Latest activity

Port access works and there are examples - most seen posted over a year ago it seems

This is where the RX<>TX jumper in place of the actual XBee board test in p#2 might give helpful feedback. If that works with .begin early in setup then the hardware is good. If it works without the XBee then points to the issue. If it then...

The idea is to make the gen5 to have the pins needed, then from that I have already planned to make a shield for the two screens that @Rezo has in his possesion. Other shields can be done by you guys as well. And of course the Gerber files and...

Hi h4yn0nnym0u5e, That is very helpful. I really appreciate you sharing your experience, and the link. I'm blown away at how awesome you all are, and how great the forum is. Teensy is much more than a product - it's a legend, with caring people...

I’m just guessing here, but is it possible to turn off “nonCopyingSend”, or some such, when sending the reply? If so, does it help? See here: https://github.com/ssilverman/QNEthernet/issues/61#issuecomment-1863672678

Boards arrived, now let's hope there will be no or minimal need for bodges.

PCM5102a datasheet tells that there is "Intelligent Muting System; Soft Up or Down Ramp and Analog Mute For 120-dB Mute SNR". Maybe it's that?

You need the WayBackMachine for that: https://web.archive.org/web/20100817235724/https://www.hardwarebook.info/S/PDIF_output Paul

A non fused multiply add has a 4 cycle throughout as I understand. Kinda unusual (but then so is an fmadd that can't be pushed every cycle) but could we interleave the fused and non fused ops to get down to 2 cycles per fmadd?

Hmm; according to this the m7 fmadd only has a 3 cycle throughput; not the 1 cycle youd like to see which is a little disappointing. Still thats only 3 cycles, not 6, but I dont see why the 3 wouldnt be possible.

Boards arrived, now let's hope there will be no or minimal need for bodges.

Hello I am planning to develop new hardware, but it needs analog output ports and the Teensy 4.0 and 4.1 do not have such ports, is there any plans for a new version with processors that have such ports? Grateful

I use an Arduino MEGA as a Motion Controller for a CNC Machine and use Port Manipulation to set all the Pins on a Port at the same time for precise Movement of multiple Motors at the same time. I’ve found examples of setting a Single Pin on the...

Not sure, might be interesting to try... Although my goal here is to use the same wiring to support as many of the Arduino/Arducam/Adafruit/Waveshare cameras that have the same breakout pinout. If my main goal is to support the 4 bit HM01b0 on...

Couldn't you just connect each input pin twice (in the order 32103210) and pretend it's regular 8-bit?

Couldn't you just connect each input pin twice (in the order 32103210) and pretend it's regular 8-bit?

Edited link above.

This link doesn't seem to work.

More info here:- http://www.hardwarebook.info/S/PDIF_output

I had that same thought and tried that before posting here. Sadly it didn't make any difference :-/

With help from the fine folks here, I was able to get TOSLINK in/out to work with a teensy 4.1. It is a "man in the middle" between source and amp and can apply filters/effects to the audio. Works quite well.

Boards arrived, now let's hope there will be no or minimal need for bodges.

Boards arrived, now let's hope there will be no or minimal need for bodges.

Soon I will be out again doing more yard/garden work... But thought I would mention that I wanted to get all of the cameras to work on T4.1 on the CSI pins, that worked on the Micromod with FlexIO. The main problem child is the Arducam HM01b0...

Doodling around with https://godbolt.org/ and pasting in the above MultMatrices, accumulating into a local variable does seem to matter; without that you do get an unnecessary write to ram inside the inner loop, that doesnt go away with...

Hey Unhappyman, are you still unhappyWithoutU4? That U4 would be the first one I would unsolder to see if the 3v3 short disappears.

Hello, Band limited waveforms requires more CPU. I don't know about memory. But for sure you must try larger buffer as Rolfdegen suggests. I am pretty sure it is going to make it. You should also use pointers to a manipulate voices, so easier...

Boards arrived, now let's hope there will be no or minimal need for bodges.

Is this from the TeensyDMX library? If so, you could use the example’s structure and most of the code and adapt a different underlying DMX library.

Hi h4yn0nnym0u5e, That is very helpful. I really appreciate you sharing your experience, and the link. I'm blown away at how awesome you all are, and how great the forum is. Teensy is much more than a product - it's a legend, with caring people...

Oh just one more thing, Would be nice if you could maybe substitute a couple 5v pins for the 3.3v pins. Would make it easier when creating a shield.

Soon I will be out again doing more yard/garden work... But thought I would mention that I wanted to get all of the cameras to work on T4.1 on the CSI pins, that worked on the Micromod with FlexIO. The main problem child is the Arducam HM01b0...

Soon I will be out again doing more yard/garden work... But thought I would mention that I wanted to get all of the cameras to work on T4.1 on the CSI pins, that worked on the Micromod with FlexIO. The main problem child is the Arducam HM01b0...

I've just been working on my own phone, have you set the volume on the SGT15000 object? Or if you have, set that to a higher value? i.e. AudioControlSGTL5000 sgt15000; sgt15000_1.enable(); sgt15000_1.inputSelect(AUDIO_INPUT_MIC)...

Don't do a waveform.begin() when you haven't changed the shape. The non-band-limited shapes don't appear to do much, but the band-limited ones have some initialisation code which you're repeatedly calling, probably very fast. In general, I'd say...

Boards arrived, now let's hope there will be no or minimal need for bodges.

I guess slapping a restrict keyword on all input arguments, on top of explicitly accumulating in a local variable, would be good practice to make sure the compiler doesnt get any silly ideas in its head.

Thanks all for your input. I'll do some breadboarding and see if I can get something to work.

I'd be tempted to be suspicious about C not being accumulated in a local variable first and the compiler being forced to be Nazi about thread safety and refusing to optimize the multiple writes away... But given the similarity to the library code...

Ah no I mixed the op count and mem count derp. If it's in DTCM im a little stumped. What I recall from the m7 docs the fmadd should be able to be executed every cycle with a pipeline depth of 3. In-between that and having two banks of TCM...

How do you figure that? I don't think it's anywhere close to 512KB. The default linker script always uses the top of DTCM for the stack. If someone was using anything else then they'd be silly not to include that information in their post.

From what I can quickly Google that seems to depend on compiler/linker flags?

Ah interesting. In the 16x16 case we have more than half an mb of ram tho so they can't all be in TCM. Yet the flops per cycle stays roughly constant. Perhaps C drops out of TCM and since it only needs writing to it doesn't stall the pipeline...

All of the arrays are on the stack which is (D)TCM.

Oh just one more thing, Would be nice if you could maybe substitute a couple 5v pins for the 3.3v pins. Would make it easier when creating a shield.

Doing some back of the envelope math; 40ms for 256x4x4x1000=4M fmadds, 0.1G fmadd per second; or about 6 core clock ticks per fmadd. That's... Not horrendous for a naive for loop, frankly. And I do think it's the lack of TCM use here which is the...

The idea is to make the gen5 to have the pins needed, then from that I have already planned to make a shield for the two screens that @Rezo has in his possesion. Other shields can be done by you guys as well. And of course the Gerber files and...

If I had to guess the main factor limiting performance here is the access to B; it isn't accessed in a contiguous fashion; so even if data caching is enabled which I'm not sure it is, it would still not do much good. Without caching, and no...

The idea is to make the gen5 to have the pins needed, then from that I have already planned to make a shield for the two screens that @Rezo has in his possesion. Other shields can be done by you guys as well. And of course the Gerber files and...

Would agree as long as the edge pin connectors are suitable to add a adapter on top it would be better. This is the adapter I made for the sdram board (note positions are off for the display and camera - have to move them fo the next iteration).