Efinix FPGA

59 views
Skip to first unread message

gnuarm.del...@gmail.com

unread,
Jul 11, 2022, 8:25:27 PMJul 11
to
Anyone using Efinix parts? They look ok, even if they don't have a lot of package offerings.

The smallest part has a 0.5A surge at power on. The list it as "minimum", I'm guessing they mean the minimum required by the supply.

They also don't provide software until you buy an eval board, so no way to check that out, up front.

Funny company, but not as "funny" as Cologne Chip. They have a part listed at Digikey, (in stock, too) but no 3.3V I/Os, only up to 2.5V.

--

Rick C.

- Get 1,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

DJ Delorie

unread,
Jul 12, 2022, 11:23:49 AMJul 12
to
"gnuarm.del...@gmail.com" <gnuarm.del...@gmail.com> writes:
> Anyone using Efinix parts? They look ok, even if they don't have a
> lot of package offerings.

I checked into them a little while back, hoping for 5v tolerance, but
they didn't have it.

> They also don't provide software until you buy an eval board, so no
> way to check that out, up front.

Their smallest board used to be dirt cheap before the chip shortage. I
got one just to test the software.

https://www.digikey.com/en/products/detail/efinix-inc/XYLONI/13535080

> Funny company, but not as "funny" as Cologne Chip. They have a part
> listed at Digikey, (in stock, too) but no 3.3V I/Os, only up to 2.5V.

3.3v I/O is listed in the datasheet for the Trion family, right on the
first page...

https://www.digikey.com/en/supplier-centers/efinix

KJ

unread,
Jul 12, 2022, 7:55:42 PMJul 12
to
On Monday, July 11, 2022 at 8:25:27 PM UTC-4, gnuarm.del...@gmail.com wrote:
> Anyone using Efinix parts? They look ok, even if they don't have a lot of package offerings.
>
We are using Efinix due to supply chain shortages to replace Cyclone 10 part.
>
> They also don't provide software until you buy an eval board, so no way to check that out, up front.
>
The software is OK, but Quartus for Intel/Altera parts is much better. Haven't used Xilinx in too long of a time to make that comparison.

Kevin Jennings

Josef Moellers

unread,
Jul 13, 2022, 2:31:30 AMJul 13
to

On 12.07.22 17:23, DJ Delorie wrote:
> "gnuarm.del...@gmail.com" <gnuarm.del...@gmail.com> writes:
>> Anyone using Efinix parts? They look ok, even if they don't have a
>> lot of package offerings.
>
> I checked into them a little while back, hoping for 5v tolerance, but
> they didn't have it.

Are there ANY FPGAs that have 5V tolerance?
I'm looking for one to replace an FDC9266.

Josef

gnuarm.del...@gmail.com

unread,
Jul 13, 2022, 9:36:46 AMJul 13
to
Unless you use a very, very obsolete device, you won't find 5V tolerance in any FPGAs. On inputs, 5V tolerance is easy to add, using just a pair of resistors as a voltage divider. On output, you will need to use at least a transistor to allow switching to 5V. But if you are working with TTL inputs, you don't need 5V drive, 3.3V drive should be enough.

There are also ways to use resistors on outputs, to allow higher rise, by giving up something on the low end. What do you need to interface to?

--

Rick C.

+ Get 1,000 miles of free Supercharging
+ Tesla referral code - https://ts.la/richard11209

Josef Moellers

unread,
Jul 13, 2022, 9:53:55 AMJul 13
to
It's a very old SingleBoardComputer with an HD64180 and lots of LS TTL
stuff. I need to interface to inputs (eg DMA requests), outputs (eg
address lines) and bidirectional (data lines). I am considering using
level shifters on the latter.

IIRC I had been asking here before and got the same answer you gave
above, but as DJ mentioned it, I thought I might be lucky today.

As I'm hopefully having lots of spare time in the near future
(retirement after almost 41 years in the IT industry), I guess I'll put
it all together and give it a try.

Thanks anyway

Josef

DJ Delorie

unread,
Jul 13, 2022, 4:11:13 PMJul 13
to
Josef Moellers <josef.m...@invalid.invalid> writes:
> On 12.07.22 17:23, DJ Delorie wrote:
> Are there ANY FPGAs that have 5V tolerance?

The Spartan 3E has an app note for how to add 5V tolerance to their I/O
with a single resistor, but 3E's are becoming scarce.

gnuarm.del...@gmail.com

unread,
Jul 14, 2022, 3:06:20 AMJul 14
to
I like switch based parts for 5-3.3V level shifting. It drops the 5V power rail to 4.3V internally, which is used to drive the gate of the pass transistors. This causes them to not conduct above 3.2V or so. So the 5V side can swing as much as it likes and the 3.3V side is protected. The 3.3V side can drive the 5V side to 3.3V which is adequate for TTL levels.

Using this with 2.5V logic doesn't work so well since 2.5V CMOS doesn't drive high enough for 5V TTL. You have to use real level shifters. Once you open that can of worms, it gets ugly with various parts for unidirectional and bidirectional. How do you control the direction with parts that actually drive?

--

Rick C.

-- Get 1,000 miles of free Supercharging
-- Tesla referral code - https://ts.la/richard11209

gnuarm.del...@gmail.com

unread,
Jul 14, 2022, 3:08:00 AMJul 14
to
That's just a matter of limiting the current when the protection diode conducts. However, it also limits speed, from the RC of the limiting resistor and the input capacitance.

--

Rick C.

-+ Get 1,000 miles of free Supercharging
-+ Tesla referral code - https://ts.la/richard11209

Anssi Saari

unread,
Jul 14, 2022, 7:33:36 AMJul 14
to
"gnuarm.del...@gmail.com" <gnuarm.del...@gmail.com> writes:

> Using this with 2.5V logic doesn't work so well since 2.5V CMOS
> doesn't drive high enough for 5V TTL. You have to use real level
> shifters. Once you open that can of worms, it gets ugly with various
> parts for unidirectional and bidirectional. How do you control the
> direction with parts that actually drive?

Here in Europe, there was a lot of RoHS conversions about two decades
ago which in practise meant swithcing to lead free components.

Anssi Saari

unread,
Jul 14, 2022, 7:48:36 AMJul 14
to
"gnuarm.del...@gmail.com" <gnuarm.del...@gmail.com> writes:

> Using this with 2.5V logic doesn't work so well since 2.5V CMOS
> doesn't drive high enough for 5V TTL. You have to use real level
> shifters. Once you open that can of worms, it gets ugly with various
> parts for unidirectional and bidirectional. How do you control the
> direction with parts that actually drive?

Second try. About two decades ago there were a lot of lead free
conversions in Europe due to the RoHS directive and again last decade
when it expanded to medical devices too. So since a lot of old
components weren't available lead free, a lot of substitution was done
and often that meant replacing 5V parts with 3.3V parts and using level
shifters. To be sure, mostly slow interfaces, old microcontrollers
running at 8 or 16 MHz and slower than that serial stuff.

In a bunch of such projects there was never any issue with the level
shifters. Never. They were completely problem free all the time.

As for the direction control, some level shifters have a direction input
and some autodetect. As I recall, the autodetecting ones were fine
already back when.

Josef Moellers

unread,
Jul 14, 2022, 10:36:04 AMJul 14
to
On 14.07.22 09:06, gnuarm.del...@gmail.com wrote:
> On Wednesday, July 13, 2022 at 9:53:55 AM UTC-4, Josef Moellers wrote:
>> On 13.07.22 15:36, gnuarm.del...@gmail.com wrote:
>>> On Wednesday, July 13, 2022 at 2:31:30 AM UTC-4, Josef Moellers wrote:
>>>> On 12.07.22 17:23, DJ Delorie wrote:
>>>>> "gnuarm.del...@gmail.com" <gnuarm.del...@gmail.com> writes:
>>>>>> Anyone using Efinix parts? They look ok, even if they don't have a
>>>>>> lot of package offerings.
>>>>>
>>>>> I checked into them a little while back, hoping for 5v tolerance, but
>>>>> they didn't have it.
>>>> Are there ANY FPGAs that have 5V tolerance?
>>>> I'm looking for one to replace an FDC9266.
>>>
>>> Unless you use a very, very obsolete device, you won't find 5V tolerance in any FPGAs. On inputs, 5V tolerance is easy to add, using just a pair of resistors as a voltage divider. On output, you will need to use at least a transistor to allow switching to 5V. But if you are working with TTL inputs, you don't need 5V drive, 3.3V drive should be enough.
>>>
>>> There are also ways to use resistors on outputs, to allow higher rise, by giving up something on the low end. What do you need to interface to?
>> It's a very old SingleBoardComputer with an HD64180 and lots of LS TTL
>> stuff. I need to interface to inputs (eg DMA requests), outputs (eg
>> address lines) and bidirectional (data lines). I am considering using
>> level shifters on the latter.
>>
>> IIRC I had been asking here before and got the same answer you gave
>> above, but as DJ mentioned it, I thought I might be lucky today.
>>
>> As I'm hopefully having lots of spare time in the near future
>> (retirement after almost 41 years in the IT industry), I guess I'll put
>> it all together and give it a try.

[...]

> Using this with 2.5V logic doesn't work so well since 2.5V CMOS doesn't drive high enough for 5V TTL. You have to use real level shifters. Once you open that can of worms, it gets ugly with various parts for unidirectional and bidirectional. How do you control the direction with parts that actually drive?

Some years ago, I have looked a bit into I²C, even wrote a 3(?) part
series for "Elektor" (German hobbyist electronics magazine, german
edition of Dutch "Elektuur") and there I learned the simple level
shifter using a FET and two pull-ups. That's what I'll (probably) use,
at least for the bidirectional signals. As only a one side is actually
driving, it should work.

Josef

gnuarm.del...@gmail.com

unread,
Jul 14, 2022, 12:48:31 PMJul 14
to
On Thursday, July 14, 2022 at 7:48:36 AM UTC-4, Anssi Saari wrote:
> "gnuarm.del...@gmail.com" <gnuarm.del...@gmail.com> writes:
>
> > Using this with 2.5V logic doesn't work so well since 2.5V CMOS
> > doesn't drive high enough for 5V TTL. You have to use real level
> > shifters. Once you open that can of worms, it gets ugly with various
> > parts for unidirectional and bidirectional. How do you control the
> > direction with parts that actually drive?
> Second try. About two decades ago there were a lot of lead free
> conversions in Europe due to the RoHS directive and again last decade
> when it expanded to medical devices too. So since a lot of old
> components weren't available lead free, a lot of substitution was done
> and often that meant replacing 5V parts with 3.3V parts and using level
> shifters. To be sure, mostly slow interfaces, old microcontrollers
> running at 8 or 16 MHz and slower than that serial stuff.
>
> In a bunch of such projects there was never any issue with the level
> shifters. Never. They were completely problem free all the time.

I'm not sure what this is in regard to. Why would you expect any problems from level shifters?

I'm referring to the proliferation of part numbers for level shifters. TI must have a dozen families of level shifters.


> As for the direction control, some level shifters have a direction input
> and some autodetect. As I recall, the autodetecting ones were fine
> already back when.

How do autodetect shifters, detect the direction? I believe this imposes some sort of limitation on the output drive, no?

--

Rick C.

+- Get 1,000 miles of free Supercharging
+- Tesla referral code - https://ts.la/richard11209

gnuarm.del...@gmail.com

unread,
Jul 14, 2022, 12:57:34 PMJul 14
to
I assume you mean a FET as a pass element with limited pull up capability? That's what I use on a board I have currently in production. TI SN74CBTD3384CPW, 10 bits, separate enables for the two 5 bit sections. Works well. This chip is designed for the job. There are smaller versions. Hacking your own out of FETs means have to select the FETs with appropriate gate thresholds and variance. This is not normally controlled carefully. You also have to generate the drive voltage for the gate. That's why they invented the such switch parts.

https://www.ti.com/lit/ml/scdb006a/scdb006a.pdf?ts=1657817753410&ref_url=https%253A%252F%252Fwww.google.com%252F

--

Rick C.

++ Get 1,000 miles of free Supercharging
++ Tesla referral code - https://ts.la/richard11209

Josef Moellers

unread,
Jul 15, 2022, 2:10:03 AMJul 15
to
On 14.07.22 18:48, gnuarm.del...@gmail.com wrote:
> On Thursday, July 14, 2022 at 7:48:36 AM UTC-4, Anssi Saari wrote:

[...]

>> As for the direction control, some level shifters have a direction input
>> and some autodetect. As I recall, the autodetecting ones were fine
>> already back when.
>
> How do autodetect shifters, detect the direction? I believe this imposes some sort of limitation on the output drive, no?

I'm only a software person (with a strong hardware interest), but I
assume that it centers around the fact that on a bidirectional
connection at any one time only one side is actually driving the
connection. I'd need to consult an ex-colleague who at that time
explained the conecpt of this FET-Level-Shifter to me.

Josef

Josef Moellers

unread,
Jul 15, 2022, 2:12:32 AMJul 15
to
As I wrote: this was in the context of the I²C bus and you only need two
shifters at the max.

> https://www.ti.com/lit/ml/scdb006a/scdb006a.pdf?ts=1657817753410&ref_url=https%253A%252F%252Fwww.google.com%252F

Thanks for the reference.
I may use it when the time comes ...

Josef

gnuarm.del...@gmail.com

unread,
Jul 15, 2022, 11:26:56 PMJul 15
to
I believe the way these work is to wire the two drivers so each one drives the other input. This makes a stable FF arraingment. The output is through a resistor to limit the current drive. Then, when one side or the other is driven, the resistor on the output (or just a current limited driver) allows the external drive to overcome the internal drive and so control the input. This is reflected on the other output, which drives the other input, which switches this side output to match the level driven by the external driver. This also acts as a bus keeper, preventing the buses from drifting to invalid voltage levels which can be an issue with excessive supply current or even damage. The current limited output can be a problem with fast switching the output.

I think I saw on some fancier parts, where, to deal with the slow switching, would drive with two drivers in parallel, a high current driver for fast switching speed and a low current driver as the "bus keeper". They used an internal timing circuit to disable the high current driver after a short time. This is a bit fuzzy to me as it was a long time ago I saw this.

--

Rick C.

--- Get 1,000 miles of free Supercharging
--- Tesla referral code - https://ts.la/richard11209

Clifford Heath

unread,
Aug 17, 2022, 7:33:49 PMAug 17
to
On 13/7/22 09:55, KJ wrote:
> On Monday, July 11, 2022 at 8:25:27 PM UTC-4, gnuarm.del...@gmail.com wrote:
>> Anyone using Efinix parts? They look ok, even if they don't have a lot of package offerings.
> We are using Efinix due to supply chain shortages to replace Cyclone 10 part.

Tell us more please. How does your number of LUTs used compare with
traditional offerings for your use case? Can you tell us more about the
kind of architecture you're implementing?

Asking because I've heard that the Trion's have minimal routing fabric,
so the tools tend to use LUTs for routing. The tradeoff is because
Efinix can radically reduce the number of mask layers, cutting both cost
and power consumption to about a third of "equivalent" devices.

I'm considering trying to port the Hermes SDR software to a Trion, and
need to know how big a device would be needed. It looks like it might
fit in a T35.

Clifford Heath
Reply all
Reply to author
Forward
0 new messages