stupid的博客

TVB的姐夫，比尔，专业建筑承包商，手戴真正的原子表HP5071B。

Agilent's Rick Karlquist helped design the legendary 5071. He's pictured here with a different clock -- a quartz "hockey puck" that can handle temperature variations of 100 degrees Celsius without distortion. This clock was designed for cell towers located in extreme-weather zones. Many time hackers favor cell-tower clocks like these because of their stability and availability.

10年前我想DIY 一个年误差不超1秒的LED数显时钟，以便发生润秒时进行调校。
这个简单的目标导致了一个有趣的旅程，经历了电子学，钟表制造，天文学，测试设备，石英振荡器，铷和铯原子钟，氢脉泽，频率计数器和相位比较器，全球定位系统，罗兰C（一种惯性导航系统），同步卫星和WWV / WWVB无线电接收机。这让我成为了一个时间坚果。
现在早已超过了这个目标，因为在我的收藏中包含了最准确的时钟（活性氢脉泽），年不准确度不超过1uS。不计国家实验室的话，我的时间实验室现在拥有世界上最准确的时钟。
也许你听说过一个人有一个时钟则知道时间，但一个人有两个时钟则永远不能确定那个是对的，我补充一句，一个有3个时钟的人肯定比有2个时钟的人确定。因为这样，我开始了时钟收集···

这是时间坚果，我宁肯译成时间狂人或极客（Geek）Tom Van Baak的个人时间网站http://www.leapsecond.com上的开篇简介。一个人烧一个东西居然能到这种地步，除了佩服，就是无限的羡慕和景仰。国内的众多发烧友在他面前简直就是小巫见大巫，最要命的是有钱有时间的人根本就不会玩这个。我能接触到的人里边只有lymex具有这种精神，而且是一个例外，但他是烧电压基准。
先来看看斯人和他的实验室。

猛男TVB。

TVB的时间实验室一角。

话说2008年的一天，TVB同学心血来潮，对Trimble Thunderbolt（国内叫天宝）的电源发生了兴趣，准确的说是考察天宝在不同外接电源下的相噪表现。于是翻箱倒柜找了一大堆不同的电源出来，再配上最好的参考时钟HP Z38011A，和专业的相噪测试仪TSC5120A，齐活以后上路啦。但在开始看他的表演之前，先来考察一下测试装置。
1. 参考时钟 HP Z3801A

首先测试一下Z3801A的相噪表现，可以看出Z3081A的相噪全面优于Thunerbolt,，比铷原子钟也要好的多，作为参考时钟再合适不过。

2. 相噪测试仪 TSC5120A

由Symmetry推出的这款仪器是TSC5110A的后续版本，可以非常直观的查看时钟的阿伦方差和相噪，傻瓜化的操作，人人均可轻易上手。关键是价不便宜。在大于10kHz偏移下，可测试低至-175dBc的相噪。
3. Trimble Thunderbolt

Thunderbolt 给出的相噪极限是-145dBc/Hz，为什么这么说，因为过了1k以后，斜率大幅减小，过了10kHz，相噪曲线就非常平稳，变化很小。对电源的要求spec中说明冷启动15W，待机10W，事实上功耗要比这个小的多。TVB也给出了实测结果。


事实上早在2007年TVB就DIY了天宝的电源，用的是Made in China的开关电源。并对天宝的供电情况进行了测试。

言归正传，来看看TVB同学对不同电源的测试情况。
Trimble Thunderbolt Power Supply Noise
28-May-2008
Introduction
It would make sense if the quality of the power supply has some effect on the frequency stability or phase noise of a GPSDO. I suspect the dominant effect is phase noise, even if DC input power is filtered or regulated inside the box.

The phase noise of a typical Thunderbolt was measured using each one of six different power supplies. All the plots below are 10 minute runs and displayed in the same scale. The 10 MHz reference was a Z3801A.

ATX PC power supply
A Trimble Thunderbolt requires three different voltages: +5, +12, and -12 volts; a set of voltages available inside almost any PC. So the first test is to see how well a Thunderbolt works when powered with the simplest, cheapest 5/12 volt PC power supply. The answer is as you might expect -- these computer-grade power supplies work fine, but add a lot of noise to the RF output. This is probably not an issue for GPS timing but there may be some GPSDO applications where this level of noise is unacceptable.


Open frame switching power supply (Mean Well)
The next test is to use a higher-grade open frame 5/12 volt switching power supply. The results are somewhat better, although even more spurs show up at lower frequencies. The spur near 1 Hz is surprising and rather unwelcome.


Compact switching supply (AULT)
Here is a small enclosed switching power supply. It's quite a bit cleaner than the above two, although some switching noise appears in the kHz region. Since this power supply was available very cheap, and in large quantity, and works with 100-240 VAC input, it was chosen for the Thunderbolt bundle.


Laboratory supply (Acopian goldbox)
I thought this excellent high-power laboratory supply would give a perfectly clean spectrum. It's much cleaner at high frequencies than the above three, but some line frequency noise still shows up in the Thunderbolt output (it was powering 4 Thunderbolts in parallel at the time).


Benchtop supply (Elentech)
This low-end adjustable bench supply was better than I expected. Less noise than the Acopian.


Laboratory linear supply (HP 6235A)
The Thunderbolt draws more current on warm-up than this little triple supply is rated for, but after the current draw drops, both the Thunderbolt and power supply are fine. This power supply turned out to be the best. Don't you just love HP?


Conclusion:
The quality of the supply does indeed make a difference in the phase noise of 10 MHz RF output. Use of PC power supplies is clearly a bad idea, at least if phase noise is a concern. It may or may not have any impact on frequency stability.

The choice of the enclosed switching power supply seems a reasonable compromise in price and performance for the TAPR group buy. However, for extremely low noise or frequency multiplication applications, you might want to see if a better power supply is necessary.

As expected mains frequency (50/60 Hz) and its harmonics appear in the output of the linear supplies. Apparently HP did the best job filtering all of this.

最后，TVB同学得出了一个这样的结论，不同质量的电源会对天宝的10MHz相噪造成显著的影响。使用电脑电源是最坏的主意，如果你要考虑相噪的话。屏蔽好的开关电源也许在你要求不是很高的时候是个好的选择，但正如文中指出的就算实验室的用的在频谱表现方面要比开关电源好的多的大功率电源还不如低端的可调台式电源。当然了最好的表现是HP一款非常古老的可调指针台式线性电源 6235A，极其平滑的曲线，几乎没有突变噪音，简直是测试相噪的理想电源。

要知道无论是HP的Z3081A，还是6235A都有着非常悠久的历史，怪不得TVB同学在做完功课以后，面对HP这些老当益壮产品的表现，不禁叹道：Don't you just love HP?

你能不爱HP，不爱TVB吗？

申明：所有测试数据均来自 http://www.leapsecond.com。