Are there any features that could be clobbered by the change, directly or indirectly?
I'm thinking about any potential for the new chips arriving and needing some other fix because of this mod?
Maybe not such a risk if you are considering a small test order of the new chip, vs going for the full 10K with hope in your heart!? (however logically fine that may seem, it also seems a bit scary...)??
We have a couple remaining internal-proto blank Eval pcbs should another globtop feature test be required. So that part is all good.
I still think it would be better to leave as-is, and recommend tying pin pairs together to get 1-bit extra resolution, or ground the pin-pair. I don't think this will cause any problems for users.
But our awesome SCOPE mode would be all compromised, since the streamer captures four channels on four contiguous pins. Fixing this also makes all pins into independent ADC's, like they ought to be.
That could be a feature for the next version or future errata. But yeah, it is AWESOME that scope feature. Will it drive sales of the first release though? I have a hunch about that, but no real evidence!
No one wants to reschedule the production order. There would probably a penalty fee too. It's either insert the change on the fly or not do it until a later revision.
No one wants to reschedule the production order. There would probably a penalty fee too. It's either insert the change on the fly or not do it until a later revision.
Right. They are going to charge us at least $5,000 for the mask change. This has no impact on the wafer build time. They won't even need this mask for another 8 weeks, or so, assuming they've already started building the bottom layers onto blank wafers.
Are there any features that could be clobbered by the change, directly or indirectly?
I'm thinking about any potential for the new chips arriving and needing some other fix because of this mod?
Maybe not such a risk if you are considering a small test order of the new chip, vs going for the full 10K with hope in your heart!? (however logically fine that may seem, it also seems a bit scary...)??
We have a couple remaining internal-proto blank Eval pcbs should another globtop feature test be required. So that part is all good.
This eliminates the PinB selection (%100010_OHHHLLL) in the ADC modes. If you select PinB in the next silicon, you will be selecting a no-connect, which will allow you to measure the ADC's balance point.
None of the ROM code uses this ADC sub-mode. The ATE test program did, but I've already rewritten it and tested it carefully. So, I see no possibility of a bad surprise.
The point was that if you insert this change on the fly, then 10k chips arrive with some other issue caused by an as-yet unknown side effect of the change, then how to mitigate that?
You seem pretty certain Chip, and the matter got clarified. Thanks.
That letter/appeal to On Semi/Wendy was one of the best letters I've ever read. Chip explained the situation well and made it as easy (painless) as possibly for On Semi to make the change. He tactfully addressed the possible concerns that they might raise. It was kind of a pre-emptive "attack" designed to head off a possible unwarranted negative initial reaction that could snowball and unnecessarily restrict things. Beautiful! So, if they are going to reject his request (which hopefully they won't), it will be due to substantive reasons that even Chip will likely agree with. And while Chip showed great respect for their experience and policies, he still came across with competence and confidence. It's not easy to write such a great letter (it could be a case study for how to make requests). The care that he took to compose it in such a strategic way shows how important this is to him. I say that we support him. Sure, we can point out the risks and ask questions, but in the end, we should show our support unless we can provide a very good reason not to. We didn't get this far by being negative.
That letter/appeal to On Semi/Wendy was one of the best letters I've ever read. Chip explained the situation well and made it as easy (painless) as possibly for On Semi to make the change. He tactfully addressed the possible concerns that they might raise. It was kind of a pre-emptive "attack" designed to head off a possible unwarranted negative initial reaction that could snowball and unnecessarily restrict things. Beautiful! So, if they are going to reject his request (which hopefully they won't), it will be due to substantive reasons that even Chip will likely agree with. And while Chip showed great respect for their experience and policies, he still came across with competence and confidence. It's not easy to write such a great letter (it could be a case study for how to make requests). The care that he took to compose it in such a strategic way shows how important this is to him. I say that we support him. Sure, we can point out the risks and ask questions, but in the end, we should show our support unless we can provide a very good reason not to. We didn't get this far by being negative.
Yes, the situation is a little delicate because we just signed their "Prototype Approval Form", which means they could dig in their heels and say, "Nope! Too late for any more changes." I wanted them to understand that the risk is very low, the reward is high, and I've already taken care of the testing program, since that's been a sensitive topic, all along.
So far, the toughest guy at ON said this in the group email circulating about the request: "Based on Chip’s description and the image of the changes needed, it will be an easy ECO." He was the one I was worried most about. Now if they will all agree and assess a reasonable price, everything will be great.
Chip, I trust your risk/benefit analysis is correct. And I see your point about the scope-mode with and without the change.
I can't believe I'm saying this, but.... I support your decision to make the change.
But, if it turns out to be more that a simple fix though, I say "Just let it be.".
I guess the absolute worst case scenario would be the ADC operating worse than currently. Nothing else is likely to result from the fix.
While it's nice to have everything consistent, I wonder if the pin pairs 56-57, 58-59, 60-61, 62-63 were left unmodified? Why, because they could be used to get the extra ADC bit of resolution. Other than the 56-57 pair, the others are unlikely to be used for ADC since they are the Serial and FLASH/SD pins. So if anyone really needed the extra bit of resolution, they could redefine how those 8 pins were used to get access to the extra resolution. Just a thought???
Could they not just tie the pins together externally?
Not if PinB is cut
It doesn't matter. You can always tie the two pins together and get an extra bit of resolution by running both of their ADCs. This M2 trace cut just means that you can run them cleanly, independently of each other. That's all it does. It's like inhibiting someone else's hair from getting in your soup.
Could they not just tie the pins together externally?
Not if PinB is cut
It doesn't matter. You can always tie the two pins together and get an extra bit of resolution by running both of their ADCs. This M2 trace cut just means that you can run them cleanly, independently of each other. That's all it does. It's like inhibiting someone else's hair from getting in your soup.
Great
Sorry I missed the point of using adjacent ADCs to get extra precision.
Users of the P2. Chip answered, "yes" below. Non-issue, if you ask me. Anyone desiring that extra bit can make the connection outside the chip and are good to go.
Looking at it this way, I favor the solution Chip is proposing.
Well - I just supported Chips plan by buying $3xx worth of stuff from Parallax I might use soon, or as soon as I have time, or later.
This is the final run (hopefully) for production. The possibility of side effects seem very low, the gain is cleaner ADC, one of the main upgrades from P1 to P2, we have ADC/DAC now.
Yes, I think we need to make the analog work as well as possible. It would be a shame to leave things compromised when the fix is so simple.
And didn't you say that changed mode is basically turned into a reference measure now? Rather than talk to the other pin, it's just not connected, meaning it's a sort of baseline?
I think it's worth it to get the analog optimal. IMHO, people will actually use it in this way too. As projects get done, some good tools will fall out of that, and already have. For some, maybe just getting into this stuff, or who maybe don't need pro grade, just learning grade, having the analog run optimally will be pretty high value. Someone can get a couple P2 chips, setup a few circuits and or connectors, and basically have a nice, quick and simple signal, generate, test, measure system. That's nice.
That, and let's be honest! If we were Chip? Knowing it's just a quick trace delete? Would drive me nuts! Seriously.
Stupid question,
Can Parallax for test, blew one connection on an actual P2 ?
I guess, put high power on pins to blew up one mask connection, just for test ?
I know I am from old school, and did such funny smoke tricks.
This is *definitely* one of the things I wanted to do with my prop2.
I've been holding off on buying a new precision scope for a while, because getting anything over 25Mhz/50Mhz cheap ($300?) either involves a firmware patch or a chinese board with overclocked ADCs.
Having a 4 channel scope under a hundred bucks at my disposal anywhere I've got a VGA or HDMI monitor, even if it's not full of features, is still worth it to someone like me that relies on a megohm resistor poked into a 12 bit ADC off a STM32 as a poor man's 2Mhz oscilloscope.
The cleaner and more useful that signal is, the less of a reason I have to spend on precision gear like a rigol for a hobby, when I'd be fine if I still had my old HP 20Mhz analog scope. May she rest her tired old CRT in peace.
I support the cut trace. If I decide I really want to eke out another bit of precision, I'm *sure* I can find room for a 0 ohm link on an acquisition PCB. But I'd rather the -70dB crosstalk figure than the -24dB crosstalk figure.
I don't think I should have to mention this -- but since there was a previous mixup with the testing program; perhaps the old and new copies should be content hashed and you should verify which test program hash was run in the testing report, so it is always clear exactly which testing procedure was being implemented at any given time.
This is *definitely* one of the things I wanted to do with my prop2.
I've been holding off on buying a new precision scope for a while, because getting anything over 25Mhz/50Mhz cheap ($300?) either involves a firmware patch or a chinese board with overclocked ADCs.
.
The inbuilt ACs will still have bandwidth limits, but the sampling rates look good on the plots above.
It may be that some external ADC sweet-spot can be found, that allows higher megasample rates ?
Often it's nice to include a Multimeter & Frequency Counter into such Oscilloscope boards, the Frequency counter P2 already nails nicely
For multimeter I see a new ADS131M04 from TI that has 4 channels simultaneous sampling 24b, includes a typ 0.1% 1.2V 7.5ppm/°C reference $3.50/100
ADC's start from simple ones like ADC08060C, with 1 chan, 8 bit, 60MSps, $2.86/100, and go up quite rapidly...
MAX19506ETM is 8b 100M 2 chan $6.49/100
MAX19516ETM is 10b 100M 2 chan $11.88/100
then, things start to thin out, as P2 needs Digital supply of 3v3
LTC2281CUP 10b 125M 2 Chan $23.08/100
ADS62P15IRGCT 11b 125M 2 chan $31.21/100
LTC2283CU 12b 125M 2 Chan $52.90/100
LTC2285CU 14b 125M 2 Chan $141/100
Comments
I'm thinking about any potential for the new chips arriving and needing some other fix because of this mod?
Maybe not such a risk if you are considering a small test order of the new chip, vs going for the full 10K with hope in your heart!? (however logically fine that may seem, it also seems a bit scary...)??
We have a couple remaining internal-proto blank Eval pcbs should another globtop feature test be required. So that part is all good.
That could be a feature for the next version or future errata. But yeah, it is AWESOME that scope feature. Will it drive sales of the first release though? I have a hunch about that, but no real evidence!
Right. They are going to charge us at least $5,000 for the mask change. This has no impact on the wafer build time. They won't even need this mask for another 8 weeks, or so, assuming they've already started building the bottom layers onto blank wafers.
This eliminates the PinB selection (%100010_OHHHLLL) in the ADC modes. If you select PinB in the next silicon, you will be selecting a no-connect, which will allow you to measure the ADC's balance point.
None of the ROM code uses this ADC sub-mode. The ATE test program did, but I've already rewritten it and tested it carefully. So, I see no possibility of a bad surprise.
You seem pretty certain Chip, and the matter got clarified. Thanks.
Yes, the situation is a little delicate because we just signed their "Prototype Approval Form", which means they could dig in their heels and say, "Nope! Too late for any more changes." I wanted them to understand that the risk is very low, the reward is high, and I've already taken care of the testing program, since that's been a sensitive topic, all along.
So far, the toughest guy at ON said this in the group email circulating about the request: "Based on Chip’s description and the image of the changes needed, it will be an easy ECO." He was the one I was worried most about. Now if they will all agree and assess a reasonable price, everything will be great.
I can't believe I'm saying this, but.... I support your decision to make the change.
But, if it turns out to be more that a simple fix though, I say "Just let it be.".
Bean
I presume you don't expect the stub trace to act as an aerial?
Otherwise, go for it. I am trying to make time to look at scope mode, plus other things
No, it's not going to do anything blowing in the local breeze. It will be good to get it disconnected from the other pin.
While it's nice to have everything consistent, I wonder if the pin pairs 56-57, 58-59, 60-61, 62-63 were left unmodified? Why, because they could be used to get the extra ADC bit of resolution. Other than the 56-57 pair, the others are unlikely to be used for ADC since they are the Serial and FLASH/SD pins. So if anyone really needed the extra bit of resolution, they could redefine how those 8 pins were used to get access to the extra resolution. Just a thought???
You're right though you'll get less crosstalk than what Chip is proposing if you do tie Pin B to Pin A. Kind of like a driven ground
Not if PinB is cut
It doesn't matter. You can always tie the two pins together and get an extra bit of resolution by running both of their ADCs. This M2 trace cut just means that you can run them cleanly, independently of each other. That's all it does. It's like inhibiting someone else's hair from getting in your soup.
Great
Sorry I missed the point of using adjacent ADCs to get extra precision.
Users of the P2. Chip answered, "yes" below. Non-issue, if you ask me. Anyone desiring that extra bit can make the connection outside the chip and are good to go.
Looking at it this way, I favor the solution Chip is proposing.
This is the final run (hopefully) for production. The possibility of side effects seem very low, the gain is cleaner ADC, one of the main upgrades from P1 to P2, we have ADC/DAC now.
I think it is the right decision.
Mike
Keep having fun. I am
And didn't you say that changed mode is basically turned into a reference measure now? Rather than talk to the other pin, it's just not connected, meaning it's a sort of baseline?
I think it's worth it to get the analog optimal. IMHO, people will actually use it in this way too. As projects get done, some good tools will fall out of that, and already have. For some, maybe just getting into this stuff, or who maybe don't need pro grade, just learning grade, having the analog run optimally will be pretty high value. Someone can get a couple P2 chips, setup a few circuits and or connectors, and basically have a nice, quick and simple signal, generate, test, measure system. That's nice.
That, and let's be honest! If we were Chip? Knowing it's just a quick trace delete? Would drive me nuts! Seriously.
Can Parallax for test, blew one connection on an actual P2 ?
I guess, put high power on pins to blew up one mask connection, just for test ?
I know I am from old school, and did such funny smoke tricks.
I've been holding off on buying a new precision scope for a while, because getting anything over 25Mhz/50Mhz cheap ($300?) either involves a firmware patch or a chinese board with overclocked ADCs.
Having a 4 channel scope under a hundred bucks at my disposal anywhere I've got a VGA or HDMI monitor, even if it's not full of features, is still worth it to someone like me that relies on a megohm resistor poked into a 12 bit ADC off a STM32 as a poor man's 2Mhz oscilloscope.
The cleaner and more useful that signal is, the less of a reason I have to spend on precision gear like a rigol for a hobby, when I'd be fine if I still had my old HP 20Mhz analog scope. May she rest her tired old CRT in peace.
I support the cut trace. If I decide I really want to eke out another bit of precision, I'm *sure* I can find room for a 0 ohm link on an acquisition PCB. But I'd rather the -70dB crosstalk figure than the -24dB crosstalk figure.
I don't think I should have to mention this -- but since there was a previous mixup with the testing program; perhaps the old and new copies should be content hashed and you should verify which test program hash was run in the testing report, so it is always clear exactly which testing procedure was being implemented at any given time.
The inbuilt ACs will still have bandwidth limits, but the sampling rates look good on the plots above.
It may be that some external ADC sweet-spot can be found, that allows higher megasample rates ?
Often it's nice to include a Multimeter & Frequency Counter into such Oscilloscope boards, the Frequency counter P2 already nails nicely
For multimeter I see a new ADS131M04 from TI that has 4 channels simultaneous sampling 24b, includes a typ 0.1% 1.2V 7.5ppm/°C reference $3.50/100
ADC's start from simple ones like ADC08060C, with 1 chan, 8 bit, 60MSps, $2.86/100, and go up quite rapidly...
MAX19506ETM is 8b 100M 2 chan $6.49/100
MAX19516ETM is 10b 100M 2 chan $11.88/100
then, things start to thin out, as P2 needs Digital supply of 3v3
LTC2281CUP 10b 125M 2 Chan $23.08/100
ADS62P15IRGCT 11b 125M 2 chan $31.21/100
LTC2283CU 12b 125M 2 Chan $52.90/100
LTC2285CU 14b 125M 2 Chan $141/100
they claim :
Pin Compatible Family
125Msps: LTC2283 (12-Bit), LTC2285 (14-Bit)
105Msps: LTC2282 (12-Bit), LTC2284 (14-Bit)
80Msps: LTC2294 (12-Bit), LTC2299 (14-Bit)
65Msps: LTC2293 (12-Bit), LTC2298 (14-Bit)
40Msps: LTC2292 (12-Bit), LTC2297 (14-Bit)
Looks like RedPitaya used LTC2145CUP-14, (125Msps 14b ~$60/100+) but that is 1.8V-only operation.
Addit: seems 3v3 digital is very much trailing edge, so maybe a scattering of 74AVCH8T245 (35~40c/3k) , which claim 380 Mbit/s is cost effective ?