first of all read the datasheet. the part takes 24mA and keeps its pin 1 at about 5V. they have used a 220ohm in their Vcc line. which drops 5.3V at 24mA. So I bet pin 1 is just below 7 V when your Vcc is 12V. At 9V you are supplying the device with about 4 V. Pretty much all of these MMIC amplifiers (any brand, any shape) usually have a constant current consumption when the voltage at vcc pins passes a certain threshold (5V here). I have only seen very few exceptions to this. second w.r.t. the multiple de caps on vcc line it's a common practice for tuning after production. by positioning the cap in the right distance you create the "shorted" quarter wavelength microstrip line (that is required for the bias T) between the pin 1 and the decoup cap. knowing the exact wavelength in the transmission line is a bit difficult so in this way they can place the cap in multiple locations and measure to find the best spot which is where the cap is quarter wave away from pin 1. (remember the rule is pin 1 of the chip must see "AC open" circuit looking into the bias line
YES, they are almost always a two transistor cascade. Common collector driving a common emitter stage. Hybrid Pi model is a good start at finding the 2cnd stage Re. You can work backwards from the load Z (50r) with roughly a gain of 40 (32dB) to approximate Re. Gain = Zl/Re if all the gain is second stage. Temperature and current and doping of the crystal are what derives Ic/Vt ~.026 Gm/~.026 = Re { small signal silicon low freq calculation here - I don't know the crystal semiconductor material - probably some type of gallium, gold, arsenic mixed in for GHz range and possibly it could be germanium } I got 'brain fog' on this stuff - haven't done it in 30 years - GO look it up to be safe Hfe is in there, too Re has to be about 1.25r to get Av = 40 into 50r load at 24ma. There is the clue to the operating temp of the die and mix of materials. Did he say 32dB 'POWER' gain or voltage gain... work the math backwards from gain, load and input Z's and operating current - you get the idea. The funky prototype micro strip / strip line long trace with cap placement merely is a cheap trick to build LC filters the easy way; the designer was implying all the possible bias T sort of arrangements ( that kind of work is typically handled by the 'new guy' that doesn't have experience of senior RF design, so he does what's adjustable - and maybe his boss is a smart cookie and told him to do it that way ) The board you bought with SMA's has resistor dropping for Vc, so it will need higher voltage like the MAR series micro X chips. And those tiny 'stubs' are series notch filters at harmonics above the freq band of the MMIC. Caps at those freqs also act as series notches; thus the multitude of cap positions/possibilties in the proto board suggestion. edit: Filters above about 500MHz become distributed elements, no longer lumped element - sooo, even tiny edge effects are capacitive to ground and the smallest of lengths must include inductance (including ground plane 'loop' distributed L - that's why only one cap position was chosen for the specs) Leaving tiny tuning stubs for notches at say 5 or 8 GHz to avoid interference becomes a matter of experience and measurements. There are free and professional grade finite element analysis software packages to design complex stripline/microstrip filters. The difference is in the resolution of element size, materials supported, and E & H field calculations (cost). Semiconductor fabs are not likely to own the SOTA software like a defense contractor would have, sooo, my guess is these guys at NXP drew the proto board 'seat of the pants'
Ya... reading data sheets gets in the way of making videos. Or... maybe IMSAI guy is just 'pretending' he doesn't read data sheets? ... you know... to get us to make comments....nah... it's IMSAI Guy.
@@willthecat3861 He's an optoelectronics PHYSICS specialist, seat of the pants self taught electronics. Sometimes he is simply clumsy and others it Is to encourage beginners 👍 Shahir over at Signal Path channel is the semiconductor physics PHD who can definitely explain the devices - look on his channel if he has done MMIC's in the past.
@@willthecat3861 Used to love going over the new data books the rep left after a visit. Then getting on the phone to try and get a few free samples to play with at home. Funny but all the predicted sales got dropped when the customers pulled out. 🙂
Need to read a paper from analog devices “MT-101:Decoupling Techniques “ and “Decoupling Capacitor Placement Guidelines” by Sierra Circuits. It’s really clever how the go about eliminating as much noise as possible by implementing this type of voodoo
Everyone loves chip of the day! They say you should design your power distribution network like a transmission line. This board has taken that to an extreme. It would be very interesting to see if there are really significant improvements over the standard practice of putting the capacitor as close to the pin as possible...
Sounds about right the Chinese PCB has the 220Ohm resisitor giving the correct voltdrop /current limit .. Note on NXPs layout there is no current limiting only that cap creating the 1/4 Wave stub.. Fun having questions to the audience Keep it up ta!!
I've encountered something vaguely similar, but using an SMT inductor and variable cap, not a pcb trace with selectable length. It's an RF z-match, Pi topoplogy "transmatch" (with an inductor horizontal, with grounded caps at either end. The IC's input capacitance would provide the second one.) But those are resonant circuits. They might be massaging the Z-in at the high end, to make it more flat for broadband. Or, maybe adding a broad peak at the high end, which gives a flat output, but with higher upper -3dB freq. I was gonna say, put it on an HP network analyzer, and hold the cap in various positions w/tweezers, to change the GHz resonant frequency of that trace. (Also, compare to the spectrum which results while connecting to the input pin, with the weird part entirely cut out.) Or, remove the chip, add a ?1? pF cap to simulate the Cinp of the input pin, and use network analyzer to see how it behaves. PS There once was a forum for vintage HP 5700 network analyzers. They had 2723 eproms for updating your older unit to add the Smith charts feature. Also, they point out that the display is color VGA with colored traces, and some people pull out the green CRT and replace it with color VGA of the same size. Me, I just dropped a 14in color VGA display on top, and plug it into the rear VGA connector. Fasten it down with duct tape, so it doesn't fall over when wheeling the whole thing around on a tilted Tektronix scopecart. Also, I found a cheap HP that was marked as a 75-ohm instrument, but someone had modded the boards to make it 50ohms.
I have the same board and I have been operating it at 12V no problem. I even once powered it prom a lead-acid battery 13-14V, still works. I remember the Aliexpress listing saying that the gain scales with voltage from 5V to 8V.
usually the Vcc line is not so critical for stubs, transmission lines etc, but its sure some kind of adjustment of impedance, also the small stub at the output is for impedance matching - "The location of the 470 pF supply decoupling capacitor (Cdec) can be precisely chosen for optimum performance. matching," this from the datasheet shows that it is important part of the schematic. Because lack of the internal schematic and and formulars, it can be done only with trial and I think wil depend on the used frequency. On the other side the stub at the output is also ciritical for impedance matching to 50 Ohm, I had similar cases with oder MMICs on 2.4GHz (Transceiver in Cubesat), very very critical positioning, but I never come upon such critical structure in the VCC part of a device, suualy as said an CLC network will do, sometimes more ciritical when the Vcc is overlayed into the output pin, but then half circle filters can be used to block frequencies. Maybe in this case the VCC transmission line and stub is part of the output circuit.
How hard were you driving it? Could it have been a compression problem at the lower voltages? The stub may have been to suppress parasitics at 10-14GHz.
I really like your chip of the day series. It introduces us to useful parts that otherwise we'd not be aware of. If I may be so bold to have a suggestion, I think it would be very interesting and informative if you did a couple of additional tests. Such as characterizing the input range. See where distortion may start, for example. Testing at frequencies not in specified range. I find sometimes a part not rated for say hf performs well there regardless. Enjoy your channel, and look forward to it every morning.
Yes, that PCB L- or Pi-filter instinctively does make sense. If it's intended for LNB applications, then it's operating around 1 to 2 GHz. Thus we're into the so-called "magic" territory, given the current era. PCB layout is part of the RF design, as is very commonly seen. Not so sure about the two stubs between the chip's Vcc and output. File under "magic". LOL 🙂
As said @makerspace533, you should insert an attenuator at the input of the amplifier or decrease the output voltage of the generator because at low supply voltage, you saturate the mmic input sooner and so the fact that the transistor doesn't accept the total peak to peak voltage is seen as an higher impedance by the generator. I hope it's clear because English is not my mother language... 😊 the input does a "burp" too when saturated 🤣 When the saturation begins, if you continue to increase the input signal on the generator, the voltage continue to increase but the current not. This is seen by the generator as an higher and higher impedance...
Based on your video I purchased two packages of the WA5VJB MMIC amp boards. They are really nice little boards and a real bargain. Have you priced the OEM eval boards for these sort of things? Wow.
i supply 12v and it has temp around 50-60c with room temp around 30c where i live. it has been running continuously for about 6 months and no signal at output anymore.
So instead of a DC input inductor they are tuning a piece of transmission line based on a fixed test frequency. You didn't mention what the trace width was or the PCB thickness and material type specified for the layout. Similar to a low-pass filter you might use on a transmitter output stage, just simpler. Also with 9V in, what was the voltage at the chip? I'm guessing that they have an RC network dropping some voltage given the chip draws about 24ma.
Pi Network - series L - shunt C - decoupling? Move the caps along the inductance to give most effective decoupling at a particular frequency? That was wild to see the input Z change so much with Vcc; that blew my mind!
If you do 'chip of the day', please include an original one, and compare that to the AliExpress chip. No we see a chip that may or may not be the 'chip of the day' chip...
first of all read the datasheet. the part takes 24mA and keeps its pin 1 at about 5V. they have used a 220ohm in their Vcc line. which drops 5.3V at 24mA. So I bet pin 1 is just below 7 V when your Vcc is 12V. At 9V you are supplying the device with about 4 V. Pretty much all of these MMIC amplifiers (any brand, any shape) usually have a constant current consumption when the voltage at vcc pins passes a certain threshold (5V here). I have only seen very few exceptions to this.
second w.r.t. the multiple de caps on vcc line it's a common practice for tuning after production. by positioning the cap in the right distance you create the "shorted" quarter wavelength microstrip line (that is required for the bias T) between the pin 1 and the decoup cap. knowing the exact wavelength in the transmission line is a bit difficult so in this way they can place the cap in multiple locations and measure to find the best spot which is where the cap is quarter wave away from pin 1. (remember the rule is pin 1 of the chip must see "AC open" circuit looking into the bias line
Was going to point out the supply resistor affecting the device vcc, but you beat me to it 😊
YES, they are almost always a two transistor cascade. Common collector driving a common emitter stage. Hybrid Pi model is a good start at finding the 2cnd stage Re. You can work backwards from the load Z (50r) with roughly a gain of 40 (32dB) to approximate Re. Gain = Zl/Re if all the gain is second stage. Temperature and current and doping of the crystal are what derives Ic/Vt ~.026 Gm/~.026 = Re { small signal silicon low freq calculation here - I don't know the crystal semiconductor material - probably some type of gallium, gold, arsenic mixed in for GHz range and possibly it could be germanium } I got 'brain fog' on this stuff - haven't done it in 30 years - GO look it up to be safe Hfe is in there, too
Re has to be about 1.25r to get Av = 40 into 50r load at 24ma. There is the clue to the operating temp of the die and mix of materials. Did he say 32dB 'POWER' gain or voltage gain... work the math backwards from gain, load and input Z's and operating current - you get the idea.
The funky prototype micro strip / strip line long trace with cap placement merely is a cheap trick to build LC filters the easy way; the designer was implying all the possible bias T sort of arrangements ( that kind of work is typically handled by the 'new guy' that doesn't have experience of senior RF design, so he does what's adjustable - and maybe his boss is a smart cookie and told him to do it that way )
The board you bought with SMA's has resistor dropping for Vc, so it will need higher voltage like the MAR series micro X chips.
And those tiny 'stubs' are series notch filters at harmonics above the freq band of the MMIC. Caps at those freqs also act as series notches; thus the multitude of cap positions/possibilties in the proto board suggestion.
edit: Filters above about 500MHz become distributed elements, no longer lumped element - sooo, even tiny edge effects are capacitive to ground and the smallest of lengths must include inductance (including ground plane 'loop' distributed L - that's why only one cap position was chosen for the specs) Leaving tiny tuning stubs for notches at say 5 or 8 GHz to avoid interference becomes a matter of experience and measurements. There are free and professional grade finite element analysis software packages to design complex stripline/microstrip filters. The difference is in the resolution of element size, materials supported, and E & H field calculations (cost).
Semiconductor fabs are not likely to own the SOTA software like a defense contractor would have, sooo, my guess is these guys at NXP drew the proto board 'seat of the pants'
Ya... reading data sheets gets in the way of making videos. Or... maybe IMSAI guy is just 'pretending' he doesn't read data sheets? ... you know... to get us to make comments....nah... it's IMSAI Guy.
@@willthecat3861 He's an optoelectronics PHYSICS specialist, seat of the pants self taught electronics. Sometimes he is simply clumsy and others it Is to encourage beginners 👍
Shahir over at Signal Path channel is the semiconductor physics PHD who can definitely explain the devices - look on his channel if he has done MMIC's in the past.
@@willthecat3861 Used to love going over the new data books the rep left after a visit. Then getting on the phone to try and get a few free samples to play with at home. Funny but all the predicted sales got dropped when the customers pulled out. 🙂
Need to read a paper from analog devices “MT-101:Decoupling Techniques “ and “Decoupling Capacitor Placement Guidelines” by Sierra Circuits. It’s really clever how the go about eliminating as much noise as possible by implementing this type of voodoo
breakfast didn't agree with you today!
I thought that was low-key commentary about the quality of the amp!
I thought the burp was an auditory rendering of the bad transfer function. 😂
You have 220R in series on VCC, at 24 mA supply current it's about 5.2V of voltage drop on it, so about 7 volts on VCC pin👌
Oh, this feels like especially sinister RF magic. Do not meddle in the affairs of wizards, for they are subtle and quick to anger.
As Landro was to learn😂😂
Everyone loves chip of the day! They say you should design your power distribution network like a transmission line. This board has taken that to an extreme. It would be very interesting to see if there are really significant improvements over the standard practice of putting the capacitor as close to the pin as possible...
Sounds about right the Chinese PCB has the 220Ohm resisitor giving the correct voltdrop /current limit .. Note on NXPs layout there is no current limiting only that cap creating the 1/4 Wave stub.. Fun having questions to the audience Keep it up ta!!
I've encountered something vaguely similar, but using an SMT inductor and variable cap, not a pcb trace with selectable length. It's an RF z-match, Pi topoplogy "transmatch" (with an inductor horizontal, with grounded caps at either end. The IC's input capacitance would provide the second one.) But those are resonant circuits. They might be massaging the Z-in at the high end, to make it more flat for broadband. Or, maybe adding a broad peak at the high end, which gives a flat output, but with higher upper -3dB freq.
I was gonna say, put it on an HP network analyzer, and hold the cap in various positions w/tweezers, to change the GHz resonant frequency of that trace. (Also, compare to the spectrum which results while connecting to the input pin, with the weird part entirely cut out.)
Or, remove the chip, add a ?1? pF cap to simulate the Cinp of the input pin, and use network analyzer to see how it behaves.
PS
There once was a forum for vintage HP 5700 network analyzers. They had 2723 eproms for updating your older unit to add the Smith charts feature. Also, they point out that the display is color VGA with colored traces, and some people pull out the green CRT and replace it with color VGA of the same size. Me, I just dropped a 14in color VGA display on top, and plug it into the rear VGA connector. Fasten it down with duct tape, so it doesn't fall over when wheeling the whole thing around on a tilted Tektronix scopecart.
Also, I found a cheap HP that was marked as a 75-ohm instrument, but someone had modded the boards to make it 50ohms.
How hot was it getting at 12 volts? It may have internal active bias control to place the gain element at the ideal level.
Ironic, a part with BGA in the name comes in a SOT package.
is there a dropping R in the vcc?
I have the same board and I have been operating it at 12V no problem. I even once powered it prom a lead-acid battery 13-14V, still works. I remember the Aliexpress listing saying that the gain scales with voltage from 5V to 8V.
5:11 the position of the cap devides the long path in two(!) inductors - one to the chip and one to the supply
correct, making it into a L-C-L lowpass filter
@ and the left/right position is for a second one for trimming
Next to the marking "5" you can see an SMD resistor. That seems to be the secret why you can apply up to 12V
wow and peace be upon you sir from me
usually the Vcc line is not so critical for stubs, transmission lines etc, but its sure some kind of adjustment of impedance, also the small stub at the output is for impedance matching - "The location of the 470 pF supply decoupling capacitor (Cdec) can be precisely chosen for optimum performance. matching," this from the datasheet shows that it is important part of the schematic. Because lack of the internal schematic and and formulars, it can be done only with trial and I think wil depend on the used frequency. On the other side the stub at the output is also ciritical for impedance matching to 50 Ohm, I had similar cases with oder MMICs on 2.4GHz (Transceiver in Cubesat), very very critical positioning, but I never come upon such critical structure in the VCC part of a device, suualy as said an CLC network will do, sometimes more ciritical when the Vcc is overlayed into the output pin, but then half circle filters can be used to block frequencies. Maybe in this case the VCC transmission line and stub is part of the output circuit.
I think those are probably WiFi and cellular interference stubs for series notch to ground - not in band SWR matching
How hard were you driving it? Could it have been a compression problem at the lower voltages? The stub may have been to suppress parasitics at 10-14GHz.
I really like your chip of the day series. It introduces us to useful parts that otherwise we'd not be aware of.
If I may be so bold to have a suggestion, I think it would be very interesting and informative if you did a couple of additional tests. Such as characterizing the input range. See where distortion may start, for example. Testing at frequencies not in specified range. I find sometimes a part not rated for say hf performs well there regardless.
Enjoy your channel, and look forward to it every morning.
0:30 LNB sometimes (oftentimes?) means Low-Noise (yes, yes....we know that...) Block downconverter, as used with satellite TV dishes.
Yes, that PCB L- or Pi-filter instinctively does make sense.
If it's intended for LNB applications, then it's operating around 1 to 2 GHz. Thus we're into the so-called "magic" territory, given the current era.
PCB layout is part of the RF design, as is very commonly seen.
Not so sure about the two stubs between the chip's Vcc and output. File under "magic". LOL 🙂
As said @makerspace533, you should insert an attenuator at the input of the amplifier or decrease the output voltage of the generator because at low supply voltage, you saturate the mmic input sooner and so the fact that the transistor doesn't accept the total peak to peak voltage is seen as an higher impedance by the generator. I hope it's clear because English is not my mother language... 😊 the input does a "burp" too when saturated 🤣
When the saturation begins, if you continue to increase the input signal on the generator, the voltage continue to increase but the current not. This is seen by the generator as an higher and higher impedance...
I don't know either😁😁
08:45
The resistor on VCC? @6:19
It has a 270 Ω resistor so you have lower voltage at PIN 1.
I would measure the voltage on pin 1 when your power supply is at 9 volts or 12 volts etc... I thought I saw a dropping resistor in the Vcc line?
Based on your video I purchased two packages of the WA5VJB MMIC amp boards. They are really nice little boards and a real bargain. Have you priced the OEM eval boards for these sort of things? Wow.
I wonder what temperature the part was at 12V.
Have some fresh coffee.
i supply 12v and it has temp around 50-60c with room temp around 30c where i live. it has been running continuously for about 6 months and no signal at output anymore.
A couple of people are mentioning about "did IMSAI Guy measure the voltage at the power pin, on the chip?" Somehow... I think not.
So instead of a DC input inductor they are tuning a piece of transmission line based on a fixed test frequency. You didn't mention what the trace width was or the PCB thickness and material type specified for the layout. Similar to a low-pass filter you might use on a transmitter output stage, just simpler. Also with 9V in, what was the voltage at the chip? I'm guessing that they have an RC network dropping some voltage given the chip draws about 24ma.
Chinese copy of the IC?
Component location ?
Stubs ?
-
Welcome to microwave RF.
microwave RF = black magic!
@zebo-the-fat yup, especially around 1GHz 😁
I actually though those were tubing stubs or for tuning caps at first 🤣
Amazing what can be done with such little parts, and that was a decade ago. Amplification range DC all the way to 2ghz, neat.
Pi Network - series L - shunt C - decoupling? Move the caps along the inductance to give most effective decoupling at a particular frequency? That was wild to see the input Z change so much with Vcc; that blew my mind!
I wonder if adjusting the Vcc decoupling will affect input/output impedance?
Was it hot at 12v ?
Maybe these are rejects that work but not to spect.
If you do 'chip of the day', please include an original one, and compare that to the AliExpress chip. No we see a chip that may or may not be the 'chip of the day' chip...