Last night I asked Warren the same question, and here is his answer. My assumption was the User Interface / LCD real estate was a major issue. Here is Warren’s answer. 73, Rob
As to how many hams make use of more than two slices or receivers, I don’t think there are very many. Adding slices/receivers is not a frequent request from the openHPSDR operators. When they see it, they think it’s kind of cool; but, for the most part, they don’t know what to do with them.
For the IC-7610, one would need to be sure the FPGA was big enough and that there was enough processing power in the DSP chip to add more slices. However, for a radio with knobs, I suspect the bigger challenge for more slices would be the user interface. For example: Does the display have enough screen real estate for more panadapters? Where do you put more knobs and buttons to control the additional slices, or, do you somehow share the ones that are there among multiple slices, etc.?
Yes, the Apache products provide more than one slice per ADC. However, only two complete software receivers and one sub-receiver are currently exposed to the operator by the UI. If the radio hardware has only one ADC, both receivers get “slices” from that ADC. If the hardware has two ADCs, the operator has the option to use one ADC for one receiver and the other for the other receiver.
In our new beta console version, Thetis, all the underpinnings are there to offer almost an unlimited number of slices. However, the user interface for that is not done currently. It’s been tested with a proof-of-concept version that works fine.
From: Elecraft-K3@groups.io [mailto:Elecraft-K3@groups.io] On Behalf Of Jeffrey griffin
Sent: Sunday, May 05, 2019 6:28 AM
Subject: Re: [Elecraft-K3] SDR
Thanks Rob, and thanks Warren for the clear explanation of what a slice is. Now what is the difference between the Icom 7610, and the Flex 6000 series? The flex 6000, except the 6300 can get four ‘slices’ per their Spectral Capture unit, while the 7610 manages only one? The Flex doesn’t seem to lose any performance advantage while doing this, so what is the difference hardware wise between the two?
73 Jeff kb2m
From: Elecraft-K3@groups.io <Elecraft-K3@groups.io>
On Behalf Of Rob Sherwood
Jeff. As promised, Warren has some answers for you. Hopefully this helps. We all get to learn. Rob, NC0B
There are many questions embedded in this discussion! I hardly know where to start. However, let me try to cover some of the basics about “ADCs”, “FPGAs”, “slices”, and a few other things.
First, the job of an ADC is conceptually very simple; although, it’s important that it does its job accurately. An ADC can be viewed as a digital voltmeter. It repeatedly and periodically measures the voltage presented to its input and reports that as a digital value. Think of your handheld digital voltmeter that periodically updates the reading that’s displayed.
Second, we have to take note of a very important principle, the Nyquist-Shannon Sampling Theorem. This theorem states that the faster the ADC repeats its voltage measurements, the more bandwidth it can capture. More specifically, the amount of bandwidth it can capture is equal to one-half of the rate at which it samples the signal (measures the voltage). So, if we had an ADC running at 122.88Mhz (measuring the voltage 122.88 million times per second), it could accurately and completely capture all the signals/information in a bandwidth of 61.44 Mhz.
If we have an ADC running at 122.88 Mhz, it can capture all the ham bands up through 6M at the same time. However, a data rate of 122.88 million samples per second is just too much for today’s conventional CPUs to absorb and process. That’s where the FPGA comes in. (Today’s direct-sampling radios such as the Apache-Labs ANAN series, Flex Radio Systems, ICOM 7300, etc., use an FPGA.) An FPGA is difficult to program and is not suitable for many general algorithms; HOWEVER, it can process data very fast. So, the FPGA is used to accept the very high-speed data from the ADC and “decimate” it to lower sample-rates that can be processed by conventional CPUs.
OK, this “decimation” to lower sample-rates is interesting. So, we can have a sample-rate of 122.88 million samples-per-second going into the FPGA. The FPGA may put out samples at a rate of only, for example, 768,000 samples-per-second by using this “decimation” process. The output rate of 768,000 samples-per-second is something that a conventional CPU can further process. However, again taking note of the Nyquist-Shannon Sampling Theorem, we must conclude that this output data stream can only transfer one-half of that, or, 384 Khz of bandwidth. That’s not so bad! 384Khz can encompass a significant part or all of each of the various HF bands. We’ll call that 384Khz of bandwidth a “slice”.
FPGAs are usually “large” enough that they have enough programmable elements that they can do this “decimation” process multiple times in parallel. So, using the same 122.88Mhz input rate (61.44Mhz of bandwidth), one decimation process could pick out part of the 80 meter band, another the entire 40M band, another the 20M band, etc., all in parallel. So, we can receive multiple “slices”, all from the same ADC. Then it’s just up to the software and user interface to make that multiple receiver functionality available to the operator.
Multiple ADCs. There are times when more than one ADC can be advantageous in a radio. These often relate to mitigation of interference. I’ll give two examples; however, there are other cases.
Example 1: ADCs can be overloaded, just like your favorite digital voltmeter. Suppose you want to simultaneously receive signals on 80M and 10M. However, you have a next-door neighbor who is transmitting on 40M. One way to deal with this is to connect your 80M antenna through an 80M bandpass filter to the first ADC – that works for the 80M reception. Then, connect your 10M antenna through a 10M bandpass filter to the second ADC. In both cases, the 40M interfering signal is filtered-out so it can’t overload an ADC.
Example 2: Two ADCs can be connected to two different antennas. We then “mix” the two outputs, on the same frequency, to null interference being received from a particular direction. This is often called “beam-forming” or “diversity reception”.
I don’t know that I can adequately answer your question as to programming an ADC. I am forwarding this thread to NR0V who programs all the DSP for Apache products. I would assume the speed (and cost) of an ADC chip affects how much horsepower is available to do multiple jobs at the same time. I’ll report back with a more precise explanation when I receive it.
Both the Flex and new Icom radios are direct sampling, so there is no debate there. Icom chose to use one ADC per receiver, be it a 7300 with one ADC chip, or the 7610 with two ADC chips and two receivers. The Flex 6700 has two ADC chips, and can do four slices per chip, or a total of 8 slices. Your 6500 has one ADC chip and as you say can handle up to 4 slices. The 6300 has one chip but can only do 2 slices with a slower less expensive ADC chip. The 6700 had several very expensive parts, and was a $7500 radios when it was introduced. On the other hand a friend of mine priced out the cost of the parts in an IC-7300 and said no electronic part cost over $20.00 in quantity. For $1000 the 7300 is quite a nice radio, and Icom has sold zillions of them. [Well maybe only (?) 30,000.]
It would seem to come down to a price performance ratio as to how much money to allocate to the ADC chip. There are also one or more FPGA chip or chips, and again there is added cost for a faster and larger FPGA. There is a thread on the Apache reflector talking about the 10E and 100D running out of space in the FPGA and barely being able to run their new operating system that is under development. The FPGA is 97% full, so choices had to be made as to what features could be offered.
My friend with a 6700 uses 7 slices to monitor 7 MARS frequencies at once. On the other hand I rarely use my 7610’s second receiver unless I am trying to work a DXpedition running split. During 2018 CQ Worldwide I only worked three stations split out of 542 contacts. Proficient SO2R operators likely make much more use of the second receiver or possibly a third or fourth receiver. I don’t multi-process well, so I leave that to others.
The new Yaesu FTdx-101D has two receivers, but they are down-conversion superhets like the K3 / K3S with the second receiver option. As I mentioned earlier, in really rough RF environments a radio like the K3S with its mode specific narrow roofing filters can handle very strong signals on the same band better than any direct sampling radio. Consider the front end L/C filters in a Flex, Icom and Apache direct sampling radio as the “roofing filter”, but that filter is as wide as each individual band. The ADC has to handle a strong signal or multiple strong signals even if these other signals are hundreds of kHz removed from a desired weak signal. Most of the time overload isn’t the performance limit for reception. My nearest ham is 13 miles away!
The operator of any direct sampling radio needs to handle total gain more carefully than the operator of a K3, TS-890S or FTdx-101D. By that I mean keep the preamp OFF unless you really need it. Use attenuation as appropriate, which means there is no reason to have band noise reading upscale on the S meter. Of course this is true of any radio if it is operated at its maximum capability. For instance during a 160m CW contest I run between 12 and 18 dB attenuation at night with any radio I own, direct sampling or superhet. I want band noise to be about 6 dB below the AGC threshold to significantly reduce operator fatigue.
If I get more answers, I’ll respond again. 73, Rob, NC0B
Hi Rob, always good information in your posts. Can you explain the differences, as you see it between the direct receivers from Flex radio vs the Icoms? I have a Flex 6500 and on occasion use the 4 available slices. I find them very useful. I don’t understand why the 7610 can receive only two separate signals. There must be a difference in how the direct receive frontend is done by both manufactures. Do you have any input on this?
Oh, and to keep this on subject I also have a K3 I bought in 2007, I have upgraded to K3S specs, I would never sell…
73 Jeff kb2m
From my perspective, the term SDR (software defined radio) today is virtually meaningless. A K3S is a superhet down-conversion radio with lots of DSP software programming in the tail end of the radio. The TS-890S is also a down-conversion superhet, but with a direct sampling bandscope and waterfall display. It also has software programming that can be upgraded by a firmware flash. The newly released FTdx-101D is a hybrid like the TS-890S.
The Flex transceivers are direct sampling radios, as are the IC-7300, IC-7610 and IC-9700. There is an embedded OS in the Icom radios and lots of firmware upgradable software.
All the Apache radios are direct sampling, and some of the software is in the radio, but a lot of the processing is in the PC.
I am not a programmer, so I don’t know if it is possible to write code to run on Windows, Apple and Linux platforms without separate binaries.
If you are looking for an Elecraft “SDR”, I assume you are meaning a direct sampling radio. Direct sampling has advantages and disadvantages. Most of the time the architecture doesn’t matter much, but if you are operating Field Day, or have another ham a mile away, a down-conversion radio like the K3S will not overload nearly as easily as a direct sampling radio, if both stations are on the same band at the same time. On the other hand, the contest DX K3S package is on sale for $5959.95, while an IC-7610 cost between $3000 and $3100. The direct sampling radio has a cost advantage, and the hybrid TS-890S and FTdx-101D are both under $4000. These are all great radios, but there are times with the K3S wins hands down. (I am assuming the FTdx-101D will test out well, and I’ll know that in a week.)
73, Rob, NC0B
I wonder if Elecraft is ever going to replace the K3S with a true SDR radio that provides the processing in the rig and uses the front controls and or a low power computer to interface with it. Similar to what Flex is doing. I would also
like the PC software to be generic platform independent so that it can run on Windows, Mac, or Linux. I'd love to see Elecraft join the competition or maybe even jump ahead of the SDR competition. It would also be nice to have a SDR that was STABLE in the
VHF/UHF bands, unlike the new Icom 9700 which drifts and can't seem to stay on frequency.