RocketBox HD500 w/9530 FETs

[blasphemy000]

You are imposing bipolar characteristics on fet's and that''s why you get confused. I have never seen any body measure the source current to balance and bias the amp because it is not necessary. As long as you gate bias is set and the FETS are of the same VGSon you are good to go. If the situation occurs where you have a fet just barely starting to conduct and another one is full on your drive circuits are screwed up not the fets. Or you have mixed fet's of a different part number and the VGSon is completely wrong. Other tha that I have to call B.S.on the rest of it.

I'm not confused at all, you seem to be missing the point. Bipolars and Mosfets do share some characteristics, load line theory calculations are one of those things. How the device is being driven doesn't particularly matter, voltage or current. Now, you are correct that as long as VGS(th) is matched between the devices then you can set the bias via gate voltage and all of the devices will be biased the same, but this is just an argument in favor of using matched devices. Also, when you are matching VGS, you do it by measuring the drain current. The devices are matched when the same gate voltage causes the same drain current to flow, so in the end, even though you're setting the gate voltage, the result is the same idle drain current following.
I also said that my example was extreme and not exactly real world, but here is a real world example. I have a lot of 25 Vishay IRF520 Mosfets, all with the same lot number, bought from a reputable supply house. The amount of gate voltage required to allow 100mA of drain current to flow varies by 0.117V between the highest and lowest. Using the 2 FETs that are the furthest apart, if I set the gate voltage to cause 100mA IDS on the lowest one, the same gate voltage on the highest one only flows 12mA. If they are paired in parallel biased like that, it creates a much higher level of IMD than if they are both biased to 100mA drain current rather than the same gate voltage. If you match the VGS(th) characteristic, then setting the same gate voltages will also cause the same drain current to flow. As I said above, this is an argument in favor of matching the FETs. You still need to know how much idle/quiescent drain current is flowing, otherwise, how do you know what class you have them biased in? Many power Mosfets can have well over 1V on the gate and still be in cutoff, class-c. To accurately set the bias and class of operation, I've only ever seen it done by measuring quiescent drain current. Here are some examples...
The Yaesu FT450 service manual specifies that the Mosfet biasing in the PA section be set as follows. The RD06HHF1 pre-driver is set to 100mA drain current. Each of the 2 RD16HHF1 drivers get set to 500mA drain current. And each of the 2 RD100HHF1 finals get set to 1A of drain current. No mention of gate voltages anywhere in the procedure. In both push-pull sections, drivers & finals, each Mosfet has it's own bias adjustment, thus negating the need to match VGS(th) between the 2 devices.
The Yaesu FT-1000MP MkV uses BLF145 driver and a pair of BLF147 finals, both 28V Mosfets. The service manual specifies a driver bias of 1.3A of drain current. The finals, in Class-AB mode each get set to 1A of drain current, and in Class-A mode each get set to 5A of drain current. Final biasing is separate for each of the 2 Mosfets here also, and again, no mention of gate voltages.
The Cobra 200GTL specifies it's final (before the 2x2290 PA section), an RD16HHF1 be biased to 1A (I believe) of idle current. The 150GTL uses 2xRD16HHF1 finals, they are biased by measuring drain current.
I could keep going through service manuals for rigs with Mosfet PA sections, but I think that I made my point. I also have manuals for standalone amplifiers, and application notes for amplifier designs that use Mosfets, every single one specifies bias settings in terms of idle/quiescent (no signal) drain current. I mean, to be completely honest here, the only place that I have ever seen Mosfet biasing specified in terms of gate voltage is in the world of cheap CB and export radios. I've had much better, and easily repeatable results by biasing them via drain current measurements.
I'm mostly into learning and building my own homebrew radios and amps. As such I've read and studied many books, manuals, application notes, and datasheets. What I've said it's just the overwhelming trend I've seen from these studies of mine and I'm sure other means exist. This seems to be the industry norm from what I've gathered. Finally, as far as the other things being BS, I'll assume you mean my mention of matching passives. At HF it doesn't seem important enough to matter unless you're doing something super special. I have seen quite a few designs and application notes for VHF and UHF circuits that will specify certain passive components be matched to 1% or something similar. It's not something that I pulled from my colon, but you'd be correct in saying it's mostly unnecessary at HF.

73s - Brad

 

[Tallman]

I'm not confused at all, you seem to be missing the point. Bipolars and Mosfets do share some characteristics, load line theory calculations are one of those things. How the device is being driven doesn't particularly matter, voltage or current. Now, you are correct that as long as VGS(th) is matched between the devices then you can set the bias via gate voltage and all of the devices will be biased the same, but this is just an argument in favor of using matched devices. Also, when you are matching VGS, you do it by measuring the drain current. The devices are matched when the same gate voltage causes the same drain current to flow, so in the end, even though you're setting the gate voltage, the result is the same idle drain current following.
I also said that my example was extreme and not exactly real world, but here is a real world example. I have a lot of 25 Vishay IRF520 Mosfets, all with the same lot number, bought from a reputable supply house. The amount of gate voltage required to allow 100mA of drain current to flow varies by 0.117V between the highest and lowest. Using the 2 FETs that are the furthest apart, if I set the gate voltage to cause 100mA IDS on the lowest one, the same gate voltage on the highest one only flows 12mA. If they are paired in parallel biased like that, it creates a much higher level of IMD than if they are both biased to 100mA drain current rather than the same gate voltage. If you match the VGS(th) characteristic, then setting the same gate voltages will also cause the same drain current to flow. As I said above, this is an argument in favor of matching the FETs. You still need to know how much idle/quiescent drain current is flowing, otherwise, how do you know what class you have them biased in? Many power Mosfets can have well over 1V on the gate and still be in cutoff, class-c. To accurately set the bias and class of operation, I've only ever seen it done by measuring quiescent drain current. Here are some examples...
The Yaesu FT450 service manual specifies that the Mosfet biasing in the PA section be set as follows. The RD06HHF1 pre-driver is set to 100mA drain current. Each of the 2 RD16HHF1 drivers get set to 500mA drain current. And each of the 2 RD100HHF1 finals get set to 1A of drain current. No mention of gate voltages anywhere in the procedure. In both push-pull sections, drivers & finals, each Mosfet has it's own bias adjustment, thus negating the need to match VGS(th) between the 2 devices.
The Yaesu FT-1000MP MkV uses BLF145 driver and a pair of BLF147 finals, both 28V Mosfets. The service manual specifies a driver bias of 1.3A of drain current. The finals, in Class-AB mode each get set to 1A of drain current, and in Class-A mode each get set to 5A of drain current. Final biasing is separate for each of the 2 Mosfets here also, and again, no mention of gate voltages.
The Cobra 200GTL specifies it's final (before the 2x2290 PA section), an RD16HHF1 be biased to 1A (I believe) of idle current. The 150GTL uses 2xRD16HHF1 finals, they are biased by measuring drain current.
I could keep going through service manuals for rigs with Mosfet PA sections, but I think that I made my point. I also have manuals for standalone amplifiers, and application notes for amplifier designs that use Mosfets, every single one specifies bias settings in terms of idle/quiescent (no signal) drain current. I mean, to be completely honest here, the only place that I have ever seen Mosfet biasing specified in terms of gate voltage is in the world of cheap CB and export radios. I've had much better, and easily repeatable results by biasing them via drain current measurements.
I'm mostly into learning and building my own homebrew radios and amps. As such I've read and studied many books, manuals, application notes, and datasheets. What I've said it's just the overwhelming trend I've seen from these studies of mine and I'm sure other means exist. This seems to be the industry norm from what I've gathered. Finally, as far as the other things being BS, I'll assume you mean my mention of matching passives. At HF it doesn't seem important enough to matter unless you're doing something super special. I have seen quite a few designs and application notes for VHF and UHF circuits that will specify certain passive components be matched to 1% or something similar. It's not something that I pulled from my colon, but you'd be correct in saying it's mostly unnecessary at HF.

73s - Brad

I'll yield on the quiescent current. My knowledge like me is old and have not designed any thing more complicated than a high frequency switch mode power supply and that was 1990. We had individual cards with IRF150 fets and they were calibrated with wave forms that did not involve source current.
So I yield to more current knowledge.

 

[loosecannon]

I'm not confused at all, you seem to be missing the point. Bipolars and Mosfets do share some characteristics, load line theory calculations are one of those things. How the device is being driven doesn't particularly matter, voltage or current. Now, you are correct that as long as VGS(th) is matched between the devices then you can set the bias via gate voltage and all of the devices will be biased the same, but this is just an argument in favor of using matched devices. Also, when you are matching VGS, you do it by measuring the drain current. The devices are matched when the same gate voltage causes the same drain current to flow, so in the end, even though you're setting the gate voltage, the result is the same idle drain current following.
I also said that my example was extreme and not exactly real world, but here is a real world example. I have a lot of 25 Vishay IRF520 Mosfets, all with the same lot number, bought from a reputable supply house. The amount of gate voltage required to allow 100mA of drain current to flow varies by 0.117V between the highest and lowest. Using the 2 FETs that are the furthest apart, if I set the gate voltage to cause 100mA IDS on the lowest one, the same gate voltage on the highest one only flows 12mA. If they are paired in parallel biased like that, it creates a much higher level of IMD than if they are both biased to 100mA drain current rather than the same gate voltage. If you match the VGS(th) characteristic, then setting the same gate voltages will also cause the same drain current to flow. As I said above, this is an argument in favor of matching the FETs. You still need to know how much idle/quiescent drain current is flowing, otherwise, how do you know what class you have them biased in? Many power Mosfets can have well over 1V on the gate and still be in cutoff, class-c. To accurately set the bias and class of operation, I've only ever seen it done by measuring quiescent drain current. Here are some examples...
The Yaesu FT450 service manual specifies that the Mosfet biasing in the PA section be set as follows. The RD06HHF1 pre-driver is set to 100mA drain current. Each of the 2 RD16HHF1 drivers get set to 500mA drain current. And each of the 2 RD100HHF1 finals get set to 1A of drain current. No mention of gate voltages anywhere in the procedure. In both push-pull sections, drivers & finals, each Mosfet has it's own bias adjustment, thus negating the need to match VGS(th) between the 2 devices.
The Yaesu FT-1000MP MkV uses BLF145 driver and a pair of BLF147 finals, both 28V Mosfets. The service manual specifies a driver bias of 1.3A of drain current. The finals, in Class-AB mode each get set to 1A of drain current, and in Class-A mode each get set to 5A of drain current. Final biasing is separate for each of the 2 Mosfets here also, and again, no mention of gate voltages.
The Cobra 200GTL specifies it's final (before the 2x2290 PA section), an RD16HHF1 be biased to 1A (I believe) of idle current. The 150GTL uses 2xRD16HHF1 finals, they are biased by measuring drain current.
I could keep going through service manuals for rigs with Mosfet PA sections, but I think that I made my point. I also have manuals for standalone amplifiers, and application notes for amplifier designs that use Mosfets, every single one specifies bias settings in terms of idle/quiescent (no signal) drain current. I mean, to be completely honest here, the only place that I have ever seen Mosfet biasing specified in terms of gate voltage is in the world of cheap CB and export radios. I've had much better, and easily repeatable results by biasing them via drain current measurements.
I'm mostly into learning and building my own homebrew radios and amps. As such I've read and studied many books, manuals, application notes, and datasheets. What I've said it's just the overwhelming trend I've seen from these studies of mine and I'm sure other means exist. This seems to be the industry norm from what I've gathered. Finally, as far as the other things being BS, I'll assume you mean my mention of matching passives. At HF it doesn't seem important enough to matter unless you're doing something super special. I have seen quite a few designs and application notes for VHF and UHF circuits that will specify certain passive components be matched to 1% or something similar. It's not something that I pulled from my colon, but you'd be correct in saying it's mostly unnecessary at HF.

73s - Brad

Blasphemy000,

I'm curious to know how much of a difference in a radio's performance one might notice from changing out a set of FETs in a radio for a set of matched FETs?

I know there will be differences from radio design to radio design, so i'll just use a common example from the CB world.
the Stryker 955's service manual states to set the bias of the FET finals using quiescent current, and there is only one adjustment for all.
If the FETs were set using this method, what would the real world performance difference be with regards to a stryker 955?

from what you stated, it sounds to me like if the FETs are set using the quiescent current, then there really wouldn't be a need for matching them, and there wouldn't be a noticeable difference in performance if you did.
LC

 

[blasphemy000]

I'll yield on the quiescent current. My knowledge like me is old and have not designed any thing more complicated than a high frequency switch mode power supply and that was 1990. We had individual cards with IRF150 fets and they were calibrated with wave forms that did not involve source current.
So I yield to more current knowledge.

Building SMPSs is fun. That's how I got into electronics and radio in the beginning. Very cool.

 

[Tallman]

from what you stated, it sounds to me like if the FETs are set using the quiescent current, then there really wouldn't be a need for matching them, and there wouldn't be a noticeable difference in performance if you did.
LC

Like I said in my earlier post I intentionally mismatched FET's and the balance resistor did not get hot. That's where the rubber meets the road. The tune up procedure could be unit specific to that model of radio.

 

[blasphemy000]

Blasphemy000,

I'm curious to know how much of a difference in a radio's performance one might notice from changing out a set of FETs in a radio for a set of matched FETs?

I know there will be differences from radio design to radio design, so i'll just use a common example from the CB world.
the Stryker 955's service manual states to set the bias of the FET finals using quiescent current, and there is only one adjustment for all.
If the FETs were set using this method, what would the real world performance difference be with regards to a stryker 955?

from what you stated, it sounds to me like if the FETs are set using the quiescent current, then there really wouldn't be a need for matching them, and there wouldn't be a noticeable difference in performance if you did.
LC

I've never had a 955 on my bench, so this will be an "on-paper" analysis based on experience with similar circuits, so a grain of salt should be added.
The 955 has 4 identical Mosfets in it's final chain, setup in a 1 pre-driver into 1 driver into 2 linear PA section. The pre-driver and driver are attached to the AM regulator and provides the modulation in AM mode. Each of these 2 stages have their own individual biasing adjustment. Now, the final PA stage is single-ended and has 2 Mosfets directly in parallel with only a single bias adjustment. These 2 should be matched together due to the single bias adjustment. Let's say (I haven't looked it up) this stage gets biased at 100mA of drain current (50mA each). With the single adjustment, if the Mosfets aren't matched, one could be biased at 80mA while the other is only at 20mA, but it will still read 100mA at the measurement point. The more they are mismatched, the worse this can be. This can result in one final running much hotter than the other, even to the point of destruction if it's bad enough, and can also be the source of distortion and IMD products. Setting the bias via measuring the drain current isn't what fixes this issue. The two ways of fixing it are either matching the Mosfets so they share the idle current equally (the easiest way in the case of the Stryker), or providing a separate biasing adjustment for each of the 2 Mosfets (this is something that would be done when designing the circuit). As for the quantifiable performance improvements, well that depends on how badly mismatched the original Mosfets were. Most of the improvements will come in the form of balanced heat output and lower distortion and IMD products. How much less would be something that would have to be measured to see if it was worth it for those benefits, but balancing the heat output would be worth it to me from a reliability standpoint, but again, how much of a mismatch was there beforehand will dictate the amount of improvement. Now, from a power output standpoint, there is essentially zero performance to be gained there. Reducing distortion and IMD will allow more of the output power to be focused within the channel bandwidth. Since distortion and IMD will cause emission outside of the desired frequency, none of that power is actually helping you talk further, but again, how much of your total power is contained in those off-frequency emissions before and after will determine how much was gained. Back to the 955 specifically though, yes, the 2 parallel finals should be matched, within reason, since there is only one bias adjustment for both together. That's my opinion on it anyhow.

 

[blasphemy000]

Like I said in my earlier post I intentionally mismatched FET's and the balance resistor did not get hot. That's where the rubber meets the road. The tune up procedure could be unit specific to that model of radio.

What circuit though? What do you mean by balance resistor? Are we talking in a power combiner? Some circuits matter and some don't care. A Class-E switching RF amplifier doesn't really care at all about matching the Mosfets because they're being switched completely on and off. It operates more like a resonant switch-mode power supply. A regular Class-AB linear amplifier is operating the Mosfets in their linear/constant-current region, and has different requirements.

 

[rabbiporkchop]

Back to the 955 specifically though, yes, the 2 parallel finals should be matched, within reason, since there is only one bias adjustment for both together. That's my opinion on it anyhow.

That's not just an opinion. That's common sense at its finest.

 

[loosecannon]

I've never had a 955 on my bench, so this will be an "on-paper" analysis based on experience with similar circuits, so a grain of salt should be added.
The 955 has 4 identical Mosfets in it's final chain, setup in a 1 pre-driver into 1 driver into 2 linear PA section. The pre-driver and driver are attached to the AM regulator and provides the modulation in AM mode. Each of these 2 stages have their own individual biasing adjustment. Now, the final PA stage is single-ended and has 2 Mosfets directly in parallel with only a single bias adjustment. These 2 should be matched together due to the single bias adjustment. Let's say (I haven't looked it up) this stage gets biased at 100mA of drain current (50mA each). With the single adjustment, if the Mosfets aren't matched, one could be biased at 80mA while the other is only at 20mA, but it will still read 100mA at the measurement point. The more they are mismatched, the worse this can be. This can result in one final running much hotter than the other, even to the point of destruction if it's bad enough, and can also be the source of distortion and IMD products. Setting the bias via measuring the drain current isn't what fixes this issue. The two ways of fixing it are either matching the Mosfets so they share the idle current equally (the easiest way in the case of the Stryker), or providing a separate biasing adjustment for each of the 2 Mosfets (this is something that would be done when designing the circuit). As for the quantifiable performance improvements, well that depends on how badly mismatched the original Mosfets were. Most of the improvements will come in the form of balanced heat output and lower distortion and IMD products. How much less would be something that would have to be measured to see if it was worth it for those benefits, but balancing the heat output would be worth it to me from a reliability standpoint, but again, how much of a mismatch was there beforehand will dictate the amount of improvement. Now, from a power output standpoint, there is essentially zero performance to be gained there. Reducing distortion and IMD will allow more of the output power to be focused within the channel bandwidth. Since distortion and IMD will cause emission outside of the desired frequency, none of that power is actually helping you talk further, but again, how much of your total power is contained in those off-frequency emissions before and after will determine how much was gained. Back to the 955 specifically though, yes, the 2 parallel finals should be matched, within reason, since there is only one bias adjustment for both together. That's my opinion on it anyhow.

I completely understand the theory behind using a set of matched transistors, but my question is specifically about the real world performance of the stryker 955.

Yes, there should be a separate bias adjustment for each transistor, and on anything higher priced than a CB radio there would be.
Do strykers have a bad reputation for blowing mosfets due to over heating?
no. on the contrary they have a great reputation.

From the information you posted, i think i can safely assume that you know the difference between heat conduction and heat convection when it comes to how semiconductors heat up when attached to a common heatsink.
My guess is that Stryker put that brass piece in front of the mosfets in an attempt to even out the heat that might be stronger on one device than the one next to it.
how well that works is up for debate and i am just guessing.

for your example of one fet pulling 80mA and another pulling 20, there would be an imbalance, but if the fets are not blowing and the radio can be operated normally as it came from the factory, who would really notice the difference from changing out the fets for a matched pair?
would it bother you every time you keyed up knowing that one of your fets is pulling more current than another?
would any receiver that is listening to your radio be able to tell the difference between how it sounded before and how it sounded afterwards?
please remember that i am talking about how this style of radio actually operates on RX and TX and not the theoretical differences.

I guess what im asking is, would the juice be worth the squeezin' with regards to someone sending a working radio to a shop to have the fets matched?
I feel like this is just another money grab by CB shops that prey on the fact that their customers trust their knowledge and don't know any better.
LC

 

[nomadradio]

If it ain't broke, just what is it that needs fixing?

Precautions and procedures to make a legitimate repair last longer should always be followed. That means checking the idle current of new RF transistors, bipolar or FET. A newly-installed part almost never just comes up adjusted correctly. Now and again, a bias trimpot will be set correctly by accident.

But if it ain't broke? Not much to be gained by cracking it open just to check that, unless there is some performance problem to justify it.

And then there's the "rule of the shade tree".

If it ain't broke, fix until it is.

73

 

[blasphemy000]

I guess what im asking is, would the juice be worth the squeezin' with regards to someone sending a working radio to a shop to have the fets matched?
I feel like this is just another money grab by CB shops that prey on the fact that their customers trust their knowledge and don't know any better.
LC

If they are already within 5-10% (usually what ham radios specify for bias current tolerances), then no, you'd never notice the difference between that and a tightly matched set of the same part number. Now if you happened to get one that was unlucky and they were way off, such as the extreme 80/20 example I gave above, sure, it would be worth swapping them. How many are actually unlucky enough to get a set that's that badly mismatched though? It has to be pretty unlikely. Just properly setting the biasing cleaned up the 2 655s that I used to have. That's way too small of a sample size to be used as gospel though.

 

[loosecannon]

thanks for clearing that up.
Shops that are charging people money to change out perfectly good FETs for matched sets claiming that there will be a noticeable change in performance are just taking advantage of people like some CB shops have always done.
LC

 

[Highway Man]

i thought the engineers really know what they are doing so why all the part changing from FETS to knobs to encoders to regulators to.....
do you follow what I'm saying?

 

[rabbiporkchop]

would any receiver that is listening to your radio be able to tell the difference between how it sounded before and how it sounded afterwards?

Just by switching out the parts you may or may not see a difference. It depends on how much Distortion you have to begin with and at what frequency that Distortion occurs.

Shops that are charging people money to change out perfectly good FETs for matched sets claiming that there will be a noticeable change in performance are just taking advantage of people like some CB shops have always done.

That's true hack shops that don't have a clue how to tune are definitely taking people's money by swapping out parts without doing a sweep and showing zero distortion anywhere. If you only have one adjustment for two components in parallel the only way to eliminate distortion is to make sure those components are perfectly matched and bias them accordingly. Plenty of idiots swapping out parts without any clue how to bias them. If they can't tune a simple Cobra 138 then they definitely can't tune something like a President Lincoln 2.

i thought the engineers really know what they are doing so why all the part changing from FETS to knobs to encoders to regulators to.....
do you follow what I'm saying?

The engineers really know how to design a radio that performs a certain way for a specific price point to allow for maximum profit. If you want an export radio that performs like an Icom 7300 and lasts a long time then you'll have to replace some components and realign the thing

 

[Onelasttime]

I almost bought a Rocketbox in kit form myself. I am glad I waited. While they look like they would be great they just do not hold up and self destruct fairly quickly and with regularity!

So I have not owned one. Plenty of people can make cheap non-rf FET's work but most can not. Well not with any kind of consumer level durability.

Sadly due to ignorance in there market I do not know if the amp is poorly designed or if it is all customer error?

I know one guy that designs amps that just plane work but his amps are not cheap and they are not insane output power amps either so not 34 Pill boxes or anything over 4 transistors as far as I know. His bipolar and FET designs all work as advertised.

In general I like external amplifiers that use output transistors intended for the production of RF and I do not really care if they are bipolar, FET, LDMOS or anything else. I think we have seen that if gain is not an issue and you are not going to push them hard that non-rf fet's can be pushed into service and survive ok but not great in consumer 2 way radios as low power finals.

The second you abuse them with either trying to get as much power as you can out of them or swr higher than 1.5:1 they pop like party poppers!

You really are best keeping them at stock output power and making sure you have a FANTASTIC swr and that the box's input and output tune are spot on.

Clearly in RF output situations matched transistors are always better but in this case with non-RF FET's I wonder if it makes that much difference? They truly are the wrong part for the job with none of the characteristics one wants in and output transistor for RF. To put it bluntly there only redeeming quality is that they are insanely CHEAP!

In fact even a cheap bipolar like DEI or PP are better than a non-rf FET! If affordably and more durable than a non-RF FET is desired purchasing a properly designed and built class B almost AB amp with HG's would be far better than any amp using non-rf FET's! "Buy once, Cry Once!" not sure who it was that taught me that saying but I think it was on this sight!