approximately 20dB. Now the saturation point improves from -26dBm to - 9dBm. This shows that the dynamic range of the LT5517 was under-utilised. At the time the IQ+ was designed this amount of gain was required to overcome the higher noise floor of then available soundcards. Using the UADC4 it is possible realise a 17dB improvement. After the modification the noise floor improves from -141dBc/Hz@1KHz to an impressive -163dBc/Hz@1KHz. However, the phase noise of the IQ+ LO (si570) limits the realised performance improvement to approximately 11dB. What you can expect if you just replace your sound card with the UADC4? The honest answer is that a simple drop-in replacement will give you just a small increase in performance. In order to take advantage of the increased performance of the UADC4 you must modify your hardware. The modifications are simple, step one is just a matter of reducing the analogue gain after the IQ mixer. If you also improve the phase noise of your LO then you will reach performance levels never previously achieved. The UADC4 opens a new frontier for ZERO IF receivers at performance levels that a DDC radio cannot compete with due to limited sampling capabilities and excessive bandwidth. This is not optimal for EME and other weak signal communication techniques.

20 The Future After the success of the UADC4 future plans are already on my bench. Independent of the IQ+ modifications to exploit the UADC4 performance Leif, George and I are convinced that there is potential for an even better IQ mixer designed specifically for the UADC4, so we have started to work on it. It will take considerable time before we attain our goals, however a new generation of IQ mixers already exists. These 5mm x 5mm packaged devices have a coverage from 20MHz to 1.4GHz. Additionally, a new generation of local oscillators with lower phase noise than ever previously available are planned. These LOs will achieve performance that DDS and PLL s cannot compete with. The core of the new IQ+ generation will be the UADC4 and the new mixer which will push the UADC4 to its limits. The final unit will be known as the IQ+Pro. It will be at least 2-4 years before the IQ+Pro reaches the production phase. Meanwhile the UADC4 will give you extraordinary performance, and for those with IQ+ receivers (+300 users) the deployment will be straightforward with minor modifications. The same applies to those who use Softrock units. The real challenge for our community is to design independent IQ mixer designs based on the UADC4. Sound cards are for music the UADC4 is for SDR radios. Stay tuned!! 73 de Alex, ZS6EME (HB9DRI) hb9dri (at) emeham (dot) com

21