tehsiggi wrote on 2025-10-28, 05:07:

The write-up instructions don't look too bad. I don't know of course the menus of your rigol, but you'll get around them I am su […]
Show full quote

The write-up instructions don't look too bad. I don't know of course the menus of your rigol, but you'll get around them I am sure of that.

I'll see today if I can find any card in my stash that uses a similar schematic, so I can see what normal operation would look like on pins for your reference.
Fingers crossed.

The similar rail resistance gives me hope as well!

Well I configured the Rigol like so (I think)...

But when I measure the SS pin, I don't get any sort of signal so I'm probably not doing something right on the Rigol.

After going through all the menu options and configuring settings, I feel the oscilloscope is like a hobby in itself.

I guess I'm hitting a dead end with my technical abilities here 🤣. A bit disappointed. I have ordered some replacement ISL6525's incase the chip itself is faulty.

Just for verification, connect the probe to the 5V of your PSU. You should see a nice flat line around the 5V mark on your scope.

I don't expect this to be a measurement issue to be honest. Perhaps we're seeing already what's bad.

shevalier wrote on 2025-10-28, 07:14:

According to the ISL6525 block diagram, its oscillator should start immediately.
Check for a sawtooth voltage on pin 1.

I checked the ISL6522 (which is basically the cousin). The RT principle is the same as for both (same formula, same functional description by either strapping it with a resistor to +12V or GND).

I measured (3!) working ISL6522 on a Radeon 9500 Pro, which all have a flat voltage on RT. I don't think it's directly part of the RC oscillating circuit, since it can also be omitted. So I wouldn't bet on that as a verification to see if the ISL runs. The noise I have on the flat line is the same as if I would just connect the probe to ground. Noise, to be expected.

Just to verify: Yes, the ISL is running at around 290kHz and my measurement setup works, here is the PHASE pin of one:

And I also measured that just with the ground wire going to the AGP slot bracket, no sweat.
Don't overengineer what you're trying to do. We should see a voltage going to around 4V on SS. Either flat (working fine) or ramping (if current limit). That's not hard to measure and works easily with even the jankiest setup. Is it the best way to measure? Of course not. Is it sufficient? Yes. Don't sweat it.

tehsiggi wrote on 2025-10-28, 09:59:

which all have a flat voltage on RT. I don't think it's directly part of the RC oscillating circuit, since it can also be omitted.

This depends on the controller's internal circuitry.
When using a current mirror, a straight line will be displayed.
Sometimes artifacts from sawtooth voltage of several mV are visible.

If there is nothing holding ISL from starting (some external power sequencing circuit), then it looks like it has totally failed.

shevalier wrote on 2025-10-28, 10:45:

This depends on the controller's internal circuitry. When using a current mirror, a straight line will be displayed. Sometimes a […]
Show full quote

tehsiggi wrote on 2025-10-28, 09:59:

which all have a flat voltage on RT. I don't think it's directly part of the RC oscillating circuit, since it can also be omitted.

This depends on the controller's internal circuitry.
When using a current mirror, a straight line will be displayed.
Sometimes artifacts from sawtooth voltage of several mV are visible.

If there is nothing holding ISL from starting (some external power sequencing circuit), then it looks like it has totally failed.

The interesting part about the ISL6525 is that it does not have the EN pin, like the ISL6522. There is a potential connection between the AGP PG signal and OCSET, however R1305 (0Ohms) is not populated, so that "over-current trigger" to inhibit operation is not in place.

So we know that the 12V are present to power the ISL (measured through R1262, having 12V on both ends). We don't see any activity on SS, which is odd (either solid 4V or ramping is to be expected (current limit)). So perhaps that ISL just gave up? If the caps are blind, at least some activity should be there.

Just for some sanity checks Zuldan, could you measure resistance of:

• GL1 (Gate, Pin 4 of Q905) against ground (non operation)
• GH1 (Gate, Pin 4 of Q210) against ground (non operation)
• Pin 4 of U831 (OCSET) against ground (non operation)
• R1072 (non operation)
• R1118 (non-operation)
• R1118 both sides during operation - one should be 5V.

Interesting side note: NVVDD_PWRGD is pulled to 5V via 10k (R1118) - if it is kept low, that means that the ISL is actually doing "something", albeit just saying power is not good. Since PGOOD is open drain, there must be somebody pulling it low and that can just be the ISL.

tehsiggi wrote on 2025-10-28, 09:59:

We should see a voltage going to around 4V on SS. Either flat (working fine) or ramping (if current limit). That's not hard to measure and works easily with even the jankiest setup. Is it the best way to measure? Of course not. Is it sufficient? Yes. Don't sweat it.

Confirmed it's 4.4v on SS

tehsiggi wrote on 2025-10-28, 10:56:

The interesting part about the ISL6525 is that it does not have the EN pin, like the ISL6522. There is a potential connection be […]
Show full quote

shevalier wrote on 2025-10-28, 10:45:

This depends on the controller's internal circuitry. When using a current mirror, a straight line will be displayed. Sometimes a […]
Show full quote

tehsiggi wrote on 2025-10-28, 09:59:

which all have a flat voltage on RT. I don't think it's directly part of the RC oscillating circuit, since it can also be omitted.

This depends on the controller's internal circuitry.
When using a current mirror, a straight line will be displayed.
Sometimes artifacts from sawtooth voltage of several mV are visible.

If there is nothing holding ISL from starting (some external power sequencing circuit), then it looks like it has totally failed.

The interesting part about the ISL6525 is that it does not have the EN pin, like the ISL6522. There is a potential connection between the AGP PG signal and OCSET, however R1305 (0Ohms) is not populated, so that "over-current trigger" to inhibit operation is not in place.

So we know that the 12V are present to power the ISL (measured through R1262, having 12V on both ends). We don't see any activity on SS, which is odd (either solid 4V or ramping is to be expected (current limit)). So perhaps that ISL just gave up? If the caps are blind, at least some activity should be there.

Just for some sanity checks Zuldan, could you measure resistance of:

• GL1 (Gate, Pin 4 of Q905) against ground (non operation)
• GH1 (Gate, Pin 4 of Q210) against ground (non operation)
• Pin 4 of U831 (OCSET) against ground (non operation)
• R1072 (non operation)
• R1118 (non-operation)
• R1118 both sides during operation - one should be 5V.

Interesting side note: NVVDD_PWRGD is pulled to 5V via 10k (R1118) - if it is kept low, that means that the ISL is actually doing "something", albeit just saying power is not good. Since PGOOD is open drain, there must be somebody pulling it low and that can just be the ISL.

• GL1 (Gate, Pin 4 of Q905) against ground (non operation) - 10m ohms
• GH1 (Gate, Pin 4 of Q210) against ground (non operation) - 11.5m ohms
• Pin 4 of U831 (OCSET) against ground (non operation) - Cannot get a reading, made sure pin was super clean and tried with another multimeter
• R1072 (non operation) - 1.5k ohms
• R1118 (non-operation) - 10k ohms
• R1118 both sides during operation - one should be 5V. - Confirmed, 1 side is 5v

Thanks for checking.

Having a solid 4.4V on SS would mean the soft-start is done and the ISL6525 is not in current limiting mode.

I assume your measurements that stage m ohms mean megaohms, so the values for GL1 and GL2 appear good. No short to ground between the gates, meaning they should be able to nicely switch.

R1072 being 1.5k is exactly the value of the schematic, which is good.
R1118 having 10kOhms is good as well. The pull-up to 5V is in place and would be working, if the ISL6525 is not pulling PGOOD low.

That leaves us a bit out in the open.
Just to state what we know:

• GPU is cold, PGOOD for NVVDD is LOW, meaning all other rails are NOT enabled and thus have no voltage
• We have 12V on the ISL6525 to supply it and the switching of the mosfets
• There is no dead short on the NVVDD rail, the reading (around 9Ohms) is in the same region as with a regular Ti4200
• Pull-up (R1118) for PGOOD is of right value and has the pull-up voltage 5V available.
• SS of the ISL6525 appears to have 4.4V, meaning the chip should be active and not current limited, but the output is not currect
• OCSET (overcurrent configuration) is set via 10k and confirmed to be correct

This is quite odd.. not sure what's going on here. And not having any activity on the gates (apparently) seems even more odd.

I just drop some thoughts / ideas what could be checked as well:

• Is NVVDD really 0.0V?
• Check if R1075 is 1.02K and connected to NVVDD and FB (pin5) of ISL6525
• Check if R1078 is 2.61K and connected to FB (pin5) of ISL6525 and GND
• If available, you could put known good low ESR caps parallel on the back of the card for the input (3.3V) and output (NVVDD) to see if that gets the ISL going (in case the caps are dead)
• What's the resistance of the 3.3V rail of the card when not installed? Typical resistance climbing because we charge a capacitor?

I also tried something on my Ti4200 earlier on:
Used a current limited lab power supply. Set it 1.5V and current limit to 300mA (0.45W overall). Inject that 1.5V to NVVDD with the card not in the system and see if the PSU goes into current limit. If it does, I'd assume something's wrong on the rail. In my case it pulled around 120mA (iirc) and did not go to current limit, thus ruling out the rail to have a real issue.

Unfortunately, I have no card with a ISL6525 in my collection. Only 6522s and 6529s 😒 so i can't get some reference readings.

tehsiggi wrote on Yesterday, 19:56:

Thanks for checking. […]
Show full quote

Thanks for checking.

Having a solid 4.4V on SS would mean the soft-start is done and the ISL6525 is not in current limiting mode.

I assume your measurements that stage m ohms mean megaohms, so the values for GL1 and GL2 appear good. No short to ground between the gates, meaning they should be able to nicely switch.

R1072 being 1.5k is exactly the value of the schematic, which is good.
R1118 having 10kOhms is good as well. The pull-up to 5V is in place and would be working, if the ISL6525 is not pulling PGOOD low.

That leaves us a bit out in the open.
Just to state what we know:

• GPU is cold, PGOOD for NVVDD is LOW, meaning all other rails are NOT enabled and thus have no voltage
• We have 12V on the ISL6525 to supply it and the switching of the mosfets
• There is no dead short on the NVVDD rail, the reading (around 9Ohms) is in the same region as with a regular Ti4200
• Pull-up (R1118) for PGOOD is of right value and has the pull-up voltage 5V available.
• SS of the ISL6525 appears to have 4.4V, meaning the chip should be active and not current limited, but the output is not currect
• OCSET (overcurrent configuration) is set via 10k and confirmed to be correct

This is quite odd.. not sure what's going on here. And not having any activity on the gates (apparently) seems even more odd.

I just drop some thoughts / ideas what could be checked as well:

• Is NVVDD really 0.0V?
• Check if R1075 is 1.02K and connected to NVVDD and FB (pin5) of ISL6525
• Check if R1078 is 2.61K and connected to FB (pin5) of ISL6525 and GND
• If available, you could put known good low ESR caps parallel on the back of the card for the input (3.3V) and output (NVVDD) to see if that gets the ISL going (in case the caps are dead)
• What's the resistance of the 3.3V rail of the card when not installed? Typical resistance climbing because we charge a capacitor?

I also tried something on my Ti4200 earlier on:
Used a current limited lab power supply. Set it 1.5V and current limit to 300mA (0.45W overall). Inject that 1.5V to NVVDD with the card not in the system and see if the PSU goes into current limit. If it does, I'd assume something's wrong on the rail. In my case it pulled around 120mA (iirc) and did not go to current limit, thus ruling out the rail to have a real issue.

Unfortunately, I have no card with a ISL6525 in my collection. Only 6522s and 6529s 😒 so i can't get some reference readings.

Love the detail in the recap you're doing. The time you're putting into this diagnosis is really appreciated.

Yes megaohms

Here is what I could do before work. Will try the cap later today.

• Check if R1075 is 1.02K (736 ohms) and connected to NVVDD and FB (pin5) of ISL6525 (confirmed)
• Check if R1078 is 2.61K (741 ohms) and connected to FB (pin5) of ISL6525 and GND (confirmed)
• What's the resistance of the 3.3V rail of the card when not installed? Typical resistance climbing because we charge a capacitor? (245 ohms)

• Is NVVDD really 0.0V?
• If available, you could put known good low ESR caps parallel on the back of the card for the input (3.3V) and output (NVVDD) to see if that gets the ISL going (in case the caps are dead)

Installed 16v 470uf (low ESR) caps on C1303, C1301, C1267 and C1263.

Voltages taken with caps installed

zuldan wrote on Yesterday, 22:18:

Love the detail in the recap you're doing. The time you're putting into this diagnosis is really appreciated.

Always welcome. I just love to see those cards getting back to life. Yes, there is no potential use apart from retro computing, but I feel it's better to have it working for that than not working at all.

zuldan wrote on Yesterday, 22:18:

Check if R1075 is 1.02K (736 ohms) and connected to NVVDD and FB (pin5) of ISL6525 (confirmed)
Check if R1078 is 2.61K (741 ohms) and connected to FB (pin5) of ISL6525 and GND (confirmed)
What's the resistance of the 3.3V rail of the card when not installed? Typical resistance climbing because we charge a capacitor? (245 ohms)

That, together with your test with the caps leads me to believe that the caps are not the issue. The in circuit measurement of the resistors is good enough I think. Especially since you verified that they are connected with the ISL6525 at the right place.

Thanks for measuring again. Perhaps it's really ISL6525 that's a goner. I've started an inquiry at my trusted vendor to see if they have some in stock as well.
I plan on building some "evaluation board" for ISL6522, ISL6525, ISL6529, SC1075, HIP6012 and some IR parts as well, so that I can "simulated" failure modes and get reference measurements of that. That would be a good collection I assume.

tehsiggi wrote on Today, 06:53:

Always welcome. I just love to see those cards getting back to life. Yes, there is no potential use apart from retro computing, […]
Show full quote

zuldan wrote on Yesterday, 22:18:

Love the detail in the recap you're doing. The time you're putting into this diagnosis is really appreciated.

Always welcome. I just love to see those cards getting back to life. Yes, there is no potential use apart from retro computing, but I feel it's better to have it working for that than not working at all.

zuldan wrote on Yesterday, 22:18:

Check if R1075 is 1.02K (736 ohms) and connected to NVVDD and FB (pin5) of ISL6525 (confirmed)
Check if R1078 is 2.61K (741 ohms) and connected to FB (pin5) of ISL6525 and GND (confirmed)
What's the resistance of the 3.3V rail of the card when not installed? Typical resistance climbing because we charge a capacitor? (245 ohms)

That, together with your test with the caps leads me to believe that the caps are not the issue. The in circuit measurement of the resistors is good enough I think. Especially since you verified that they are connected with the ISL6525 at the right place.

Thanks for measuring again. Perhaps it's really ISL6525 that's a goner. I've started an inquiry at my trusted vendor to see if they have some in stock as well.
I plan on building some "evaluation board" for ISL6522, ISL6525, ISL6529, SC1075, HIP6012 and some IR parts as well, so that I can "simulated" failure modes and get reference measurements of that. That would be a good collection I assume.

Happy to do any testing you need. Learning heaps! Got some ISL6525's arriving in a week or so, I'll swap the old one out and see what happens.

isl chip most likely is ok.

check with oscilloscrope if there is any initial "attepmpts" to start mosfets. there must be 3 pulses and isl will go to protection mode.

from my expirience, a caps in feedback/ comp chain is very important. failed caps will cause isl not functioning
also check voltage on OCSET pin.

Previous measurement attempts on GL, GH showed no activity. However they didn't also show anything on SS, which measured with a DMM shows 4.4V now, so perhaps worth a shot measuring again.

To check if the probe setup works, go measure one of the rails, like 3.3V or 5V first. That should give you a solid flat line at the voltage of that rail on the scope. If that works, go ahead with trying to measure on GL and GH.