[time-nuts] HP5370 power supply measurements
phk at phk.freebsd.dk
Mon Jul 11 14:46:23 EDT 2016
Inspired by the recent thread about power trouble, I spent an hour
to actually measure why the HP5370 gets so warm as it does.
I'm running on 230VAC here, with the 230/240VAC power setting.
With A6 and A7 removed and turned off: PAC= 4.2 W
That is the loss in the transformer and rectifiers, since the "power"
switch on the HP5370 really isn't that.
Turning it on, PAC= 19.9W
So the fan uses 15.7W. (it's also noisy, I should find a replacement)
I also measured the four unregulated DC voltages:
-20V -> 28.9V
-10V -> 13.4V
+10V -> 13.4V
+20V -> 28.9V
Inserting A6 does not change the OFF state PAC measurably.
Turning on however: PAC = 140 W
-20V -> 27.5V
-10V -> 10.9V
+10V -> 11.9V
+20V -> 27.1V
+15V -> 14.99V
-15V -> 15.04V
+5V -> 5.05V
+10V -> 10.0009V
-20V -> 123mV ripple
-10V -> 2.02V ripple
+10V -> 960mV ripple
+20V -> 312mV ripple
The unregulated busses are still waay over what the schematics
would have you belive (10/20V). The ripple on the -10V seems
pretty high, but the four electrolytic capacitors seem healthy
Next I tried measuring the current on the four regulated voltages,
which all have short-circuit measuring resistors:
A6 pin 5- 6 188mV, 0.07R 3% -> 2.68A +5V
A6 pin 7- 8 498mV, 0.07R 3% -> 7.11A -5.2V
A6 pin 9-10 109mV, 0.4R 1% -> .2725A +15V
A6 pin 11-12 55mV, 0.4R 1% -> .1375A -15V
Since I measured this on the bottom of the motherboard, the contact
and PCB trace resistance to/on A6 is also included so these currents
are high estimates.
Measuring with a clamp milliamp-meter instead I find:
+5V -> 2.2A
-5.2V -> 5.5A
Doing the math:
Trafo: 4.2 W 4.2 W
Fan: 15.7 W 15.7 W
OCXO: 14.0 W 14.0 W Netto Brutto
-20: 3.8 W 2.1 W 2.1 W 3.8 W
-10: 60.0 W 29.0 W 29.0 W 60.0 W
+10: 31.9 W 13.4 W 13.4 W 31.9 W
+20: 7.4 W 4.1 W 4.1 W 7.4 W
137.5 W 82.5 W 48.6 W 103.1 W
Modern efficient DC/DC converters should therefore be able to
reduce the PAC by about 50W or one third.
I tried load-testing the unregulated busses, and it looks like all
four of them can be loaded up to about 5A individually before the
voltage falls out of spec (10/20 V)
So I've been playing around with KICAD and come up with a design
with three Traco TEN40WIR DC/DC converters (2x2411 + 2423) running
off the +/- 10V rails, and I'm seriously tempted to try order a
PCB just to try it out...
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
More information about the time-nuts