RA8875 display problems with T4.x - a visual study
The subject of problems with the SPI interface between the T4.x & RA8875 displays comes up quite often on the Teensy forum.? I would like to point out that the SPI interface is not anything that PJRC has much control over (it's baked into the processor that NXP created).? I'm not defending PJRC, just suggesting that any angst should almost certainly be directed to NXP rather than PJRC.
I personally am using an RA8875 display from buydisplay.com in my TeensyMIDIPolySynth (TMPS) project, & my experience with incorporating this display included some significant challenges.? I initially wire-wrapped a prototype of my TMPS & everything worked like a champ.? When I took the design to PCB, I transitioned through multiple (very expensive, with the PCB being 5.5" x 8.5" !!) versions of the PCB (with additional decoupling caps, with additional ground plane, with no ground plane, with two ground planes sandwiching everything else & a very large number of thru-vias stitching them closely together, with shorter SPI signal traces, with orthagonal SPI signal traces, etc.).
I was very frustrated until I found a few posts on the Teensy forum suggesting placing small value resistors (50-100ohm) inline with some of the SPI signals.? So, I cut traces on one of my non-working PCBs & patched in some resistors.? Doing so actually resolved my specific display problems.? I determined (thru experimentation & by observing the signals on my o-scope) that 100ohm resistors resulted in the cleanest SPI signals for my particular setup, so that's what the final PCB version for my TMPS successfully includes.? I am using the SPI interface connected to the RA8875 display at native speeds, with no software slowdown.? I have not had any RA8875 display problems of any kind since.
To me, this is very counter-intuitive: how could a very sloppy wire-wrapped board work just fine, but a very clean looking PCB not allow the display to work ??? Signal ringing & signal cross-talk were the major contributors to my display failures.? I know that there's some controversy with this "solution" of using inline small value resistors, but, for reference, here's a very recent Teensy forum post that visually shows the difference in the quality of the SPI signals before & after the use of inline small value resistors:
For comparison, I'd be interested in testing one of my non-working TMPS PCBs with the K9HZ buffer chip, if that's still an available option.
Mark J Culross KD5RXT
P.S. As a side note (since the subject of using serial interconnects between processors has also recently been discussed), my TMPS makes use of two Teensy boards: T4.1 runs the display & manages traditional 5-pin DIN serial MIDI, USBmidi, & USBhostMIDI, while a T4.0 manages all of the audio thru a Teensy Audio Adapter.? The two Teensy boards communicate over a shared 500Kbps serial port (settings flow T4.1-to-T4.0, audio/performance status flows T4.0-to-T4.1, & debug info flows both ways).? The device is a 3-voice, multi-waveform, multi-modulator (selectable FM & PM), 12-poly synth, with filtering & envelope control of all generators, where all of the controls are digitally displayed as buttons & sliders on the RA8875 display, switching between & among multiple menu screens, & everything can be adjusted in real time with no degradation of the resulting sound while it is being generated, & no induced delays in the MIDI response.? I am continually impressed with what a beast the T4.x actually is.? MJC
|
Re: Oscillations on high bands
In ver 50.0 in SSB and CW modes, TX_BPF_SEL
and TR_BPF_SEL both switch states when PTT or Key engaged In ver 66.4 in CW mode, TX_BPF_SEL
and TR_BPF_SEL both switch states when Key engaged. In SSB mode, only
TX_BPF_SEL
switches state.
I am going to sweep the RX and TX paths to try to find blown
MASWSS0179's
dave, n3ds
By comparing your final plot (Isolation PA Out to PA In, LPF Cntl SN1, 25 MHz SSB, ver 50.0) to the plot of the same configuration with oscillating software loaded (Isolation PA Out to PA In, LPF Cntl SN1, 25 MHz SSB, ver 66.4), we see that the isolation is different between these two versions even though they have nominally identical configurations. Are you able to probe the control signals on the LPF board to see how they might differ?
I did notice that the Ver 66.4 CW isolation (which does not oscillate, IIRC) is identical to the Ver 50.0 SSB isolation (which also does not oscillate). So that is the configuration we should be seeking to replicate.
Regardless, the isolation is a lot worse than it should be, regardless of the configuration. Something is not right. We should expect about 48 dB of isolation from the MASWSS0179 switch alone:
This makes it seem like the TR switch is not working -- we should measure the TR switch performance on its own outside of the control board circuit.
On Monday, March 24th, 2025 at 3:55 PM, D Solt via <davesolt= [email protected]> wrote:
I ran some isolation tests today. Below is the first page of my report (attached).
I have had an oscillation at ~35MHz in the K9HZ 20-watt
power amp in my T41 V12 (SW ver 66.4) when I press PTT in the higher
bands. It appears to be caused by low
isolation between the PA Output, J8 and PA Input, J7, on the LPF-Control
board. Below are several plots from my
NanoVNA showing the isolation between 21MHz and 25MHz. On my two systems, the feedback oscillations
only occur above 22MHz.
The first plot shows the isolation with C18 removed from the
PCB. This capacitor is the path between
the transmit and receive RF switching circuits on the LPF-Control board. So is C7, but I didn¡¯t bother removing that. Isolation is better than 65dB with C18
removed.
The next two plots show the isolation at 25MHz in the CW
mode (key down) and SSB mode (PTT) engaged.
Isolation is in the 35dB range.
Additionally, the poor return loss in the SSB mode indicates that the
LPF is not being connected properly.
The next plot shows isolation at 50MHz in SSB mode with PTT
engaged. This is so bad (<25dB) that
I must have a bad RF switch.
The final two plots are done with version 50.0 at 21 and
25MHz. Performance in CW and SSB is
identical.
Conclusions:
¡¤
I may have a hardware problem or a test
equipment problem. Could someone else
duplicate this test?
¡¤
There is a software change in how CW and SSB are
controlled between version 50.0 and 66.4, but that bug doesn¡¯t seem to affect
isolation.
I found that the 25 MHz and 30 MHz bands oscillate, but the lower bands don't, matching what others have found.
To figure out why I measured the insertion loss from the output of the PA to the input of the PA. i.e., I connect port 0 of my NanoVNA to J8, and port 1 to J7. What this measures is the magnitude of the feedback loop.
What I found is that this is a lot higher than it should be, particularly at the higher bands.
| 7 MHz | 14 MHz | 21 MHz | 24 MHz | 28 MHz |
|-------|--------|--------|--------|--------|
|<-80dB | -63 dB | -53 dB | -48 dB | -47 dB |
I would expect roughly 40dB of insertion loss each through the T/R switch and the BPF selection switch U8. Clearly, I'm not getting that. It seems that the feedback amplitude is getting high enough at the higher bands to cause oscillations. What I would like to do next is:
1) Repeat this measurement for a known-good version of the code as identified by Jerry. Is this feedback path amplitude different?
2) Find the reason for the change in the feedback amplitude path by examining the LPF control signals between known-good and known-bad code.
Unfortunately, I am tied up for the week and won't be able to do this until next weekend.
|
Might consult Al. ?He may have stuck some code back in to work with the legacy V11 LPF. ?I recall there was some discussion of that several months ago. ?
Dr.?William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ ? Owner - Operator Big Signal Ranch ¨C K9ZC Staunton, Illinois ? Owner ¨C Operator Villa Grand Piton - J68HZ Soufriere, St. Lucia W.I. Rent it: www.VillaGrandPiton.com
email:??bill@... ?
toggle quoted message
Show quoted text
On Mar 24, 2025, at 7:32?PM, jerry-KF6VB <jerry@...> wrote:
?Working with the Version 66.9 T41 software.
This particular weirdness affects nobody but me....
I use four unused i/o bits to control my homebrew linear amplifier. ?They tell the amplifier what band the transceiver
is on.
These bits ( 28,29,30,31 ) were used in the V11 system to control its four-band LPF. ?There was an 80M pin, a 40M pin etc.
They are not used at all in the V12 system.
So I hooked them up to some transistor buffers and hence out to the linear.
V12 does have internal Yaesu codes in the LPF control card, but they are the wrong polarity. ?Instead of adding hardware
inverters, I just used those other four pins.
Now - with the V66.9 software, I am unable to write to those pins! ?The linear stays at its default - 6M. ?The reason
for that particular default is that it minimizes the risk of transmitting on the wrong band. ?If you use an LPF that is
too high, you run the risk of annoying the FCC. ?If OTOH you transmit through an LPF that's too low, you run the risk of
frying the expensive LDMOS final. ?I do have protection against that in the linear, but I am a "belt & suspenders" guy...
Back to the problem: ?I have verified that the code to write to those pins is being executed. ?ALSO, I wrote separate code at the top of loop() to write to them. ?No joy, no output.
BUT - I wrote a throwaway sketch - not a radio at all - to just write to those pins - actually an incrementing count. ?Yup, they work fine. ?And the linear cycles through its bands.
SO; ?Something in the code is preventing me from writing those pins. ?Gotta be one of two things: 1. ?Something else is writing to them. ?or 2) ?Something is pinMode-ing them to INPUTs.
Darned if I can find it though. ?There are some legacy places where those pins are written, but I'm pretty sure I found them all. ?It has worked on other versions, just not this one.
????????????????- Jerry, KF6VB
|
Re: boards and kits and stuff
SPI on the Teensy today is in fractions of the (600 MHz) clock speed and that can be very high (eg 50 MHZ). ?There is a way to set it in software but i have not experimented with it. ?That certainly has the ability to interfere with some of the operations of the radio. ?
Dr.?William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ ? Owner - Operator Big Signal Ranch ¨C K9ZC Staunton, Illinois ? Owner ¨C Operator Villa Grand Piton - J68HZ Soufriere, St. Lucia W.I. Rent it: www.VillaGrandPiton.com
email:??bill@... ?
toggle quoted message
Show quoted text
On Mar 24, 2025, at 9:14?PM, ken WA2MZE <wa2mze@...> wrote:
?Years ago I worked for Niles Audio, and we produced a product that used plug in cards on a short bus.? The interconnect was via SPI between AVR ATmega128 and ATmega32 processors.? The bus had a buffer at the master side (mega128), the mega32 mcus on the cards were connected unbuffered.? The bus had the usual 3 SPI signals plus one? CS for each of the cards, IIRC there were a max of 4 cards on the bus.? We never had any issues with communication on the bus, I don't recall what the clock speed was, I think it was around 500khz.
|
Re: Oscillations on high bands
SWAG is a technical term:? "Scientific Wild Ass Guess".
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
toggle quoted message
Show quoted text
On Tue, Mar 25, 2025 at 9:19?AM Greg KF5N via <greg.electricity= [email protected]> wrote: So that would be a reason for the isolation to degrade versus increasing frequency.
Because the module has common ground inductances on the input and output.
Just a SWAG here, it's hard to tell without having a good visualization of the hardware.
But it's a common problem with any RF thing which has an input and an output and connectors.
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
Bill, what I was trying to get at was the measurements of the signal path.? The behavior versus frequency.
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
Since its a SWAG¡ explain why it happens with V69 of the software but not with V50. ?
Dr.?William J. Schmidt - K9HZ J68HZ 8P6HK ZF2HZ PJ4/K9HZ VP5/K9HZ PJ2/K9HZ ? Owner - Operator Big Signal Ranch ¨C K9ZC Staunton, Illinois ? Owner ¨C Operator Villa Grand Piton - J68HZ Soufriere, St. Lucia W.I. Rent it: www.VillaGrandPiton.com
email:??bill@... ?
toggle quoted message
Show quoted text
On Mar 25, 2025, at 9:19?AM, Greg KF5N via groups.io <greg.electricity@...> wrote:
? So that would be a reason for the isolation to degrade versus increasing frequency.
Because the module has common ground inductances on the input and output.
Just a SWAG here, it's hard to tell without having a good visualization of the hardware.
But it's a common problem with any RF thing which has an input and an output and connectors.
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
So that would be a reason for the isolation to degrade versus increasing frequency.
Because the module has common ground inductances on the input and output.
Just a SWAG here, it's hard to tell without having a good visualization of the hardware.
But it's a common problem with any RF thing which has an input and an output and connectors.
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
toggle quoted message
Show quoted text
On Mar 25, 2025, at 8:45?AM, Greg KF5N via groups.io <greg.electricity@...> wrote:
? This switch is a plug-in module?
--
73 Greg KF5N
|
Re: Oscillations on high bands
This switch is a plug-in module?
--
73 Greg KF5N
|
Re: Oscillations on high bands
It is here:
https://github.com/DRWJSCHMIDT/K9HZ/blob/main/K9HZ_LPF_Module/K9HZ_LPF_V1.00_Schematic_PDFs/K9HZ_LPF-Carrier_Relay_Board_Schematic_V1.00_07-25-24_KiCad.pdf
-------- Original Message -------- On 3/25/25 5:59 AM, Greg KF5N via groups.io wrote: Hi Oliver, where is the schematic of the TR switch?
I would like to see the circuit topology.? A few decades ago(!) I was designing TR switches and I'm curious about this one.
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
Hi Oliver, where is the schematic of the TR switch?
I would like to see the circuit topology.? A few decades ago(!) I was designing TR switches and I'm curious about this one.
?
--
73 Greg KF5N
|
Re: Oscillations on high bands
By comparing your final plot (Isolation PA Out to PA In, LPF Cntl SN1, 25 MHz SSB, ver 50.0) to the plot of the same configuration with oscillating software loaded (Isolation PA Out to PA In, LPF Cntl SN1, 25 MHz SSB, ver 66.4), we see that the isolation is different between these two versions even though they have nominally identical configurations. Are you able to probe the control signals on the LPF board to see how they might differ?
I did notice that the Ver 66.4 CW isolation (which does not oscillate, IIRC) is identical to the Ver 50.0 SSB isolation (which also does not oscillate). So that is the configuration we should be seeking to replicate.
Regardless, the isolation is a lot worse than it should be, regardless of the configuration. Something is not right. We should expect about 48 dB of isolation from the MASWSS0179 switch alone:
This makes it seem like the TR switch is not working -- we should measure the TR switch performance on its own outside of the control board circuit.
On Monday, March 24th, 2025 at 3:55 PM, D Solt via groups.io <davesolt@...> wrote:
I ran some isolation tests today. Below is the first page of my report (attached).
I have had an oscillation at ~35MHz in the K9HZ 20-watt
power amp in my T41 V12 (SW ver 66.4) when I press PTT in the higher
bands. It appears to be caused by low
isolation between the PA Output, J8 and PA Input, J7, on the LPF-Control
board. Below are several plots from my
NanoVNA showing the isolation between 21MHz and 25MHz. On my two systems, the feedback oscillations
only occur above 22MHz.
The first plot shows the isolation with C18 removed from the
PCB. This capacitor is the path between
the transmit and receive RF switching circuits on the LPF-Control board. So is C7, but I didn¡¯t bother removing that. Isolation is better than 65dB with C18
removed.
The next two plots show the isolation at 25MHz in the CW
mode (key down) and SSB mode (PTT) engaged.
Isolation is in the 35dB range.
Additionally, the poor return loss in the SSB mode indicates that the
LPF is not being connected properly.
The next plot shows isolation at 50MHz in SSB mode with PTT
engaged. This is so bad (<25dB) that
I must have a bad RF switch.
The final two plots are done with version 50.0 at 21 and
25MHz. Performance in CW and SSB is
identical.
Conclusions:
¡¤
I may have a hardware problem or a test
equipment problem. Could someone else
duplicate this test?
¡¤
There is a software change in how CW and SSB are
controlled between version 50.0 and 66.4, but that bug doesn¡¯t seem to affect
isolation.
I found that the 25 MHz and 30 MHz bands oscillate, but the lower bands don't, matching what others have found.
To figure out why I measured the insertion loss from the output of the PA to the input of the PA. i.e., I connect port 0 of my NanoVNA to J8, and port 1 to J7. What this measures is the magnitude of the feedback loop.
What I found is that this is a lot higher than it should be, particularly at the higher bands.
| 7 MHz | 14 MHz | 21 MHz | 24 MHz | 28 MHz |
|-------|--------|--------|--------|--------|
|<-80dB | -63 dB | -53 dB | -48 dB | -47 dB |
I would expect roughly 40dB of insertion loss each through the T/R switch and the BPF selection switch U8. Clearly, I'm not getting that. It seems that the feedback amplitude is getting high enough at the higher bands to cause oscillations. What I would like to do next is:
1) Repeat this measurement for a known-good version of the code as identified by Jerry. Is this feedback path amplitude different?
2) Find the reason for the change in the feedback amplitude path by examining the LPF control signals between known-good and known-bad code.
Unfortunately, I am tied up for the week and won't be able to do this until next weekend.
|
|
|
|
|
Re: boards and kits and stuff
Years ago I worked for Niles Audio, and we produced a product that used plug in cards on a short bus.? The interconnect was via SPI between AVR ATmega128 and ATmega32 processors.? The bus had a buffer at the master side (mega128), the mega32 mcus on the cards were connected unbuffered.? The bus had the usual 3 SPI signals plus one? CS for each of the cards, IIRC there were a max of 4 cards on the bus.? We never had any issues with communication on the bus, I don't recall what the clock speed was, I think it was around 500khz.
|
Working with the Version 66.9 T41 software.
This particular weirdness affects nobody but me....
I use four unused i/o bits to control my homebrew linear amplifier. They tell the amplifier what band the transceiver
is on.
These bits ( 28,29,30,31 ) were used in the V11 system to control its four-band LPF. There was an 80M pin, a 40M pin etc.
They are not used at all in the V12 system.
So I hooked them up to some transistor buffers and hence out to the linear.
V12 does have internal Yaesu codes in the LPF control card, but they are the wrong polarity. Instead of adding hardware
inverters, I just used those other four pins.
Now - with the V66.9 software, I am unable to write to those pins! The linear stays at its default - 6M. The reason
for that particular default is that it minimizes the risk of transmitting on the wrong band. If you use an LPF that is
too high, you run the risk of annoying the FCC. If OTOH you transmit through an LPF that's too low, you run the risk of
frying the expensive LDMOS final. I do have protection against that in the linear, but I am a "belt & suspenders" guy...
Back to the problem: I have verified that the code to write to those pins is being executed. ALSO, I wrote separate code at the top of loop() to write to them. No joy, no output.
BUT - I wrote a throwaway sketch - not a radio at all - to just write to those pins - actually an incrementing count. Yup, they work fine. And the linear cycles through its bands.
SO; Something in the code is preventing me from writing those pins. Gotta be one of two things: 1. Something else is writing to them. or 2) Something is pinMode-ing them to INPUTs.
Darned if I can find it though. There are some legacy places where those pins are written, but I'm pretty sure I found them all. It has worked on other versions, just not this one.
- Jerry, KF6VB
|
Re: Oscillations on high bands
I ran some isolation tests today.? Below is the first page of my report (attached).
I have had an oscillation at ~35MHz in the K9HZ 20-watt
power amp in my T41 V12 (SW ver 66.4) when I press PTT in the higher
bands.? It appears to be caused by low
isolation between the PA Output, J8 and PA Input, J7, on the LPF-Control
board.? Below are several plots from my
NanoVNA showing the isolation between 21MHz and 25MHz.? On my two systems, the feedback oscillations
only occur above 22MHz.
The first plot shows the isolation with C18 removed from the
PCB.? This capacitor is the path between
the transmit and receive RF switching circuits on the LPF-Control board.? So is C7, but I didn¡¯t bother removing that.? Isolation is better than 65dB with C18
removed.?
The next two plots show the isolation at 25MHz in the CW
mode (key down) and SSB mode (PTT) engaged.?
Isolation is in the 35dB range.?
Additionally, the poor return loss in the SSB mode indicates that the
LPF is not being connected properly.
The next plot shows isolation at 50MHz in SSB mode with PTT
engaged.? This is so bad (<25dB) that
I must have a bad RF switch.
The final two plots are done with version 50.0 at 21 and
25MHz.? Performance in CW and SSB is
identical.
?
Conclusions:
¡¤????????
I may have a hardware problem or a test
equipment problem.? Could someone else
duplicate this test?
¡¤????????
There is a software change in how CW and SSB are
controlled between version 50.0 and 66.4, but that bug doesn¡¯t seem to affect
isolation.
I found that the 25 MHz and 30 MHz bands oscillate, but the lower bands don't, matching what others have found.?
?
To figure out why I measured the insertion loss from the output of the PA to the input of the PA. i.e., I connect port 0 of my NanoVNA to J8, and port 1 to J7. What this measures is the magnitude of the feedback loop.
?
?
What I found is that this is a lot higher than it should be, particularly at the higher bands.
?
| 7 MHz | 14 MHz | 21 MHz | 24 MHz | 28 MHz |
|-------|--------|--------|--------|--------|
|<-80dB | -63 dB | -53 dB | -48 dB | -47 dB |
?
I would expect roughly 40dB of insertion loss each through the T/R switch and the BPF selection switch U8. Clearly, I'm not getting that. It seems that the feedback amplitude is getting high enough at the higher bands to cause oscillations. What I would like to do next is:
?
1) Repeat this measurement for a known-good version of the code as identified by Jerry. Is this feedback path amplitude different?
2) Find the reason for the change in the feedback amplitude path by examining the LPF control signals between known-good and known-bad code.
?
Unfortunately, I am tied up for the week and won't be able to do this until next weekend.
?
?
?
|
Re: boards and kits and stuff
On 2025-03-24 09:29, K9HZ wrote: Well the lesson i get from this category of devices is that mounting
the processor on the display board is preferred. But thinking about
it, all of my commercial have the processors mounted within an inch of
the processor that drives it. I guess there is some degree of
obviousness here. *** A main board bolted to the back of the display would get that cable length effectively to zero. We could use the parallel interface. Wonder if it's faster than SPI? OTOH, that would marry us to buydisplay... not sure that's where we want to go. HDMI? - Jerry, KF6VB
|
K9HZ