Topics

Matching LDMOS device


Clive, G3GJA
 

Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA



HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.


Robin Szemeti - G1YFG
 

To stand a realistic chance of a predictive match using a Smith chart you really need all 4 S parameters.  I use SimSmith, as it makes things a bit easier ... but remember the input match "sees" a bit of the output match through S12 so  its not just a question of finding S22 and thats all you need, there is considerable interplay between the input and the output.

You do need the S parameters right at the device terminals though to produce a design from scratch,  VNA's with an S parameter test set come with sets of calibrations that allow you to calibrate and set the reference plane right up to the device terminals, compensating for all the cable lengths.

If you look on Page 7 of the datasheet, you'll find all the S parameters .. you could do worse than use SimSmith with an "S" block to represent the device ... you could simulate the exiting matching in SimSmith and then see what you need to add/take away to sort it out ...



On Sun, 14 Feb 2021 at 21:49, Clive, G3GJA <clive@...> wrote:

Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA



HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.


--
Robin Szemeti - G1YFG


geoffrey pike
 

Hi Clive,
You dont say what the original device was in this case. The VNA reading is what not the old device as its broken? anyway the if its the old device its too high and the sign is probably wrong.
Anyway what to do, i recently did something at 2.4 GHz with a LDMOS device on a blank canvas Rogers 4003 i think
 I used an old DOS program called Puff from CalTech (there must be a modern equivalent)
So you obviously need to have some impedance info from the the data sheet and the substrate Er.
The process i use is to take the the complex input impedance and initially match it to the geometric mean of the source impedance and device impedance, in this case about 6 ohms, then match this to 50 ohms. So both the input and output are transformed in 2 stages and not directly from 0.7 ohms to 50 ohms or 1.6 ohms to 50 ohms.
The Puff program will give you the length of the low Z transmission line (try around 10 to 12 Ohms to give reasonable line dimensions) that you need to take the input or output to this level and real. Then transform to 50 ohms with another line.
All very vague but it will do it. Puff will take the complete s-parameter file and then you can try it over the F range you are interested in, at this point S12 will raise its head and the values you got for the input and output match need to be fine tuned with either flakes or strips previously etched onto the pcb. As S12 is 0.007 you may see only a small effect.
Puff runs under Windows XP but i haven't yet managed to get it to run under Windows 10.


Inline image


Inline image
These 2 quick look examles use a single low Z line to do the transformation and not 2 as mentioned previously, i used the data for 1GHz at 1,3GHz. Worth a try ! ( the output line was shortened to allow a trimmer to finish the transformation)
cheers
Geoff
GI0GDP

On Sunday, 14 February 2021, 21:49:42 GMT, Clive, G3GJA <clive@...> wrote:


Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA



HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.


Greg - ZL3IX
 

I'm not a PA specialist, but have designed one or two in the distant past. I believe that you need to use the impedance characteristics under drive at a typical power level, and these are not usually the same as would be the case under small signal conditions, ie when measured with a VNA. Making impedance measurements under drive is not a simple matter, and you may do well to take the manufacturer's data, here https://www.nxp.com/docs/en/data-sheet/MRF184.pdf.  Admittedly it only goes to 1 GHz, so you would need to attempt extrapolation up to 1.3 GHz. Even at 1 GHz the plot looks like around 1 ohm for Zin, which is way different from the 13 ohms you measured on your VNA.


Clive, G3GJA
 

Thank you Greg, Geoff and Robin for your helpful replies.

 

The original device was a BLF6G15L-40BR, internally matched for use between 1450-1550MHz. Rated at 40W RF out it was designed for ultra-linear operation at 2.5W output with ACPR of -45dB. It even has an on-die sense FET that looks as though it was used to stabilize the bias, judging by the PCB tracks.

 

I presume the impedance I’m seeing on theVNA will be substantially modified by the matching for the original device; the substrate is very thin and the tracks wide so I’m guessing it’s very low Z, causing it to be a long way from the 1 ohm Greg suggests that the MRF184 might present at 1.3GHz. I’m not competent to work out the Er of the substrate from the matching of the original device from the track dimensions I can see and the known characteristics of the original device from its datasheet, but somebody with a lot more experience might be able to work it backwards.

 

Geoff, the Z I’m seeing is with the new MRF184 fitted. It doesn’t make much difference if it’s biased of not (Vds 28v, Ids 300mA)

 

The original device is only characterized at its intended frequency as here:

 

Table 9. Typical impedance per section

IDq = 330 mA; main transistor VDS = 28 V

 

F                          ZS                               ZL

(MHz)

1480                     3.2- j6.3                4.6- j4.5

1510                     4.4 -j6.5               4.6 - j4.5

 

I’ll have a look at Puff and Robin’s suggestion, SimSmith. It should be possible to run Puff on a guest VM running 32bit Windows in one of the Windows 10 machines.

 

Clive G3gja

 

 

 

 

 

 

From: UKMicrowaves@groups.io <UKMicrowaves@groups.io> On Behalf Of Greg - ZL3IX
Sent: 15 February 2021 19:44
To: UKMicrowaves@groups.io
Subject: Re: [UKMicrowaves] Matching LDMOS device

 

I'm not a PA specialist, but have designed one or two in the distant past. I believe that you need to use the impedance characteristics under drive at a typical power level, and these are not usually the same as would be the case under small signal conditions, ie when measured with a VNA. Making impedance measurements under drive is not a simple matter, and you may do well to take the manufacturer's data, here https://www.nxp.com/docs/en/data-sheet/MRF184.pdf.  Admittedly it only goes to 1 GHz, so you would need to attempt extrapolation up to 1.3 GHz. Even at 1 GHz the plot looks like around 1 ohm for Zin, which is way different from the 13 ohms you measured on your VNA.



HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.


Paul Randall G3NJV
 

May I add a couple of related questions, hopefully still on topic? Firstly,  with LDMOS often giving power gains far in excess of what might be required,  is there any need to accurately match the input for max power transfer? Input swr seen by the driver can be controlled by resistive padding?
Second,  when using the familiar series/ parallel coax loops at the output,  I've not been able to find information on the length of coax required. I understand using 25 ohm coax makes a nice match from 50 ohms to 12.5 ohms (or 6.25 ohms per drain) and the relationship between output power, supply voltage and drain impedance, but many designs use 12 ohm coax for higher power, how does this work?  
Advice greatly appreciated as I have some very nice LDMOS PA boards built for around 270MHz that I'd like to bring down to 2m by changing the input and output cables.

Regards to all,  Paul.



Sent from my Galaxy



-------- Original message --------
From: "geoffrey pike via groups.io" <gi0gdp@...>
Date: 15/02/2021 17:35 (GMT+00:00)
To: "UKMicrowaves@groups.io" <ukmicrowaves@groups.io>
Subject: Re: [UKMicrowaves] Matching LDMOS device

Hi Clive,
You dont say what the original device was in this case. The VNA reading is what not the old device as its broken? anyway the if its the old device its too high and the sign is probably wrong.
Anyway what to do, i recently did something at 2.4 GHz with a LDMOS device on a blank canvas Rogers 4003 i think
 I used an old DOS program called Puff from CalTech (there must be a modern equivalent)
So you obviously need to have some impedance info from the the data sheet and the substrate Er.
The process i use is to take the the complex input impedance and initially match it to the geometric mean of the source impedance and device impedance, in this case about 6 ohms, then match this to 50 ohms. So both the input and output are transformed in 2 stages and not directly from 0.7 ohms to 50 ohms or 1.6 ohms to 50 ohms.
The Puff program will give you the length of the low Z transmission line (try around 10 to 12 Ohms to give reasonable line dimensions) that you need to take the input or output to this level and real. Then transform to 50 ohms with another line.
All very vague but it will do it. Puff will take the complete s-parameter file and then you can try it over the F range you are interested in, at this point S12 will raise its head and the values you got for the input and output match need to be fine tuned with either flakes or strips previously etched onto the pcb. As S12 is 0.007 you may see only a small effect.
Puff runs under Windows XP but i haven't yet managed to get it to run under Windows 10.


Inline image


Inline image
These 2 quick look examles use a single low Z line to do the transformation and not 2 as mentioned previously, i used the data for 1GHz at 1,3GHz. Worth a try ! ( the output line was shortened to allow a trimmer to finish the transformation)
cheers
Geoff
GI0GDP

On Sunday, 14 February 2021, 21:49:42 GMT, Clive, G3GJA <clive@...> wrote:


Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA



HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.


geoffrey pike
 

Hi Paul,
Yes the input could be done with a resitive padder attenuator or what ever you like but it's a tad untidy but yes you could as the gain in LDMOS is at times quite high,
 OK the coax length for 4:1 output coax transformers is "normally" no shorter than 1/8 wavelength at the design F. You can make it longer but it doesn't really help much but only increase costs.
As you mentioned the 25 ohm cable does a neat job of setting the drain load to the devices, the 12 ohm cable does the same but presents a lower load to a device that can develop the higher power into that. Most of this is for 48 V devices.
Guanella type devices are inherently broadband so you wont get much advantage in changing the coax geometry you would need to look elsewhere to get your boards to work, have you tried them at 144 MHz?
cheers
Geoff
GI0GDP



On Saturday, 20 February 2021, 11:24:27 GMT, Paul Randall G3NJV <paulfrandall@...> wrote:


May I add a couple of related questions, hopefully still on topic? Firstly,  with LDMOS often giving power gains far in excess of what might be required,  is there any need to accurately match the input for max power transfer? Input swr seen by the driver can be controlled by resistive padding?
Second,  when using the familiar series/ parallel coax loops at the output,  I've not been able to find information on the length of coax required. I understand using 25 ohm coax makes a nice match from 50 ohms to 12.5 ohms (or 6.25 ohms per drain) and the relationship between output power, supply voltage and drain impedance, but many designs use 12 ohm coax for higher power, how does this work?  
Advice greatly appreciated as I have some very nice LDMOS PA boards built for around 270MHz that I'd like to bring down to 2m by changing the input and output cables.

Regards to all,  Paul.



Sent from my Galaxy



-------- Original message --------
From: "geoffrey pike via groups.io" <gi0gdp@...>
Date: 15/02/2021 17:35 (GMT+00:00)
To: "UKMicrowaves@groups.io" <ukmicrowaves@groups.io>
Subject: Re: [UKMicrowaves] Matching LDMOS device

Hi Clive,
You dont say what the original device was in this case. The VNA reading is what not the old device as its broken? anyway the if its the old device its too high and the sign is probably wrong.
Anyway what to do, i recently did something at 2.4 GHz with a LDMOS device on a blank canvas Rogers 4003 i think
 I used an old DOS program called Puff from CalTech (there must be a modern equivalent)
So you obviously need to have some impedance info from the the data sheet and the substrate Er.
The process i use is to take the the complex input impedance and initially match it to the geometric mean of the source impedance and device impedance, in this case about 6 ohms, then match this to 50 ohms. So both the input and output are transformed in 2 stages and not directly from 0.7 ohms to 50 ohms or 1.6 ohms to 50 ohms.
The Puff program will give you the length of the low Z transmission line (try around 10 to 12 Ohms to give reasonable line dimensions) that you need to take the input or output to this level and real. Then transform to 50 ohms with another line.
All very vague but it will do it. Puff will take the complete s-parameter file and then you can try it over the F range you are interested in, at this point S12 will raise its head and the values you got for the input and output match need to be fine tuned with either flakes or strips previously etched onto the pcb. As S12 is 0.007 you may see only a small effect.
Puff runs under Windows XP but i haven't yet managed to get it to run under Windows 10.


Inline image


Inline image
These 2 quick look examles use a single low Z line to do the transformation and not 2 as mentioned previously, i used the data for 1GHz at 1,3GHz. Worth a try ! ( the output line was shortened to allow a trimmer to finish the transformation)
cheers
Geoff
GI0GDP

On Sunday, 14 February 2021, 21:49:42 GMT, Clive, G3GJA <clive@...> wrote:


Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA



HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.


Paul Randall G3NJV
 

Geoff many many thanks for that.

Untidy at the input, yes you’d never say elegant But with 10’s of watts from a transceiver giving far too much power for the gates many will need an attenuator anyway. I just wondered if there was a special reason to do a full input match regardless.

 

Actually no I have not tried these boards in anger yet – just a quick and dirty low volts power up to check the LDMOS turned on around the correct voltage and were not duff. I was attracted to them because the Teflon pcb looked like a “universal” design with no printed lines or obvious microstrip matching, just coax lines easily changed. High quality, made by NEC with a pair of separate BLF574 amplifiers operating at 50V so possibly a conservative 2x 300w. The layout has an input coax section feeding the classic crossed 25ohm coax transformer. The output is almost identical in reverse order. Coax lengths vary a bit but are around  6.5cm in teflon hardline. There is provision on the PCB for capacitors at all the usual places you might want to put one for matching.

 

Looking at other 2m amplifiers (eg W6PQL) the coax sections appear to be around 12cm which is more like 1/16 wavelength? Have I got this right?

 

Paul.

 

Sent from Mail for Windows 10

 

From: geoffrey pike via groups.io
Sent: 20 February 2021 17:35
To: UKMicrowaves@groups.io
Subject: Re: [UKMicrowaves] Matching LDMOS device

 

Hi Paul,

Yes the input could be done with a resitive padder attenuator or what ever you like but it's a tad untidy but yes you could as the gain in LDMOS is at times quite high,

 OK the coax length for 4:1 output coax transformers is "normally" no shorter than 1/8 wavelength at the design F. You can make it longer but it doesn't really help much but only increase costs.

As you mentioned the 25 ohm cable does a neat job of setting the drain load to the devices, the 12 ohm cable does the same but presents a lower load to a device that can develop the higher power into that. Most of this is for 48 V devices.

Guanella type devices are inherently broadband so you wont get much advantage in changing the coax geometry you would need to look elsewhere to get your boards to work, have you tried them at 144 MHz?

cheers

Geoff

GI0GDP

 

 

 

On Saturday, 20 February 2021, 11:24:27 GMT, Paul Randall G3NJV <paulfrandall@...> wrote:

 

 

May I add a couple of related questions, hopefully still on topic? Firstly,  with LDMOS often giving power gains far in excess of what might be required,  is there any need to accurately match the input for max power transfer? Input swr seen by the driver can be controlled by resistive padding?

Second,  when using the familiar series/ parallel coax loops at the output,  I've not been able to find information on the length of coax required. I understand using 25 ohm coax makes a nice match from 50 ohms to 12.5 ohms (or 6.25 ohms per drain) and the relationship between output power, supply voltage and drain impedance, but many designs use 12 ohm coax for higher power, how does this work?  

Advice greatly appreciated as I have some very nice LDMOS PA boards built for around 270MHz that I'd like to bring down to 2m by changing the input and output cables.

 

Regards to all,  Paul.

 

 

 

Sent from my Galaxy

 

 

 

-------- Original message --------

From: "geoffrey pike via groups.io" <gi0gdp@...>

Date: 15/02/2021 17:35 (GMT+00:00)

To: "UKMicrowaves@groups.io" <ukmicrowaves@groups.io>

Subject: Re: [UKMicrowaves] Matching LDMOS device

 

Hi Clive,

You dont say what the original device was in this case. The VNA reading is what not the old device as its broken? anyway the if its the old device its too high and the sign is probably wrong.

Anyway what to do, i recently did something at 2.4 GHz with a LDMOS device on a blank canvas Rogers 4003 i think

 I used an old DOS program called Puff from CalTech (there must be a modern equivalent)

So you obviously need to have some impedance info from the the data sheet and the substrate Er.

The process i use is to take the the complex input impedance and initially match it to the geometric mean of the source impedance and device impedance, in this case about 6 ohms, then match this to 50 ohms. So both the input and output are transformed in 2 stages and not directly from 0.7 ohms to 50 ohms or 1.6 ohms to 50 ohms.

The Puff program will give you the length of the low Z transmission line (try around 10 to 12 Ohms to give reasonable line dimensions) that you need to take the input or output to this level and real. Then transform to 50 ohms with another line.

All very vague but it will do it. Puff will take the complete s-parameter file and then you can try it over the F range you are interested in, at this point S12 will raise its head and the values you got for the input and output match need to be fine tuned with either flakes or strips previously etched onto the pcb. As S12 is 0.007 you may see only a small effect.

Puff runs under Windows XP but i haven't yet managed to get it to run under Windows 10.

 

 

Inline image

 

 

Inline image
These 2 quick look examles use a single low Z line to do the transformation and not 2 as mentioned previously, i used the data for 1GHz at 1,3GHz. Worth a try ! ( the output line was shortened to allow a trimmer to finish the transformation)

cheers

Geoff

GI0GDP

On Sunday, 14 February 2021, 21:49:42 GMT, Clive, G3GJA <clive@...> wrote:

 

 

Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA

 

HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

 

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.

 


geoffrey pike
 

HI Paul,
That wouldn't be far off if the Vp =0.666 for the coax and the wavelength is 200cm then 0.125 x0.666x 200 =16.5 cm
cheers
Geoff
GI0GDP

On Saturday, 20 February 2021, 19:22:50 GMT, Paul Randall G3NJV <paulfrandall@...> wrote:


Geoff many many thanks for that.

Untidy at the input, yes you’d never say elegant But with 10’s of watts from a transceiver giving far too much power for the gates many will need an attenuator anyway. I just wondered if there was a special reason to do a full input match regardless.

 

Actually no I have not tried these boards in anger yet – just a quick and dirty low volts power up to check the LDMOS turned on around the correct voltage and were not duff. I was attracted to them because the Teflon pcb looked like a “universal” design with no printed lines or obvious microstrip matching, just coax lines easily changed. High quality, made by NEC with a pair of separate BLF574 amplifiers operating at 50V so possibly a conservative 2x 300w. The layout has an input coax section feeding the classic crossed 25ohm coax transformer. The output is almost identical in reverse order. Coax lengths vary a bit but are around  6.5cm in teflon hardline. There is provision on the PCB for capacitors at all the usual places you might want to put one for matching.

 

Looking at other 2m amplifiers (eg W6PQL) the coax sections appear to be around 12cm which is more like 1/16 wavelength? Have I got this right?

 

Paul.

 

Sent from Mail for Windows 10

 

From: geoffrey pike via groups.io
Sent: 20 February 2021 17:35
To: UKMicrowaves@groups.io
Subject: Re: [UKMicrowaves] Matching LDMOS device

 

Hi Paul,

Yes the input could be done with a resitive padder attenuator or what ever you like but it's a tad untidy but yes you could as the gain in LDMOS is at times quite high,

 OK the coax length for 4:1 output coax transformers is "normally" no shorter than 1/8 wavelength at the design F. You can make it longer but it doesn't really help much but only increase costs.

As you mentioned the 25 ohm cable does a neat job of setting the drain load to the devices, the 12 ohm cable does the same but presents a lower load to a device that can develop the higher power into that. Most of this is for 48 V devices.

Guanella type devices are inherently broadband so you wont get much advantage in changing the coax geometry you would need to look elsewhere to get your boards to work, have you tried them at 144 MHz?

cheers

Geoff

GI0GDP

 

 

 

On Saturday, 20 February 2021, 11:24:27 GMT, Paul Randall G3NJV <paulfrandall@...> wrote:

 

 

May I add a couple of related questions, hopefully still on topic? Firstly,  with LDMOS often giving power gains far in excess of what might be required,  is there any need to accurately match the input for max power transfer? Input swr seen by the driver can be controlled by resistive padding?

Second,  when using the familiar series/ parallel coax loops at the output,  I've not been able to find information on the length of coax required. I understand using 25 ohm coax makes a nice match from 50 ohms to 12.5 ohms (or 6.25 ohms per drain) and the relationship between output power, supply voltage and drain impedance, but many designs use 12 ohm coax for higher power, how does this work?  

Advice greatly appreciated as I have some very nice LDMOS PA boards built for around 270MHz that I'd like to bring down to 2m by changing the input and output cables.

 

Regards to all,  Paul.

 

 

 

Sent from my Galaxy

 

 

 

-------- Original message --------

From: "geoffrey pike via groups.io" <gi0gdp@...>

Date: 15/02/2021 17:35 (GMT+00:00)

To: "UKMicrowaves@groups.io" <ukmicrowaves@groups.io>

Subject: Re: [UKMicrowaves] Matching LDMOS device

 

Hi Clive,

You dont say what the original device was in this case. The VNA reading is what not the old device as its broken? anyway the if its the old device its too high and the sign is probably wrong.

Anyway what to do, i recently did something at 2.4 GHz with a LDMOS device on a blank canvas Rogers 4003 i think

 I used an old DOS program called Puff from CalTech (there must be a modern equivalent)

So you obviously need to have some impedance info from the the data sheet and the substrate Er.

The process i use is to take the the complex input impedance and initially match it to the geometric mean of the source impedance and device impedance, in this case about 6 ohms, then match this to 50 ohms. So both the input and output are transformed in 2 stages and not directly from 0.7 ohms to 50 ohms or 1.6 ohms to 50 ohms.

The Puff program will give you the length of the low Z transmission line (try around 10 to 12 Ohms to give reasonable line dimensions) that you need to take the input or output to this level and real. Then transform to 50 ohms with another line.

All very vague but it will do it. Puff will take the complete s-parameter file and then you can try it over the F range you are interested in, at this point S12 will raise its head and the values you got for the input and output match need to be fine tuned with either flakes or strips previously etched onto the pcb. As S12 is 0.007 you may see only a small effect.

Puff runs under Windows XP but i haven't yet managed to get it to run under Windows 10.

 

 

Inline image

 

 

Inline image
These 2 quick look examles use a single low Z line to do the transformation and not 2 as mentioned previously, i used the data for 1GHz at 1,3GHz. Worth a try ! ( the output line was shortened to allow a trimmer to finish the transformation)

cheers

Geoff

GI0GDP

On Sunday, 14 February 2021, 21:49:42 GMT, Clive, G3GJA <clive@...> wrote:

 

 

Can someone explain how to approach matching an LDMOS PA transistor. I’m repairing a 23cm PA that used a PCB cut from a power amplifier designed around a device internally matched for use around 1485MHz. The PCB had been snowflaked on the output line and a trimmer added to the input matching to get the amp to work at 23cm.

 

The original device was overdriven resulting in the gate insulation breaking down. The replacement, an MRF184, is not ideal but should still be usable.

 

I can look at the input with my miniVNA and extract the input impedance at 23cm; it’s 13.3 -j42.4. Is this valid? Is there an easy way to use an online Smith Chart simulator to work out what I need to do to get the input to look like 50R?

 

I presume that the reading is not much help as it is what is presented to the input socket and not the device itself. The problem isn’t helped by not knowing exactly what frequency the amplifier was originally designed for, so I’ve no idea what the existing matching is doing.

 

Where do I start?

 

TIA

 

Clive G3GJA

 

HESH Typographic Logo

HESH Computer Services

hesh.co.uk
01482 324936

8 Louis Pearlman Centre, Goulton Street, Hull, HU3 4DL

 

Registered in England No. 3003479. The information contained in this email may be legally privileged and/or confidential. If you are not the intended recipient, use of this information (including disclosure, copying or distribution) or any action taken, or omitted to be taken, in reliance on it is prohibited and may be unlawful; therefore please inform the sender and delete the message immediately. The views expressed in this email are those of the originator and not necessarily those held by HESH Computer Services Ltd, who do not accept liability for any action taken in reliance on the contents of this message or for the consequences of any computer viruses which may have been transmitted by this email.