Needing more than a spark test?

[graham-xrf]

I have not had experience with angular contact bearings (yet), but I did check out Robin Renzetti's spindle rebuild. It seems angular contact are the ultimate for spindles, and they can also be obtained with pre-set preload built in. I thought taper rollers are the choice for things like wheel bearings. At least on wheels, one is supposed to take up all the play, and then "back off" some part of a turn, which allows for them to heat up. That is stuff I am just not sure about. You will find angular contact types are just excellent!

 

[homebrewed]

I have not had experience with angular contact bearings (yet), but I did check out Robin Renzetti's spindle rebuild. It seems angular contact are the ultimate for spindles, and they can also be obtained with pre-set preload built in. I thought taper rollers are the choice for things like wheel bearings. At least on wheels, one is supposed to take up all the play, and then "back off" some part of a turn, which allows for them to heat up. That is stuff I am just not sure about. You will find angular contact types are just excellent!

From what I've read concerning tapered roller bearings vs. angular contact bearings, 7x owners are about evenly divided in what they use for an upgrade. Little Machine Shop sells all three types - that's including the deep groove version - so they aren't favoring one vs. the other.

The AC bearings LMS is selling have no shields, but the lathe hardware has something to serve that purpose.

 

[graham-xrf]

Re: Easy Integration
I know I am here getting into some old-school analog, but it seems to me that for us, extremely accurate real-time integration can be done with the same opamp integrator circuit as was (maybe still is) used to control missile guidance feedback loops, Saturn 5 rocket engine gimbals, LEM landers, etc. This is just another passing notion that I thought might eliminate a whole bunch of sampling and integrating software. All the program need do is listen to the trigger, fetch the answer from ADC, and reset the integrator. The number is ready to be used.

Sure - I do know that Moore's Law is reaching it's limit for for digital stuff [Ref: 1], because transistors are becoming nearly the same size as atoms, and analog computing is making a comeback , but this is much simpler stuff. An opamp integrator delivers at the speed of electrons, with infinite resolution on the time axis. If started by a trigger set a bit above the noise level, it will ride up to the value analogue of the photon energy in the pulse, effectively, the area under the curve. Follow it with an inverter buffer for a positive output. The circuit needs a FET analog reset switch for after the pulse is deemed to be over.





Opamp integrator is good for all amplified photon current pulses from TIA's, stretched or not. Yes, the final circuit will have a few more parts, there needing to be an enable-reset and all, but the bits are tiny and cost only cents. One of the simplest just uses a FET to short out the capacitor.

The low pF value capacitors seen across the feedback resistors in the Pocket Geiger amplifiers were there for stability. This is different, and such an integrator would come after the gain stages. Having a capacitor be the opamp main feedback, makes an integrator, and any resistor across it turns it into a low pass filter. For accurate computing, you make the resistance value so high it can't affect the computation, or you just don't need have any resistor there at all. For the duration the pulses have, low leakage capacitors, and special substrate materials, etc. is just not needed.

There are, of course, dozens of integrator circuits described on the internet. Obviously, they were very popular!

[Ref 1: The Most Powerful Computers You've Never Heard Of ]

 

[homebrewed]

Long ago one of the engineering classes I took was on control theory. It included a lab, which used analog computers to simulate control systems. You could combine integrators, differentiators, gain blocks, summers etc. using banana plugs. Plots of the system response were done on chart recorders. Not a digital computer in sight! No calculators, either: the class predated the now famous HP35.

 

[graham-xrf]

Long ago one of the engineering classes I took was on control theory. It included a lab, which used analog computers to simulate control systems. You could combine integrators, differentiators, gain blocks, summers etc. using banana plugs. Plots of the system response were done on chart recorders. Not a digital computer in sight! No calculators, either: the class predated the now famous HP35.

Funny you should say that! Since Christmas, I have a brand new HP35.

 

[WobblyHand]

Are you looking for an integrate and dump? So you can reset the integrator? How can we expect the average person to deal with charge injection of the dump/reset switch, or for that matter to make sure one isn't integrating on input offsets, or even dc bias? An integrate and dump is an optimal circuit, but they are tricky in practice, especially for the less experienced. Does our target audience have this expertise? Not trying to rain on the parade, but merely trying to bring this back to technology that could be replicated en mass without having years of analog experience.
Current (heh, bad pun,) path is fine for me, but suspect folks with less analog experience will get less than a satisfactory experience. Circuits for the masses may have to include a lot of trimpots, or require resistor substitution boxes to prevent saturation, along with setup instructions. Analog tweaking is becoming a lost art.

Sad to say, (as someone who likes analog stuff,) it may be a lot better if we digitize at a reasonable rate and do the heavy lifting in the software. Lots easier to make a software change and host on your git site of choice. Just my thoughts on this. What say you?

 

[WobblyHand]

Funny you should say that! Since Christmas, I have a brand new HP35.

What do you do for batteries? The LED displays tended to run down the batteries pretty quick. I have an old HP45, with the leather case. I'm not sure where the heck the charger is! For general use, when not at a computer, or in the shop, I use a HP15C. Three button cells, lasts for years, if not 10 years. Long live RPN!

 

[graham-xrf]

Are you looking for an integrate and dump? So you can reset the integrator? How can we expect the average person to deal with charge injection of the dump/reset switch, or for that matter to make sure one isn't integrating on input offsets, or even dc bias? An integrate and dump is an optimal circuit, but they are tricky in practice, especially for the less experienced. Does our target audience have this expertise? Not trying to rain on the parade, but merely trying to bring this back to technology that could be replicated en mass without having years of analog experience.
Current (heh, bad pun,) path is fine for me, but suspect folks with less analog experience will get less than a satisfactory experience. Circuits for the masses may have to include a lot of trimpots, or require resistor substitution boxes to prevent saturation, along with setup instructions. Analog tweaking is becoming a lost art.

Sad to say, (as someone who likes analog stuff,) it may be a lot better if we digitize at a reasonable rate and do the heavy lifting in the software. Lots easier to make a software change and host on your git site of choice. Just my thoughts on this. What say you?

I get it about charge dump, and offsets and so on. I would not expect folk to have to work up an integrator circuit that is prone to that stuff. If I propose one, it would already have been tested. In my circuit, there are no DC bias issues. The base line of the amplitude is expected to be within microvolts of zero, basically from the input offset voltage, and even that I try to take care of. I don't want any stuff like clamp circuits for dc restoration, nor undershoot as it returns to zero. Also, not the arcane software approaches to "correct for" or "process out" stupid stuff like waveform baseline return overshoots. This is one of the motivations I have for using gain-bandwidth products sufficient to accurately reproduce the pulse shape. I see it at something precious, to be preserved until we get it measured.

We can use quite powerful digital approaches, but they require quite powerful digital computing. Analog integration seems a gorgeously simple alternative for a KISS approach that is also accurate. 300kHz shape-defining content in a waveform is easily within the capability of our 16-bit ADC's to deal with, and still OK for Raspberry Pi's, and smartphones, but one is pushing the limits.

In my kit, the ADC costs $40 bucks. Use about $2 worth of analog bits, and change the ADC for a slower one costing about £15, and pick up a sample every 15uS instead of every 500nS, and there is time to do a reset. I was just feeling out another way, trying to knock off cost here and there. It's bad enough the diode alone costs $65 if we get it on a Pocket-Geiger.

As it happens, my Raspberry Pi can easily add up ADC samples in a software integration, though it would be going a whole lot faster than working on stretched pulses. I am still tempted to try a hardware integrator, though you do correctly point out the temptation is to save some components, and write more code instead. I will end up going the line of least resistance.

 

[graham-xrf]

What do you do for batteries? The LED displays tended to run down the batteries pretty quick. I have an old HP45, with the leather case. I'm not sure where the heck the charger is! For general use, when not at a computer, or in the shop, I use a HP15C. Three button cells, lasts for years, if not 10 years. Long live RPN!

So far - no real test yet. I did try to find out about battery life. It uses two CR032 button cells. User comments said it can work in anger for months. We shall see!

 

[WobblyHand]

I get it about charge dump, and offsets and so on. I would not expect folk to have to work up an integrator circuit that is prone to that stuff. If I propose one, it would already have been tested. In my circuit, there are no DC bias issues. The base line of the amplitude is expected to be within microvolts of zero, basically from the input offset voltage, and even that I try to take care of. I don't want any stuff like clamp circuits for dc restoration, nor undershoot as it returns to zero. Also, not the arcane software approaches to "correct for" or "process out" stupid stuff like waveform baseline return overshoots. This is one of the motivations I have for using gain-bandwidth products sufficient to accurately reproduce the pulse shape. I see it at something precious, to be preserved until we get it measured.

We can use quite powerful digital approaches, but they require quite powerful digital computing. Analog integration seems a gorgeously simple alternative for a KISS approach that is also accurate. 300kHz shape-defining content in a waveform is easily within the capability of our 16-bit ADC's to deal with, and still OK for Raspberry Pi's, and smartphones, but one is pushing the limits.

In my kit, the ADC costs $40 bucks. Use about $2 worth of analog bits, and change the ADC for a slower one costing about £15, and pick up a sample every 15uS instead of every 500nS, and there is time to do a reset. I was just feeling out another way, trying to knock of cost here and there. It's bad enough the diode alone costs $65 is we get it on a Pocket-Geiger.

As it happens, my Raspberry Pi can easily add up ADC samples in a software integration, though it would be going a whole lot faster than working on stretched pulses. I am still tempted to try a hardware integrator, though you do correctly point out the temptation is to save some components, and write more code instead. I will end up going the line of least resistance.

I tend to agree with your feelings with baseline return overshoots and rubbish like that. Have had awful times dealing with it. When I had a chance at making changes, I eliminated those sorts of circuits.

My concern is we build a few of these circuits (whatever they turn out to be) and they are ok. But, we may see quite a few issues as they get built up in quantity. I had high volume experience and saw this effect. We used to say, "anyone can build a couple, but it takes skill to make millions". (Lots of engineering!) Or our participants will use products from different lots, (or manufacturers,) or inadvertently get counterfeit parts which maybe externally indistinguishable from real ones. Causes a lot of heartache. Any way to eliminate analog tweaking in this X-RF unit is probably a good idea.

This is not to say that software never has problems! Seen lots of issues. But overall, it is usually faster and cheaper to scale up production of digital designs than analog ones. That's why we see digital overtaking analog.