Digital logic gate needed

[blueskull]

Is there an open drain 8-bit or 4-bit DFF in BGA package?
I tried to search TI and DigiKey, with no result. If not, I will consider MAX II or MAX V CPLDs.
My end application is a high speed (12.288Msps), 4-bit, full differential DAC, designed as output stage of a high end audio sigma delta DAC.
Jitter performance of global clock tree is crucial, and so do output driving capability (SSO, matching, etc.) and digital noise.
Therefore I prefer a little logic device rather than a CPLD, and I will not consider FPGA (even small ones like iCE40 or MAX 10) at all.
Thanks for any suggestions.

[David Hess]

I am not aware of any open drain D flip-flops.  A tri-state buffer can be used to add open drain capability to any logic signal if the logic family does not include open drain parts.

[Zero999]

I am not aware of any open drain D flip-flops.  A tri-state buffer can be used to add open drain capability to any logic signal if the logic family does not include open drain parts.

What about adding Schottky diodes? It's a bit of a hack but they're cheap and come in two per package.

[David Hess]

I am not aware of any open drain D flip-flops.  A tri-state buffer can be used to add open drain capability to any logic signal if the logic family does not include open drain parts.

What about adding Schottky diodes? It's a bit of a hack but they're cheap and come in two per package.

Schottky diodes will work if the voltage drop is acceptable which it will not be with the lowest voltage CMOS.  Keep in mind that a schottky diode will not allow switching a voltage higher or lower than the supply and ground.

A bipolar or MOSFET common emitter/source inverter on the output could also be used to avoid the forward voltage drop of the schottky diode but if you were going to do that, then an IC filled with open drain buffers or inverters would save space.

[Zero999]

I am not aware of any open drain D flip-flops.  A tri-state buffer can be used to add open drain capability to any logic signal if the logic family does not include open drain parts.

What about adding Schottky diodes? It's a bit of a hack but they're cheap and come in two per package.

Schottky diodes will work if the voltage drop is acceptable which it will not be with the lowest voltage CMOS.  Keep in mind that a schottky diode will not allow switching a voltage higher or lower than the supply and ground.

A bipolar or MOSFET common emitter/source inverter on the output could also be used to avoid the forward voltage drop of the schottky diode but if you were going to do that, then an IC filled with open drain buffers or inverters would save space.

I see what you mean about the voltage drop of Schottky diodes being too high for very low voltage CMOS, but as far as switching higher voltage is concerned, don't most open drain devices have internal ESD protection diodes from the output to +V? Do you know of any devices which don't?

Rather than a common emitter/source inverter, how about common base/gate configuration? That's non-inverting and there's no Miller effect so could be faster too. The only downside is the extra voltage drop.

[David Hess]

Also keep in mind that I'm talking 12.288MHz with output pulse shape correction, which means at least 24.576MHz, and I plan to use 49.152MHz switching speed or even double that. At that speed, I have no idea how schottky diodes will behave.

Schottky diodes are faster than switching diodes and will behave just fine at those frequencies.  The chief advantage in this application is their lower forward voltage drop than a switching diode.

[David Hess]

I see what you mean about the voltage drop of Schottky diodes being too high for very low voltage CMOS, but as far as switching higher voltage is concerned, don't most open drain devices have internal ESD protection diodes from the output to +V? Do you know of any devices which don't?

Bipolar open collector outputs do not include them and I think for CMOS, the protection diodes are part of the body structure of the MOSFETs so with an open drain output, the top body diode is missing.  The bipolar parts were often used and specified for switching voltages higher than Vcc.

I just checked the datasheet for a CMOS open drain output and it deliberately shows no high side protection diode and there is no requirement that the output voltage be limited to below the supply voltage.  They were often used to translate low voltage logic levels to higher voltage logic levels.

Rather than a common emitter/source inverter, how about common base/gate configuration? That's non-inverting and there's no Miller effect so could be faster too. The only downside is the extra voltage drop.

I actually started my post discussing this but then just kept the common emitter/source part because it is simpler to understand.

The common base level shifter works great for this and is non-inverting.  If base bias is provided by a resistor to say 1.2 volts, then the base current can be controlled driving the transistor into saturation allowing an output voltage down to ground so there is no forward voltage drop.  In slower applications this is better than using a diode and it can still operate pretty quickly if a fast saturated switch is used.  (1) (2) It is more difficult to do with MOSFETs because of their higher gate threshold voltage but I have done it with 5 volt logic.

(1) Storage time still applies in common base mode if the transistor saturates which is what makes fast saturated switches so convenient in this application.

(2) Saturation can be prevented with a Baker clamp but now the high collector to emitter voltage makes the transistor have no advantage over using a diode.  I ran across this trying to replace fast saturated switches with RF transistors that had Baker clamps.

[Zero999]

I see what you mean about the voltage drop of Schottky diodes being too high for very low voltage CMOS, but as far as switching higher voltage is concerned, don't most open drain devices have internal ESD protection diodes from the output to +V? Do you know of any devices which don't?

Bipolar open collector outputs do not include them and I think for CMOS, the protection diodes are part of the body structure of the MOSFETs so with an open drain output, the top body diode is missing.  The bipolar parts were often used and specified for switching voltages higher than Vcc.

I just checked the datasheet for a CMOS open drain output and it deliberately shows no high side protection diode and there is no requirement that the output voltage be limited to below the supply voltage.  They were often used to translate low voltage logic levels to higher voltage logic levels.

Please post the data sheet. I was aware that old TTL and slightly newer LS devices could be used for this kind of level shifting but the CMOS devices I've found such as the 74HCT05 and 74ACT05 still include that pesky diode, even if it isn't explicitly shown on the schematic!
http://www.sycelectronica.com/semiconductores/74HCT05.pdf
http://www.ti.com/lit/ds/symlink/cd74act05.pdf

[David Hess]

Please post the data sheet. I was aware that old TTL and slightly newer LS devices could be used for this kind of level shifting but the CMOS devices I've found such as the 74HCT05 and 74ACT05 still include that pesky diode, even if it isn't explicitly shown on the schematic!
http://www.sycelectronica.com/semiconductores/74HCT05.pdf
http://www.ti.com/lit/ds/symlink/cd74act05.pdf

https://assets.nexperia.com/documents/data-sheet/74HC_HCT03.pdf

Notice that on your TI datasheet, the maximum output clamp current is specified as -50 milliamps maximum only for output voltages below 0.  There is no positive clamp!

That other part looks defective to me.

[julian1]

> Jitter performance of global clock tree is crucial,

Is an fpga a problem from the point of view of jitter? Presumably, chaining together discrete devices will introduce varying propagation delay. Otherwise if it's the fpga PLL, is it possible to disable/bypass, or just not use the internal global clocking lines? I think the ice40 does open-drain via the led IO block.
 
Obviously there are other reasons to avoid an fpga - overkill, price, etc.

[T3sl4co1l]

Can't do a '173 + '126?

Tim

[David Hess]

With FPGA, core voltage is shared by all logic resources, and can't be effectively bypassed. With discrete devices in DSBGA package, I can decouple VCC at every single logic gate.

From datasheet of ProASIC3, the jitter, even without PLL or anything, from a global clock tree, is in hundreds of picosecond range, their rated worst case is 400ps.

With a logic gate, though not specified, estimated jitter from a standard package is 40ps, quoted from a TI employee's answer in E2E. I believe with better decoupling and output current biasing, this can be reduced to a couple ps.

Exactly, modulation of the ground and power supplied to each gate changes the switching thresholds increasing jitter.  For synchronous logic, this is irrelevant because each clocked register stage retimes the previous logic.  Unfortunately this applies to the output stages of the FPGA as well so in applications where jitter is important, the external clock is used with a register to retime the signals.

Extremely low jitter discrete logic is very careful about power and ground decoupling and may use low noise local regulation, differential signaling, or a separate reference for the logic threshold.  Gates may be distributed to specific packages to prevent interference.

I have read about two cases now, one from an old Tektronix digital sampling oscilloscope and one with a modern FPGA evaluation board, where a blinking LED had to be disabled because it was introducing significant jitter into the logic.  The supply and ground currents for the LEDs were the problem.

[Zero999]

Please post the data sheet. I was aware that old TTL and slightly newer LS devices could be used for this kind of level shifting but the CMOS devices I've found such as the 74HCT05 and 74ACT05 still include that pesky diode, even if it isn't explicitly shown on the schematic!
http://www.sycelectronica.com/semiconductores/74HCT05.pdf
http://www.ti.com/lit/ds/symlink/cd74act05.pdf

https://assets.nexperia.com/documents/data-sheet/74HC_HCT03.pdf

Notice that on your TI datasheet, the maximum output clamp current is specified as -50 milliamps maximum only for output voltages below 0.  There is no positive clamp!

That other part looks defective to me.

Why is the maximum recommended output voltage VCC? Sorry, I'm not convinced there's no diode on the output. There are some 74HC05s in the inventory at work I'll have to have a play with after hours to test this.

[David Hess]

Why is the maximum recommended output voltage VCC? Sorry, I'm not convinced there's no diode on the output. There are some 74HC05s in the inventory at work I'll have to have a play with after hours to test this.

On the NXP part I linked and the TI part, the maximum output voltage is *not* Vcc.  It is the same as the maximum Vcc voltage of 6 volts but typically 5 volts and Vcc may be lower than the output voltage.  That is what makes these parts useful for level translation.

Based on the other datasheet you linked, I suspect some manufacturers added the output clamp diode and others did not.  Based on the datasheets, the NXP and TI parts lack it.

[Zero999]

Why is the maximum recommended output voltage VCC? Sorry, I'm not convinced there's no diode on the output. There are some 74HC05s in the inventory at work I'll have to have a play with after hours to test this.

On the NXP part I linked and the TI part, the maximum output voltage is *not* Vcc.  It is the same as the maximum Vcc voltage of 6 volts but typically 5 volts and Vcc may be lower than the output voltage.  That is what makes these parts useful for level translation.

Based on the other datasheet you linked, I suspect some manufacturers added the output clamp diode and others did not.  Based on the datasheets, the NXP and TI parts lack it.

I've just used my diode tester on a TI part I found in the inventory at work and there's definitely a diode junction between pin 2 (anode) and 14 (cathode). The NXP data sheet confirms this. If you want to level shift beyond 0.5V above the positive rail, the 74HC05 is no good. You should use the 74LS05 instead.
https://assets.nexperia.com/documents/data-sheet/74HC05.pdf

[David Hess]

I've just used my diode tester on a TI part I found in the inventory at work and there's definitely a diode junction between pin 2 (anode) and 14 (cathode). The NXP data sheet confirms this. If you want to level shift beyond 0.5V above the positive rail, the 74HC05 is no good. You should use the 74LS05 instead.
https://assets.nexperia.com/documents/data-sheet/74HC05.pdf

The NXP datasheet I linked was unambiguous about no diode being present but this would not be the fist time datasheets are completely wrong.  This is one of those thing which would require qualified parts.

[MK14]

I've just used my diode tester on a TI part I found in the inventory at work and there's definitely a diode junction between pin 2 (anode) and 14 (cathode). The NXP data sheet confirms this. If you want to level shift beyond 0.5V above the positive rail, the 74HC05 is no good. You should use the 74LS05 instead.
https://assets.nexperia.com/documents/data-sheet/74HC05.pdf

The NXP datasheet I linked was unambiguous about no diode being present but this would not be the fist time datasheets are completely wrong.  This is one of those thing which would require qualified parts.

But your datasheet (in your earlier post) was for the HC'03 part, NOT the HC'05 part. Maybe the HC'03 part is the one WITHOUT the top diode ?
But the HC'05 DOES seem to have the top diode.
I've only quickly skimmed this thread, so maybe I'm mixed up.

[Zero999]

I've just used my diode tester on a TI part I found in the inventory at work and there's definitely a diode junction between pin 2 (anode) and 14 (cathode). The NXP data sheet confirms this. If you want to level shift beyond 0.5V above the positive rail, the 74HC05 is no good. You should use the 74LS05 instead.
https://assets.nexperia.com/documents/data-sheet/74HC05.pdf

The NXP datasheet I linked was unambiguous about no diode being present but this would not be the fist time datasheets are completely wrong.  This is one of those thing which would require qualified parts.

But your datasheet (in your earlier post) was for the HC'03 part, NOT the HC'05 part. Maybe the HC'03 part is the one WITHOUT the top diode ?
But the HC'05 DOES seem to have the top diode.
I've only quickly skimmed this thread, so maybe I'm mixed up.

I suspect the 74HC03 also has a diode from the output to Vcc. The ON Semiconductor data sheet shows it and the TI data sheet specifies the output clamping current can be positive as well as negative. The NXP data sheet only specifies a negative output clamp current, but the maximum recommended output voltage is Vcc, so I suspect it still has the diode. If anyone has a Phillips/NPX 74HC03/05/09 device, it would be interesting to know whether its output has a diode to Vcc.
https://www.onsemi.com/pub/Collateral/MC74HC03A-D.PDF
http://www.ti.com/lit/ds/symlink/sn74hc03.pdf

[MK14]

I suspect the 74HC03 also has a diode from the output to Vcc. The ON Semiconductor data sheet shows it and the TI data sheet specifies the output clamping current can be positive as well as negative. The NXP data sheet only specifies a negative output clamp current, but the maximum recommended output voltage is Vcc, so I suspect it still has the diode. If anyone has a Phillips/NPX 74HC03/05/09 device, it would be interesting to know whether its output has a diode to Vcc.
https://www.onsemi.com/pub/Collateral/MC74HC03A-D.PDF
http://www.ti.com/lit/ds/symlink/sn74hc03.pdf

The (NXP==>>new name) datasheet does seem to show that they have left the top output clamp diode out (probably intentionally). There is not necessarily a problem with (NXP==>>new name) being different to some of the other manufacturers.
I would have thought that such features can help them sell more chips. So I can understand why they do things like this.
There is at least one application note (which does NOT necessarily apply to HC) as well (it does NOT seem to specifically mention the 74HC03). But it does explain some of the reasons why they are sometimes removing clamp diodes.

E.g. Plugging/unplugging still live devices, WITHOUT causing damage because of the way clamp diodes can work.

But I agree with you, it is best to make sure with the real/actual device. Before committing yourself to using the parts and/or making a PCB. Because datasheets can sometimes be wrong or apply to older (i.e. your datasheet is an old version, you picked up on the internet), or newer (but you are using OLD parts), or even unreleased parts.
(NXP==>>new name) should be ok. But with companies such as Maxim, there is a definite risk of the parts either never coming out or becoming obsolete, soon after release.