Recently, there have been a lot of chip shocks. From the initial price increase of chips, many chips cannot be bought now. Chip prices can still continue to bite the bullet, as long as the procurement cost is lower than the cost of redesign and proofing. The worst thing is not being able to buy chips, and they are directly forced to fail to produce. If it’s an analog chip, it’s okay, only the circuit needs to be changed, but if the digital chip also involves changing a bunch of software, it’s a headache.
HackRF should have encountered similar problems recently. A friend told me that a part model was changed in the BOM. In the latest version of the BOM, several sky13350 are replaced by sky13453, and the PCB silk screen is changed to the 2021 version. I checked github and found that the firmware and drivers have been modified. Suddenly panicked, is the new version of firmware and driver going to abandon the old one? What about my PortaPack? Although I can upgrade the firmware of my PortaPack according to the latest firmware of HackRF, what should I do for buyers who are still using the 2014 HackRF? Are you going to abandon them and stop updating them?
I immediately downloaded the latest hackrf repository. First searched for words such as sky, 13350 and 13453 with grep in the firmware directory, and found that neither comments nor variable names were mentioned. At this time, I was relieved a lot. It seems that this chip has nothing to do with the driver in the firmware. It means that it is at least not a digital chip, and it is probably only an analog chip. But I am still not at ease, if the pin logic of the digital chip that controls it changes, there will be problems, but this is not easy to check through a simple search.
Later, I went to the hardware directory and looked at the PCB, and it was really different. At this time, the worry value increased by a few points.
Until I patiently looked at the schematic. Found the location of the sky chip, in the frontend part. In fact, there are several, but there are certain rules.
1. The topology of the circuit has not changed.
2. There is no difference between the functions corresponding to the names of the PINs. They are all 1 input, 2 outputs, and the control signal VCTL.
3. The old version on the left has 2 logic control ports (1 and 5), and the new version on the right only has 1 logic port (5), and the other (1) is changed to a power supply port.
4. Judging from the network label name, the logic of the old version of the 5-port and the new version of the 5-port is opposite. It is the relationship between RX and TX, or the relationship between LP and HP. There is no screenshot but I can see it. No relationship with ! (programmers understand what it means).
So I just find the datasheets for both chips and check them out
A. Their functions are indeed switches, switching one radio frequency port to the other two radio frequency ports respectively
B. The new chip can be controlled with 1 pin, and the old chip needs to be controlled with 2 pins.
C. The control pins of the new chip are just the opposite of the logic of the corresponding pins of the old chip.
I first searched on the Easy Mall, and the old chip was out of stock, even the data sheet was gone, and the new one still had a data sheet. I had no choice but to look for sky13350 on digikey, and I found its data sheet, and I found a substitute at the bottom, which was written as sky13453. It seems that the function is indeed the same.
Open the data sheet and find that the two are indeed similar. It's just that the old chip on the left supports wifi standards up to g, and the new chip supports up to n. I think this just depends on the release year, and has nothing to do with performance and bandwidth. Of course, we can look at the specific radio frequency parameters later. The description in the lower right corner is very clear, the functions are the same, they are all RF switches, then A confirms that there is no problem.
Looking down, several pins of 2346 are indeed corresponding on the two chips, 24 are the two controlled channels, 3 is the ground, and 6 is the input RF channel. 1 On the old chip, VCTL1 corresponds to the power supply port of the new chip, which is consistent with the hackrf schematic diagram. Port 5 on the old chip is VCTL2 corresponding to the new chip VCTL. This means that B is also confirmed. The new chip replaces the 2 control pins of the old chip with 1 control pin. This is also logical, because the two control pins of the old chip are mutually exclusive, and the information that can be provided is redundant, so just keep one.
All that remains now is to confirm that pin 5 has the opposite effect on the two chips. At this time, we can look down, Table 5 of the old chip and Table 4 of the new chip have explanations.
We only look at VCTL2 on the left, assuming it is 0, corresponding to OUTPUT1 is insertion loss, OUTPUT2 is isolation, if you do not understand radio frequency, it can be understood that when VCTL2 is pulled low, output 1 is on, output 2 is off of. Then, if it corresponds to a new chip, due to the opposite logic, when this pin VCTL is pulled low, output 1 should be off, and output 2 should be on. Look at the table on the right, RFC to RF2 means to output 2 channels, so this is also confirmed. As for VDD, it is now the power supply port, and it should be pulled high all the time.
After this, it is confirmed that the two chips have the same function and can be interchanged. All that needs to be modified is that one logic pin of the new chip needs to be changed to a power supply port, and the other logic pin needs to be reversed. This has been changed in the schematic diagram, which means that the hardware connection on the PCB has been changed as required, so there is no need to change the logic in the program. The same logic can be used.
This looks like a false alarm, the old and new HackRF can use the same set of code, whether it is new firmware or old firmware. Just make sure that the new driver corresponds to the new firmware, and the old driver corresponds to the old firmware. This is also in line with what Michael Ossman said on github, he did not mention that new hardware must use new firmware.
As for the difference in performance between the two chips, I will find time to check the indicators in detail.
First of all, to explain the meaning of several indicators.
Insertion loss represents the signal loss during transmission, the smaller the better. Return loss represents the reflected signal, generally a negative value, the larger the absolute value, the better. Isolation represents the influence between channels, and the bigger the better. IP3 is the third-order intercept point indicating linearity, the bigger the better. The switching speed is represented by time, the smaller the better. EVM represents the impact on the digital signal constellation, the smaller the better.
We compared the table, the old chip is on the left, and the new chip is on the right.
Older chips have better insertion loss and switching speed. The return loss, isolation, and IP3 of the new chip are better. EVMs are all the same.
Look at the scope of work
Older chips have higher operating and storage temperature limits than newer chips.
So the two chips have their own advantages and disadvantages. Whether the new one is better than the old one depends on which indicator your project cares more about.
So the benefits of open source are here. Whether it is good or not, you can completely evaluate it yourself, and there is no need to completely listen to the propaganda of the merchants.
Later, in line with the principle of detail, I checked the BOM and schematic diagram of the two versions.
C169 and C170 only have the new version on the BOM, the old version does not, but the new and old versions are in the schematic diagram, so they are the same.
Several power ports 1V8 VCC VAA look different on the two BOMs, but they are actually the same on the schematic diagram, except that the old one uses PCB flying wires to bypass the jumper, and the new one is directly connected.
The only difference is that the voltage dividing resistance of the USB power supply lead changes when it is pulled to the MCU power supply port.
Pay attention to the difference between R65 and R62, but I think this has little effect. If there is a difference, then one of the boards must not even light up the LED. This is obviously impossible, and it is easy to test to verify.
Another point is that the old model has a P42 connected to TX, and the new model has removed it, only keeping RX. This is understandable, because the new model has changed the switch, and RX can be used to represent TX, so there is no need for too much redundant information. The problem is which chip pin P42 is. I haven't figured it out yet. If readers know, please tell me, so I can compare. At present, it is estimated that this chip is the original P42, and now this pin should be floating.