Getting down to 'fine hairs', are ICs in the same family like tubes? Usually an Amperex 12AX7 with short plates sounds the same in the same circuit. Changing to an RCA short plate 12AX7 will sound different. Yet, both share the same part number.
Are ICs with same part number from different manufactures sound different in the same circuit? How about a late 1990s IC vs a 2005 IC from same manufacture? Identical sonics?
What brings up the question is using power transistors over the years. Same part number has an HFE that will vary from say an HFE of 20 to a HFE of 150. Kind of like jumping from a 12AU7 to a 12AX7, yet the transistor shares the same part number.
Apparently, solid-state has huge variances in manufacturing. I assume ICs are much closer tolerance. But, I still wonder about the ICs as mentioned above.
Are ICs with same part number from different manufactures sound different in the same circuit? How about a late 1990s IC vs a 2005 IC from same manufacture? Identical sonics?
What brings up the question is using power transistors over the years. Same part number has an HFE that will vary from say an HFE of 20 to a HFE of 150. Kind of like jumping from a 12AU7 to a 12AX7, yet the transistor shares the same part number.
Apparently, solid-state has huge variances in manufacturing. I assume ICs are much closer tolerance. But, I still wonder about the ICs as mentioned above.
Good question,
I do know that in some cases popular chips are redesigned because the newer processes allow for smaller transistors and therefore less silicon. I'm not saying that is a bad thing. There are many parts that carry the old base number (with a few new letters appended) that have vastly improved characteristics... LM324XXXXX comes to mind.
😉
I do know that in some cases popular chips are redesigned because the newer processes allow for smaller transistors and therefore less silicon. I'm not saying that is a bad thing. There are many parts that carry the old base number (with a few new letters appended) that have vastly improved characteristics... LM324XXXXX comes to mind.
😉
amperex said:
Yes...identical sonics , because op amps unlike tubes use heavy feedback and can't be used open loop as tubes... and the feedback help to make differences to vanish .
Yep...but in the transistor , even if the HFE change , the output impedance will be roughly the same (in common collector configuration ), unlike a change from a 12AX7 to a 12AU7 where the change in output impedance is drastic .
Not so!...But as ICs can't be used without feedback , they behave almost ideally , at last , on paper... 😉
When it comes to transistors, they can vary in spec from manufacturer to manufacturer. In the case of JEDEC (2NXXXX numbered devices) devices parameters were assigned but, there is no absolute rule that says a device with that number has to meet all of the specs. A little different for the Japanese transistors since the numbers are assigned to the manufacturer.
In the case of IC's the same is roughly true. The differences are usually due to circuit topologies. For instance, a 1458 opamp from 3 different manufacturers, can have completely different layouts from each other. They are also dependent on the photolithography resolution(usually in uM or nM), that the manufacturer is using. Packaging material variations can also play a part as well.
Revisions to the die making process usually force revisions to popular devices. Most translate well but everybody has there klinkers.
Also be aware that, when reading the datasheet on a device, what is not in the data is as important as what is not. An example of this is the TL072. At the top they state a really attractive GBW and slew rate. Elsewhere in the document, you see that this is for small signal conditions with a 5 volt rail. Performance decreases considerably with +/- 15 volt rails, and even more with large signal response. This is no surprise since all amplifiers have reduced performance with large signal response. But, to those unfamiliar with the finer point of amplifier behavior, dissapointments may occasionally arise.
The upshot is, read the datasheets carefully and question any specs that seem to be to good to be true. Datasheets are ultimately written by marketeers, not engineers. Understand your critical design parameters. Look for things like voltage and current self-noise, large signal reponse, unity gain stability and, single supply stability.
Simple, isn't it?
In the case of IC's the same is roughly true. The differences are usually due to circuit topologies. For instance, a 1458 opamp from 3 different manufacturers, can have completely different layouts from each other. They are also dependent on the photolithography resolution(usually in uM or nM), that the manufacturer is using. Packaging material variations can also play a part as well.
Revisions to the die making process usually force revisions to popular devices. Most translate well but everybody has there klinkers.
Also be aware that, when reading the datasheet on a device, what is not in the data is as important as what is not. An example of this is the TL072. At the top they state a really attractive GBW and slew rate. Elsewhere in the document, you see that this is for small signal conditions with a 5 volt rail. Performance decreases considerably with +/- 15 volt rails, and even more with large signal response. This is no surprise since all amplifiers have reduced performance with large signal response. But, to those unfamiliar with the finer point of amplifier behavior, dissapointments may occasionally arise.
The upshot is, read the datasheets carefully and question any specs that seem to be to good to be true. Datasheets are ultimately written by marketeers, not engineers. Understand your critical design parameters. Look for things like voltage and current self-noise, large signal reponse, unity gain stability and, single supply stability.
Simple, isn't it?
Apparently, solid-state has huge variances...... Comparing simple transistors to vacuum tubes it does vary a lot.
But, I am not being silly here. OP-Amps & other solid-state IC devices can perform tasks tubes could never perform. I see ther future will have continued growth in op-amp designs.
But, I am not being silly here. OP-Amps & other solid-state IC devices can perform tasks tubes could never perform. I see ther future will have continued growth in op-amp designs.
And the power transistotrs that vary their gain? It depends on how much current they are conducting, in the sense that your car will generate different amounts of torque at different engine speeds. You will usually see the gain of a transistor specified at a certain level of collector current. If you don't specify the current, then the gain is given as a range. It has absolutely nothing to do with any sloppy manufacturing tolerances.
We used to built linear power supplies for the US Military. Believe me, transistors vary a lot at any given current. Power supplies providing 130 amperes at 13.8 VDC uses a $hit load of pass transistors.
The basic answer is that different manufacturers IC's don't really sound (or perform) any differently. As already suggested, IC's and transistors have wide spreads of specifications - but you should design to make those irrelevent - this is what good design is all about.
In the (very!) distant past I've repaired equipment where you have to select resistor values when you replace a faulty transistor - this is incredibly poor design!, and isn't something you should ever see these days!.
In the (very!) distant past I've repaired equipment where you have to select resistor values when you replace a faulty transistor - this is incredibly poor design!, and isn't something you should ever see these days!.
If you are refering to our design......
A group of pass transistors in a large lot in a high current linear type power supply is not a poor design. The extra heat must be spread across a large area, thus one or two pass transistors would not perform even with copper heat sinks or a heat sink with a copper spreader.
After all, specs having less than 3 mv ripple at 130 amperes @ 13.8 vdc as measured on the output & .2 volt overshoot at start-up is as good as it gets without a large battery on the output.
Cycling a contactor to a 'chatter type' condition five times a second from no load to full load with severe arcing has no effect on the linear power supply design either. Others tested lost regulation & tripped the crowbar circuit.
That is most likely the best high current power supply ever manufactured. 'Switcher' types do not come close either with regard to same performance under the same conditions.
A group of pass transistors in a large lot in a high current linear type power supply is not a poor design. The extra heat must be spread across a large area, thus one or two pass transistors would not perform even with copper heat sinks or a heat sink with a copper spreader.
After all, specs having less than 3 mv ripple at 130 amperes @ 13.8 vdc as measured on the output & .2 volt overshoot at start-up is as good as it gets without a large battery on the output.
Cycling a contactor to a 'chatter type' condition five times a second from no load to full load with severe arcing has no effect on the linear power supply design either. Others tested lost regulation & tripped the crowbar circuit.
That is most likely the best high current power supply ever manufactured. 'Switcher' types do not come close either with regard to same performance under the same conditions.
No, he didn't say using a large number of parts is a bad design. He said designs that are vulnerable to part variations are bad designs.
Exactly, linear PS low voltage & high current design requires large amounts of pass transistors & are vulnerable to failure if one uses non-matched pass transistors. No other way to design it, yet nothing short of a PS with a battery will outperform linear type design.
amperex said:
You use low value emitter resistors in each transistor to balance the load - it's a standard technique, and would be FAR better than attempting to use matched transistors (because there's no way they will be matched accurately enough).
Did you discount that method for some reason? (although I can't think of one?).
8,000,000 audio amplifiers can't be wrong. Those fractional ohm resistors are called "ballast" resistors in that use. They allow transistors to share current. the transistors all ought to be the same type, but they don't then have to be exactly matched.
Disagree........
13.8 vdc @ 130 amperes regulated power supply here- hello.
Typical bipolar power transistors vary in hfe from say 20 to 150. We match closely for a lot of pass transistors. Do some simple ohms law math & you will find to use any hfe from far less a spread of 20 to 150 would require such a high value load sharing lot of resistors that little current would be available at the output. Like I said, we supplied the Military & NASA.
You obviously have no experience in high current linear power supplies. Enzo, your fired.
13.8 vdc @ 130 amperes regulated power supply here- hello.
Typical bipolar power transistors vary in hfe from say 20 to 150. We match closely for a lot of pass transistors. Do some simple ohms law math & you will find to use any hfe from far less a spread of 20 to 150 would require such a high value load sharing lot of resistors that little current would be available at the output. Like I said, we supplied the Military & NASA.
You obviously have no experience in high current linear power supplies. Enzo, your fired.
Re: Disagree........
I suggest YOU try applying ohms law, I can't believe you didn't use emitter resistors! - and (from what you've posted!) didn't even consider doing it that way? (the CORRECT way!).
It's an incredibly simple concept, the resistors basically give local negative feedback. To give a simple example - based on just two transistors, although it makes no difference how many you use, the principle is identical.
OK, two transistors in parallel, with 0.1 ohm emitter resistors. because transistors aren't perfectly matched (even with your selected ones!), one will turn ON before the other one - fairly obviously. You pick the value of the resistors to reduce dissipation and voltage loss, the lower the value the more current the 'best' transistor will provide on it's own at low loads.
So we'll assume that one turns ON at 0.7V, and the other at 0.75V (nice round figures). As the load increases from zero the first transistor will provide ALL of the power to the load, as it's 0.7V Vbe is preventing the second transistors 0.75V Vbe being reached.
However, the effective Vbe of the first transistor is 0.7V PLUS the voltage drop across it's 0.1 ohm emitter resistor. So once 500mA is flowing through it then it's effective Vbe becomes 0.75V, enough for the second transistor to start doing some of the work.
This bypasses some of the current from the first transistor, which then takes more of the current back. In this way every transistor takes it's share of the load (except at very small currents when you only need one transistor anyway).
The gain of the individual transistors is irrelevent, as long as the drivers have enough current capacity to feed them all - making the circuit device insensitive, as any good design should be.
I'm absolutely staggered that a commercial manufacturer would try and match devices rather than using standard design techniques, and I'm even MORE staggered that the US Military and NASA would use such designs?.
BTW, it's common practice to provide a current meter by measuring the voltage drop across any one of the resistors - obviously this won't work for very small loads - but you choose the resistors accordingly (no point measuring 0.5A with a 130A PSU).
amperex said:
I suggest YOU try applying ohms law, I can't believe you didn't use emitter resistors! - and (from what you've posted!) didn't even consider doing it that way? (the CORRECT way!).
It's an incredibly simple concept, the resistors basically give local negative feedback. To give a simple example - based on just two transistors, although it makes no difference how many you use, the principle is identical.
OK, two transistors in parallel, with 0.1 ohm emitter resistors. because transistors aren't perfectly matched (even with your selected ones!), one will turn ON before the other one - fairly obviously. You pick the value of the resistors to reduce dissipation and voltage loss, the lower the value the more current the 'best' transistor will provide on it's own at low loads.
So we'll assume that one turns ON at 0.7V, and the other at 0.75V (nice round figures). As the load increases from zero the first transistor will provide ALL of the power to the load, as it's 0.7V Vbe is preventing the second transistors 0.75V Vbe being reached.
However, the effective Vbe of the first transistor is 0.7V PLUS the voltage drop across it's 0.1 ohm emitter resistor. So once 500mA is flowing through it then it's effective Vbe becomes 0.75V, enough for the second transistor to start doing some of the work.
This bypasses some of the current from the first transistor, which then takes more of the current back. In this way every transistor takes it's share of the load (except at very small currents when you only need one transistor anyway).
The gain of the individual transistors is irrelevent, as long as the drivers have enough current capacity to feed them all - making the circuit device insensitive, as any good design should be.
I'm absolutely staggered that a commercial manufacturer would try and match devices rather than using standard design techniques, and I'm even MORE staggered that the US Military and NASA would use such designs?.
BTW, it's common practice to provide a current meter by measuring the voltage drop across any one of the resistors - obviously this won't work for very small loads - but you choose the resistors accordingly (no point measuring 0.5A with a 130A PSU).
to Nigel
I did not say we do not use emitter balancing resistors of fractional ohms. For you to suggest matching hfe is not good design practice means you also obviously have no experience designing or manufacturing high power linear power supplies.
Not one idiot would not close match pass transistors AND use emitter resistors. You suggest using whatever transistor comes out of the freakin box be it an hfe of 15 or 150 & simply toss a lot of 16 of these in a high current PS?
You have no experience in high current power supplies. Do you have any idea of what is required to be Mil-Spec Cage certified?
Your out to lunch here big time here. Go back to audio applications as I do believe you are well versed.
BTW- The original discussion was the tolerance of solid-state devices varies & wildly so with discrete transistors compared to vacuum tubes.
I did not say we do not use emitter balancing resistors of fractional ohms. For you to suggest matching hfe is not good design practice means you also obviously have no experience designing or manufacturing high power linear power supplies.
Not one idiot would not close match pass transistors AND use emitter resistors. You suggest using whatever transistor comes out of the freakin box be it an hfe of 15 or 150 & simply toss a lot of 16 of these in a high current PS?
You have no experience in high current power supplies. Do you have any idea of what is required to be Mil-Spec Cage certified?
Your out to lunch here big time here. Go back to audio applications as I do believe you are well versed.
BTW- The original discussion was the tolerance of solid-state devices varies & wildly so with discrete transistors compared to vacuum tubes.
Re: to Nigel
So WHY has it taken you so long to confirm that?, my entire point of view has been that these are essential, and you have always given the impression that didn't use any, instead relying on matching the transistors!.
As for your assertion that you need to match the transistors as well?, I would still say that correct design would prevent that requirement!. However, as your application was to Military spec (with a corresponding massive price tag), there's no harm to a 'belt and braces' approach.
amperex said:
So WHY has it taken you so long to confirm that?, my entire point of view has been that these are essential, and you have always given the impression that didn't use any, instead relying on matching the transistors!.
As for your assertion that you need to match the transistors as well?, I would still say that correct design would prevent that requirement!. However, as your application was to Military spec (with a corresponding massive price tag), there's no harm to a 'belt and braces' approach.
I have been designing high current low noise power supplies since 1973. Why would I not use current balancing resistors? Many PS I designed up to 150 amperes output from the 1970s are still functioning in automotive plants today without failure.
Why would one used unmatched pass transistors & triple the wasted heat thru those transistors to force matching with large value matching resistors?
Should one design & produce a KW of wasted heat? Match the transistors, document the installed hfe right on the PS itself & make for a more efficient design.
Per Motorola, for every 10 degree C drop in junction temperature doubles the life of a transistor. One can not afford improper matching with one transistor operating at a junction temperature of over 200 degree C in the lot.
Anyways, this is really an audio site & I am sure you know electronics design as applied to audio. Time to move on, good day.
Why would one used unmatched pass transistors & triple the wasted heat thru those transistors to force matching with large value matching resistors?
Should one design & produce a KW of wasted heat? Match the transistors, document the installed hfe right on the PS itself & make for a more efficient design.
Per Motorola, for every 10 degree C drop in junction temperature doubles the life of a transistor. One can not afford improper matching with one transistor operating at a junction temperature of over 200 degree C in the lot.
Anyways, this is really an audio site & I am sure you know electronics design as applied to audio. Time to move on, good day.
- Status
- Not open for further replies.