PE425x switches noise at 900 kHz and harmonics (from the EMRFD group)

[in3otd]

Hello,
FYI, there is currently an interesting discussion on the EMRFD group about IC switches for front-end filter selection where Graham / KE9H said:

> The problem with the PE425x series switches is that they contain an internal bias generator, which is a square wave oscillator driving a charge pump at about 900 kHz.

> This fundamental frequency and its odd harmonics bleed into the signal path, and can cause spurious signals in the low HF range. 
> It is not a linearity problem in the normal sense.

> This is not documented in the spec sheet. 
> It is not a problem in the cellular band frequencies, which is where this switch was designed to operate.

Gian, I7SWX asked the Peregrine support, and they replied:

> This is true. The charge pump fundamental is about ~-120dBm  at ~ 1Mhz ( plus harmonics) so mostly isn’t a problem except in a rx path at the front end.
> You can use the 4239 which has no charge pump, or the 4257 which has  the external bias option.

You can see the whole discussion at
https://groups.yahoo.com/neo/groups/emrfd/conversations/topics/12434

I did not have the time to test this with the H-L to see how relevant the noise is, but seems an interesting thing to know anyway - I didn't see this reported anywhere before.

73 de Claudio, IN3OTD / DK1CG

[ZL2APV]

Great catch Claudio!

73, Graeme zl2apv

[Steve Haynal]

Hi Claudio,

Thanks for the very helpful information. I am a member of EMRFD but don't always read. Peregrine gave me the same warning about the PE4259 charge pump about 6 months ago. See the quoted e-mail below. This spur, along with how well it works at the lowest frequences of 136 kHz and 474.2 kHz WSPR bands, are tests I will definitely run on the PE4259 before committing to using it in V2. I can do those tests with my last frontend board. Some items to consider and discuss:

• All Peregrine devices seem hard to find in small quantities. Digikey lists a minimum order size of 3000.
• The analog switch devices listed in the EMRFD post may work okay for receive, but I don't see how they will handle the voltage swings for TX even at 20 dBm.
• Skyworks and M/A Com make similar devices that are more readily available which we might want to consider.
  

73,

Steve

KF7O

Hi Steve,

Thanks for the information. The difference in low frequency specification was more a function of marketing. For our older parts, we just specified DC-X GHz to be consistent with the rest of the markets' convention. However both parts will experience large signal parameter derating below 10MHz. Having said that, the PE4259 is actually a better performer at low frequency compared to the 4283. Both parts can be used down to 100KHz (and lower), and neither part should be used to pass actual DC.

S-parameters will not degrade at low frequency.

How sensitive is your application to low level spurs? One difference I've seen between the two parts, is that the PE4259 has higher level spurs (generated by the internal charge pump) compared to the PE4283. The spur fundamental of the former is at around 1MHz and level is around -105dBm. 

Regards,

Rohan

[ZL2APV]

In addition I thought I would mention that Skyworks also make a nice PIN diode set for HF switching in the up to 50 Watt bracket in either series or shunt mode. I was looking at doing the HF Amp switching with PIN diodes instead of a relay (not much more expensive) and of course silent plus allowing full break in keying and matching the Peregrine speeds. Of course this is future gazing but my experiments with GY10 diodes to test this are working well and even switch my 100 watt rig but only 50 dB isolation at 50 MHz so the Skyworks PIN diodes would be much better.

73, Graeme

[Malcolm G4DMH]

I have taken an interest in this thread because I was thinking of using PE4259 switches in octave filter BPFs for a general coverage receiver. They would be switched by a rotary switch providing 3.3v to the appropriate pairs of pin 4.

This should give better all-round performance compared to switching diodes, I think.

Looking at the Peregrine Semiconductor Product description, pe4259ds.pdf, (page 1 - Functional Diagram) when CTRL (pin 4)  is high, RFC (pin5) is disconnected from RF2 (pin 3), but RF2 is then grounded internally.

Looking at the front end BPF circuit diagram, Softerhardware Hermes-Lite pre-V2 RX Filter board PDF, this would appear to ground the common in/out lines.

I must have misunderstood the operation of the PE4259. Is it the suffix? The circuit diagram lists PE4259-53. DigiKey sell the PE4259-63, but these suffixes are not mentioned in the Peregrine Datasheet.

It looks like DigiKey do sell the PE4259-63 in small quantities, under part number 1046-1011-1-ND. I agree that at first glance on the product page, the minimum quantity is 3000. Click on the 'Alternate Package' a little to the right and lower down.


Best wishes


Malcolm   G4DMH
Doncaster UK
The birthplace of the Flying Scotsman

[John Williams]

Malcolm. The pre-V2 RX design is preliminary and untested. Thus you may have discovered a design defect in the design. I will take a closer look at the datasheet.

John W9JSW

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[John Williams]

Malcolm, you have indeed uncovered a defect in my design! I will have to rectify this error. Good catch! I think I will just flip the sides and use RF1 for signal and ground RF2. It was my intent to ground the filter at both ends if not selected and your sleuthing has shown that this can be done within the device.

Thanks again for bringing this up.

John - W9JSW

[John Williams]

Actually I will have to leave RF2 open...

[Steve Dick]

Hi Malcolm and John. You might want to consider an alternative to the PE4259 and that is the FSA4157. See https://www.fairchildsemi.com/datasheets/FS/FSA4157.pdf  I like these for receiver filter switching because the PE4259 has the following disadvantages:

1. effectively single source

2. Not specified below 10 MHz operation, probably due to the fact that the datasheet does not report the  PE425x contain an internal bias square wave generator, driving a charge
pump at about 900 kHz and generating spurs in the low HF.

The FSA4157 is available in the SC-70 package.  Its advantages are:

1. Operates at standard voltage of 5V

2. At 5V, max on resistance of 1.15 ohm and specified 0.3 ohm maximum Ron flatness. This should provide must better linearity than many of the FSA series devices

3. Many other footprint compatible devices in the FSA series parts not limiting you to one vendor

4. data sheet specifies loss as a function of frequency. Looks like a bit over 0.6 dB at up to 50 MHz

5. low cost ($.65 from Mouser)

 

I just selected this for homebrew bandpass filter switching for my general coverage receiver, which is set up for active low filter activation.  (File attached)

 

Regards,

“Digital Steve”, K1RF

--

You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

This email has been sent from a virus-free computer protected by Avast. www.avast.com

[Steve Haynal]

Hi Malcolm,

The reflective switches will short/ground the unselected input at HF frequencies. This may be what you are noticing? Anyway, there is a good tutorial document on minicircuits (google reflective rf switch minicircuits) that covers reflective and absorbtive switches.

73,

Steve

KF7O

[Malcolm G4DMH]

John,

I still don't know if that will work.

RF1 (pin1) is grounded internally when CTRL (pin4) is low. So the common in/out lines will still be shorted at some point.

This may be academic following the charge-pump RFI issue, but I think RFC (pin5) should connect to the common in/out lines, with either RF1 (pin1) or RF2 (pin3) connected to the individual filters, depending on whether you are switching with a high on CTRL (pin 4) or holding CTRL high and switching low, respectively. In that way, the individual filters are grounded at both ends when not in use.

I hope I have this correct, please check and check again before proceeding with a design.

As far as my Hermes-Lite build is concerned, I have assembled the PCBs and ethernet adaptor connector, now I have the long wait for the BeMicro CV, which is on back-order, to arrive around the middle of the year.

Best wishes

Malcolm  G4DMH
Doncaster UK

[John Williams]

Malcolm,

I am going to investigate the FSA4157 part that Steve Dick suggested.

John

[ZL2APV]

Hi John,

The bit that concerns me with the FSA4157 switches is the off port's 12pF capacitance. If you are switching say 6 filters these ports will be commoned at 72pF and will connect to the filter when switched. Port A is 40pF so you could have a total of 100 odd pF presented to the filter. Of course this value could be included in the filter itself and a series parallel conversion done to absorb it. Bus switches could be worth a look and there are 1 to 8 fanouts which would be ideal but speed, through resistance and capacitance were the limiting factors when I last looked at them a couple of years ago. There could be advances now so I should look again.

73, Graeme ZL2APV

[Steve Dick]

You are right Graeme. The capacitance of the switches must be taken into consideration.  The tradeoff for this FSA4157 is very good linearity across the full input voltage range vs a relative high input active (40pF) and input off (12pF) capacitance. If a large number of filters are being switched, this is problematic. My apologies for not taking this into consideration. The FSA3157, by comparison, has a lower capacitance (18.5pF on, 6.5pF off) with a higher Ron (varies from 3 to 7 ohms over full voltage input range vs the FSA4157’s much lower variation (.06 ohms to .95 ohms as shown in specified Ron curves) and a 5X worse higher harmonic distortion (.011 percent vs the FSA4157’s .002%). I need to think about this some more.

 

“Digital Steve”, K1RF

 

From: ZL2APV

Sent: Monday, March 07, 2016 2:09 PM

To: Hermes-Lite

Subject: Re: PE425x switches noise at 900 kHz and harmonics (from the EMRFD group)

 

[Steve Dick]

I thought about it and don’t have a good solution for the RF switch issues.  One approach is to actually do measurements on the PE4259 and see what noise really looks like at low frequencies (<10 MHz).  It may be acceptable since the lower frequencies atmospheric noise will likely drown out any internally generated noise. There are other RF switches around that will work down to DC and have low input capacitance but they cost about $3.50 rather than the 50 or 60 cent variety.  Analog Devices, for one, makes some. They are basically n-channel MOSFET switches rather than transmission gates, which use back to back n-channel and p-channel transistors. See app note AN 952: http://www.analog.com/media/en/technical-documentation/application-notes/AN_952.pdf  Data sheet for the ADG919, for example, shows an input capacitance of only 1.6pF. These chips require a non-standard power supply voltage of 1.65 to 2.75 volts and optionally biasing of the chip at the output.  The single NMOS switch will have worse linearity than transmission gates. Relays anyone?

 

“Digital Steve”, K1RF

[John Williams]

Can we use the PE4239?

[John Williams]

I went ahead and revised the V2RX design to correct the issue with RF1/RF2 internal grounding. Please review.

I am tending toward having a few boards made so we can test out the issue with charge pump noise to see if it is a serious factor. The Electraft KX3 uses the PE4283 and it is reported to have excellent RX performance. Peregrine stated that the PE4259 is a direct replacement for that device. Perhaps we should devise a exhaustive test scenario to see if this charge pump issue is a practical concern for the HF bands.

Joe WA9CGZ has been using his Hermes Lite in RX mode for the 2 experimental bands below the BC band. I am wondering if we should revise the DNI filter position to hold a BPF for those bands. Joe - Is there much BCI interference to warrant a BPF or is the antenna good enough at rejecting out of band QRM.

In a separate thread I will discuss whether we should now be considering 6M support, since Alan has it working using undersampling.

John

[Steve Dick]

The PE4239 looks interesting. However there’s an anomoly on Pegrine’s web site.  All of their tables say it goes down to only 10 MHz but their data sheet says it goes down to DC with degraded performance.  Someone needs to contact them and ask them. The insertion loss is not great but acceptable. Also note that none of their data sheets specify input capacitance of any of their chips.  That’s another good question.  It is probably low because they have some sort of proprietary process, probably silicon on insulator. I’d be happy to contact Peregrine and ask them these questions. This may be the best solution, though single source

 

“Digital Steve”, K1RF

[John Williams]

Please do contact them...

[Peter Johnson]

Hi John and Steve.
Steve KF70 has already spoken to them.
And he mentions it on the wiki under Hermes Lite V2.

73 Peter F1VKK

[John Williams]

I read that article Peter. Our questions go a bit deeper...

John

[Steve Dick]

left a message this morning. Awaiting email or phone call back

[Steve Dick]

I just got this reply from Peregrine Semi:

“Hi Steven,

Both the 4283 and the 4259 have on-board negative bias generators. The 4259's spur fundamental is at around 1MHz at -105dBm. The 4283's was about 15dB better. The 4239 can be used at 1MHz, but there may be some large signal performance degradation. Small signal parameters however will not get degraded at low frequencies. This part does not have an on board bias generator, so you won't seen any spurs here. 
The s-parameters for all our parts are available on our website, so you should be able to deduce the input impedance from those. Else just drop the files into your simulation.
If you are very sensitive to spurs, you can also check out our Test and Measurement switches. Any PE425XX part falls in this category. They all have an external negative Vss option, that disables the internal generator and along with it the spurs.
Let me know if you need anything else.Sorry for the delay in responding as I was on vacation earlier this week. (I did get your voicemail as well.)
Regards,

Rohan (Netto)”

0 viruses found. www.avast.com

[John Williams]

Steve,

So, in a RX only environment, would you think the 4239 would be the better part? Or would we potentially overload the switch in a strong signal environment? As to the -105dbm spec, at 1MHz that is outside of any ham bands and would then show much lower, if not immeasurable spurs at the harmonics, correct?

What are your thoughts. I am inclined to stick with the 4259...

Any others care to opine?

John

[Steve Dick]

When all is said and done, I’d stick with the PE4259 because of the following reasons:

Insertion loss: PE4259 – 0.5dB max. PE4239 – 0.9dB max

IIP3: PE4259 – 55dBm,  PE4239-45dBm

 

It is unfortunate that the PE4283 is no longer available as it was spec’d to go down to DC and had a 15dB lower bias generator noise. It turns out that the PE4259 actually has better IIP2 and IIP3 specs than the obsolete PE4283 even down to 1 MHz based on prior discussions I had with PSemi.  But the –105dBm spec corresponds to 1.26 microvolts RMS which sounds pretty sizeable until you realize it will be swamped out by atmospheric noise at the low end of the HF spectrum.

 

So if it were me, I would stay with the PE4259 with its known bias generator noise limitation.  It would be good if someone verified its operation in actual circuit before its committed to.

[John Williams]

Thanks Steve, I will push on to get some sample boards and we will give it to the guys to give it a good examination.

Thanks so much for following up with Peregrine on this.

John

[John Williams]

I think I will build a 3 filter mini-board using the layout I plan for the full board. A much smaller board would be quite inexpensive at oshpark and would give me sample boards to distribute to those that have the proper test equipment. I could forgo the external logic for level matching and automatic band switching to make it pretty compact. Just use a 0.1in DIP header to jumper 3v3 to each of the select pins to activate each bank.

John

On 3/11/2016 4:33 PM, Steve Dick wrote:

[Joe]

All,

With all the talk about the correct switch for above what ever happened to a similar wide band noise from the AD9866 in

transmit mode? I posted a spectrum analyzer post showing a similar curve as shown in the AD9866 data sheet.

This wide band noise also has a peak about 900KHZ above and below the signal however this noise is about -75db

in reference to max TX out and kind of at my limit for proper measurement as I remember this noise gets the the above levels

above 15 meters. There was some talk that caused by a internal PLL which was used on the clock oscillator

on when TX. Since it's such a low level and peaks above and bellow the Ham bands I just ignore it.

Has anyone else taken a look at the TX output?

Joe   wa9cgz  

[John Williams]

Well, I worked on the layout and finished it this morning. The complete design. was just as easy. Will give it a review and then send off for some boards.

John

[John Williams]

Attached are pics of the top and bottom, gerbers and schematic. Please comment...

I will have to adjust connector placement when we work out final fitting of this board with the V2 board. My desire is to have 3 stacked 80x100mm boards with plug-in interconnects, if possible. The board stacking is constrained by the need for the PA to be on the bottom so it has a heat sink to the case. Then we will have to determine if we want the RX board next or the V2 board then the last board on the top. My hope is that we can have appropriate male and female connectors so that we can build the stack, then slide it into the enclosure. Preferably we can devise a way for the 2x07 connector to ride across all three boards in a single stack. The RX in and RX out will some how have to interject itself in the path between the V2 RX connector and the PA RX connector.

John - W9JSW

[Glenn P]

John

Is the board 2 or 4 layer? If 2 layer, the in and out '50R' tracks look very narrow.  I would be inclined to reduce the length of the stubs into each of the switches also if possible?

glenn
vk3pe

[Glenn P]

also, the pad size on the vias looks very very small. Some pcb makers will specify a  minimum annulus of 14mils for example. 

 You run the risk of holes being off centre in a via, if there is any slight runout, when making the boards.

[John Williams]

2 layer board. Will do on the width and length suggestions.

[John Williams]

The min via size is 35 mils with a 20 mil hole so the min annulus is 15 mils. This has worked for all of my boards for the last year or so.

[Glenn P]

Hi John, OK, that just meets the most common specs. It just appears smaller in the picture !

glenn

[John Williams]

Glenn,

Reworked the traces for RX in and RX out to be close to 50R (used a microstrip calculator using PCB parameters), also rerouted all controls so no uninterrupted grounds under the microstrips.

Let me know what you think. Will order some boards if you think we are good to go.  These boards will be really reasonable on cost at Elecrow since they are 2 layer boards. 

John

[John Williams]

Darn, forgot the attachments...

[Steve Dick]

John, finally got around to looking at the schematic. I would recommend additional at least a few additional power supply bypassing on the 3.3V. The only bypass on the whole board is C81. The PE4259’s don’t have any power supply bypassing at all. You don’t need one on every chip, but maybe at least two on each PE4259 string.  The power lead doesn’t take much current on these since it is basically switches handling low level signals.

 

-Steve K1RF

 

From: John Williams

Sent: Monday, March 14, 2016 10:00 AM

To: herme...@googlegroups.com

Scanned for viruses. www.avast.com

[John Williams]

Will do, thanks for taking the time to look at the board.

John

[Glenn P]

Looks pretty good John.

I wonder though if its worth changing the links for polarity selection to 0R 'jumpers instead, as its tidier and they won't be changed after build anyway, normally.

glenn
vk3pe

[Steve Haynal]

Hi Joe,

This is clock phase noise due to the AD9866 always using the internal PLL during TX. It is unavoidable but fortunately very low level. See https://github.com/softerhardware/Hermes-Lite/wiki/Clock-Phase-Noise for more discussion.

73,

Steve

KF7O

[Steve Haynal]

Hi Group,

Since Jim's 1.42 board has an option for a PE4259-based TR switch, I finally got around to measuring the RX noise due to the charge pump. Here are several Quisk screen shots of the RX noise I see. All measurements, except when noted, were with a 50 Ohm dummy load and with the LNA set to +22 dB.

Below is the charge pump fundamental, at 929 kHz for my part. This probably varies from part to part. I do not see this birdie when using a frontend card without the PE4259. Notice also the two birdies around 1 MHz. These are from the two switching regulators using  two MP2467s in my current power supply setup. These have a switching frequency of around 500 kHz and what you see is the second harmonic. As a reminder, the switching regulators speced for V2 have fundamentals between 2 and 2.5 MHz, and my early experiments showed no noticeable birdies from them. 

Below is what the second harmonic looks like with the same exact Quisk and other variable settings. Notice that the power has dropped significantly as compared to the fundamental. All you can see is a faint line on the waterfall at 1858 kHz.

The odd harmonics are a different story though. Below is the third harmonic. Again, this noise is not present when using a basic frontend without the PE4259.

I have screenshots of odd and even harmonics for all the HF harmonics, but they don't add much to the overall picture. The even harmonics above the second are not visible. The odd harmonics are visible and diminish just slightly as the frequency increases, The first odd harmonic that hits an amateur radio band is the 23rd harmonic and is pictured below. Note the spread of the power across a wider bandwidth as expected. The peak power is lower but still visible with RX and dummy load only.

Once you connect an antenna, natural and man made noise make it hard to see the noise from the PE4259. Below is a screenshot of the 23 harmonic in the 15M band when connected to my 20/15/10M attic dipole. Note that I have a particularly noisy environment as I live in close proximity to several neighbors and my antenna is in the attic. The noise I see from switching power supplies, monitors, etc. that I pick up over the antenna is much worse than the PE4259. I also looked for the PE4259 noise with an antenna connected on all lower odd harmonics and could not pick any noise out. My antennas do not cover the low bands which means that often my coax is a very bad antenna. Still, at some of the lower harmonics, I do see a significant drop in noise picked up by the antenna input since there is such a great mismatch with the antenna/coax. Even in those cases I could not discern the PE4259 noise. I suspect that if someone were to run very narrow bandwidth, such as WSPR, on these odd harmonic frequencies, they would still be able to discern the PE4259 noise though.  

I am interested in hearing what others think.

73,

Steve

KF7O

On Sunday, February 28, 2016 at 1:06:25 PM UTC-8, Steve Haynal wrote:

Hi Claudio,

Thanks for the very helpful information. I am a member of EMRFD but don't always read. Peregrine gave me the same warning about the PE4259 charge pump about 6 months ago. See the quoted e-mail below. This spur, along with how well it works at the lowest frequences of 136 kHz and 474.2 kHz WSPR bands, are tests I will definitely run on the PE4259 before committing to using it in V2. I can do those tests with my last frontend board. Some items to consider and discuss:

• All Peregrine devices seem hard to find in small quantities. Digikey lists a minimum order size of 3000.
• The analog switch devices listed in the EMRFD post may work okay for receive, but I don't see how they will handle the voltage swings for TX even at 20 dBm.
• Skyworks and M/A Com make similar devices that are more readily available which we might want to consider.
  

73,

Steve

KF7O

Hi Steve,

Thanks for the information. The difference in low frequency specification was more a function of marketing. For our older parts, we just specified DC-X GHz to be consistent with the rest of the markets' convention. However both parts will experience large signal parameter derating below 10MHz. Having said that, the PE4259 is actually a better performer at low frequency compared to the 4283. Both parts can be used down to 100KHz (and lower), and neither part should be used to pass actual DC.

S-parameters will not degrade at low frequency.

How sensitive is your application to low level spurs? One difference I've seen between the two parts, is that the PE4259 has higher level spurs (generated by the internal charge pump) compared to the PE4283. The spur fundamental of the former is at around 1MHz and level is around -105dBm. 

Regards,

Rohan

[Steve Dick]

Steve and all, hams have a mindset of wanting everything “perfect” and it is annoying that what appears to be a simple switch should introduce spurs – any spurs, especially if information about these is not mentioned in the data sheet, other than saying don’t use the part below 10 MHz.  However, if I were designing this into a product that I were to sell, I would make an engineering judgment that the spurs, for all practical purposes, are not significant enough to not use the part.  Here’s my reasoning:

 

1. The fundamental spur is in the middle of the broadcast band. It will never be a problem in this very strong signal, noisy band.

2. The second harmonic is faint and at the edge of the 160m band. Again, the 160m band is very noisy and you will never see this spur under any normal conditions with a reasonable antenna.

3. The third harmonic, at about 2790, is strong but not in any amateur band.

4. The 23rd harmonic is the first to fall in an amateur band. Fairly weak and would almost certainly be masked by antenna noise with any reasonable antenna.

 

So unless an alternative switch is found that has acceptable performance at low cost without a built-in charge pump, I would not rule out the PE425X.  Relays are not a good option in this application because of their larger footprint and higher cost.

 

“Digital Steve”, K1RF

 

Virus-free. www.avast.com

[Steve Haynal]

Hi Steve,

Thanks for the feedback. I agree that in most cases this switch is just fine although there will always be some with objections. For HL2, I have enough room to include an option for these switches. It will also be possible to leave them off and have separate single RX and TX connectors. I will run two wires to each switch as I'd like to try some of the other similar switches in the same form factor as another one may have better properties for HF.

73,

Steve

KF7O