PDA

View Full Version : Pitastic



Man in a van
2019-06-24, 01:25
https://www.raspberrypi.org/

Haven't yet read the puff:eek:

d6jg
2019-06-24, 02:08
https://www.raspberrypi.org/

Haven't yet read the puff:eek:


I saw the announcement the other day. The increased RAM versions should be ideal for an LMS server with storage via USB3

Apesbrain
2019-06-24, 06:47
I wish they'd add a S/PDIF output. That 1/8" jack could easily be a combo mini-optical/analog.

d6jg
2019-06-24, 07:14
I wish they'd add a S/PDIF output. That 1/8" jack could easily be a combo mini-optical/analog.

I use one of these for that job
https://hifime.uk/DACs/sabre-dac-uae23

Apesbrain
2019-06-24, 14:55
I use one of these for that job
https://hifime.uk/DACs/sabre-dac-uae23
Sure, just as I use a HiFiBerry HAT, but how much more elegant it would be to just put S/PDIF on the board!

cliveb
2019-06-25, 01:34
I saw the announcement the other day. The increased RAM versions should be ideal for an LMS server with storage via USB3
My Pi3 works perfectly as an LMS server. And I've got an old Pi1 running as a player.
The extra capability of the Pi4 is welcome for lots of applications, but not really relevant to us.

d6jg
2019-06-25, 01:42
My Pi3 works perfectly as an LMS server. And I've got an old Pi1 running as a player.
The extra capability of the Pi4 is welcome for lots of applications, but not really relevant to us.

I slightly disagree. The weak point of the Pi to date has been the shared bus for the LAN & USB. In normal use it doesn't show itself but if you run 24/7 and have a bunch of synced players connected to it then you will notice some "drift" that you don't get with more traditional hardware. If I read correctly the Pi4 has 1. a true Gigabit LAN and 2. USB3 which I don't think share the same bus (yet to see it stated conclusively though). If this is the case a Pi4 with 2GB RAM and some USB3 attached storage will make a great LMS server. Overkill as a player without doubt and anything more than 2GB RAM would be a waste of money for a server. I'll be interested to see how the guys who develop piCorePlayer decide to deal with the increased RAM available.

bpa
2019-06-25, 02:17
I slightly disagree. The weak point of the Pi to date has been the shared bus for the LAN & USB. In normal use it doesn't show itself but if you run 24/7 and have a bunch of synced players connected to it then you will notice some "drift" that you don't get with more traditional hardware. If I read correctly the Pi4 has 1. a true Gigabit LAN and 2. USB3 which I don't think share the same bus (yet to see it stated conclusively though). If this is the case a Pi4 with 2GB RAM and some USB3 attached storage will make a great LMS server. Overkill as a player without doubt and anything more than 2GB RAM would be a waste of money for a server. I'll be interested to see how the guys who develop piCorePlayer decide to deal with the increased RAM available.

I agree USB / LAN has been a weak point and the following blog post ( https://blog.hackster.io/meet-the-new-raspberry-pi-4-model-b-9b4698c284 ) seems to indicate that this has been fixed old shared LAN /USB via LAN7515 chip has been replaced by a PCiE bus separating LAN chip from USB bus.

slartibartfast
2019-06-25, 03:59
My Pi3 works perfectly as an LMS server. And I've got an old Pi1 running as a player.
The extra capability of the Pi4 is welcome for lots of applications, but not really relevant to us.I use MusicIP which can use a lot of RAM so an increase would be very welcome if swap space can be avoided.

Sent from my SM-G900F using Tapatalk

Daverz
2019-06-27, 02:14
My Pi3 works perfectly as an LMS server. And I've got an old Pi1 running as a player.
The extra capability of the Pi4 is welcome for lots of applications, but not really relevant to us.

If you have time to check, it would be interesting to know how big your library is (tracks and playing time) and and how long it takes to do a full clear and rescan.

Man in a van
2019-06-27, 05:00
Mine arrived this morning, 4 GB.

I only have a small library;

on my old Shuttle with its Intel(R) Atom(TM) CPU D2700 @ 2.13GHz


The server has finished scanning your media library.
Total Time: 00:02:22 (Wed, 12 Jun 2019 / 8.53)


on the new pi 4B


The server has finished scanning your media library.
Total Time: 00:00:36 (Thursday, June 27, 2019 / 12:06 PM)

I'm running LMS on the Raspbian Desktop OS, remote access via VNC (a little hacking at first, to get it working) I don't have the required HDMI cable :eek:

ronnie

mherger
2019-06-27, 05:46
> on my old Shuttle with its Intel(R) Atom(TM) CPU D2700 @ 2.13GHz
>>
>> The server has finished scanning your media library.
>> Total Time: 00:02:22 (Wed, 12 Jun 2019 / 8.53)
>
> on the new pi 4B
>
>> The server has finished scanning your media library.
>> Total Time: 00:00:36 (Thursday, June 27, 2019 / 12:06 PM)

Is this using the same set of plugins?

--

Michael

kidstypike
2019-06-27, 05:54
Mine arrived this morning, 4 GB.

Where did you get your 4GB Pi from? :p

Jeff07971
2019-06-27, 06:36
Where did you get your 4GB Pi from? :p

CPC have 1,2 and 4 GB Pi 4's

Edit: My Bad ! They did have yesterday !

https://cpc.farnell.com/MarketingProductList?storeId=10180&catalogId=15002&langId=69&orderCode=SC15183,SC15184,SC15185

mherger
2019-06-27, 06:55
> CPC have 1,2 and 4 GB Pi 4's

Not on stock, do they?

--

Michael

Man in a van
2019-06-27, 06:59
Where did you get your 4GB Pi from? :p


OKdo (a sub branch of RS)

RS say they will have stock later in July, I think.

CPC, RS (and OKdo) seem to have better prices :p

Man in a van
2019-06-27, 07:06
@mherger

Actually running more Plugins on the pi.

I ran the scan on the pi before installing any Plugins

I have just a "clear and rescan" on the fully loaded pi


The server has finished scanning your media library.
Total Time: 00:01:45 (Thursday, June 27, 2019 / 3:03 PM

Much more representative, I think, but still good :)

ronnie

is the chocolate running down the mountains?

Man in a van
2019-06-27, 07:08
Showing in stock

https://www.okdo.com/product-choice/pi4models/

slartibartfast
2019-06-27, 07:16
Showing in stock

https://www.okdo.com/product-choice/pi4models/Yeah strange that the bundle with the power supply is in stock but the board only is out of stock.

https://www.okdo.com/shop/raspberry-pi/other-accessories/99130100-pi-sbcs/raspberry-pi-4-model-b-board-with-4gb-lpddr4-sdram-board-only/

Sent from my SM-G900F using Tapatalk

bpa
2019-06-27, 07:18
Showing in stock

https://www.okdo.com/product-choice/pi4models/

What the Irish branch of RS Components (who get deliveries overnight from UK) says

There is always hype and expectation when it comes to any Raspberry Pi launch, therefore we'd like to make our customers aware that RS Components will have no stock at the time of launch and are not expecting to have stock until September 2019. If you are purchasing for personal use, contact sales@okdo.com and a sales contact will be in touch.

kidstypike
2019-06-27, 07:19
Showing in stock

https://www.okdo.com/product-choice/pi4models/

Ordered!

27591

mrw
2019-06-27, 09:20
OKdo (a sub branch of RS)

Thanks for bringing OKdo to my attention. I never saw an announcement from Radiospares (as I still think of them) concerning the launch.

cliveb
2019-06-27, 10:23
If you have time to check, it would be interesting to know how big your library is (tracks and playing time) and and how long it takes to do a full clear and rescan.
I very rarely do a full clear & rescan - usually I just scan for changes when I add new music to the server.

But FYI, I just fired one off:

Tracks: 14190
Albums: 1306
Playing Time: 1112:24:19
Clear+rescan time: 05:57

This is on a stock RPi 3 model B+ running piCorePlayer with the library on a WD Blue SSD attached via USB.

(Now waiting in trepidation for my wife to complain that the "New Music" list has been trashed :-)

chill
2019-06-27, 11:16
Some more numbers for comparison.

My RPi 4 4GB arrived yesterday, so I've had a play with it this evening. I've done a direct comparison against a 3B+. I installed stretch-lite on the 3B+ and buster-lite on the 4. I did a bare minimum install of LMS 7.9.2, with only the default plugins. I used a Sandisk SSD with a USB3 interface adapter.

With the 3B+, a 'clear and rescan' took 6 minutes 30 seconds:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi3Bplus%20Music%20Scan_800.jpg

On the 4, with the SSD connected to one of the USB2 sockets, this came down to 4:09*:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi4%20USB2%20Music%20Scan_800.jpg

And with the SSD connected to one of the USB3 sockets, this came down to 3:52:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi4%20USB3%20Music%20Scan_800.jpg

*I'm a bit puzzled by the 'Find updated coverart files' step in the RPi4 USB2 case - I though that was the MAI plugin, but that's definitely not loaded.

d6jg
2019-06-28, 07:29
4GB back in stock at The Pi Hut - just ordered one

kidstypike
2019-06-28, 07:43
4GB back in stock at The Pi Hut - just ordered one

I emailed Lincoln Binns asking if they intended manufacturing a PI-BOX Pro for the Pi4, MD's reply was:

We sure do, The new Pi is arriving today (Tuesday) and we just need to check measurements and make a prototype before its full launch.

d6jg
2019-06-28, 07:52
I emailed Lincoln Binns asking if they intended manufacturing a PI-BOX Pro for the Pi4, MD's reply was:

We sure do, The new Pi is arriving today (Tuesday) and we just need to check measurements and make a prototype before its full launch.

Excellent news. That is still the best case I have found by some way.

kidstypike
2019-06-28, 08:00
Excellent news. That is still the best case I have found by some way.

I also mentioned the problem with the end plate screws, reply was:

We have sourced a “New” Thread forma which should cut better into the extrusion.

d6jg
2019-06-28, 08:18
I also mentioned the problem with the end plate screws, reply was:

We have sourced a “New” Thread forma which should cut better into the extrusion.

👍

NeverSimple
2019-06-28, 14:16
I emailed Lincoln Binns asking if they intended manufacturing a PI-BOX Pro for the Pi4, MD's reply was:

We sure do, The new Pi is arriving today (Tuesday) and we just need to check measurements and make a prototype before its full launch.

Interesting case, didn't know about that one. I have a Rock64 in an aluminium housing that uses the case as a heatsink, a little extruded part of the case makes contact with the cpu and memory (through heat conducting foil). It really helps to keep the temps down. Would the PI4 need anything like that (or benefit from it)? The Lincoln Binn casing looks like it could do something like that, with the internal plate. Performance and heat production still go togeteher, even on a 'simple' SBC.

Daverz
2019-06-28, 20:20
Some more numbers for comparison.

My RPi 4 4GB arrived yesterday, so I've had a play with it this evening. I've done a direct comparison against a 3B+. I installed stretch-lite on the 3B+ and buster-lite on the 4. I did a bare minimum install of LMS 7.9.2, with only the default plugins. I used a Sandisk SSD with a USB3 interface adapter.

With the 3B+, a 'clear and rescan' took 6 minutes 30 seconds:

On the 4, with the SSD connected to one of the USB2 sockets, this came down to 4:09*:

And with the SSD connected to one of the USB3 sockets, this came down to 3:52:

*I'm a bit puzzled by the 'Find updated coverart files' step in the RPi4 USB2 case - I though that was the MAI plugin, but that's definitely not loaded.

So it seems like USB 2 is not that big a bottleneck for scanning.

Daverz
2019-06-28, 20:21
(Now waiting in trepidation for my wife to complain that the "New Music" list has been trashed :-)

It was for science!

Roland0
2019-06-29, 03:48
I have a Rock64 in an aluminium housing that uses the case as a heatsink, a little extruded part of the case makes contact with the cpu and memory (through heat conducting foil). It really helps to keep the temps down. Would the PI4 need anything like that (or benefit from it)?

Yes:


To reduce thermal output when idling or under light load, the Pi4B reduces the CPU clock speed and voltage. During heavier load the speed and voltage (and hence thermal output) are increased. The internal governor will throttle back both the CPU speed and voltage to make sure the CPU temperature never exceeds 85 degrees C.
The Pi4B will operate perfectly well without any extra cooling and is designed for sprint performance – expecting a light use case on average and ramping up the CPU speed when needed (e.g. when loading a webpage). If a user wishes to load the system continually or operate it at a high temperature at full performance, further cooling may be needed.


Won't be relevant just for running LMS, but will be for other scenarios.

Roland0
2019-06-29, 03:56
So it seems like USB 2 is not that big a bottleneck for scanning.

Access time is much more relevant for scanning performance than throughput, and USB3 won't help with that.
However, the Pi4 supports UASP (https://en.wikipedia.org/wiki/USB_Attached_SCSI), which may explain the small improvement.

chill
2019-07-13, 03:17
I've been running the pCP6 beta on a 4GB 4B inside my amplifier enclosure. Some first impressions on the heat issues:

Running LMS and Squeezelite seem to provide only a very light load to the 4B. CPU temperature seems to be dominated by the ambient temperature and limited airflow inside my amplifier enclosure. So with the amplifier powered up and playing music from Squeezelite, the CPU sits in the mid- to high-60s. I can easily get it into the low 70s by doing something more demanding, like transcoding a FLAC album with oggenc, but for the majority of the time when it's just being a server and a player the temperature seems under control.

I realise this is not an ideal environment for a device that may be marginal on heat, and I may end up swapping the 3B+ back in, but the extra responsiveness of the server, quicker library scans, and even the faster reboots are all nice to have. I suppose I don't really need the server to be running on this device, so I could even put an older Pi in there and just run Squeezelite on it, and run LMS on the 4B externally, but my target was always to have an all-in-one integrated amp.

As an aside, the amplifier enclosure is incredibly efficient at cooling. I'm using a 'Dissipante' enclosure, which has sides that are giant aluminium 'wave' profile heatsinks. I have my UcD400OEM modules mounted directly to these heatsinks, and even when pumping out the volume the heatsinks are never more than vaguely warm to the touch.

https://cdn2.shopify.com/s/files/1/1006/5046/products/1PD03300N_4521dce5-9c10-4389-84ef-5ee22b88be8a_1024x1024.jpg
https://cdn2.shopify.com/s/files/1/1006/5046/products/DSC_9912-900x600_1024x1024.jpg

Jeff07971
2019-07-13, 03:28
Connect the CPU of the Pi4 to the heatsinks, That'll warm them up :D :D :D

chill
2019-07-13, 03:42
Connect the CPU of the Pi4 to the heatsinks, That'll warm them up :D :D :D

Indeed it might!

I have the Pi mounted to the front panel via my button PCB: the button PCB mounts to the chassis front panel, and the Pi mounts behind this PCB via the GPIO pins, hat-style. I currently use nylon stand-offs/spacers, but I'm thinking I might swap to brass to provide some degree of thermal conductivity. The front panel will be 10mm aluminium, so if I can get at least some of the heat into that it should dissipate quite nicely. It would be nice to have some sort of direct thermal connection between the CPU and the chassis - I might just use a strip of aluminium, say 3mm thick, formed into a dog-leg shape and a bit of thermal paste to hold it in place.

Jeff07971
2019-07-13, 03:49
Indeed it might!

I have the Pi mounted to the front panel via my button PCB: the button PCB mounts to the chassis front panel, and the Pi mounts behind this PCB via the GPIO pins, hat-style. I currently use nylon stand-offs/spacers, but I'm thinking I might swap to brass to provide some degree of thermal conductivity. The front panel will be 10mm aluminium, so if I can get at least some of the heat into that it should dissipate quite nicely. It would be nice to have some sort of direct thermal connection between the CPU and the chassis - I might just use a strip of aluminium, say 3mm thick, formed into a dog-leg shape and a bit of thermal paste to hold it in place.

Its a pity that Raspberry did not flip the processor to the "underside" of the board, then you could have good cooling AND a HAT !
From reports I've read the board itself is a significant part of the thermal design, maybe they'll change to an aluminium substrate though this is an expensive option
Hopefully someone will come up with somthing, maybe a CNC ali plate might work well IE Between the Pi and Hat to allow a heatsink of resonable size to be fitted

chill
2019-07-13, 04:15
That's interesting if the board is part of the thermal design. If the heat is intended to reach the mounting holes, then brass stand-offs should help to pull that heat away. Another option is to use longer stand-offs and make the 'hat' connection via a short ribbon cable. That'll create a bit more space around the CPU. Yes, it would have been good to have the CPU on the underside (for my installation certainly), but I suspect they were conscious that this could affect compatibility with a lot of earlier installations/cases/fans etc.

Man in a van
2019-07-13, 04:38
I saw this mentioned (https://shop.pimoroni.com/products/fan-shim)

over on

https://www.raspberrypi.org/forums/



There are lots of threads about the "thermal" problem (and some solutions offered or suggested).

There are quite a few reasonably priced solutions appearing already and more, no doubt, to come.:)

chill
2019-07-13, 05:03
There are quite a few reasonably priced solutions appearing already and more, no doubt, to come.:)

I suspect you're right - there'll be more of that type of solution in no time. I think I'll work on making space for such a solution. I'm sure those fans are quiet, but it would be nice to keep it passive, given that it's an amplifier. I suspect that just making a bit of space to breathe will be enough in my situation, and if not then a bigger heatsink or a thermal connection to my enclosure will be plenty.

Here's a quick mock-up:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Long%20brass%20stand-offs_640.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Long%20brass%20stand-offs%202_640.jpg

The sharp-eyed among you will spot that that's not a 4B - that's because I haven't yet drilled out the mounting holes to 3mm in my spare 'BNIB' 1GB 4B. And the ribbon connectors aren't clamped down (no point wasting them), but as a proof of concept it'll do. Unfortunately, when there's a ribbon connector on the pins there isn't room for stand-offs on the two adjacent mounting holes, but the two on the opposite edge of the board seem to be rigid enough.

Jeff07971
2019-07-13, 08:10
Yes, it would have been good to have the CPU on the underside (for my installation certainly), but I suspect they were conscious that this could affect compatibility with a lot of earlier installations/cases/fans etc.

They seem to have broken compatability with all other Pi's anyway I don't think one connector is in the same place.

Man in a van
2019-07-14, 03:25
Yesterday I imaged a new Raspbian Buster on my rpi4B (4gb). There is an updated image on the raspberry site.

I ran the updates and install a sceensaver and PiShrink, but kept the install as standard.

Right now I'm building a librespot Spotify-Connect client.

There is a neat little thermometer on the rpi desktop.

I ran


sudo vcgencmd measure_temp

and got 82.0 C :rolleyes:

ronnie

Man in a van
2019-07-14, 03:37
The build completed in just over 26 minutes.

The warning thermometer disappeared and the temp dropped to 71.0 C.

phew! what a scorcher :p

chill
2019-07-14, 04:14
and got 82.0 C :rolleyes:


Presumably it was throttling at that stage, or was it just a quick temperature spike?

I keep reading that a heatsink or fan is not needed for the safety of the RPi, because the throttling is the failsafe. My impression is that pCP running just LMS and Squeezelite isn't going to load the 4B enough to cause throttling. But if something else causes the temperature to spike and cause throttling, my guess is that that wouldn't have very much effect on LMS, given that it isn't a heavy load.

I implemented the longer brass stand-offs to create a bit of space around the CPU. I went for three 11mm lengths rather than 4 as in my earlier photo.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Long%20brass%20stand-offs%203_640.jpg

This has dropped the temperature by 2-3 degrees. The key seems to be the extra airflow. My small heatsink is arranged so that the fins are vertical, so that air flows between the fins. The ribbon cable still blocks the airflow a bit, but it's improved with the longer stand-offs. And hopefully when those stand-offs connect to a big 10mm aluminium front panel that will help a bit too.

The extra space would allow me to fit a fan or a bigger heatsink. A bigger heatsink provides extra thermal mass, so can absorb more heat energy and smooth out short spikes in temperature. But unless it also benefits from extra airflow then it won't necessarily do much to control temperature in the longer term - once it heats up the benefit is lost unless it can also be cooled. So my aim is to search for a heatsink that has a greater area than the CPU area, i.e. so that it presents a wider profile to the rising airflow and benefits from more of that airflow. The extra space also allows me to consider some kind of direct thermal connection to the amplifier chassis. I'm going to experiment with some folded-over aluminium foil to act as a thermal bridge.

Man in a van
2019-07-14, 04:33
Presumably it was throttling at that stage, or was it just a quick temperature spike?



Throttling I guess, or just a warning, it was present for the whole of the build.

Apparently it also appeared in Jessie.

It was the first time I had seen it as normally I do the build from my Windows Desktop with no monior on the pi.

Apparently switching of the 4k setting and some other tweaks help to reduce the heat (also running properly headless, to which I may return...we'll see..sigh).

chill
2019-07-14, 05:24
My improvised thermal bridge - a few folds of aluminium foil wedged between the fins of the small heatsink. The foil doesn't grip very well, so it's held in place with clear tape, but if I go ahead with this then a bit of thermal paste should do the job:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Aluminium%20foil%201_640.jpg

Not yet bridged to the chassis heatsink, just flapping around in free air:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Aluminium%20foil%202_640.jpg

Surprisingly it's knocked another 4-5 degrees off. Here's a plot of temperature from a 4 minute 'oggenc'. Previously this took the temperature into the low 70s. Here it seems to have levelled off at 63 degrees before the oggenc process finished and it started to cool down again immediately.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Aluminium%20foil%20temp_640.jpg

chill
2019-07-14, 05:54
With the foil wedged so that it's making good contact with the enclosure the performance is a couple of degrees worse than the 'free air' example, but I think a few more folds of foil, some thermal compound, and the 10mm aluminium front panel should all help to claw a bit back. I'm quite happy that it'll stay comfortably cool in normal operation.

kidstypike
2019-07-15, 11:20
I emailed Lincoln Binns asking if they intended manufacturing a PI-BOX Pro for the Pi4, MD's reply was:

We sure do, The new Pi is arriving today (Tuesday) and we just need to check measurements and make a prototype before its full launch.

Hmm . . . I hope they come up with something workable, the Pi4 gets hot in a case. :(

First picture below shows the temp graph after 64 minutes, for a Pi3B+ in a complete PI-BOX Pro case, serving 2 synced players with a playlist of mixed mp3 and flac files.

Second picture shows the temp graph after 64 minutes for a Pi4 (4GB) serving just one player, (same SSD, same library, same playlist), in a PI-BOX Pro case without the back plate fitted (it needs some work to allow for the different USB and ethernet port positions). The internal carrier plate is easily filed to fit around the ethernet port. The converter board wasn't used (doesn't fit), without the back plate fitted the power USB (C) can be plugged directly into the Pi.

27685

27686

chill
2019-07-15, 13:49
I think 80 degrees is the temperature at which throttling kicks in to control the temperature. Did you notice any effect on LMS - was it any less snappy?

kidstypike
2019-07-15, 13:58
I think 80 degrees is the temperature at which throttling kicks in to control the temperature. Did you notice any effect on LMS - was it any less snappy?

No, I wasn't actually doing anything, just listening to music and watching the temp graph (like you do :)).

Man in a van
2019-07-16, 04:39
Well yesterday I ordered one of these,

https://shop.pimoroni.com/products/fan-shim?variant=29210095812691&currency=GBP&utm_source=google&utm_medium=cpc&utm_campaign=google+shopping&gclid=EAIaIQobChMI-uy276m54wIVx7HtCh1IxwM7EAQYASABEgKOWvD_BwE


then I ordered the header (sigh) just in case and a case,

https://shop.pimoroni.com/products/pibow-coupe-4?variant=29210100170835


they were pushed through the letterbox this morning.

I have a Raspnian Desktop image


Raspbian Buster with desktop Image with desktop based on Debian Buster

Version: July 2019
Release date: 2019-07-10

Kernel version: 4.19





with a librespot Spotify-Connect client.

I assembled fan, header and case and powered up.

The fan began to spin and was just audible at about half a metre in a quiet room (faint noise, I would call it).

Started to play Music and ran a temerature check, got 42C. Played a Ry Cooder You Tube via the Chromium browser, same temp.

Swapped the sd card for a pCP install and started fip radio, temp given as 44C, I just checked again and it's 43C.

Seems to work.

Pimoroni offer a script for raspbian which allows the fan to run automatically to a preset temerature.

ronnie

I have no idea how long the thing will last though :p:rolleyes:

Man in a van
2019-07-16, 05:30
I just ran the Pimoroni install instructions for automatic control of the fan and set it at the second "example" given.


pi@raspberrypi:~ $ vcgencmd measure_temp
temp=51.0'C
pi@raspberrypi:~ $ sudo systemctl status pimoroni-fanshim.service
● pimoroni-fanshim.service - Fan Shim Service
Loaded: loaded (/etc/systemd/system/pimoroni-fanshim.service; enabled; vendor preset: enabled)
Active: active (running) since Tue 2019-07-16 13:17:16 BST; 7min ago
Main PID: 591 (python3)
Tasks: 2 (limit: 4915)
Memory: 9.9M
CGroup: /system.slice/pimoroni-fanshim.service
└─591 python3 /home/pi/fanshim-python/examples/automatic.py --threshold 55 --hysteresis 2.5 --delay 2

Jul 16 13:17:16 raspberrypi systemd[1]: Started Fan Shim Service.
pi@raspberrypi:~ $ vcgencmd measure_temp
temp=54.0'C
pi@raspberrypi:~ $


I have installed LMS and I'll pop Squeezelite on later today.

ronnie

chill
2019-07-16, 05:50
Well yesterday I ordered one of these,

https://shop.pimoroni.com/products/fan-shim?variant=29210095812691&currency=GBP&utm_source=google&utm_medium=cpc&utm_campaign=google+shopping&gclid=EAIaIQobChMI-uy276m54wIVx7HtCh1IxwM7EAQYASABEgKOWvD_BwE



That seems quite effective. Will be interesting to see if it can hold 55 degrees under load. Remind me - is your 4B in any sort of case?

I've ordered a larger heatsink (40mm x 40mm x 11mm), just to see if the extra side length is enough to make any difference. It'll need a small adjustment to fit around the camera connector, but if it works to dissipate more heat energy (rather than just storing more) it'll be simpler to use than my Heath-Robinson thermal link to the amplifier chassis.

In the meantime I wrote my first ever Python script (yay!), to plot a longer term CPU temperature log.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/RPi4%20CPU%20Temperature%20Log%20test_640.jpg

Man in a van
2019-07-16, 06:44
That seems quite effective. Will be interesting to see if it can hold 55 degrees under load. Remind me - is your 4B in any sort of case?




and a case,

https://shop.pimoroni.com/products/pibow-coupe-4?variant=29210100170835




Still showing 55C


pi@raspberrypi:~ $ vcgencmd measure_temp
temp=55.0'C
pi@raspberrypi:~ $

chill
2019-07-16, 06:50
then I ordered the header (sigh) just in case and a case,

https://shop.pimoroni.com/products/pibow-coupe-4?variant=29210100170835



D'Oh!

Jeff07971
2019-07-16, 06:51
Still showing 55C


pi@raspberrypi:~ $ vcgencmd measure_temp
temp=55.0'C
pi@raspberrypi:~ $


Has the ambient temperature changed while the measurement is being made ?

Do not forget that heatsink "effeciency" is above ambient IE if ambient goes down 10 C so will the chip temp all things being equal.

THe chip temperature will be an absolute temperature

Man in a van
2019-07-16, 08:56
Hi Jeff

No major change in the ambient, it may have increased a couple of degrees since I started this morning.

I got the command from here

https://linuxhint.com/raspberry_pi_temperature_monitor/


Just checked again


pi@raspberrypi:~ $ vcgencmd measure_temp
temp=54.0'C
pi@raspberrypi:~ $ vcgencmd measure_temp | egrep -o '[0-9]*\.[0-9]*'
56.0
pi@raspberrypi:~ $


I found a little tutorial (https://www.cyberciti.biz/faq/linux-find-out-raspberry-pi-gpu-and-arm-cpu-temperature-command/) and now have a bash script :)




pi@raspberrypi:~ $ ./my-pi-temp.sh
Tue 16 Jul 16:49:24 BST 2019 @ raspberrypi
-------------------------------------------
GPU => temp=54.0'C
CPU => 55'C



The area where the pi is can get fairly warm in the morning if the sun is up, I'll just monitor the readings for a a day or so.

chill
2019-07-17, 13:50
My 40mm x 40mm x 11mm heatsink arrived today. It just about fits without having to be trimmed around the camera connector. I used a thermal pad to fix it to the CPU.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Heatsink1_640.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/Heatsink2_640.jpg

It seems to do the job. I ran approx 20 minutes of oggenc jobs, and simultaneously did a library 'clear and rescan'. The temperature got up to the low 70s, and then cooled back down to ~63 degrees. The 'background' temperature inside my amplifier seems to be the dominant factor. I think this is simpler than trying to connect the CPU to my amplifier chassis.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/underload_640.jpg

Jeff07971
2019-07-17, 14:22
Chill,

Actually looks like quite a good fit.

I think its becoming clear that without "active" cooling a Pi4 working hard (and not throttling) in a closed case is going to be difficult !

chill
2019-07-19, 01:20
Chill,

Actually looks like quite a good fit.

I think its becoming clear that without "active" cooling a Pi4 working hard (and not throttling) in a closed case is going to be difficult !

The fit was much more convenient than I expected - it slid in between the two ribbon cable sockets, touching both but without forcing them in any way.

I think you're right - any serious continuous load will require active cooling if the aim is to avoid throttling. Fortunately for this discussion, LMS+Squeezelite don't seem to represent a serious continuous load. And I suspect that even if throttling does kick in, Squeezelite would be unaffected and LMS only marginally, if at all.

For a bit of fun I logged the CPU temperature on the 4B in my amplifier for a 24 hour period:

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/RPi%20Board/pcp_cpu_temp_18JUL_640.jpg

I can see:
- the gradual cooling of the amplifier after it was shut down at the end of the previous day.
- a brief spike at 2am where the cron script runs to take a backup of the SD card.
- a slight rise in temperature just after 10am. I think this could be when the sun hits the amplifier for a short period, and then as the sun moves on the amplifier cools down again.
- a very clear warm up just before 6pm, when the amplifier was powered up. During the evening the amplifier variously played Squeezebox music and TV sound, and there's no clear difference in temperature - it just sits at around 65 degrees, which must be dominated by the ambient temperature inside the amplifier.

cliveb
2019-07-21, 10:08
Been on holiday for a couple of weeks, and am perplexed by the direction this thread has taken while I was away.
It would seem that various people are trying out solutions to potential overheating problems when running LMS on a Pi4.
I guess using a Pi4 then looking to fix any issues that arise is a case of trying something for the fun of it,
but from a practical perspective, the solution is patently obvious: don't use a Pi4.
A Pi3 is more than man enough for the task (even a Pi2 will do the job), and it doesn't get hot.

d6jg
2019-07-22, 05:58
Been on holiday for a couple of weeks, and am perplexed by the direction this thread has taken while I was away.
It would seem that various people are trying out solutions to potential overheating problems when running LMS on a Pi4.
I guess using a Pi4 then looking to fix any issues that arise is a case of trying something for the fun of it,
but from a practical perspective, the solution is patently obvious: don't use a Pi4.
A Pi3 is more than man enough for the task (even a Pi2 will do the job), and it doesn't get hot.

You say that Clive but all Pi variants before 4 feature a shared bus for the USB and LAN interfaces which is known to be problematic. The Pi4 splits them and provides a full gigabit LAN and also includes USB3 ports enabling users to attach a full USB3 drive which is much faster than USB2. Apart from the overheating issue or non issue depending on how you look at it a Pi4 + USB3 drive should be much quicker than a Pi3 + USB3 drive attached to a USB2 port.

Man in a van
2019-07-22, 09:44
Been on holiday for a couple of weeks,

I'm still waiting for your card, did you lick the stamp properly? :p

cliveb
2019-07-22, 10:15
You say that Clive but all Pi variants before 4 feature a shared bus for the USB and LAN interfaces which is known to be problematic. The Pi4 splits them and provides a full gigabit LAN and also includes USB3 ports enabling users to attach a full USB3 drive which is much faster than USB2. Apart from the overheating issue or non issue depending on how you look at it a Pi4 + USB3 drive should be much quicker than a Pi3 + USB3 drive attached to a USB2 port.
All of the good things you mention about why the Pi4 is better than the Pi3 are perfectly valid when considered in an appropriate context. I applaud the fact they've decoupled the LAN from USB, and that it now has USB3. (FWIW, I still think an onboard SATA controller would be the icing on the cake, but we're still waiting for that).

But none of the Pi3's limitations you mention are significant in the context of an LMS server. My LMS running on a Pi3 with library on an SSD attached via USB2 has never, not once, ever, given me any trouble. It supports 5 players no problem. You don't need gigabit ethernet and/or USB3 speeds to stream the mickey-mouse data rates consumed by audio.

Using a more capable device when it's not justified, and suffering other issues as a result, would be considered pretty daft in most walks of life.
It's a bit like choosing to go to Tesco (Walmart for our US cousins) in a Ferrari instead of a Ford - no advantage, and you end up burning more fuel.

d6jg
2019-07-22, 10:23
All of the good things you mention about why the Pi4 is better than the Pi3 are perfectly valid when considered in an appropriate context. I applaud the fact they've decoupled the LAN from USB, and that it now has USB3. (FWIW, I still think an onboard SATA controller would be the icing on the cake, but we're still waiting for that).

But none of the Pi3's limitations you mention are significant in the context of an LMS server. My LMS running on a Pi3 with library on an SSD attached via USB2 has never, not once, ever, given me any trouble. It supports 5 players no problem. You don't need gigabit ethernet and/or USB3 speeds to stream the mickey-mouse data rates consumed by audio.

Using a more capable device when it's not justified, and suffering other issues as a result, would be considered pretty daft in most walks of life.
It's a bit like choosing to go to Tesco (Walmart for our US cousins) in a Ferrari instead of a Ford - no advantage, and you end up burning more fuel.

All true but I submit it might depend on why you went to Tesco. You might have gone in your Ferrari - which model have you got by the way - in order to impress someone and not just for shopping. Mind you I suppose you’d go to Waitrose in a Ferrari so the point is academic. 😀

w3wilkes
2019-07-22, 11:01
It's a bit like choosing to go to Tesco (Walmart for our US cousins) in a Ferrari instead of a Ford - no advantage, and you end up burning more fuel.

The Ferrari does make the journey a lot more fun and if I had the Ferrari I'd probably take the long way both to and from the store!

edwin2006
2019-07-22, 11:06
Stel depend on the weather. When it's raining ING and roads are wet I would prefer something else beside a Ferrari [emoji6]

Jeff07971
2019-07-22, 11:16
Stel depend on the weather. When it's raining ING and roads are wet I would prefer something else beside a Ferrari [emoji6]

The wet makes it even more fun, believe me

cliveb
2019-07-23, 02:45
The Ferrari does make the journey a lot more fun and if I had the Ferrari I'd probably take the long way both to and from the store!
I'd guess that any journey in the wide open spaces of Utah is more fun in a Ferrari.
I can assure you that it isn't in Watford!
Driving in the UK is no fun any more - it's now just a necessary chore to get from A to B.
I gave up on sports cars a long time ago.

toby10
2019-07-23, 06:06
What? Every time I see a James Bond movie he is *zipping* around British streets in his Aston without issue! You must be on the wrong streets......

chill
2019-08-09, 02:18
There seems to be a fair amount of resistance to using the Pi4 for LMS, based on heat issues, and reports seem to suggest that in an enclosed case with no airflow it can run very hot. I should stress that in my application, with a 40mm heatsink and a bit of convection airflow, running LMS and Sqeezelite and serving various players doesn't cause any heat problems at all.

So I was wondering whether anyone has tried underclocking the Pi4 to the point where the temperature is similar to a 3B+. Would the performance running LMS then end up similar to a 3B+ (which of course is more than adequate), or would the new processor, extra ram, USB3 and gigabit ethernet all contribute to an improvement over the 3B+?

Jeff07971
2019-08-09, 04:02
There seems to be a fair amount of resistance to using the Pi4 for LMS, based on heat issues, and reports seem to suggest that in an enclosed case with no airflow it can run very hot. I should stress that in my application, with a 40mm heatsink and a bit of convection airflow, running LMS and Sqeezelite and serving various players doesn't cause any heat problems at all.

So I was wondering whether anyone has tried underclocking the Pi4 to the point where the temperature is similar to a 3B+. Would the performance running LMS then end up similar to a 3B+ (which of course is more than adequate), or would the new processor, extra ram, USB3 and gigabit ethernet all contribute to an improvement over the 3B+?

I think that there should not really be so much resistance to using the Pi4, cooling in enclosed spaces is somewhat of an issue but the benefit of running a 4 may outweigh this. Its really just another thing to think about, like do I have enough disk space.

Regarding the underclocking I think this is a good idea though finding the balance may be time consuming but ultimatey usefull

Untill then a fan if probably the way to go as you and Paul- have found

Edit: you use a large H/S and Paul- uses a fan

Jeff

chill
2019-08-09, 05:54
Underclocking seems to make remarkably little difference in my setup - perhaps 3 degrees under load. I don't think my setup is completely representative though, because my Pi lives inside my amplifier, where the ambient temperature is higher than normal. So I set it the task of a clear and rescan on my library - when it's doing that I think the temperature caused by the load on the CPU is less affected by the ambient temperature. But I do suspect the results will be slightly different in other people's setups.

At the default 1500MHz
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Rescan%20at%201500MHz_480.jpg

Underclocked to 1000MHz
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Rescan%20at%201000MHz_480.jpg

Underclocked to 600MHZ
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Rescan%20at%20600MHz_480.jpg

The scan times did take longer:
1500MHz: 6:04
1000MHz: 7:10
600MHz : 9:50

The interface still seems very responsive even at 600MHz. My memory isn't good enough to compare properly with how it was with my 3B+ though. And I haven't tried it with music playback yet. I think I'll leave it at 600MHz for a while to see how well it performs.

chill
2019-08-09, 06:19
Underclocking seems to make remarkably little difference in my setup - perhaps 3 degrees under load.

Actually, LMS isn't really much of a load. So I ran some consecutive oggenc jobs while LMS was rescanning. Whereas this would have taken the temperature into the low 70s, it now peaks at about 67 degrees. So maybe 5ish degrees improvement.

Playback to 5 synced players still seems perfect at 600MHz.

In my case there's no real benefit in underclocking - the temperature savings are minimal and I wasn't struggling with heat at the default clock speed. But maybe there'd be bigger gains for Pis in a tighter space, or with a smaller heatsink. Worth a try if you're up for some testing. To underclock, edit the config.txt file on the boot partition, and uncomment the #arm_freq= line, e.g:


arm_freq=600

Greg Erskine
2019-08-09, 06:35
Have you had a play with the CPU Governors on the [Tweaks] page?

By default, pCP uses ondemand.

chill
2019-08-09, 06:44
Have you had a play with the CPU Governors on the [Tweaks] page?

By default, pCP uses ondemand.

I haven't - didn't even think to look there!

Plenty more there to investigate, thank you. I think the 'Powersave' option will do the same as I did when I set 600MHz, but it allows the setting to be made without a reboot. And there's also a separate underclocking option, which evidently slows down a few other settings.

chill
2019-08-10, 13:10
I tried to find an underclocking speed that would make my 4B 4GB perform similarly to my 3B+. It's very difficult to find a test that eliminates all the variables. I settled on a single oggenc job on a 90-minute flac file, as I figured that involved a bit of disk read and write, and a lot of CPU activity. I used the same hard disk on both machines, and I used a USB2 socket on the 4B so as to match the 3B+.

On the 3B+ the task took 11m 07s.
On the 4B at the standard 1500MHz it took 4m 13s - quite an improvement.
I had to underclock the 4B to 600MHz to get that time up to 10m 12s - still 9% faster than the 3B+.

The temperature conditions were far from equal, so the only thing I can tell is that underclocking the 4B does lower the temperature, but only by a couple of degrees.
The 3B+, sealed inside the official case without a heatsink or fan, got up to 63 degrees during this task.
The 4B, mounted inside my amplifier enclosure (powered down, so at room temperature) with a 40mm heatsink and no fan, got up to 57 degrees at 1500MHz, and 55 degrees at 600MHz.

So the surprising things for me were:
1) the 4B is very much faster than the 3B+ at this task with their default clock speeds
2) the temperature savings by underclocking the 4B are minimal

So I shall be running my 4B at the default 1500MHz, and based on these and earlier tests I'm happy that the conditions inside my amplifier are fairly kind to the 4B.

Greg Erskine
2019-08-10, 14:42
Thanks for documenting this.

I wonder, as the CPU is practically doing nothing, it is probably throttled back to 600MHz most of the time. The periods of increased speed are so short it is not captured by temperature sampling period.

Man in a van
2019-08-11, 02:34
Just made another investment

27895


27896

kidstypike
2019-08-11, 02:39
Just made another investment

You're a nut! :D

kidstypike
2019-08-11, 02:54
But seriously, these (https://www.amazon.co.uk/gp/product/B07TVLTMX3/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&psc=1) are good.

27897

27898

chill
2019-08-11, 03:19
But seriously, these (https://www.amazon.co.uk/gp/product/B07TVLTMX3/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&psc=1) are good.


Agreed. I needed a case for my spare 4B 1GB, so I ordered one of these on the strength of your earlier post. It arrived this morning. I paid £1 more and got the version with an SD card reader and a 3A power supply (https://www.amazon.co.uk/gp/product/B07TTRVM4P) :-)

I ran the same oggenc job that took my amplifier-mounted 4B 4GB up to 57 degrees.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Oggenc%203.3V%20Fan_480.jpg

Wired for 5V the fan is impressively quiet, but I wired it for 3.3V, and it's effectively inaudible - I can only hear it if I put it close to my ear. Not that I need it, but I'm now toying with the idea of mounting such a fan over my amplifier's 4B, since it's so completely quiet.

chill
2019-08-11, 05:08
I made a mount for the fan, to hopefully slot in between my RPi and my button board.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Printing%20support_480.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20on%20support_480.jpg

Here it's slotted into a mockup, over a 3B+
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20in%20stack_480.jpg.jpg

I may have to adjust the standoff heights to make sure nothing clashes.

kidstypike
2019-08-11, 05:13
I made a mount for the fan, to hopefully slot in between my RPi and my button board.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Printing%20support_480.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20on%20support_480.jpg

Here it's slotted into a mockup, over a 3B+
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20in%20stack_480.jpg.jpg

I may have to adjust the standoff heights to make sure nothing clashes.

But you'll need good ventilation into and out of your amp, no good just blowing hot air onto the Pi4 CPU?

chill
2019-08-11, 05:55
But you'll need good ventilation into and out of your amp, no good just blowing hot air onto the Pi4 CPU?

That's true, and it has. The baseplate has a grid of mounting holes:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Enclosure/Base%20plate_480.jpg

And the top plate has ventilation slots right above where I've mounted the RPi:
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Enclosure/Top%20plate_480.jpg

And the front and back plates don't seal against the baseplate. So plenty of ventilation.

I've now added the fan to the RPi4 in the amp, and I'm waiting for the amp to warm up to see what effect the fan has. This is more out of curiosity than need, and I've left the fan's power connection accessible so that I can disconnect it easily.

chill
2019-08-11, 06:36
The amp got up to operating temperature, and I connected the fan at 14:16. It seems to drop the temperature by ~5 degrees. I'm guessing it's no greater because it's mostly just moving warm air around inside the amplifier, as kidstypike suggested. The fan's not really big enough, nor suitably ducted, to draw in fresh air. I don't think it's worth the effort.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Pi4%20with%20fan_480.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/PI4%20stack_480.jpg

chill
2019-08-11, 10:29
I wonder, as the CPU is practically doing nothing, it is probably throttled back to 600MHz most of the time. The periods of increased speed are so short it is not captured by temperature sampling period.

Sorry Greg - I missed this.

I'm not so sure about that. You can apparently show the current CPU frequency with:

sudo cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq

On my RPi4, running pCP (LMS & Squeezelite) I've never seen it show anything but the currently set maximum. So apparently there's a difference, certainly in terms of heat generation, between an idling CPU @1500MHz and one under load at the same frequency. Or maybe that command isn't doing what I think it is - that might explain why I never saw it report a different frequency when I briefly played with the governor settings in the tweaks menu.

kidstypike
2019-08-14, 01:30
Just noticed this (https://www.raspberrypi.org/forums/viewtopic.php?f=28&t=243500&start=375#p1518239).

27907

chill
2019-08-23, 03:16
I was interested to see how to control a simple fan in software. To date my RPi4 seems to run comfortably cooler than the throttling temperature, but having developed a way to mount a fan in my 'stack' inside my amplifier, I was keen to see if I could leave it permanently connected but only have it running when needed.

My simple 2-wire fan draws about 45mA on 3.3V, which is too much to drive directly from a GPIO pin. So I wired up a little transistor to switch the GND side of the fan according to the state of a GPIO pin. I made it so that it simply plugs into the existing fan wiring.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Transistor%20wiring_480.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Transistor%20fan%20control_480.jpg

Then I made a script to turn the fan on and off according to temperature. I set the 'off' temperature one degree below the 'on' temperature to reduce the frequency that it turns on and off.

Here's the result from an idling pCP (no LMS, no Squeezelite) in the open air. When running at 3.3V the fan takes about 5 minutes to pull the temperature down from a resting 56 degrees to the 40 degrees that I set, then it sits at that temperature, with the fan coming on and going off as necessary. It struggles to get much below that. Running on 5V it cools faster and lower, but it becomes audible.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20with%20GPIO%20control_480.jpg

Jeff07971
2019-08-23, 03:25
I was interested to see how to control a simple fan in software. To date my RPi4 seems to run comfortably cooler than the throttling temperature, but having developed a way to mount a fan in my 'stack' inside my amplifier, I was keen to see if I could leave it permanently connected but only have it running when needed.

My simple 2-wire fan draws about 45mA on 3.3V, which is too much to drive directly from a GPIO pin. So I wired up a little transistor to switch the GND side of the fan according to the state of a GPIO pin. I made it so that it simply plugs into the existing fan wiring.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Transistor%20wiring_480.jpg

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Transistor%20fan%20control_480.jpg

Then I made a script to turn the fan on and off according to temperature. I set the 'off' temperature one degree below the 'on' temperature to reduce the frequency that it turns on and off.

Here's the result from an idling pCP (no LMS, no Squeezelite) in the open air. When running at 3.3V the fan takes about 5 minutes to pull the temperature down from a resting 56 degrees to the 40 degrees that I set, then it sits at that temperature, with the fan coming on and going off as necessary. It struggles to get much below that. Running on 5V it cools faster and lower, but it becomes audible.

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20with%20GPIO%20control_480.jpg


Nice Job !

chill
2019-08-24, 05:15
A quick observation about heatsinks and their connection to the CPU. When I first fitted my 40mm x 40mm heatsink I used some 'thermally conductive tape (https://www.amazon.co.uk/gp/product/B075R9RGXF)'. It seemed to work, but I was suspicious because it seems to be about 1mm thick, and internally it seems to have a foam construction. I can't see how that can be good for thermal transmission.

So I reaffixed the heatsink using a small blob of silicone heatsink compound (https://www.amazon.co.uk/gp/product/B00YCMR04Q). I wasn't sure it would work in my application, as my RPi is mounted on edge and the heatsink is fairly heavy, and I don't think the compound 'sets'. But it seems to be holding perfectly, and the better thermal contact has had a noticeable effect.

With my fan turned on at 12:47, the temperature dropped by about 10 degrees in 4 minutes - that's about twice the drop I recorded when the heatsink was fixed with thermal tape. The non-fan temperature of 61 degrees is also about 3-4 degrees lower than the same test with thermal tape.
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/40mm%20with%20compound_480.jpg

This is the earlier test with the heatsink held on with thermal tape. Identical test conditions - if anything, the ambient temperature is warmer today.
http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Pi4%20with%20fan_480.jpg

Perhaps the thermal conductive tape should be renamed 'thermal insulating tape'.

Jeff07971
2019-08-24, 07:51
Had exactly the same experience with some small (Pi2/3) heatsinks that had "Genuine 3M heat tranfer tape"
Does not sound like the stuff you had though as it was very thin (Like sellotape almost) but was BAD at heat transfer

Got some HY910 thermal adhesive from ebay for a couple of quid, works much better

Jeff

chill
2019-08-27, 04:50
I updated my fan control script to keep a log of the amount of time that the fan is 'on', and then went a bit nerdy (nerdier?), and worked out the duty cycle for several different set temperatures. So on a hot summer's day, inside a hot amplifier, my RPi4 4GB running LMS and Squeezelite sits naturally at about 67 to 68 degrees. My little fan running off 3.3V can pull it down to about 52 degrees

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20Duty_480.jpg

So on the basis that 68 degrees is still some way short of the 80 degrees that causes throttling, and is therefore considered safe (?) by the manufacturers, that's where I've set my fan control. This means that it's not being triggered under normal conditions, but if the load on the Pi increases then it kicks in.

d6jg
2019-08-27, 12:42
I updated my fan control script to keep a log of the amount of time that the fan is 'on', and then went a bit nerdy (nerdier?), and worked out the duty cycle for several different set temperatures. So on a hot summer's day, inside a hot amplifier, my RPi4 4GB running LMS and Squeezelite sits naturally at about 67 to 68 degrees. My little fan running off 3.3V can pull it down to about 52 degrees

http://www.cjh.me.uk/MyPhotobucket/cache/DIYHifi/Fan%20Duty_480.jpg

So on the basis that 68 degrees is still some way short of the 80 degrees that causes throttling, and is therefore considered safe (?) by the manufacturers, that's where I've set my fan control. This means that it's not being triggered under normal conditions, but if the load on the Pi increases then it kicks in.

Is the fan audible when it kicks in? You said it was inside an amp - I take that to mean a hifi amp and therefore probably in plain view / hearing?

chill
2019-08-27, 12:53
Yes, it's a hifi amp - a power amp and my primary music source in one. In a silent room I can hear the fan from about a foot away - no further. If there's music playing at any volume, the fan is inaudible. It's a 5V fan, which I'm running on 3.3V. On 5V it is more audible (and more effective). I think with it set at 68 degrees it's there for emergencies only - a little peace of mind. To be fair, my scenario is quite a harsh test, and even here the fan isn't really needed. I think the main thing it shows is that the 4B is perfectly fine running LMS and Squeezelite with nothing more than a little passive cooling, i.e. a heatsink and some free air.

d6jg
2019-08-27, 13:03
Yes, it's a hifi amp - a power amp and my primary music source in one. In a silent room I can hear the fan from about a foot away - no further. If there's music playing at any volume, the fan is inaudible. It's a 5V fan, which I'm running on 3.3V. On 5V it is more audible (and more effective). I think with it set at 68 degrees it's there for emergencies only - a little peace of mind. To be fair, my scenario is quite a harsh test, and even here the fan isn't really needed. I think the main thing it shows is that the 4B is perfectly fine running LMS and Squeezelite with nothing more than a little passive cooling, i.e. a heatsink and some free air.

Does it have a connected SSD or have you mounted a NAS drive?

chill
2019-08-27, 13:30
I've extended one of the USB3 ports to the back panel, as well as the ethernet and HDMI ports. Currently I have an SSD connected to that rear panel USB3 extension, but I'm not fixed on keeping it that way. I do also have a 3B+ that runs Jivelite on my TV, as well as handling nightly backups, and I may move the SSD to that machine and share it via Samba. I've also considered running LMS on another device, and swapping out the 4B for a Zero to run just Squeezelite, but I do like the idea of having everything in one box. My target for the amp was always for it to be a fully integrated device. But it's good to have options!