U-blox 7 and PCIe RS232 card....new GPS+PPS project

Hi all,

New project, I bought me a mini U-Blox GPS+PPS…

1: The receiver, U-Blox 7M…about 7 euro/usd :slight_smile:


2: The RS-232 card, StarTech dual port.


Looks like this:

Why the Startech card? Well it can do RS-232 in all variants, but it has also jumpers to supply 0V/5V/12V on DB9 pin 9.

This means you can connect any GPS that needs power to work and you do not need an extra powersupply/usb-plug to give current.

So you only need 5 wires to the RS-232 connector and all is provided.

I have not tested this yet, but it’s in the planning to do soon, at least for testing.

Total costs, about 36 euro/usd…

You can connect an outdoor antenna, but then the Garmin with 5m wire is a better option, however, with this RS-232 card you should be able to use a single connector and power it.

I want to see what PCIe RS-232 does for timing, does it differ from a normal port?


1 Like

Nice find.

I can’t seem to find any specs on the current capability on pin 9 or if it is fuse protected.

If it is tied directly to the 5/12 volt bus, a short might damage the card, motherboard, or power supply.

Great receiver, and amazing price. How much was the shipping and tax?

ublox makes timing receivers that know they’re being used in a fixed location, so can average out much of the error, and have a precision oscillator to provide much less jitter and better holdover. I picked up a few 8T pulls from cellular base stations with a $25 offer from eBay: U-BLOX ublox LEA-M8T-0-10 HUAWEI GPS Module New | eBay. $3 shipping and 6% Maryland sales tax. Given the value of your time, perhaps the higher price is justified.

Pretty slick. Personally, I’d rather take 5V from USB and be future-proofed rather than tied to PCIe and a particular serial card. Also your solution could be used by others who may not have a PCIe slot.

That’s a real apples-and-oranges comparison in my opinion. The Garmin receiver is a 20-year-old design, ublox 7 is 10 years old, Garmin is GPS-only vs GPS+Galileo+GLONASS+QZSS, 12 channel vs. 56 channels, and I could go on and on about the improvements during those 10 years.

What is a “normal port”?


Paid 10 Euro in total, no taxes. As it’s below 20 euro in total, then it’s taxfree.

You can do that as well, nobody stops you from doing that.

A normal port is on the motherboard and part of the south-bridge of the chip-set.
However, Intel CPU’s (don’t know about AMD) have direct PCIe channels, not passing the south-bridge.
So in theory an PCIe RS-232 card has a shorter route to the CPU and thus faster signal-processing.

In theory; so I’m going to test that. Need to take the server down for that, so have to find a good moment, is a bit of work.

For timekeeping that shouldn’t matter, as you only need 1 sat for time and PPS, so if you see 5 or 50 sats, it shouldn’t matter…again in theory :slight_smile:

It’s directly tied to the DB-9 connector and the jumpers decide if you use the 12V or 5V line or 0V…at 0V it should behave as normal.

If something happens to the voltage, the powersupply will burn the lines on the PCIe card doing that, but it will not pass any chips or any connection with PCIe.

I have checked the lines on the PCB.

It will not damage the PSU or motherboard, sure of that.

That’s what a dedicated timing receiver like the M8T puts into practice. Non-timing receivers typically can’t do that, that’s one of the aspects making dedicated timing receivers special.

1 Like

I venture to hypothesize that the crucial part are not the physical signal travel/processing times, but rather the interrupt handling latency.

1 Like

They are correlated.

1 Like

Not sure. The time from when a hardware event occurs (DCD signal asserted) until the interrupt is generated may be smaller, but the time for the software to pick up and handle the interrupt is going to be the same. And that time is likely to be somewhat larger, and likely will have higher jitter so as to mostly mask the benefit of faster handling on the hardware side. So I’ll expect that a careful look will be needed to see the difference. I can’t quantify it, but just a hunch on the qualitative level.

But let’s see, it is a hypothesis, let’s get it tested.

1 Like

Well I soldered the U-blox and installed the RS-232 card.

In a testing machine…looks like the U-blox is broken or won’t function at 5V.

As it will not connect or respond.

So I plugged another (non-PPS) GPS that I have and it started right away giving NMEA.

So, I’m going to install the card in my server next week, connect the Garmin and see what it does.

The card is supported and working under Linux, so that is good.

As for the IRQ matter, it’s not the software to pickup an IRQ as the IRQ tells the CPU to stop what it’s doing and look at the RS-232 port.

Because the path is shorter, the IRQ should be received faster at the CPU and the CPU should respond faster. In theory…

That could make the PPS timing more precise, maybe not faster, but it shouldn’t run the same path as other data. Again theory :slight_smile:

Just look at the way the CPU is connected…that is why I think an PCIe card is faster then the onboard.


Lot of stuff connected to the H55…ah and note to self:

Connect it to 16x lane, not x1!!! Didn’t know that :slight_smile:

Yes, at least the signal lines on those boards are typically 3V. Whether power is 5V tolerant or 3V only as well depends on the board. To be on the safe side, I usually use 3V for power as well ever since I fried my first module.

That’s what I was referring to as usually being the same for either way of device attachment.

Yep, though the hypothesis I mentioned is that the speedup on this path is likely to be negligible in comparison to the time it takes the CPU to actually react to the interrupt and look at the port.

Well I don’t care about the U-blox, as I only wanted to play with one.
It wasn’t expensive, just 10 euro/usd.

For time it doesn’t matter as my Garmins work fine.

Is it?

As we talk nano-seconds here. I have read several stuff from Steve and he states it does matter.
As e.g. USB uses polling and not IRQ handling, also he states that USB3 is faster then USB2. But RS-232 is fast…yet others say PCIe is faster for RS-232.

I argued this stuff before…the thing is: I want to know if it matters!

The rest is irrelevant. If it doesn’t matter, I wasted some money…if it does work…we all can do this.

I just want to know :grinning:

Because if it’s better on the 1PPS Garmin, I’m going to test the 5PPS Garmin with it.

Like I said…I WANT TO KNOW :rofl:

Just in case you are curious, you could try plugging it in via USB to see whether it still works or not. The device would appear as ttyACMx under Linux.

Agree, I’d be curious about the outcome as well. The purpose of a hypothesis is to be tested to see whether it is true or false.

I could do that…but I’m not interested in that.
It was never my intention to use this el-cheapo for the pool anyway.

I know what the Garmin does…as such they are my default GPS receivers to measure against.

My point too…what is the best bus to use on a PC to get the most accurate time…

Next week I will change my server RS-232 port.

Yes it will take a bit of time, as I have 1 port to put a card in, but it’s connected to the 16x connector.
But it’s taking a hardwarestamp nic at this moment…that has to go.

Then the RS-232 card can take it’s place. And taken on Ask’s his graph we can see if it’s better or not.

Plan is to use it in this server:


It seems to be 15ms…if the PCIe card does better, we know the answer :slight_smile:

Good luck then with the testing next week!

Thank you…as I have long-covid, it’s not easy to quickly change stuff.

I need to plan everyting or I’m out of breath and can’t finish the job.

As such, sorry people, my plans take time…that is also the reason I dropped the U-blox not responding.

I simply do not have the energy…

I will report back…