PDA

View Full Version : Speed of Optos on BoBs



boldford
20-08-2012, 06:39 PM
What sort of clock rate can I sensibly get through a BoB fitted with EL817s? If too slow, would it be sensible to replace the STEP channels with something like a TLP521 leaving the DIR and ENABLE as is?

Jonathan
20-08-2012, 06:44 PM
The opto-isolators are very unlikely to be the limiting factor on a breakout-board as most parallel ports struggle stably to output more than around 35kHz.

irving2008
20-08-2012, 07:24 PM
I agree with Jonathan, though it does depend on how they've been used. The EL817 are good to 50kHz+, if driven hard enough, which is better than the parallel port in typical use. The TLP521 isn't that much faster (they both have a 'typical' rise/fall time of 4uS with a 100ohm collector load).

JAZZCNC
20-08-2012, 07:55 PM
The opto-isolators are very unlikely to be the limiting factor on a breakout-board as most parallel ports struggle stably to output more than around 35kHz.

Actually you'd be surprised how many problems come from slow opto's and I've seen plenty of people on the Yahoo forum have all sorts of troubles with cheap Boards using slow opto's.!!

Must admit Thou personally I've never had any problems.!

John S
20-08-2012, 08:09 PM
I have, breakout boards have been the bane of my life.
Old desktop computer on the router died over the weekend, grabbed a 'new' Pc from up in the hayloft, got a stack up there all bagged and clean genuine install of XP on it.

Got it all setup, nice HP Compact and wouldn't work the router, router was working fine on Friday. Threw it to one side and grabbed a Trusty Dell SX270,used loads of these over the years, absolutely bomb proof.

Still dead.

Go looking in the router box, no charge pump light. Check the output from the Dell with a scope, 4.5 volts dead steady at 17kh.

Short enable down to ground to bypass the charge pump and voilà, all working. only problem is if the router is on whilst it's booting up it jumps all over and the spindle starts and stops on it's own 3 times.

This is about the sixth time I have had different breakout board with charge pump problems. So basically 4 hours wasted because someone who is supposed to know better when they are designing can't be arsed on checking properly.

When you query problems all you get is "Well I have sold loads of these and not had problems "

In the past I have sent 5 boards back that didn't work correctly and they got returned with link wires and extra diodes soldered on over the back and an input stolen so it could amplify the charge pump.
If they are that good why the mods ?

irving2008
20-08-2012, 08:46 PM
Because there's a difference between designing for mass production and designing to a specification and a price... I can design and build a one-off that'll be perfect...but it'll be a one-off... if I want to design it so it can be assembled and tested by a team of trained monkeys and is bullet proof to its environment it needs a lot more rigour. I cut my teeth on troubleshooting manufacturing problems at GEC-Marconi on head-up displays and the like... it was rare to get a batch that actually had 100% good ones, in fact there were on occasions batches that were close to 100% failures... With logic circuits its obviously easier to get it right (though the number of SystemX telephone exchange cards I've seen with spaghetti all over the back suggests otherwise.) Designing analogue circuits seems to be a dying art these days. I had a couple of electronics students on work experience not that long ago who hadn't done anything of note with op-amps or even discrete transistors :(

Jonathan
20-08-2012, 09:15 PM
I had a couple of electronics students on work experience not that long ago who hadn't done anything of note with op-amps or even discrete transistors :(


Sadly that doesn't surprise me any more. In the first week on my course (Electrical/Electronic eng.) for a bit of fun they gave us a little car thing to put together which involved soldering components on to a PCB. The vast majority had never even used a soldering iron. I can only think of one person, besides myself, out of about 100 on my course in my year who does any electronics at home which really showed when it came to actually making something. That was quite some time as there's not much practical work... Also noticed that the people who had done A-level electronics weren't much better off and didn't seem to know things I'd consider basic, so I'm glad my school didn't offer it since, as it was the subject I was most interested in I would probably have naïvely taken it in instead of further maths.

Anyway, I reckon I'll make a breakout board soon...bought the components ages ago but keep putting it off. How hard can it be!

irving2008
20-08-2012, 09:23 PM
Not hard at all :) tho not necessarily cheaper than buying one...

John S
20-08-2012, 09:43 PM
Because there's a difference between designing for mass production and designing to a specification and a price...

Sorry doesn't wash with me, although I hate the term in that it's so misused there is a "Fit for Purpose" clause in that it SHOULD be fit for purpose.

OK anyone can drop a bollock but to keep dropping back on the old excise of "We have sold loads " when you know full well they had had to modify the design three times for faults that I have found and they are still not right.

As son as they hit the marketplace they should work correctly without having to resort to soldering jump wires all over and extra components.

irving2008
20-08-2012, 10:03 PM
I totally agree... my point was that designing something as a hobbyist then taking it to market is a different ball game to an experienced designer designing for mass production and manufacturing tolerances etc.

One of the key differences is testing. As a 'hobby designer' I'd consider ten or so prototypes tested by friends (or worse, the first dozen or so unlucky punters - were you one of those I wonder) as adequate. As a manufacturer I'd want to build, and evaluate to destruction if need be, several hundred at least before letting them loose...

boldford
21-08-2012, 09:10 AM
The particular BoB I'm playing with doesn't seem to be defective as such. (It's almost too simple not to work). I'm commenting on the difference between I/P and O/P 'scope wave forms. To test it I'm feeding it with a 555 astable on a breadboard producing close to a 50/50 square wave. I'm varying the frequency with a simple change of C. What comes out is 'orrible!
What I find odd is the channels outbound from the PC pass through 74HC14s before they get to the optos. No such buffering/inverting on the inbound direction.

irving2008
21-08-2012, 09:24 AM
not odd at all... those optos need driving and the parallel port on some PCs isnt up to the job. The question I would ask is where are the LEDs and the 74HC14 being powered from?

Also what are the pull up resistors either side of the opto?

boldford
21-08-2012, 09:17 PM
... those optos need driving and the parallel port on some PCs isnt up to the job. Of course! I was forgetting some crappy PCs only output 3.3V
The question I would ask is where are the LEDs and the 74HC14 being powered from?

Also what are the pull up resistors either side of the opto?At the present they are being powered from the 5V rail of a smorgasbord bench PSU. I'll confirm the R values tomorrow.

boldford
22-08-2012, 10:28 PM
.........Also what are the pull up resistors either side of the opto?The channels outbound from the PC have 1K pull-up resistors.

irving2008
22-08-2012, 11:29 PM
on which side of the opto? R1 or R2 below? I'm guessing you meant R2?

To get any reasonable slew rate R2 needs to be closer to 100ohm not 1k.

6641

boldford
23-08-2012, 08:20 AM
on which side of the opto? R1 or R2 below? I'm guessing you meant R2?

To get any reasonable slew rate R2 needs to be closer to 100ohm not 1k.

6641
I did mean R2. I'm sorry for not being clearer. Many thanks for the information. I'll try with a 120R in parallel using ugly construction and report my findings. If that works I might try find out if I can still solder SMT devices.

What would you recommend for the I/P side? I.e R1.

irving2008
23-08-2012, 09:03 AM
I did mean R2. I'm sorry for not being clearer. Many thanks for the information. I'll try with a 120R in parallel using ugly construction and report my findings. If that works I might try find out if I can still solder SMT devices.

What would you recommend for the I/P side? I.e R1.
Whats in there already? and what version on the opto is it? and whats the isolated supply Vcc?

The reason being, when you reduce R2 to a lower resistance, you increase the collector current therefore you'll need to increase the drive current (If) to the LED. Current I'd guess R1 is 1K, which gives a forward current of (5 - 1.4)/1000 = 3.6mA. with the lowest spec'd opto the transfer ratio (CTR) is 50 which will give a collector current of 130mA but obviously this is limited by R2 to (assuming Vcc = 5v and R2 = 100R) (5 - 0.4)/100 = 46mA so R1 won't need to be changed. However you may find slew rate improves if you reduce it to increase If to, say, 10mA, which needs R1 = (5 - 1.4)/.01 = 360R, so if R1 = 1k try a 560R in parallel.

The fact that these are both 1k may well be an example of offsetting cost of manufacture v breadth of applicability. by making them all 1k it reduces inventory, handling and piece cost, but means there are performance limits that are probably OK for 95+% of cases... you haven't actually said if this is giving you issues, you only asked about the likely maximum speed.

boldford
23-08-2012, 09:34 AM
Whats in there already? and what version on the opto is it? and whats the isolated supply Vcc?

The reason being, when you reduce R2 to a lower resistance, you increase the collector current therefore you'll need to increase the drive current (If) to the LED. Current I'd guess R1 is 1K, which gives a forward current of (5 - 1.4)/1000 = 3.6mA. with the lowest spec'd opto the transfer ratio (CTR) is 50 which will give a collector current of 130mA but obviously this is limited by R2 to (assuming Vcc = 5v and R2 = 100R) (5 - 0.4)/100 = 46mA so R1 won't need to be changed. However you may find slew rate improves if you reduce it to increase If to, say, 10mA, which needs R1 = (5 - 1.4)/.01 = 360R, so if R1 = 1k try a 560R in parallel.

The fact that these are both 1k may well be an example of offsetting cost of manufacture v breadth of applicability. by making them all 1k it reduces inventory, handling and piece cost, but means there are performance limits that are probably OK for 95+% of cases... you haven't actually said if this is giving you issues, you only asked about the likely maximum speed.

Once again, many thanks for your information and explanation. I will try and check the I/P R values later today.
At present I'm not having issues with this kit as I'm presently trying to get acceptable (in my book "best possible") performance out of this equipment.

This takes us back to my original question where I sought the collective wisdom about the maximum clock frequency I could reasonably expect with the active devices fitted.

boldford
23-08-2012, 05:20 PM
Whats in there already? and what version on the opto is it? and whats the isolated supply Vcc?

The reason being, when you reduce R2 to a lower resistance, you increase the collector current therefore you'll need to increase the drive current (If) to the LED. Current I'd guess R1 is 1K, which gives a forward current of (5 - 1.4)/1000 = 3.6mA. with the lowest spec'd opto the transfer ratio (CTR) is 50 which will give a collector current of 130mA but obviously this is limited by R2 to (assuming Vcc = 5v and R2 = 100R) (5 - 0.4)/100 = 46mA so R1 won't need to be changed. However you may find slew rate improves if you reduce it to increase If to, say, 10mA, which needs R1 = (5 - 1.4)/.01 = 360R, so if R1 = 1k try a 560R in parallel.

The fact that these are both 1k may well be an example of offsetting cost of manufacture v breadth of applicability. by making them all 1k it reduces inventory, handling and piece cost, but means there are performance limits that are probably OK for 95+% of cases... you haven't actually said if this is giving you issues, you only asked about the likely maximum speed.
Hi irvine2008
I overlooked to reply to your question previously about the isolated supply Vcc. It is 5v.
I've now checked the value of R1 on both the inbound at outbound channels. I've found these to both be 330R.
The 74HC14 inputs have 10K pull-ups to stop them floating.

Using ugly construction, I've just tried paralleling the 1K resistor with various values between 100R and 220R. The O/P waveform shape is much improved although reduced in magnitude to circa 3.5v pk-pk.

boldford
24-08-2012, 08:40 PM
6656This is the BoB I've been referring to. (I said it was a simple design).

The plan now is to either obtain some 150R 1206 SMT resistors to mount on top of the existing 1K ones or simply replace them with 120R. With this mod applied tests indicate the O/P voltage swing will now only be approximately 3.5V. This suggests the next step will require a veroboard hack to introduce 74LS14s on all channels ensuring the I/P and O/P isolation isn't compromised.

irving2008
24-08-2012, 09:45 PM
Personally, unless its causing you problems I'd leave it alone.... if you need more speed you might want to consider a different card. When you say a 3.5v swing, between what and what? 3.5v is a good logic one level...

boldford
25-08-2012, 09:13 AM
. . . . . When you say a 3.5v swing, between what and what? 3.5v is a good logic one level...Many thanks for sticking with this one. My 'scope is years out of cal' but I'd say the swing on the modified channel is roughly between +1.6v and +4.8v which suggests the ON resistance of the opto is the limitation.

irving2008
25-08-2012, 09:59 AM
Many thanks for sticking with this one. My 'scope is years out of cal' but I'd say the swing on the modified channel is roughly between +1.6v and +4.8v which suggests the ON resistance of the opto is the limitation.

Ah, so the transistor is not saturating. with 100R load, the collector current therefore is (5 - 1.6)/100 = 34mA, whereas it needs to be (5-0.5)/100 = 45mA, so the LED needs to be driven harder as previously discussed. However the 74HC14 clamps at 20mA, so there may be a limiting factor here... try reducing R1 from 1k to around 450R or so, theres no point in going lower... if that doesnt improve things then its not going to get better without significant rework

boldford
25-08-2012, 06:30 PM
Ah, so the transistor is not saturating. with 100R load, the collector current therefore is (5 - 1.6)/100 = 34mA, whereas it needs to be (5-0.5)/100 = 45mA, so the LED needs to be driven harder as previously discussed. However the 74HC14 clamps at 20mA, so there may be a limiting factor here... try reducing R1 from 1k to around 450R or so, theres no point in going lower... if that doesnt improve things then its not going to get better without significant reworkAs stated earlier R1 is 330R, so it looks like I need to have a think if I'm going to use this BoB in the way I had intended.

boldford
03-09-2012, 11:47 PM
A week on and spare time now permits the process of clean-up the O/Ps with a few 74HC14s, It ain't gonna be petty but at least the waveforms, both inbound and outbound, are gonna be sharp. With a bit of care I will retain the complete opto-isolation on the BoB. Elimination of the now un-necessary optos at the front end of the single axis TB6560 will follow blow by blow.
I think it's fair to say "although any CNC electronics system is a series of steps it needs to be integrated as whole to get the performance expected"
The overall plan continues to be to derive the PC side supply from a spare HDD power connector within the PC. The O/P side of the BoB will come from the 5v rail of the homebrew smorgasbord driver PSU. Or have I got it totally wrong?