EMRFD Message Archive 14635

Message Date From Subject
14635 2018-04-17 19:46:39 lawrence_joy PA3AKE "Holy Grail" RCVR Frequency Architecture

In the Down Conversion Receiver Overview Martin, PA3AKE, states "It is a conventional 'single conversion superhet', down converting to a 9 MHz IF with a high side local oscillator.

For the 10.1 MHz (30 m) band and higher frequency bands, down converting to 9 MHz IF with high side LO will give spectrum inversion, that is USB will become LSB.

For the 7 MHz (40 m) band and lower frequency bands you would have to do an up conversion, not a down conversion, to the 9 MHz IF. With the LO on the high side, or low side, there is no spectrum inversion.

For the QSD (Tayloe detector) how is I minus Q performed to get the correct sideband?

9V1MI/WN8P, Larry

14636 2018-04-17 20:39:40 John Marshall Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
If the LO is higher than the receive frequency spectrum inversion will occur regardless of the intermediate frequency. Look at the math - subtracting 7 MHz USB from 16 MHz LO gives 9 MHz LSB, same as subtracting 10 MHz USB from 19 MHz LO.

John, KU4AF
Pittsboro, NC

14637 2018-04-18 20:07:31 lawrence_joy Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
Your first sentence is incorrect. If fLO is higher than fRF and the fIF is higher than fLO, that is fRF + fLO = fIF, there is no spectrum inversion. See and . I think you just proved my point. For the 7 MHz band and frequencies below, LSB is used and thus with the spectrum inversion this becomes USB.

I am trying to get my head around this, please help me to understand. I disliked in school when they said "proof of which is left to the student". Okay, I think I just figured it out. There is another spectrum inversion in the QSD (Tayloe detector) when converting to baseband, and thus the sidebands are upright due to there being two spectrum inversions.

Did I get it right? Thanks for your input.

9V1MI/WN8P, Larry
14638 2018-04-22 03:39:45 billw77aaz Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
The frequency of a high side LO is always higher than both the RF and IF frequencies.  Written as a math statement, high side LO means fLO - fRF = fIF,
or  with simple algebra:  fLO  =  fRF + fIF.  Any other case is NOT high side LO.

For receiving 1.8 MHz with an 9 MHz i.f. a high side LO  10.8 MHz.
For receiving 30 MHz with an 9 MHz i.f. a high side LO is 39 MHz.
In both cases USB on the rf frequency always produces sideband inversion (LSB) at the 9 MHz i.f.

The i.f. frequency is immaterial as long as the LO is on the high side:
If the i.f. is 100 MHz, a high side LO for 1.8 MHz will be at 101.8 MHz
for 30 MHz the high side LO will be at 130 MHz.
In both cases, again, USB on the rf frequency always produces sideband inversion (LSB) at the 100 MHz i.f.

Forget some pat formula for just a second. If the LO is 101.8 MHz, and the signal is at 1.8 MHz, USB, say with a 1 KHz tone then the sideband is at 1.801. 101.8 MHz LO minus the 1.801 MHz USB tone is 99.999 MHz. Lower than the 100 MHz carrier at the i.f.

Do this same numerical computation with a 30 MHz carrier and an upper sideband 1 KHz tone therefore at 30.001 MHz. For 9 MHz and 100 MHz i.f. you will get 8.999 and 99.999 MHz respectively. Lower sideband. Or do the computation with the relationship that

14639 2018-04-22 06:52:11 Ashhar Farhan Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
The IF of 9 mhz is very close to 10 mhz and 7 mhz band. The thing to avoid is receiving 9 mhz IF from the antenna. The doubly balanced mixer balance will, at best, suppress the 9 mhz input by 40 db. The bandpass filter of 10 mhz may, at best, contribute 20 db suppression of a 9 mhz signal. A total of just -60 db of IF. Martein had a target of 100 db. He added a 9 mhz notch filter to achieve 100 db suppression.
- f 

14640 2018-04-23 07:44:12 kb1gmx Re: PA3AKE "Holy Grail" RCVR Frequency Architecture

I have many radios with 9mhz IF that are good and clean on 40M, three are solid state..  None do 30M but that's more because 30M band for hams didn't exist at that time.  A good bandpass filter will easily strip any 9mhz component of done well.  Also a front end that has a band pass will add to that.  If a balanced mixer is used you can add that as well.  However for somethings there is a point of no significant return.

Consider a radio hooked to a 40M dipole. if the dipole is at reasonable height the mismatch loss and bandwidth
will limit the amount of 9mhz in and out.  Then we have a bandpass input that adds say 20DB (maybe more)
so we have 9mhz further decreased.  Add a balanced mixer and selectivity after that and 9mhz is 60 and more likely 70 db down.  That's good enough for out of band stuff.  This is more true now that SW BC is less 
the issue it was 30 years ago when a 100KW station was close by and nearly in band (or in the case of 
40M in the band).

Now we need better for the guy 2 miles away running a full gallon 30khz away because he is in band and 
a decent filter in the primary selectivity may need a ultimate rejection of easily 100db and maybe more.
This leads to the generation of radios with "roofing filters" as the typical upconverting radio back when had 
a filter maybe 15-20khz wide and the 2khz filter was in the low IF.  Roofing filters go  back to the early 80s
thinking that selectivity close as possible to the antenna is a good thing, indeed it was and still is.

What funny is with fairly basic radios the biggest issue I actually hear is loud and very dirty (high IMD at higher power and wide bandwidth audio).  This is something you cannot fix at the receiver.

Better receivers are a challenge and worth pursuing as it aids in understanding the limits and the problems. 


14641 2018-04-23 09:00:41 Ashhar Farhan Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
You have been illuminating in your comments. Let me build this next month and report.
- f

14642 2018-04-23 09:43:15 lawrence_joy Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
Okay I am belaboring the point and maybe we need W7ZOI and KK7B to chime in on this. I was always taught that low-side or high-side injection (LO frequency) was only in relation to the RF frequency. Not only is fLO - fRF = fIF but fLO + fRF = fIF. The first being down conversion and the 2nd being up conversion. I refer you to the RF Cafe info, links to which I referred to in my earlier posts. My ICOM IC-738 transceiver has a 1st IF of approximately 70 MHz (actual frequency depends on the mode, according to the instruction manual). I don't have a service manual for this rig and I can not find any information on the Internet as to the receiver frequency architecture, but I doubt that the LO is above the 70 MHz IF.

9V1MI/WN8P, Larry
14643 2018-04-23 15:14:48 kb1gmx Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
I haven't seen consistent math used yet for the simple arithmetic and I suspect the idea of a result of 
say -5mhz is somehow verboten and leads to the confusion.  

With modern radios you don't have to live with the problem of crystals for both sidebands, am insertion,
and CW being expensive.  We have DDSS and NCOs for that.

The 738 was like many the high first IF allowed a higher LO systems typically PLL based at VHF and 
was 70 to 100mhz for "DC to Daylight" or about 10khz to about 30mhz.    That meant fewer switched oscillators
in the PLL system or in some cases only one as it possible to make a single VCO tune a 2:1 range (70 to 100mhz is less than 2:1).  They weren't as concerned with sideband inversion as the low if (70mhz was converted to a low (5-11mhz) and sometimes another lower still) IF as they provided CW, AM, SSB upper and SSB lower BFO insertion as needed with a simple micro or a lot of switches and diodes to select the correct combo for the mode and frequency.  Its not that complex but it needs a scoresheet to keep track of everyone.
Also if you loo inside the large portion of the radio volume is dedicated to that PLL system, its display and needed hardware.

The uBITX is a great example of upconverting RX for near DC-30mhz.  The 45mhz first IF allows the 1st
low to be over or under (45mhz being near DC and 15mhz being near 30mhz or 45mhz being near DC and 75mhz being near 30mhz), the second IF at about 12mhz allows for the 2nd LO to be over or under (~33 or ~57mhz) and the BFO can be over or under (~12mhz +- a few khz).   For that radio there are more than a few combinations for each possible sideband and frequency, though some are less desirable due to spurs.

I consider it the fun part of receiver design now as in earlier designs the lack of a crystal or stable LO
at some high frequency made choices more difficult or resulted in far more hardware for the same topology.
With the greater amount of hardware being in the LO generation.  Where before I had to build a PLL system that was often as complex as the radio now a DDS or NCO like SI570 or SI5351 replaces that at the cost of a simple microprocessor encoder and display. 

14644 2018-04-23 19:16:44 tim Re: PA3AKE "Holy Grail" RCVR Frequency Architecture

The math is pretty simple math.

When you mix two frequencies you get a plus and minus product.

If you have generated a 5mhz LSB signal using a 1000hz tone and a
2000hz tone then the frequencies of those tones get added or subtracted
from your mixing signal.

The frequencies for the tones are 4,999,000 for the 1000hz tone and
4,998,000 for the 2000hz tone.

If the mixing frequency is 12Mhz you get -

12,000,000 + 4,999,000 = 16,999,000 and 12,000,000 + 4,998,000 =
16,998,000. In other words a LSB signal with a 17Mhz carrier freq

The difference is 12,000,000 - 4,999,000 = 7,001,000 and 12,000,000
-4,998,000 = 7,002,000. In other words an USB signal with a carrier
freq of 7Mhz (12-5).

The same arithmetic works for a mixer frequency lower than 5Mhz, eg.
1Mhz. Work out the sums and differences and see what kind of signal you
get. The cxr freqs will be 6Mhz and 4Mhz.

I hope I'm not just restating what everyone else is saying!

tim ab0wr

14646 2018-04-24 09:17:14 kb1gmx Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
Tim, your absolutely right and totally wrong.

You forgot the IF is narrow so you must fall inside of it and then there is the conversion to baseband.

Your math is absolutely correct.  Its also incomplete.    You left out the BFO.

So the factors are Signal, LO, BFO  and you can have any possible sim and difference of the three save 
for only a few possible ones end up in the IF pass band with the correct BFO frequency.

Its why the 80m and 20M with 9mhz IF and 5 mhz VFO radio will mess you  up.  Which crystal is used
for the BFO 8.9985 or 9.0015?

14647 2018-04-24 11:56:01 kb1gmx Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
The QSD is used in a receiver as the spectrum to baseband conversion.  Ignoring the device you have two channels I and Q that contain the same spectrums with a shift in phase (phase also has a relationship to time).
So for that radio all of the selectivity and sideband selection is not in the "QSD detector"

So how do you get which sideband you do trig on the signals because they differ in time there is an added delay called an all pass network an analog circuit can do this ( or in computer there are several ways to 
simulate this).  When you add the signals you get onesideband  or if you subtract them you get the other.  Thats a simplified and slightly inaccurate form.  

its also very different from analog superhet radios though once you wrap your head around spectrum it makes some sense.

For superhet first we convert the spectrum to fit in the filter (crystal or some collection of coils and caps).
After that we then mix it again to get audio  So there are two conversions a LO and the BFO.  So yes we can invert the spectrum so that if the signal we want increases in frequency it decreases within the IF, but we have the BFO that can be on either side of the IF and by selection we can get the right one to make the signal be hear as increasing in frequency.  The math works but only if you include all of it and the BFO counts!


14648 2018-04-24 16:28:22 tim Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
I wasn't trying to make it complete. I was just trying to show the
math. Once you can do the math then it doesn't matter how many
conversions you do, the math all works the same way.

For receive, the BFO is supposed to be at the carrier frequency for
proper demodulation of the signal. So it should be at 9Mhz since that
is what the IF is and should be where the RF is mixed to.

tim ab0wr

14649 2018-04-25 08:58:31 kb1gmx Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
Yes, the math works but the answer can still be wrong if you use 
the wrong assumptions.

The initial question was about sideband inversion or not.

9mhz in the center of the filter or at the band pass edges?

It makes a difference which edge.  Which is why the whole inversion question has 
to take that into account.   

The classic 80/75M with 5mhz VFO rig using 9mhz IF:

Working from the modulator out.
IF  3khz wide filter centered at 9mhz.  BFO is 9.0015mhz  Modulating signal is 1khz.

9.0015-.001 = 9.0005  Output of the IF (filtered and lower sideband)

The output tone of the IF is 9.0005mhz and the sideband will be lower.  There is no 9.0025
as the filter stripped it.

If I take the IF output and mix it with 5mhz and filter the result for 4mhz what sideband is the result.
5.0-9.0005= -4.0005   (lower!)  

How do we know change the modulating tone to 2khz.
9.0015-.002= 8.9995 (still in the filter pass band)
5.0-8.9995=-3.9995 (seems to go down so its lower.)

For the 13mhz LO case and still for 4mhz...

13-9.0005=3.9995  (upper sideband, (1khz tone)
13-8.9995=4.0005  ( 2khz tone, increasing frequency so definitely upper.

In both cases I kept the order of the math the same and the results have a minus(-)
sign which show the inversion of the spectrum (aka sideband).

So for the 13mhz case to get lower sideband we need the carrier oscillator to be
8.9985mhz.  That assume the same filter.

If you do not do all the math the answer is only partial at best and you cannot determine 
if there is sideband inversion.  You may guess there is spectrum inversion but that 
is not exactly the same as you have an end you not sure of.

For this we needed to know the tone(signal) the CIO, VFO and the passband of the filters 
used to determine what products are passed on to the next stages.  If you leave out one 
there is ambiguity.


14651 2018-04-26 07:07:22 tim Re: PA3AKE "Holy Grail" RCVR Frequency Architecture
Sorry to take so long in replying.

The answer can't be wrong if the math is ok. If you input the wrong
values then "garbage in - garbage out". That's not the fault of the

The math will tell you if you get sideband inversion or not. It does
*not* depend on the filter or whether you are at the center of the
filter or at the edges.

It does not make a difference in which edge you are at, not for
sideband inversion.

If you input a double-sideband signal into a filter then you *can*
determine which sideband comes out of the filter depending on whether
the carrier is below the filter or if the carrier is above the filter.
But this has nothing to do with sideband *inversion*.

You cannot determine inversion by using only one audio tone. All you
are determining there is where you have to put the signal in order to
filter it out or to allow it through.

If you apply a 1khz modulating signal to a carrier of 9.0015mhz then
you will get a tone at 9.0015Mhz - 1khz and another one at 9.0015Mhz +
1khz. I.e. you will have tones at 9.0005Mhz *and* another one at

Your filter will then decide which audio tone you choose to transmit,
lower sideband would be 9.0005Mhz and upper sideband would be

If your filter is actually *centered* at 9Mhz you will likely pass both
frequencies if you have a typical 2.5khz crystal filter since both
would be within the passband of the filter which would run from 9Mhz -
1.25khz to 9Mhz + 1.25khz (i.e. 9.00125Mhz). The higher tone would be
attenuated more but would still remain significant.

You are mixing up the concept of sideband inversion with filtering of a
double sideband signal to get either lower sideband or upper sideband.
A filter cannot invert a sideband because it is not a mixer, it is a
filter. A filter can *select* the lower sideband or the upper sideband,
it cannot invert it.

You also jumped from a 9Mhz/5Mhz IF/BFO combination to a 13Mhz/9Mhz
VFO/IF combination.

As I stated, 5Mhz is *below* 9Mhz. When you subtract 5Mhz from 9Mhz you
get no sideband inversion. When you subtract 9Mhz from 13Mhz you do get
a sideband inversion.

Take an upper sideband signal with tones at 9.0005Mhz and 9.001Mhz.
When you subtract 5Mhz from both you get 4.0005Mhz and 4.001Mhz, still
an upper sideband signal.

Now take a lower sideband signal with tones at 8.999.5Mhz and at
8.999.000Mhz. Subtract 5Mhz from each and you get 3.999.500Mhz and
3.999Mhz. A lower sideband signal.

Exactly the opposite happens when you use 9Mhz and 13Mhz. You get a
sideband inversion. 13Mhz - 9.0005Mhz = 3.999.500hz and 13Mhz -
9.001Mhz = 3.999Mhz. You just converted an upper sideband signal,
9.005Mhz to 9.001Mhz, to a lower sideband signal.

Don't conflate mixing with filtering. They *are* different. And the
math always works. At least for the last 50 years since I studied it in

tim ab0wr