|
taqueso posted:I drew a circuit that is supposed to turn on a "voltage OK" LED when voltage is close to 5V, and turn on a 2nd "overvoltage" LED if voltage is over ~5.3V. Standard TL431 parts need ~1mA to stay in regulation - i.e. they will draw around 1 mA at all times, even when "off". You also might want to add some hysteresis - what happens when your input voltage is precisely at 4.5V? edit: Also, it looks like D1 will turn off when D2 is turned on since you're connecting the regulating terminal of U2 to U1. Is this what you want? SnoPuppy fucked around with this message at 01:54 on Jan 12, 2012 |
#
?
Jan 12, 2012 01:42
|
|
|
|
| # ? May 21, 2024 11:46 |
|
|
It's a cookbook circuit out of the TL431 datasheet. Without the recommended resistor around D1 you might have some issues.
|
|
#
?
Jan 12, 2012 04:06
|
|
|
Has anyone got any advice on PCB routing. The rough overview of the circuit has a 100pin lqfp .5mm pitch in the center, a 40 pin LCD/touch header, 2.56mm pitch and a 50 pin header. I'm using eagle. I don't think the auto-router is a realistic option. I haven't been able to find much on routing. Has anyone got any recommendation on any good sites or books on routing.
|
|
#
?
Jan 13, 2012 13:08
|
|
|
Unparagoned posted:Has anyone got any advice on PCB routing. The rough overview of the circuit has a 100pin lqfp .5mm pitch in the center, a 40 pin LCD/touch header, 2.56mm pitch and a 50 pin header. Uh, could you be more specific? Unless the application is something specific (very high speed digital, RF, high power, etc) there isn't really any magic behind trace routing. It's often a repetitive, trial and error process, and not an exact science (in general). For a 0.5mm pitch part, you'll need decent DFM rules (like 8 mil clearance and 10 mil traces), and if it's 100 pins then you may end up needing 4 layers. What is the IC? The difficulty in routing traces is impacted greatly by component placement. Try to partition the design into sections related to their functions and what other parts of the system they need to interface with. Again it's not an exact science.
|
|
#
?
Jan 13, 2012 14:41
|
|
|
ANIME AKBAR posted:Uh, could you be more specific? Unless the application is something specific (very high speed digital, RF, high power, etc) there isn't really any magic behind trace routing. It's often a repetitive, trial and error process, and not an exact science (in general). It's a basically a breakout board for the STM32F103VC. 72Mhz. Will have a 16bit LCD interface, SDIO, a couple PWM outputs, and some other other digital outputs. Component placement isn't much of a problem since there are relatively few components. I'm trying to stick to a 2 layer PCB. The problem with trail and error is that there are ~100 signals, which leads to lots and lots of errors. So I need more of a system or set of basic rules that will reduce the errors. I'm currently looking at similar types of PCBs seeing how others route the signals and am trying to come up with my own system, but I'm just repeating what others have already done. I plan on doing something like having vertical signals on the top layer and horizontal signals on the bottom layer. Near the horizontal headers I plan to reserve a parallel horizontal strip for each pin, so routing would be easy. I'd then optimise once all the signals had been routed.
|
|
#
?
Jan 13, 2012 17:09
|
|
|
I'm really excited. Money has been tight for a couple of years, and I have had to lay the iron down for a while. But yesterday when I was at Radio shack buying thermal paste for a GPU reflow project, I was reminded that they carry Arduino stuff now. Picked up an Uno and have successfully annoyed the wife with blinking LED's and speaker beeps already. These things are a ton of fun! I have a Big Project That I've Wanted To Do in the back of my head, and now I can get started. I am going to take one of my spare Icom PCR-100 receivers (it's basically a wideband radio controlled via serial) and build a 44780LCD + couple-of-buttons interface to it. Just something proof of concept to tune around and mute it and maybe put a couple of presets in. The final phase will be to put Arduino + Icom into a stripped and finished vintage radio cabinet, do a Nixie frequency display and add IR remote, replace a couple of the knobs with rotary encoders for tuning and volume. Modern radio tech in old package, more or less.
|
|
#
?
Jan 14, 2012 08:42
|
|
|
Not to spoil your fun but http://pjrc.com/teensy/teensyduino.html It's an arduino with a healthy dose of Better™ (and a dash of Cheaper™) And two pinches of Fits On A Breadboard™
|
|
#
?
Jan 14, 2012 08:52
|
|
|
Yeah, I'll be using a Teensy or something similar when I build the final product. I'm not going to invest the time in building up the whole AVR circuit side from scratch for this project - it's more efficient for me to just buy a board and wire it up - but this was just a fun instant-gratification surprise to kind of motivate me a bit.
|
|
#
?
Jan 14, 2012 09:15
|
|
|
I just like to use every opportunity to offer up the teensy as a drop-in replacement for the arduino. And it really is that, most you'll likely have to do is change some pin numbers. I was a big fan of arduinos for a long time, and then I realized that everything I didn't like about the arduino was fixed with the teensy. They're pretty freakin' sweet, and all my old LED-blinking code runs on the teensy. 
|
|
#
?
Jan 14, 2012 10:31
|
|
|
Bad Munki posted:Not to spoil your fun but http://pjrc.com/teensy/teensyduino.html Eh, I've always thought those looked really poorly designed, but maybe that's just me
|
|
#
?
Jan 14, 2012 14:20
|
|
|
Unparagoned posted:It's a basically a breakout board for the STM32F103VC. 72Mhz. Will have a 16bit LCD interface, SDIO, a couple PWM outputs, and some other other digital outputs. Only after doing that stuff, should you think about routing traces. It may or may not be feasible, depending on how things shake out. Try to avoid using the bottom layer as much as possible. For signals that are high speed (like >30MHz), try to avoid having them switch layers. If necessary use jumpers (0 ohm resistors), they can be lifesavers, both for routing power supplies and signals.
|
|
#
?
Jan 14, 2012 17:00
|
|
|
In something like a SWR bridge where you couple three microstrips together (thru, forwards, reverse) how do you size them? I'm mainly interested in maintaining Zo on the thru line. If the coupled lines are a different impedance, that's fine, but I would need to know what impedance it is so I could properly terminate them.
|
|
#
?
Jan 14, 2012 17:10
|
|
|
ANIME AKBAR posted:Okay, so if you have a high end microcontroller, first thing you should do is add its necessary external components, mainly bypass capacitors, oscillators, maybe a couple ferrite beads, etc. From glancing at your datasheet, your device has about six VDD pins, each of which needs at least one bypass cap (datasheet recommends two: one 100nF ceramic, and one 4.7-10uF ceramic/tantalum). You can get the 100nF in a 0603 package. If you don't need a lot of output current on your I/Os, then the 4.7-10uF caps probably are not necessary, or you could drop it down to 1uF, which can also be had in a 0603 package. Put those guys on the PCB, as close to the VDD and VSS as possible. Then you will probably start to worry, because those capacitors are now in the way of a lot of signals. Having to route the Vdd supply will make things worse. This approach should apply to each IC. Thanks. I'm mainly concerned about how to do the stuff after that part. Most of the VDD and VSS pins are in the corners so that helps out allot. Well except for one side, which also has all the oscillator, reference pins, etc so while that side looks like hell the rest should be much cleaner. The power supply stuff is explained pretty much everywhere but I'm kind of getting stuck on the stuff afterwards. So I want to know more about stuff like: What's the most optimal way to get access to pins from the MCU, at some point a via will likely be required, whats are the common patters you can use to give enough space for the lines. e.g. Going horizontal different lengths then going at a 45 degree angle gives separation in the lines for the vias. Fan-out. Do people use them with this sort of chip or is that just for BGA. What are the basics of a fanout. Is it a bad idea to generally have many pins being accessed by a trace going under the MCU to a via? Is the horizontal on top and vertical on the bottom a good rule for boards like this, or would it be better to split the board into different regions and set up more local orientations. (It would be very hard and/or consume allot of space to either get all the top or bottom pins going horizontal on the top layer or vertical on the bottom layer). When routing a signal should you start from one end to the other or work both ends to the middle. What sort of patterns can be used to utilise the criss cross design without using ending up using a zillion vias. I'm guessing allot of the above depends on the situation and that there are no steadfast rules. I just felt that there must be more advice on the actual routing than just, component position, re-assign pins, horizontal on top, vertical on the bottom and 0 ohm resistors for jumps. Maybe if I put it another way. If I route all the power, oscillator and other crucial signals and then let the auto-router go wild, what would be the downsides? How and what would a person routing it all do better and how? What patterns, techniques or tricks do people use to be able to successfully route something like this?
|
|
#
?
Jan 15, 2012 00:25
|
|
|
Unparagoned posted:So I want to know more about stuff like: quote:Fan-out. Do people use them with this sort of chip or is that just for BGA. What are the basics of a fanout. quote:Is it a bad idea to generally have many pins being accessed by a trace going under the MCU to a via? quote:Is the horizontal on top and vertical on the bottom a good rule for boards like this, or would it be better to split the board into different regions and set up more local orientations. (It would be very hard and/or consume allot of space to either get all the top or bottom pins going horizontal on the top layer or vertical on the bottom layer). quote:When routing a signal should you start from one end to the other or work both ends to the middle. quote:What sort of patterns can be used to utilise the criss cross design without using ending up using a zillion vias. quote:Maybe if I put it another way. If I route all the power, oscillator and other crucial signals and then let the auto-router go wild, what would be the downsides? How and what would a person routing it all do better and how? What patterns, techniques or tricks do people use to be able to successfully route something like this? With purely digital design, there aren't really any tricks, except to keep things simple and short. With mixed signal designs, you have to start thinking about board partitioning, current interference paths, EMI, etc, but from what I gather you don't have a mixed signal design.
|
|
#
?
Jan 15, 2012 16:38
|
|
|
Speaking of mixed signal on a pcb. For making a 50 ohm transmission line, does the soldermask over the top affect things much? Dielectric losses I suppose, but capacitance and the like should be insignificant compared to the capacitance to the ground plane below... The only other option I guess would be to use ENIG gold plating and leave the soldermask off the transmission line. On a related note... any recommendations for a 50-ohm coax driver (all properly terminated etc..etc.. so true resistive 50 ohm load) that can do <= 1 nanosecond rise times? At most no higher than 2 nanosecond. For the source I can provide simple CMOS, LVDS, or LVPECL from a clock. But the driver must provide a very clean/flat square wave (either 1Mhz or 10Mhz, likely) with extremely fast rise/fall time.
|
|
#
?
Jan 15, 2012 23:41
|
|
|
Rescue Toaster posted:Speaking of mixed signal on a pcb. It will definitely change things, but we're talking <10 mils of different dielectric constant. Also the capacitive effects from the topside to ground plane have to be the least significant by far. Probably safe to ignore unless you're talking microwave (or possibly UHF) frequencies.
|
|
#
?
Jan 16, 2012 00:09
|
|
|
Well if we're talking rise times under a nanosecond, we're at least bordering on it.
|
|
#
?
Jan 16, 2012 01:10
|
|
|
Rescue Toaster posted:Well if we're talking rise times under a nanosecond, we're at least bordering on it. With some napkin calculations treating a solder masked microstrip as an embedded microstrip (epoxy solder mask is probably similar enough to typical PCBs), you could see a -5% change in impedance with 10 mils of mask.
|
|
#
?
Jan 16, 2012 01:25
|
|
|
As someone whose circuits operate solidly below RF, all this microstrip talk is blowing my mind.
|
|
#
?
Jan 16, 2012 01:54
|
|
|
Rescue Toaster posted:On a related note... any recommendations for a 50-ohm coax driver (all properly terminated etc..etc.. so true resistive 50 ohm load) that can do <= 1 nanosecond rise times? This is one of the fastest (80 ps risetime; I actually use this quite often): http://www.analog.com/en/other-products/automatic-test-equipment/adcmp580/products/product.html Or for the more affordable, something like this (claims <200 ps risetime): http://www.micrel.com/_PDF/HBW/sy89311u.pdf Anything with < 1 ns of risetime is going to be a CML/ECL driver, and if you need single ended output, you can use the positive output and terminate the negative output to ground (for CML) or VTT (usually VCC-2V). sixide posted:With some napkin calculations treating a solder masked microstrip as an embedded microstrip (epoxy solder mask is probably similar enough to typical PCBs), you could see a -5% change in impedance with 10 mils of mask. If your solder mask is 10 mils in thickness, you should really find another PCB manufacturer! Krenzo fucked around with this message at 03:13 on Jan 16, 2012 |
|
#
?
Jan 16, 2012 03:02
|
|
|
Rescue Toaster posted:Speaking of mixed signal on a pcb. Use solder mask unless you have a drat good reason not to. If you're asking the question, you should use solder mask. In fact, solder mask doesn't even really start to be noticeable (from a loss perspective) unless you're above ~6 GHz. At that point, you should be more concerned about fiberglass weave (you do know the weave structure of your prepreg/core, right?) than you should with the solder mask effects. And ENIG is actually worse than solder mask due to the nickel layer - at the frequencies solder mask would make an effect, the skin depth is so small that the Ni layer carries most of the signal. This is not good because Ni has a much higher resistivity than Cu (the Au only replaces the outer layer of Ni atoms) and will result in much higher loss. Ask yourself this question: Do motherboards leave off solder mask on the PCIe traces (running at 8 GBit)? Or the DRAM (running at 1.6 GBit)? To your other question, why do you care about the flatness of the square wave? Usually you only care about the edges, since they contain the information you care about (timing). Regardless, you should be terminating at the destination to provide the cleanest, sharpest edge with minimal ringing. You should be able to get away with any modern buffer - just double or tripple them up to provide more drive current. I have to ask though, what is your source? There's not much benefit to using a super high end buffer if you already have high phase noise/jitter. And if you do have a very clean clock you're distributing, you should look into buffers that are specific to clocking, like the ADCLK905, and couple that with a balun to get a single ended output. edit: sixide posted:With some napkin calculations treating a solder masked microstrip as an embedded microstrip (epoxy solder mask is probably similar enough to typical PCBs), you could see a -5% change in impedance with 10 mils of mask. Most solder mask is ~0.5 mil thick. SnoPuppy fucked around with this message at 03:18 on Jan 16, 2012 |
|
#
?
Jan 16, 2012 03:05
|
|
|
I need to generate a *very* clean square wave in the 1Mhz-10Mhz range, ie with as fast a rise time as possible (1ns pref, 2ns OK) into a 50 ohm coax (properly terminated at the end with 50 ohms). Only single ended, so with the ECL type drivers, I'll just have to terminate the inverting output. But, the top and bottom should be as flat as possible when properly terminated. This is to verify a 50-ohm 1Ghz oscilloscope input. And once rise time is verified on a terminated 50-ohm input, then it will be used to compensate high speed (300-500Mhz) passive oscilloscope probes. Source will probably be a MEMS oscillator, with CMOS (1-2nS) or LVPECL (250pS) output. I know MEMS is not frequency accurate (doesn't matter), but they're cheap and available with very good rise times. Rescue Toaster fucked around with this message at 03:40 on Jan 16, 2012 |
|
#
?
Jan 16, 2012 03:37
|
|
|
Rescue Toaster posted:I need to generate a *very* clean square wave in the 1Mhz-10Mhz range, ie with as fast a rise time as possible (1ns pref, 2ns OK) into a 50 ohm coax (properly terminated at the end with 50 ohms). Only single ended, so with the ECL type drivers, I'll just have to terminate the inverting output. Once you buffer the signal, the rise/fall time of the source won't matter. You should be able to get <=1ns edges from regular buffers (like the TI SN74AVC series), provided you double them up and drive at 3.3V. I'm still not clear on if you can destination terminate to 50 ohms - that will give you the best signal integrity. If you have to source terminate, you will run the risk of reflections bouncing around and messing up your compensation. I wouldn't bother with fancy buffers - your scope won't be able to measure a 80 ps edge, and it will probably cause more problems for you with reflections. What type of compensation are you going to be performing, anyway?
|
|
#
?
Jan 16, 2012 04:20
|
|
|
Yes, the destination will always be terminated with 50 ohms. The passive oscilloscope probes have special multiple compensation adjustments for high frequency, for rising edge, mid freqs, and then low freq. The output of the oscillator will run through RG-58 coax to a T-junction, into the 1Ghz 50 ohm input of the scope (which terminates), and then the T will go through a BNC-adapter for the probe. Comparing the input from the 1Ghz 50 ohm vertical amplifier to the 300Mhz 10M (x10) scope probe into 225 Mhz 1 Meg vertical amplifier will allow me to compensate the passive probe properly, since I have a 'reference' view of the square wave on the much faster 50 ohm input. I know that doing a T-junction into a 10meg ~15pF input will mess up the termination slightly, but I will be able to see the effect with and without the passive probe tapping the line, also.
|
|
#
?
Jan 16, 2012 05:40
|
|
|
Rescue Toaster posted:Speaking of mixed signal on a pcb. I haven't worried about soldermask yet on high-speed boards, and I've laid out PCIe, SATA, HT, QPI etc. on them with no issues. Maybe on much higher-frequency boards/protocols you'd have to start factoring in the effects of soldermask. At those frequencies too you need to pay attention to the crossweave of your PCB materials as well. I've seen some designers actually just "tilt" their design by some number of degrees to avoid issues with the crossweave. Regular circuit theory falls apart at high freqs because your physical circuit elements start to approach the wavelength in size. At 8GHz the wavelength is only ~40mm.
|
|
#
?
Jan 16, 2012 05:51
|
|
|
movax posted:I haven't worried about soldermask yet on high-speed boards, and I've laid out PCIe, SATA, HT, QPI etc. on them with no issues. Maybe on much higher-frequency boards/protocols you'd have to start factoring in the effects of soldermask. I've also heard that about rotating or zig-zagging traces to mitigate cross weave, but I usually don't bother. In my opinion, it's better to just spec in a tighter fiberglass, like 2116/3116. It also helps to use wider traces so that the effects of a single fiber weave are less noticeable, not to mention the reduced loss. A bigger problem with most FR4 type materials is the moisture absorption, which can change the dielectric constant and loss by a significant amount depending on the relative humidity. That's one of the main reasons why Rogers/Arlon material is really nice for RF - consistency. You don't want your dielectric oscillator or notch filter shifting by a few hundred MHz because it rained!
|
|
#
?
Jan 16, 2012 06:30
|
|
|
SnoPuppy posted:I've also heard that about rotating or zig-zagging traces to mitigate cross weave, but I usually don't bother. In my opinion, it's better to just spec in a tighter fiberglass, like 2116/3116. It also helps to use wider traces so that the effects of a single fiber weave are less noticeable, not to mention the reduced loss. Oh yeah, it was just given to me as an example by the PCB sales guys that gave us a presentation. They definitely recommended buying/speccing tighter fiberglass than tilting the design, hah. quote:That's one of the main reasons why Rogers/Arlon material is really nice for RF - consistency. You don't want your dielectric oscillator or notch filter shifting by a few hundred MHz because it rained! We have a legacy product, a charge amplifier, that deals in pC of charge. The legacy system has air-wired passives that are painstakingly tweaked by a line guy to achieve calibration. We started looking at the ultra high resistive materials from Rogers...good stuff, but oh-so-pricey. movax fucked around with this message at 07:08 on Jan 16, 2012 |
|
#
?
Jan 16, 2012 06:47
|
|
|
SnoPuppy posted:Most solder mask is ~0.5 mil thick. Yes, 10 mils was some worst case I found by google. I assume that's when you're using hand assembled boards from 1930 or something, because I've never seen mask thicker than a couple mils.
|
|
#
?
Jan 16, 2012 19:37
|
|
|
So, I'm hoping someone here can help me with something I've been banging my head over... I'm building a simple 555+Op-Amp PWM circuit which will operate at 20Khz (~5-100% Duty) and eventually go through a Power Mosfet to power a high performance 12v DC brushed motor (specifically, a slot-car). My problem is finding a mosfet driver that meets my requirements (or perhaps a misunderstanding of the requirements that I need). Primarily, the driver needs to operate fast enough to function at 20Khz PWM speeds and I would really like to have a current protection mechanism built in (there tend to be shorts and spikes often on the racetrack). This is the best candidate I've found: The Micrel 5021 ( http://www.micrel.com/page.do?page=/product-info/products/mic5021.shtml ) Is fast, has simple over-current protection and seems like it would work. The only problem is that even though it has super fast rise/fall times at the gate output and claims to work at up to 100Khz it also says "5kHz PWM for 2% to 100% duty cycle". With ~500nS rise/fall times why can't it function with 20Khz PWM? Any other suggestions on how to drive the mosfet at these speeds? I don't know why it's so hard to find something that'll work, but maybe I'm just looking for the wrong specs. Thanks FetusPorn fucked around with this message at 22:42 on Jan 16, 2012 |
|
#
?
Jan 16, 2012 22:13
|
|
|
I want to amplify a ~500mV signal at 10 MHz up to TTL levels and use this to drive a microcontroller and various other future things, how hard can it be? This is the circuit I've tested in PSPICE, it does work but it's pretty close and the duty cycle is about 75% (the MCU will run at 5 MHz anyway so the D-latch fixes that). Am I doing something stupid or is 10 MHz at the practical limit for those transistors? In the simulator pretty much anything would work until about 5 MHz. I'd prefer a more unconditionally working design that won't require very careful biasing to do anything at all... Opamps are mostly out at these frequencies, what are my other options? longview fucked around with this message at 23:43 on Jan 17, 2012 |
|
#
?
Jan 17, 2012 23:41
|
|
|
longview posted:I want to amplify a ~500mV signal at 10 MHz up to TTL levels and use this to drive a microcontroller and various other future things, how hard can it be? Step up transformer and AC couple would be one option (probably want to buffer it after the AC coupling though). Or use a differential buffer and tie one of the inputs to a ~250 mV reference voltage. I think the FIN1002 should have a wide enough common mode.
|
|
#
?
Jan 18, 2012 00:00
|
|
|
FetusPorn posted:So, I'm hoping someone here can help me with something I've been banging my head over... That part might work, but you're right that it's strange they call out 5KHz and spec 100KHz. Your best bet might be to try it and see how well it performs. Do you really need to do high side switching? If not, you might also look into low switching threshold MOSFETs and drive them directly with logic level signals. It wouldn't get you the over current measurement, but it would be simple and fast. I'd also look through TI, National, and IRF - all of those companies make a lot of power products, so I'd be surprised if they didn't have an option.
|
|
#
?
Jan 18, 2012 00:17
|
|
|
Thank you very much SnowPuppy, I'll probably just give it a shot.... I don't actually know that I need High-side switching. I suppose I could adjust the mosfet in the portion of the circuit I attached to the low side and expect the same results.... I'm just following the diagram on page 7 of this doc: http://www.jaygeeracing.com/files/MOSFET_Technology_Paper.pdf This guy sells his PWM slot car controller for $400! It seems insane to me given how simple it is. To be honest, I don't know why high-side or low-side is preferred for certain situations, I just follow instructions and try to make things that work. When it gets overly technical, my eyes tend to glaze over. The only concern with picking a Low-side alternative would be that I also need a secondary PWM signal driving a similar mosfet across the motor leads (for variable breaking). Nevermind that I need to include a drat-certain switch to make sure neither are on at the same time... but would driving the motor from the low side impact the breaking design? I really appreciate your insight. Thanks again.
|
|
#
?
Jan 18, 2012 02:50
|
|
|
SnoPuppy posted:Step up transformer and AC couple would be one option (probably want to buffer it after the AC coupling though). Transformer was something I hadn't considered at all, I'll look into that if my solid state experiments prove unreliable, thanks! That buffer looks like it's designed for 3.3V operation, and this needs to be 5V (interfacing with a lot of 74LS circuits)... Also not a bad idea but I'd prefer to use standard components like transistors (prefer to avoid surface mount as well). I just ran simulations on a similar design using 2N7000s that seems to work pretty well so I'll try building that before doing anything else.  There's about 90 degrees of phase shift due to slow rise times but as long as it's stable that won't matter. longview fucked around with this message at 10:42 on Jan 18, 2012 |
|
#
?
Jan 18, 2012 10:39
|
|
|
longview posted:I want to amplify a ~500mV signal at 10 MHz up to TTL levels and use this to drive a microcontroller and various other future things, how hard can it be? The transistors are fast enough, but this circuit is pretty crude, and probably won't work properly unless the biasing is tuned very carefully. But if you want a square wave-ish output, then you'll probably need to change the circuit to something like a diff amp or cascode structure, because once BJTs saturate they start slowing down quite a bit. That's probably what's happening to you. Replacing the BJTs with a a small FET (like the 2n7000) will probably help, since they don't suffer that issue so much, but then you'll have to change the biasing (edit: looks like you already tried that, good). If I were you I would adjust the gain down a bit to give more of a sine wave output. It should be enough to drive the logic input properly. Or get a high speed comparator and use that.
|
|
#
?
Jan 18, 2012 15:32
|
|
|
ANIME AKBAR posted:The transistors are fast enough, but this circuit is pretty crude, and probably won't work properly unless the biasing is tuned very carefully. But if you want a square wave-ish output, then you'll probably need to change the circuit to something like a diff amp or cascode structure, because once BJTs saturate they start slowing down quite a bit. That's probably what's happening to you. Replacing the BJTs with a a small FET (like the 2n7000) will probably help, since they don't suffer that issue so much, but then you'll have to change the biasing (edit: looks like you already tried that, good). Yeah I don't have a lot of experience with designing transistor amplifiers, especially at high speeds (second year EE/TV repairman) so a lot of these problems are new to me (never run into saturation before, though I had heard of it). Thanks for the pointers, the only comparator I have here is the LM339 which I suspect might be a bit slow  , I did consider using one but I'd like to keep things through-hole so that makes finding components pretty hard for high speed stuff.I only need to make one of these shaping circuits so I'll just build and tweak it until it works reliably (I think the nmos design will work reasonably well, it simulated well up to about 40 MHz). Now a question about oscillators, I have ordered one of the rubidium frequency standards that EEVBlog did a few videos about, the accuracy is listed as something like 2E-9 per year for drift. If I build a digital clock and run it off the oscillator, how do I use that number to get an estimate for how much the clock will be off after a year? If I multiply the drift figure by the number of seconds in a year I get approx 6 ms, is that a reasonable expectation or am I way off?
|
|
#
?
Jan 18, 2012 17:52
|
|
|
longview posted:Now a question about oscillators, I have ordered one of the rubidium frequency standards that EEVBlog did a few videos about, the accuracy is listed as something like 2E-9 per year for drift. If I build a digital clock and run it off the oscillator, how do I use that number to get an estimate for how much the clock will be off after a year? That number is the amount of frequency drift, usually given in PPM (your is not, it looks to be in Hz/Hz). Multiply your frequency by that number to get the maximum amount that your clock will drift, in Hz, per year. So if it's a 10 MHz clock, it might drift by up to +/-0.02 Hz per year. This works out to a period difference of 2e-16 seconds, or 0.2 fs, per period. Now you can multiple the error of 0.2 fs by the number of clock cycles you expect to have in a year, and come up with an error of about 63 ms.
|
|
#
?
Jan 18, 2012 20:05
|
|
|
SnoPuppy posted:That number is the amount of frequency drift, usually given in PPM (your is not, it looks to be in Hz/Hz). Multiply your frequency by that number to get the maximum amount that your clock will drift, in Hz, per year. Thanks, I thought I might have missed a step in doing the calculation. Still, 63 ms per year offset isn't bad for a digital clock.
|
|
#
?
Jan 18, 2012 20:42
|
|
|
I'm looking at taking a CML output pair and using it to drive two or more ICs. I was able to find information on doing this with LVDS where it suggested having the farthest IC be the one with the actual 100 ohm termination and leave the closer ICs unterminated. Can anyone point me to more information on this subject?
|
|
#
?
Jan 19, 2012 00:02
|
|
|
|
| # ? May 21, 2024 11:46 |
|
|
Quick question, hopefully some of you guys also have experience with these. Designing a control system for some lasers (art display), was debating between CAN and DMX as the network to link all the nodes together. Right now I have a dsPIC33 running each node, and they're all linked together via CAN. Any opinions on the two? Right now I think a DMX<->USB interface would be cheaper than CAN<->USB, but can't think of anything else offhand. Krenzo posted:I'm looking at taking a CML output pair and using it to drive two or more ICs. I was able to find information on doing this with LVDS where it suggested having the farthest IC be the one with the actual 100 ohm termination and leave the closer ICs unterminated. Can anyone point me to more information on this subject? If I understand correctly, you have a single driver and want to drive two separate endpoints with that driver? Sounds like a transmission line/high-speed to me (LVDS mentioned), probably not a good idea.
|
|
#
?
Jan 19, 2012 06:37
|
|