电子水平仪拆解一（Talyvel）
毫无疑问TaylorHobson是第一家做出机械用的电子水平仪的，这件事情大约发生在65年前，更有意思的是从Talyvel1开始，其核心部件就再也没有变化过，我虽然没有拆过Talyvel6，但是从体积和参数判断内部仍然是一样的。第一代的整机照片如下实物的图是借的，我的到手就给拆掉了。这图中所有的东西加在一起150欧元被拍卖掉了。右边那个有旋钮的就是传感器，中间的带表头的是电箱，图中还有很多附件，都是用于做不同测量时使用的
官方的原理说明如下，传感器中有两个电感线圈，中间悬挂一个金属片，在水平位置的时候，左右线圈的电感量相等，指针指零，当发生倾斜的时候，中间悬挂的金属片就会向一侧靠近，同时远离另一侧，这就打破了平衡，使用电箱将这个变化量测量出来就构成了电子水平仪的基本原理。
上海水平仪厂曾经仿制过，电路图如下，原版的talyvel电路图我没有找到，但是我估计基本上是一样的。
官宣的指标如下，如数时间为1秒，这是最有意义的地方，比起传统的水平泡，读书时间减少了至少10倍。分辨率是1秒，这是角度单位，换算下来大约是5um/m，也就是0.005mm/m，相比之下合像水平仪也就只有0.01mm/m，更重要的是这个表头很大，实物有巴掌那么大，也就是说可以估读，
英国人做东西很讲究，他们把电箱做的很大，但是里面几乎是空的，用来放置传感器，甚至还做了收线的骨架。
传感器打开如下，白色的部分都是骨架部分，黄色的部分是铜质的摆，选用铜的原因是避免外部的磁性吸引导致读书偏差，这样铜块就只受重力。焊线的位置左右各有一个线圈，中间有个金属片就是原理示意图中画出来的那个用来影响线圈电感量的零件。材质未知。
这是另一面，依然是白色的骨架部分和黄色的铜摆部分。
这张图可以看的更清楚
最精密的地方是悬挂部分，talyvel使用了5根铜丝来做悬挂，巧妙地使整个摆的运动被限制在了一个自由度上，铜丝直径大约是0.05mm
正面一根铜丝在中间
两个侧面各有两根铜丝，一共5根
这就已经是所有的秘密了，顶部的两个微分头可以用来做一定范围内的角度调整，以及拓宽水平仪的可使用场景（比如需要在一个2度的平面上使用），但是对于如此小的角度而言，这个微分头的动作实在是太大了，所以更多的时候需要使用电箱上的电位器来实现电子调零，精度会有一些变化，但是整体的量程范围内都可以保证小于1秒的精度，请注意这是角度的精度，也就是整个量程范围内的角度测量精度，而如果我们只用来做直线度的测量和调整，我们只会用上很小的测量范围（和机床的精度要求有关），这时就更需要关注重复性和漂移指标了，这也是很多朋友容易误解的地方。
另一个同样的误区是准直仪的最大可工作距离，有些准直仪会说5M，10M，30米的工作距离，指的是在这个距离下可以进行全量程的测量（比如常见的300秒量程），超过这个距离会发生什么事情呢？并不是就不能成像了，十字架就不见了，而是最大量程就变小了，如果我们只使用靠经0附近的量程（测量机床轨道的直线度一般不会超过300秒，除非是废品），是可以正常使用的，这时的最大工作距离将由灯泡的亮度、靶镜的光洁程度、环境光的亮度、工作环境中的气流温差影响，这些因素越差，工作距离越远，目镜中看到的十字架就会越来越模糊，对比度下降，直至最终无法有效进行瞄准。
有一种光路的准直仪除外，但那不符合阿贝原理，我国并没有大量引进生产。
这个真的不难，本人已经仿制成功，以下是一些制作过程中的照片和测试结果，有问题的朋友请跟帖，本人将尽力解答
静态测试1小时漂移大约1um/m
重复性测试最大5um/m，这张图时间比较早了，后来优化到了1um/m
由于这玩意实在是太贵，所以我将其原理贴出，希望有能力的朋友都可以自己做出来，用上更好的量具，做出更精美的作品，谢谢，
感谢分享。
侧面的2根铜丝，是捞过铜摆上的小销钉吊着，并不固定，是这样么？
另一种想法，小水槽里放水银，左右两端各一电极悬于水银液面上方。电极与水银液面间距变化，对应电容变化，然后测量LC回路的振荡频率，估计亦能实现高分辨率。
顺便一问，玻璃管式的水平仪，玻璃管内的弧度是怎么生产的？
1、侧面的V字型悬丝底部压板需要紧固
2、水银虽然是金属，但是热膨胀系数是液体级别的，比固体的要大得多。其二是它的表面张力很大，所以液面不平，除非面积做大。其三是水银的粘度很小，不容易实现阻尼。目前为止做电容传感器的我了解到的材料一般都是石英玻璃镀金或者金属片镀金，用得多的都是微晶玻璃类的，平面光洁度好，热膨胀系数小。
3、据我所知的，这个水泡的内弧面早期是磨制的，像磨制内径规一样的方法，后来变成压制的了，具体的压制方法我也不太清楚 本帖最后由 xyz_123 于 2023-1-31 16:26 编辑
「侧面的V字型悬丝底部压板需要紧固」，这样的话，想不透中间那根吊丝的作用。左右两侧各2根吊丝、前后向叉开吊，不就已经限制铜摆不能前后动、扭转，只能左右动了么？
若侧面的2根吊丝是捞过铜摆上的小销钉吊着，并不固定，则还以为是当水平仪机身前后向略有倾斜时，吊丝捞过小销钉可滑动，铜摆可前后倾，整体上是左、右、中三个吊点确定铜摆这个平面块。
用水银，设想的是如一根铅笔大小的水槽，电极悬于两头、距槽端一两厘米，避开边缘处水银表面太弯曲的部分即可。
热膨胀系数无妨，测量的是差动电容。
流动性好是优点。阻尼估计加个磁场即可，感生涡流应该能令其很快稳定。
水平泡仅筷条粗细，内部弧度很浅，若是斜伸一细砂轮进去磨制的话，砂轮弧形摆动的机构想象起来颇庞大。
一种设想，弧度很浅，可以不摆动砂轮来实现，而是平行伸缩砂轮，通过停留在玻璃管某段的时间来控制磨削量。
炼制玻璃管时，通过外模具、内加气压，令管中段略微鼓胀，这个想着蛮困难的样子。
本帖最后由 fobyellow 于 2023-1-31 17:45 编辑
1、两个V型的悬挂的确可以达到约束自由度的作用，但是那就要求悬挂点的中心和质心在竖直平面内重合，这样在V的底部正切值切下来的相对于质心的有效作用力臂太短，对于在正交方向的自由度约束强度不够。这种设计下的正交方向上的不平冗余度会比较小。
增加一根丝后，第5根丝的力臂等于半个边长，另外4根丝的等效力臂长度也变成了半个边长，约束度就可以极大的提高。后面我会拆解mahr的悬挂设计，他们用的是簧片，比细丝的强度高得多，就可以做对称设计了，但这只是理论，后来马尔在几年后也改成了非对称的设计
在另一个领域的固体潮观测装置里面，他们几乎只约束1个自由度，观测量程也只有几秒，而且不用移动，也能够满足要求，精度爆棚到0.0001秒
2、目前我了解到的差动电容检测方案的极限值大约是埃法左右的检出率，也就是对应的检测距离是nm级别，这个想下都头皮发麻。基于电磁阻尼的设计中，阻尼只会出现在有速度的条件下，而且是速度越快阻尼越大，在最后阶段阻尼会越变越小，我猜这是新泻精机的读数时间大约为10秒钟的原因，当然另一个原因是进行噪声的平均。
还有个原因是正交方向下的不平会导致极板不平行，产生读数变化，除非你设计成十字架的造型，然后做矢量求解
另外差动电容不是万能的，这个温度导致的共模的干扰信号是巨大和非线性的，我看做传感器的厂家都在选择低热材料。目前商用的能够做好现场补偿的电容传感器还真的只有美国雄狮，其余的补偿都不行，德国货也不行，这一点请参考各大拼接镜面天文望远镜的位移传感器方案，能搞电容的只有美利坚，其余国的几乎都是电涡流。
非要用水银的话，我只想到了准直光源下的测角原理，这个对于液面高度的变化不敏感，但是小型化会是个问题
3、研磨的工艺是成熟的，不是硬质砂轮，应该是比较软的材质，用压力和转速加时间来控制磨削量。我估计外模具不行，这个只能影响外表面，内表面要反过来做，比如用低热膨胀的抛光磨具伸到玻璃管内部，然后加热加压，感觉还是很复杂的样子 Second posting without attached pictures:
Making a 10 second spirit level vial for a mirror artificial horizon.
To use a mirror as an artificial horizon it must be carefully leveled out.Machinist levels are available to 20 seconds of arc but they are expensive, starting at around $70.00 on e-bay and rather large for the purpose. You can buy a mirror a.h. with level vials already included, from Freiberger, but it will set you back around $900.00. The cheaper route is to make your own spirit levels. With a home made level vial sensitive to about 10 seconds of arc, it is possible to get very respectable results using a mirror artificial horizon. (Routinely, intercepts from 0 to 1.5 miles using a Davis Mark 15 sextant.) I have made more than a thousand observations this year using a mirror a.h. and would recommend them to anyone for accuracy, ease of use and speed of set-up. What follows is a ‘how-to’ on building a jig to grind glass level vials. The design is a simplified version of a design by Bill Morris and he generously provided a very detailed article he had written on the subject some years ago. (Article by B. Morris in another posting) I realize there is too much detail for most Nav-list members but I wanted anyone to be able to follow it. (Photos attached below)
A very sensitive spirit level vial is just a glass tube ground on the inside, into a minutely tapered barrel shape. This grinding is done by spinning the tube on a slightly curved round steel bar that has been smeared with a grinding paste. The trick is getting the correct curve on the bar and holding that very small curve as you spin the tube by hand.Mr. Morris’s article shows such a jig made by a skilled machinist with access to sophisticated tools.
The jig is made from a section of 1 x 1 mild steel square tube cut to about 18 inches length. Centered on the bar, drill a pair of 5/16 in. holes with centers exactly 340mm apart. This distance is important as it is part of the math for getting the correct arc on the grinding bar. Drill another pair of holes outside these, spaced about two inches outside the first holes. This distance is not critical.The holes can be done by “eye” with a hand drill since it will not matter if they are tilted a little. Four, five inch long, 5/16 eye bolts are inserted through these holes, with nuts/washers above and below the tube. A pair of bar clamps is used as a base to hold the jig as the vial is spun. A round 3/8 inch diameter, 18 inch long steel rod is inserted through the four eye bolts with the glass tube slid on in the middle. You must have the glass tube slid over the bar before set up. The outer pair of eye bolts is then tightened; pulling the bar ends downward forcing the bar to bend. Bend the bar in this way until it is displaced from straight by only 0.28 mm, checking with a feeler gauge/spacer combination. (For a 10 second vial) This is not much of a bend, roughly 10 times the thickness of a human hair. It is difficult to verify this bend with the feeler gauge. A large bolt can be used as a spacer, in conjunction with the feeler gauge, setting the ends of the bar at the height of the bolt, moving the bolt to the center, and bending until the feeler gauge just fits between the bolt and . When the correct bend is achieved, tighten both the top and bottom bolts to lock the curve of the bar in place. During grinding you may stop and verify the correct bend by sliding the tube aside and putting the bolt and feeler gauge in the center.Having too much bend is the greater risk so under-bend rather than over-bend.More bend means decreased sensitivity and you can always grind more after testing.
To start the grinding, the curved bar is smeared with a coarse automobile valve grinding paste and the tube spun back and forth with a wooden paddle to shape the inside of the tube.The grinding paste is that used to polish engine valves and both a coarse and fine grade is needed. They are available in a two-pack from Amazon for $12.11 as “Powerbuilt Valve Lapping Compound. To spin the tube by hand, the wooden paddles are covered with a thin sheet of rubber.To make the paddles for spinning the tube use ¼ inch thick wood and cut this into one1.5 width and another 6 inch width. The narrow paddle was just a rectangle and the wider was cut with a handle for easier grip. In a test run it took 15-20 minutes of doing push-pull cycles with the small paddle to get the ground area out to the edges of the vial. (In his article Mr. Morris recommends a preliminary grind on a straight bar to ensure a consistent internal diameter.) Initially use a 1 inch wide paddle in the center of the vial, then as the ground ‘frosted” area reaches the edges of the vial switch a paddle that is the width of the vial. Only use enough pressure to get the tube to spin because excessive pressure will tend to flatten out the desired curve of the bar. After the desired internal shaped is achieved, the vial and bar are cleaned of coarse paste and then given a final polish with a fine grind/polishing paste to finish the inside of the tube. Clean the tube thoroughly as any residue will be floating in your vial afterwards. It is possible to change from coarse to fine paste without removing the tube form the jig. Slide the vial to one side and wipe the bar clean.You may then smear on the fine paste and resume grinding. After the first run of grinding you should test the curvature by filling and using temporary rubber caps.
The glass tubing for the vial was purchased from an on-line glass-blowers supply shop for less than $5.00 in a five foot length. This length will yield about nine 12.5 mm long vials. Shipping was about $11.00. The glass company did ask if it was OK to cut the tubing into 3 sections to decrease shipping charges. Tubing wall thickness of 2.5 mm wall was used, so enough material is available in the wall to keep it from losing strength as glass is ground away from the inside. (In retrospect I could have used 1.8 to 2.0 mm wall thickness.) The glass tube can be cut to length with a Dremel-type diamond cut-off wheel. To avoid the need for machining the end caps, a glass tube of 16 mm OD was chosen that would allow use of stock brass pipe fittings for the caps.
Caps are made from brass ‘compression fittings’ that are threaded, have six flats and cost less than $3.00 each. The ID of these caps is 16.01 mm or so.It may be necessary to touch the inside of the caps with a small sanding drum if the fit is too tight. Before gluing the caps in place a 6/32 inch hole is drilled and tapped into one of them then plugged by a screw. This is allows the vial to be filled/refilled as needed.A tiny strip of plumbers tape is wrapped around the machine screw stopper to assure sealing. Measure the thickness each of the six flats on both caps and try to match up a pair of sides of equal thickness, to be on the lower side of the vial as it is glued in place. If there are small differences, remember that the level scale is applied and adjusted after construction and can make up for slight discrepancies in the shape/thickness of the end caps.The caps are glued in place using an epoxy such as Araldite or other solvent resistant two-part clear epoxy, available for around $5.00. The curing time can be up to 12 hours. The end cap threads fill in with epoxy giving extra surface area for a proper seal.Roughening the gluing surface of the glass tube ends before gluing may also be a good idea.
The tubes are filled with denatured alcohol, leaving enough space for an air bubble 3-4 cm long. A drop of food coloring can be added for tint to make the bubble more visible. If fluid leaks over time it can be refilled in the same way.
To test the accuracy of the level during and after completing grinding, first clean the vial, fill appropriately and cap with rubber stoppers. Attach the capped vial to a thick straight edge and shim up one side with a feeler gauge of ‘T’ mm to form a triangle underneath. It is advised that you “ concentrate on getting symmetrical bending and then, when you have even grinding all around, check the vial using feeler blades, a straight edge and simple trigonometry. Some simple math/trig is used, T/L = tan A, so A = inverse tan A. If the tested sensitivity is correct you are ready to polish and assemble with brass caps.If not, you may do additional lapping as needed. The distance in mm the bubble moves when you insert the feeler, is divided by 2.5 to give the number of divisions of movement and divide the angle A by that figure to give the sensitivity of the level per division. But you will not be able to measure these distances without feeler gauges and they turn out to be only a very rough guide. You measure the distance the bubble moves when you insert the feeler, divide by 2.5 to give the number of divisions movement and divide the angle A by that figure to give the sensitivity of the level per division. (see 1st paragraph, p 41).”(Italics by B. Morris)
After testing and more grinding if needed, clean the vial very thoroughly, glue on the brass caps and fill with denatured alcohol. You must attach the scale to the vial and detailed instructions and here are detailed instructions:
“1) Print out the scale and get it copied to an overhead projection film. Cut it to a strip 100 long and 40 mm wide, which is about 7 mm less than the circumference of the tube.
2) Place the level on a flat, smooth and approximately level surface. If your hospital instrument maker has a surface plate, that would be ideal, or the top of a milling machine table.
3) On this surface will be some line which is level, so you swing the level around until the bubble is about central and make an ink mark at each end.
4) Lay a heavy straight edge along side it to mark the position accurately and reverse the level end for end. IF the bubble stays between the marks, you’ve been lucky. GOTO 6) ELSE...
5) If not, mark the ends again and swing the level to split the difference. GOTO 4)
6) Lay the strip on top of the level so the bubble is centrally between the marks and mark the position of the strip at each end on the tube. Clean other marks off the tube.
7) Carefully wrap the strip around the tube, printed side down, and secure it with a long strip of sellotape. Check that the ends are aligned with the marks ELSE remove it and GO TO 5)
8) To check, swing the level until the bubble is central on the scale, Lay the straight edge alongside and reverse end for end. The bubble should stay central.” (Italics by B. Morris)
With his kind permission, I will be posting Mr. Morris's detailed article soon. My thanks to him for his generous assistance.
Randy 上面的文档是一个爱好者自己制作水平泡的方法，印象中看到过一篇文章介绍圆柱角尺的圆柱度研磨，也是金刚砂一边检一边磨的 Talyvel的电气原理并不是LVDT的原理，甚至也不是常见的半桥传感器，严格意义来讲是左右两个电涡流传感器，说得再白一点就是两个电感器，电感中被加以高频信号，周边就会产生交流的磁场信号，靠近的金属表面就会感应出相应的电流信号，这个感应电流信号产生的磁场和电感传感器线圈自身的磁场有抵消作用，所以就会改变这个线圈的电感量。
当左右两个线圈对称放置时，中间的金属片会随着水平仪工作面的倾斜而发生移动，向一侧靠近的同时远离另一侧，这样就相当于一侧的电感减小，另一侧的电感增大，把这两个电感进行串联，然后测量中间抽头的电压值，就可以映射为金属片的位移量，从而测量出水平仪的水平读数。
测量这个交流值的方法很多，参考电感测微仪可以得到大量的参考电路，我推崇的电路结构是双路模拟开关相敏检波，一路检出抽头送ADC，一路检出激励信号作为ADC_REF,这样大部分和连接电阻以及温度漂移的变化量会同时出现在分子和分母上，做出的电路绝对温漂会很小，用4层GT150板或者导热胶粘合模拟部分何以降低板上的温差温漂。
假如有人做的话，我可以帮助审图
页:
[1]