Elecraft

K4 DSP bit width

Classic

List

Threaded

5 messages Options

Bill Frantz

K4 DSP bit width

A friend and I were discussing the K4HD and he said that it
sounds a lot like the K3S. Well, we both agreed that since the
K3S has world class receiver specs, that would not be a bad thing.

I started comparing the information about the K4HD with the K3S
manual and found:

K3S K4HD Feature
8.215MHz ~8MHz First IF
15KHz ? Second IF (if used in the K4)
24 16 ADC width
~30KHz? 122MHz ADC sample rate (WAG for the K3S)
5 2+1 Crystal filters (one empty slot is from a post
on this list.)

Note: In the K4 FAQ there are forward references to a K4HD
section, but I couldn't find it on the web site.

Analysis and wild speculation

Both DSPs and ADC technology have had at least 10 years to
improve between the K3S and the K4. WHile I don't track ADC
tech, DSPs are computers and we all know what has happened in
computers. The basic processors are somewhat faster, and there
are many more cores on a chip. The K4 certain to have taken
advantage of these improvements.

The K3S uses a 32 bit floating point DSP, and I don't see any
reason to change that specification for any of the K4 models.
Not changing means that much of the K3 DSP code should easily
port to the new DSP.

I don't understand how the basic K4 can get good dynamic range
with a 16 bit DAC. The K3's 24 bit DAC seems a better choice,
although getting high speed and wide bit width at the same time
is hard and expensive.

In the wild speculation department, there seem to be several
approaches for the K4HD. (1) Run a K3 like superhet with an
~8MHz IF and digitize that into the DSP. (2) Duplicate the K3's
2nd IF and use the 16 bit ADC at 15Khz. (3) Use a wider ADC at
either 8MHz or at 15KHz. Note that one of the features of the K4
is ease of upgrade to new ADCs.

73 Bill AE6JV

---------------------------------------------------------------------------
Bill Frantz |"Web security is like medicine - trying to
do good for
408-356-8506 |an evolved body of kludges" - Mark Miller
www.pwpconsult.com |

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

Jeff Scaparra

Re: K4 DSP bit width

Comparing a 24 bit ADC at a 32 KHz sample rate to a 16 Bit ADC at 122 Msps
and saying it is less bits and therefore will have worse performance isn't
a great comparison. The high speed ADC would be highly oversampling the
signal at the second IF. This allows the ADC to use some neat tricks to
effectively get more bits.

https://en.wikipedia.org/wiki/Oversampling
https://www.silabs.com/documents/public/application-notes/an118.pdf

I don't think we can determine theoretically if the K4 can perform better
or worse than the K3s from the current data that we have.

On Wed, May 29, 2019 at 12:08 AM Bill Frantz <[hidden email]> wrote:

> A friend and I were discussing the K4HD and he said that it
> sounds a lot like the K3S. Well, we both agreed that since the
> K3S has world class receiver specs, that would not be a bad thing.
>
> I started comparing the information about the K4HD with the K3S
> manual and found:
>
> K3S K4HD Feature
> 8.215MHz ~8MHz First IF
> 15KHz ? Second IF (if used in the K4)
> 24 16 ADC width
> ~30KHz? 122MHz ADC sample rate (WAG for the K3S)
> 5 2+1 Crystal filters (one empty slot is from a post
> on this list.)
>
>
> Note: In the K4 FAQ there are forward references to a K4HD
> section, but I couldn't find it on the web site.
>
>
> Analysis and wild speculation
>
> Both DSPs and ADC technology have had at least 10 years to
> improve between the K3S and the K4. WHile I don't track ADC
> tech, DSPs are computers and we all know what has happened in
> computers. The basic processors are somewhat faster, and there
> are many more cores on a chip. The K4 certain to have taken
> advantage of these improvements.
>
> The K3S uses a 32 bit floating point DSP, and I don't see any
> reason to change that specification for any of the K4 models.
> Not changing means that much of the K3 DSP code should easily
> port to the new DSP.
>
> I don't understand how the basic K4 can get good dynamic range
> with a 16 bit DAC. The K3's 24 bit DAC seems a better choice,
> although getting high speed and wide bit width at the same time
> is hard and expensive.
>
> In the wild speculation department, there seem to be several
> approaches for the K4HD. (1) Run a K3 like superhet with an
> ~8MHz IF and digitize that into the DSP. (2) Duplicate the K3's
> 2nd IF and use the 16 bit ADC at 15Khz. (3) Use a wider ADC at
> either 8MHz or at 15KHz. Note that one of the features of the K4
> is ease of upgrade to new ADCs.
>
> 73 Bill AE6JV
>
> ---------------------------------------------------------------------------
> Bill Frantz |"Web security is like medicine - trying to
> do good for
> 408-356-8506 |an evolved body of kludges" - Mark Miller
> www.pwpconsult.com |
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

W2xj

Re: K4 DSP bit width

In reply to this post by Bill Frantz

The K4 is a direct sampling radio. It has no IF. It samples directly to IQ baseband and performs all filtering and processing in the digital domain. That requires a very high speed ADC. In this case it is just above the Nyquist limit for 6 meters. Getting a 16 bit converter that fast approaches the limit of available technology. These converters are pipeline architecture and have better than 90 db dynamic range. Looking at S units, there is 88 db from S1 to 40 over S9 before any attenuation or AGC is applied prior to the ADC. The problem comes when there are signals exceeding that dynamic range like working someone who might be S1 or S2 while someone else is transmitting nearby as a direct conversion radio has no front end selectivity. As Eric already explained, this is the reason for the K4HD option. This appears to essentially be a K3S superhet single conversion front end that feeds an approximate 8 MHz IF into the ADC. It would then provide a very narrow band reception o
ption for the K4.

Looking at K3s architecture, the ADC is a PCM1804 which is a delta sigma converter which is basically a one bit ADC clocked at 768 times the sample rate or somewhere between approximately 24 and 36 MHz depending on what sample rate Elecraft selected. In the mode used in the K3S it can sample up to 48 KHz although there are other modes permitting up to 192 KHz. The one bit sample is decimated to 24 bits. This ADC is a fairly typical high end stereo audio converter.

The two approaches are very different and each fulfills a different need. Fortuneately, the K4 can provide both with the right options installed.

Sent from my iPad

> On May 29, 2019, at 1:08 AM, Bill Frantz <[hidden email]> wrote:
>
> A friend and I were discussing the K4HD and he said that it sounds a lot like the K3S. Well, we both agreed that since the K3S has world class receiver specs, that would not be a bad thing.
>
> I started comparing the information about the K4HD with the K3S manual and found:
>
> K3S K4HD Feature
> 8.215MHz ~8MHz First IF
> 15KHz ? Second IF (if used in the K4)
> 24 16 ADC width
> ~30KHz? 122MHz ADC sample rate (WAG for the K3S)
> 5 2+1 Crystal filters (one empty slot is from a post on this list.)
>
>
> Note: In the K4 FAQ there are forward references to a K4HD section, but I couldn't find it on the web site.
>
>
> Analysis and wild speculation
>
> Both DSPs and ADC technology have had at least 10 years to improve between the K3S and the K4. WHile I don't track ADC tech, DSPs are computers and we all know what has happened in computers. The basic processors are somewhat faster, and there are many more cores on a chip. The K4 certain to have taken advantage of these improvements.
>
> The K3S uses a 32 bit floating point DSP, and I don't see any reason to change that specification for any of the K4 models. Not changing means that much of the K3 DSP code should easily port to the new DSP.
>
> I don't understand how the basic K4 can get good dynamic range with a 16 bit DAC. The K3's 24 bit DAC seems a better choice, although getting high speed and wide bit width at the same time is hard and expensive.
>
> In the wild speculation department, there seem to be several approaches for the K4HD. (1) Run a K3 like superhet with an ~8MHz IF and digitize that into the DSP. (2) Duplicate the K3's 2nd IF and use the 16 bit ADC at 15Khz. (3) Use a wider ADC at either 8MHz or at 15KHz. Note that one of the features of the K4 is ease of upgrade to new ADCs.
>
> 73 Bill AE6JV
>
> ---------------------------------------------------------------------------
> Bill Frantz |"Web security is like medicine - trying to do good for
> 408-356-8506 |an evolved body of kludges" - Mark Miller
> www.pwpconsult.com |
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html
>

Alan Bloom

Re: K4 DSP bit width

> Getting a 16 bit converter that fast, approaches the limit of available technology.

Right. At least at reasonable cost.

> These converters are pipeline architecture and have better than 90 db dynamic range.

Theoretically the dynamic range (ADC overload to noise floor) of an ADC
is a little better than 6 dB * number of bits, which would be over 96
dB for a 16-bit ADC. But they are never that good in practice. I
believe 90 dB is way optimistic, especially when you take into account
the analog circuitry that is needed in front of the ADC.

For example, the 16-bit LTC2107 from Analog Devices has 80 dBFS
signal-to-noise ratio. Note that is a theoretical figure in that it
assumes no degradation at all by external circuitry. This
state-of-the-art device draws 1.28W of power and costs $99 in 1000-piece
quantity.

> Looking at S units, there is 88 db from S1 to 40 over S9 before any attenuation or

> AGC is applied prior to the ADC.

>> K3S K4HD Feature
>> 24 16 ADC width

As others have noted, that's an "apples and oranges" comparison because
the 16-bit direct-sampling ADC is way oversampled compared to the 24-bit
ADC at baseband or low-frequency IF. The S/N ratio of an ADC is
improved by

10 * log(sample_rate/receive_bandwidth) dB

For example, if the sample rate is 130 MHz and the receive bandwidth is
500 Hz, that's about a 54 dB improvement. If the ADC dyanamic range,
including all external circuitry, is (let's say) 75 dB then the 500 Hz
dynamic range is 129 dB. I think with current technology at reasonable
cost, the high 120's dB is about the best you can do with a
direct-sampling receiver and many don't do that well.

By constrast, I believe the blocking dynamic range of a K3S is about 150
dB in a 500 Hz bandiwdth. I assume the K4 high-dynamic-range option
will be similar. Whether you need the extra 20-30 dB of blocking
dynamic range depends on your situation and what you plan to use the
radio for.

Alan N1AL

On 2019-05-29 09:37, W2xj wrote:

> The K4 is a direct sampling radio. It has no IF. It samples directly to IQ baseband and performs all filtering and processing in the digital domain. That requires a very high speed ADC. In this case it is just above the Nyquist limit for 6 meters. Getting a 16 bit converter that fast approaches the limit of available technology. These converters are pipeline architecture and have better than 90 db dynamic range. Looking at S units, there is 88 db from S1 to 40 over S9 before any attenuation or AGC is applied prior to the ADC. The problem comes when there are signals exceeding that dynamic range like working someone who might be S1 or S2 while someone else is transmitting nearby as a direct conversion radio has no front end selectivity. As Eric already explained, this is the reason for the K4HD option. This appears to essentially be a K3S superhet single conversion front end that feeds an approximate 8 MHz IF into the ADC. It would then provide a very narrow band receptio!
n o

> ption for the K4.
>
> Looking at K3s architecture, the ADC is a PCM1804 which is a delta sigma converter which is basically a one bit ADC clocked at 768 times the sample rate or somewhere between approximately 24 and 36 MHz depending on what sample rate Elecraft selected. In the mode used in the K3S it can sample up to 48 KHz although there are other modes permitting up to 192 KHz. The one bit sample is decimated to 24 bits. This ADC is a fairly typical high end stereo audio converter.
>
> The two approaches are very different and each fulfills a different need. Fortuneately, the K4 can provide both with the right options installed.
>
> Sent from my iPad
>
>> On May 29, 2019, at 1:08 AM, Bill Frantz <[hidden email]> wrote:
>>
>> A friend and I were discussing the K4HD and he said that it sounds a lot like the K3S. Well, we both agreed that since the K3S has world class receiver specs, that would not be a bad thing.
>>
>> I started comparing the information about the K4HD with the K3S manual and found:
>>
>> K3S K4HD Feature
>> 8.215MHz ~8MHz First IF
>> 15KHz ? Second IF (if used in the K4)
>> 24 16 ADC width
>> ~30KHz? 122MHz ADC sample rate (WAG for the K3S)
>> 5 2+1 Crystal filters (one empty slot is from a post on this list.)
>>
>> Note: In the K4 FAQ there are forward references to a K4HD section, but I couldn't find it on the web site.
>>
>> Analysis and wild speculation
>>
>> Both DSPs and ADC technology have had at least 10 years to improve between the K3S and the K4. WHile I don't track ADC tech, DSPs are computers and we all know what has happened in computers. The basic processors are somewhat faster, and there are many more cores on a chip. The K4 certain to have taken advantage of these improvements.
>>
>> The K3S uses a 32 bit floating point DSP, and I don't see any reason to change that specification for any of the K4 models. Not changing means that much of the K3 DSP code should easily port to the new DSP.
>>
>> I don't understand how the basic K4 can get good dynamic range with a 16 bit DAC. The K3's 24 bit DAC seems a better choice, although getting high speed and wide bit width at the same time is hard and expensive.
>>
>> In the wild speculation department, there seem to be several approaches for the K4HD. (1) Run a K3 like superhet with an ~8MHz IF and digitize that into the DSP. (2) Duplicate the K3's 2nd IF and use the 16 bit ADC at 15Khz. (3) Use a wider ADC at either 8MHz or at 15KHz. Note that one of the features of the K4 is ease of upgrade to new ADCs.
>>
>> 73 Bill AE6JV
>>
>> ---------------------------------------------------------------------------
>> Bill Frantz |"Web security is like medicine - trying to do good for
>> 408-356-8506 |an evolved body of kludges" - Mark Miller
>> www.pwpconsult.com [1] |
>>
>> ______________________________________________________________
>> Elecraft mailing list
>> Home: http://mailman.qth.net/mailman/listinfo/elecraft
>> Help: http://mailman.qth.net/mmfaq.htm
>> Post: mailto:[hidden email]
>>
>> This list hosted by: http://www.qsl.net
>> Please help support this email list: http://www.qsl.net/donate.html
>
> ______________________________________________________________
> Elecraft mailing list
> Home: http://mailman.qth.net/mailman/listinfo/elecraft
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:[hidden email]
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html

Links:
------
[1] http://www.pwpconsult.com
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html

K9FD

K3 HI RFI burnout

In reply to this post by Bill Frantz

I have had a tech ticket in with Elecraft for a few days and have had no
reply,
thats another story and not addressed in this posting..

I was sitting here at the op desk last week and listening to aM
broadcast band
on one of my K3 and suddenly the HI RFI was on the VFO B display, it was
still
receiving fine, but checking it has no RF out, and will not shut down
with the
power button.   I checked the manual and understand why the error came
on and what
the solution is, but see nothing as to what the problem can be, what
burned out
etc.   I checked on line and see almost nothing on the subject, just the
normal
off topic replies to the question.

Has anyone had this problem and repaired it themselves? I saw one
posting that
it was the BAR64-05 burned out.   I looked visually and it looks ok,
but not sure
if it can be checked in circuit and what the readings should be.
Looking for help, perhaps Elecraft will come through this week? Like to
order parts
if necessary and get the radio back in service.
Appreciate any help, thanks 73 Merv K9FD
______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]