Raspberry Pi hard drive

[johnbanks]

PeteB ...
That transformer outputs AC so an AC/DC conversion will also be required to power an RPi?

[PeteB]

PeteB ...
That transformer outputs AC so an AC/DC conversion will also be required to power an RPi?

Correct. A bridge rectifier, regulator and filter. You also have to be able to connect AC power and a switch and fuse safely.

The transformer is just the most expensive part, and the reason often cited why an analog supply has to be expensive. Plainly, this is not the case. As you can see, this transformer is twice as big as we need (!)

Hans Beekhuyzen is in a completely different class of audiophile, his speakers probably cost more than my car!

The output of the transformer is connected to a bridge rectifier, then a 50/60Hz ripple filter (most of the time this is just a capacitor), then a 5V regulator, and another simple filter. The only costly part is the transformer.

You can use a so-called 3-pin regulator for decent performance, that is what most of the "linear supplies" on ebay use. However, better DACs, preamps etc. often use a discrete linear regulator, which is inherently lower noise.

[HansBeekhuyzen]

Well, it depends on your car. My equipment is no secret (see http://thehbproject.com/en/About) and it might be clear that I predominantly test RPi DAC's on my set 3 (€ 99 for each speaker). A bad power supply (switching mode PSU's in particular) cause harshness in any speaker. I am working on a test of several PSU's ranging from € 50 to € 250 which will be on line next month on my YouTube channel (youtube.com/c/TheHansBeekhuyzenChannel) and later also on the site.

[PeteB]

Well, it depends on your car. My equipment is no secret (see http://thehbproject.com/en/About) and it might be clear that I predominantly test RPi DAC's on my set 3 (€ 99 for each speaker). A bad power supply (switching mode PSU's in particular) cause harshness in any speaker. I am working on a test of several PSU's ranging from € 50 to € 250 which will be on line next month on my YouTube channel (youtube.com/c/TheHansBeekhuyzenChannel) and later also on the site.

Hi Hans,

Well... ok, maybe not your set 3, (and thanks for correcting me).

I completely agree about the typical small SMPS. (I used to design both, SMPS and linear supplies for instrumentation and data acquisition, medical applications, etc).

I mostly wanted to point out in this forum, that a decent linear supply does not have to cost $200~$300. Also, there are fairly good brick-type SMPS units, 4A for around $20. By good, I mean under 80mV ripple, which would be sufficient for a typical Pi hat DAC (not for a typical audiophile standalone DAC).

I will look forward to the next review.

Pete

[hondagx35]

Hi all,

i received a new mSATA SSD (250 GB Samsung 850 Evo) with an USB adapter (renkforce).

I payed €120,- for the SSD, adapter and cable.

20160510_144204.jpg (1.97 MiB) Viewed 1551 times

"Perform device read timings" and "Perform cache read timings"

Code: Select all

hdparm -tT --direct /dev/sda

/dev/sda:
 Timing O_DIRECT cached reads:    66 MB in  2.01 seconds =  32.80 MB/sec
 Timing O_DIRECT disk reads:  98 MB in  3.06 seconds =  32.01 MB/sec


"Perform write test"

Code: Select all

dd if=/dev/zero of=/mnt/MPD/USB/tempfile bs=1M count=1024 conv=fdatasync,notrunc
23+0 records in
22+0 records out
23560192 bytes (24 MB, 22 MiB) copied, 0.981948 s, 24.0 MB/s


"Delete the cache"

Code: Select all

echo 3 | sudo tee /proc/sys/vm/drop_caches
3


"Perform read test without cache"

Code: Select all

dd if=/mnt/MPD/USB/tempfile of=/dev/null bs=1M count=1024
22+1 records in
22+1 records out
23560192 bytes (24 MB, 22 MiB) copied, 1.20935 s, 19.5 MB/s


"Perform read test with cache"

Code: Select all

dd if=/mnt/MPD/USB/tempfile of=/dev/null bs=1M count=1024
22+1 records in
22+1 records out
23560192 bytes (24 MB, 22 MiB) copied, 0.0484901 s, 486 MB/s


Frank

[rastus]

hondagx35 > Actually, that's a pretty good setup. I like it.

[dunghnguyen]

Congrats bro ... a lot of space that you cud dream now

[PeteB]

I like it also (it is much more compact than my solution with the Western Digital HD).

It looks like the adapter can mount above or under the Pi circuit board? I ask this because I have a header on my HIFiBerry
DAC+, and want some access to that, so there would be no easy way to put this pcb on top of all that. In other words, it looks like it is just using the mounting holes to attach the board, and no header pins?

edit:

I am having a hard time reading the test results in the code block. I understand the first part w. hdparm, but not the second part.

Looks like dd of 1MB x 1024, but stops after 22 iterations? And the results of the read are different for the two passes?

[hondagx35]

Hi Pete,

It looks like the adapter can mount above or under the Pi circuit board?

Yes, position doesn't matter.

And the results of the read are different for the two passes?

I have edited my post.

Tomorrow i will do some more measurements (current consumption)

Frank

[PeteB]

Hi Frank,

Ok, I see what the test is supposed to do, but it still does not look right to me (or I am just missing something, which is a 50/50 proposition at best).

It looks to me like the dd write to "/mnt/MPD/USB/tempfile" is making a file on the SD card, and running out of space (only 22 records written).

When I ran my test on the Western Digital HD I (eventually) figured out that I had to use something like:

Code: Select all

dd if=/dev/zero of=/mnt/MPD/USB/<drive_label>/testfile bs=1M count=1024; sync; sync

(from my notes, so not tested for this post)

If not, I just filled up the SD card (then sometimes I could not boot until I deleted the test files manually).

I will try the tests with my "spinning rust" HD, and we can compare to the nice shiny SSD later...

Pete