I honestly don't know - probably used whatever they could get decent stocks of batch to batch as you suggested. My 34s are serial numbers 7328 (brown, DIN socketed) and 24943 (grey, phono sockets) and the former is TL071 entirely, the latter TL071/72s. The DaDa revision manual(s) make no mention of TLE2071/2s. In fact the only place I've seen them mentioned is on a schematic that Keith Snook's has redrawn, so had assumed he'd added it purely as the latest/last iteration of that series. Interesting that Quad did indeed use the TLE series too then.
There was the leaking cap problem in the C77/78 positions, post-tantalum to electrolytic change, which some argue was due to polarity, and others that it was simply a large batch of capacitors that unfortunately for Quad had electrolyte problems & failed, often eating into the pcb. I notice in the later DaDa manuals they no longer mention the polarity as incorrect anymore, with the exception of C16/17 I believe.
There was the leaking cap problem in the C77/78 positions, post-tantalum to electrolytic change, which some argue was due to polarity, and others that it was simply a large batch of capacitors that unfortunately for Quad had electrolyte problems & failed, often eating into the pcb. I notice in the later DaDa manuals they no longer mention the polarity as incorrect anymore, with the exception of C16/17 I believe.
Interesting indeed. I found schematics for all 3 issues in the service manual listed at Hifiengine and Electrotanya, though the upload is not the best and is misleading in that some caps in the output section appear to show polarity but I realized the symbol was inconsistent and more likely a scanning flaw: Quad 34 Manual - 4 Input Pre Amplifier - HiFi Engine. I don't think it really shows polarity of any electrolytics or tantalums, not even the main power supply electrolytic, so I presume this is marked on the PCB if at all.
https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
This document by Analog Devices provides information on the question as well as an example with the AD8099 high performance amplifier.
In addition, I would like to say what I do, yet, I do not pretend to be correct.
I would use three types of parallel capacitors as near the positive and negative pins of the amplifier as possible :
1. From the positive supply voltage pin of the amplifier to ground.
2. From the positive supply voltage pin of the amplifier to ground.
3. From the positive supply voltage pin of the amplifier to the negative supply voltage pin of the amplifier.
Now, Type 1 and Type 2 are the same. Type 3 may be the same in capacitances, but, higher in voltage.
For each type I would use these in parallel :
* One electrolytic AUDIO capacitor with as high value as physically normal, definitely >= 10uF. This capacitor MUST be : Made by Nichicon AND Made in Japan and no other country AND of as high temperature as Nichicon provides AND the best type ( lowest ESR and ESL ) of all Nichicon Audio Capacitors.
* One Tantalum capacitor of lowest ESR and ESL, >= 10uF. Best be Made in Japan by a huge company, such as TDK.
* One Ceramic capacitor with the lowest ESR. Usually, of every brand of Ceramic capacitors, the lowest ESR is at around 1uF. Must be Made in Japan by a huge company, which publishes ESR's. TDK is the only such company in the world I know of. ( Illinois Capacitor is an exotic company which makes the lowest ESR Ceramic capacitors, but, at high voltage. ) Also, THE HIGHER THE VOLTAGE THE LOWER THE ESR. Thus, 100V can be used, instead of the standard 50V. I have been immensely happy with NTE, Ceramic, 100V, probably X7R ( markings cannot be read even with a magnifying glass ). TDK may be better.
* One X7R, standard, 0.1uF. Again, Made in Japan by a huge company, lowest ESR, probably higher voltage to get lower ESR, etcetera.
* More lower values in case one likes to go capacitor mad.
The key to success is to find the lowest ESR's possible. Usually, for a given brand, these are, again, Made in Japan and not other country. Do NOT use capacitors made by huge Japanese companies, but, in another country, such as US or China. Must be Made in Japan. This ensures not only lowest ESR's and ESL's, but, the highest reliability and guaranteed rated voltage.
Do NOT use Chinese Electrolytic Audio capacitors of high values, such as 22mF, 16V, except when derated heavily. 16V means < 12V in Chinese.
Higher temperature brings longevity and reliability.
Do NOT use bipolar electrolytic capacitors ( back to back electrolytic capacitors ). These have higher ESR than their polar counterparts.
Now, in regards to Type 3 : Note : All other designers prefer NOT to use these. They prefer to use more or bigger value Type 1 and Type 2. This is because the noise, filtered by Type 3 has to go through the two supplies ( two secondary coils of the transformer ), the positive and the negative as opposed to Type 1 and Type 2, which go directly to ground, through only one of the supply ( coil ). Obviously, two coils have higher internal impedance than one. I think, Type 3 introduces stability, but, again, everyone else prefers NOT to have Type 3.
For those who make professional PCB's and not Vero boards, of course, a ground plate and vias as per Analog Devices are a must. Such can, probably, be made manually on Vero boards, but with a great difficulty and questionable quality.
Again, I do not pretend to be right as noise does not follow logic.
This document by Analog Devices provides information on the question as well as an example with the AD8099 high performance amplifier.
In addition, I would like to say what I do, yet, I do not pretend to be correct.
I would use three types of parallel capacitors as near the positive and negative pins of the amplifier as possible :
1. From the positive supply voltage pin of the amplifier to ground.
2. From the positive supply voltage pin of the amplifier to ground.
3. From the positive supply voltage pin of the amplifier to the negative supply voltage pin of the amplifier.
Now, Type 1 and Type 2 are the same. Type 3 may be the same in capacitances, but, higher in voltage.
For each type I would use these in parallel :
* One electrolytic AUDIO capacitor with as high value as physically normal, definitely >= 10uF. This capacitor MUST be : Made by Nichicon AND Made in Japan and no other country AND of as high temperature as Nichicon provides AND the best type ( lowest ESR and ESL ) of all Nichicon Audio Capacitors.
* One Tantalum capacitor of lowest ESR and ESL, >= 10uF. Best be Made in Japan by a huge company, such as TDK.
* One Ceramic capacitor with the lowest ESR. Usually, of every brand of Ceramic capacitors, the lowest ESR is at around 1uF. Must be Made in Japan by a huge company, which publishes ESR's. TDK is the only such company in the world I know of. ( Illinois Capacitor is an exotic company which makes the lowest ESR Ceramic capacitors, but, at high voltage. ) Also, THE HIGHER THE VOLTAGE THE LOWER THE ESR. Thus, 100V can be used, instead of the standard 50V. I have been immensely happy with NTE, Ceramic, 100V, probably X7R ( markings cannot be read even with a magnifying glass ). TDK may be better.
* One X7R, standard, 0.1uF. Again, Made in Japan by a huge company, lowest ESR, probably higher voltage to get lower ESR, etcetera.
* More lower values in case one likes to go capacitor mad.
The key to success is to find the lowest ESR's possible. Usually, for a given brand, these are, again, Made in Japan and not other country. Do NOT use capacitors made by huge Japanese companies, but, in another country, such as US or China. Must be Made in Japan. This ensures not only lowest ESR's and ESL's, but, the highest reliability and guaranteed rated voltage.
Do NOT use Chinese Electrolytic Audio capacitors of high values, such as 22mF, 16V, except when derated heavily. 16V means < 12V in Chinese.
Higher temperature brings longevity and reliability.
Do NOT use bipolar electrolytic capacitors ( back to back electrolytic capacitors ). These have higher ESR than their polar counterparts.
Now, in regards to Type 3 : Note : All other designers prefer NOT to use these. They prefer to use more or bigger value Type 1 and Type 2. This is because the noise, filtered by Type 3 has to go through the two supplies ( two secondary coils of the transformer ), the positive and the negative as opposed to Type 1 and Type 2, which go directly to ground, through only one of the supply ( coil ). Obviously, two coils have higher internal impedance than one. I think, Type 3 introduces stability, but, again, everyone else prefers NOT to have Type 3.
For those who make professional PCB's and not Vero boards, of course, a ground plate and vias as per Analog Devices are a must. Such can, probably, be made manually on Vero boards, but with a great difficulty and questionable quality.
Again, I do not pretend to be right as noise does not follow logic.
Addition to the Post
The lowest ESR capacitors I had ever seen were Illinois Capacitor with ESR = 1.2mOhms. However, these seem to have been discontinued. Also, Illinois Capacitor is a tiny US company, based in Chicago, which, automatically means they are not trustworthy. So, what they said in their datasheets may not have been what was really true. I do NOT trust such companies. Analog Devices, Texas Instruments, yes, these are trustworthy companies. But not Illinois Capacitor.
Thus, the only real company I have ever seen to make low ESR Ceramic ( the lowest ESR possible of all types ) is TDK. Some of the interesting capacitors may be their Multilayer Ceramic Capacitors, CLL and C series. There may be of lower ESR. The C series have ESR <= 0.2Ohms, 20mOhms at 10 to 100MHz. These are designed to filter CPU and other digital equipment frequencies, though.
Please, note : there are not many companies which publish ESR data at all.
The lowest ESR capacitors I had ever seen were Illinois Capacitor with ESR = 1.2mOhms. However, these seem to have been discontinued. Also, Illinois Capacitor is a tiny US company, based in Chicago, which, automatically means they are not trustworthy. So, what they said in their datasheets may not have been what was really true. I do NOT trust such companies. Analog Devices, Texas Instruments, yes, these are trustworthy companies. But not Illinois Capacitor.
Thus, the only real company I have ever seen to make low ESR Ceramic ( the lowest ESR possible of all types ) is TDK. Some of the interesting capacitors may be their Multilayer Ceramic Capacitors, CLL and C series. There may be of lower ESR. The C series have ESR <= 0.2Ohms, 20mOhms at 10 to 100MHz. These are designed to filter CPU and other digital equipment frequencies, though.
Please, note : there are not many companies which publish ESR data at all.
I think the two are very important. However, while some companies publish data on their ESL or specify their devices a low or ultra low ESL, not very many provide data on ESR's.
I think the lower the ESR, the better as the noise will be better filtered. Low ESL is, of course, important for the same reason.
I prefer to have ringing than noise. Of course, the best is to have neither.
Again, I do NOT pretend to be correct.
Some companies also publish impedance Z as a function of frequency. TDK does only in some cases, though.
Anyway, any information of low ESR and ESL capacitors will be welcome.
This is an interesting discussion of decoupling. I have some thoughts:
1) One should only decouple between each supply pin and ground.
2) One should have a very quiet and intelligently laid out ground. Whether that's a star ground, a ground plane, or some combination is up to the designer. Probably a plane common to all op amps or circuits that do not have some kind of differential or quasi-differential connection between them.
If the ground is not quiet, then that needs to be fixed.
3) Local decoupling assumes there are large supply bypass caps somewhere on the board.
4) Do not decouple from the V+ supply pin to the V- supply pin. A droop spike on the V+ pin will make the V- pin spike more negative. That might even cause an overvoltage problem, if running close to the datasheet limits. The supply of the whole op amp will shift downward, probably causing a negative spike on the op amp inputs and outputs.
1) One should only decouple between each supply pin and ground.
2) One should have a very quiet and intelligently laid out ground. Whether that's a star ground, a ground plane, or some combination is up to the designer. Probably a plane common to all op amps or circuits that do not have some kind of differential or quasi-differential connection between them.
If the ground is not quiet, then that needs to be fixed.
3) Local decoupling assumes there are large supply bypass caps somewhere on the board.
4) Do not decouple from the V+ supply pin to the V- supply pin. A droop spike on the V+ pin will make the V- pin spike more negative. That might even cause an overvoltage problem, if running close to the datasheet limits. The supply of the whole op amp will shift downward, probably causing a negative spike on the op amp inputs and outputs.
Thank you very much for your answer.
I concentrated on the amplifier decoupling and forgot to mention the star connection. Regardless on whether there is a ground plate or not, I always have a star connection at the overall outputs. Ideally, all grounds go there and only there, yet, this is not always easily possible with home made Vero boards. Star connection is a must for mixed, digital and analogue.
Yes, I do have huge capacitors at the power supply as well as at a device plugged to a power supply ( when separate ). Many capacitors in parallel.
Now, the most important thing : as I have mentioned, the correct way, expressed by all, is NOT to put capacitors between the rails, but, only between each rail and ground ( the ground is the reference to everything ) for the reason you have outlined well. However, such a capacitor ( s ) is very tempting, like a drug to a drug ( electronics ) addict. The only thing I hoped is to directly clear the noise between the two lines. However, yes, because of the higher source impedance, this noise would appear on the capacitors between each rail and noise.
Anyway, looks like capacitors between rails and ground only is the correct and better way.
Can you all, please, express more opinions on the topic. Is there any advantage at all for capacitors between the rails in addition to capacitors between each rail and ground? Or only disadvantages? Are there other disadvantages?
For example : What happens when not the power supply but the load ( something on the board, say an amplifier or an unlike digital circuit connected between the two rails ), something on the board makes a particular noise between the two rails, for example, a circuit which works only or mainly between the two rails. In case of such a noise, wouldn't capacitors between the two rails help get rid of most of the noise better?
Please, post.
Last edited:
I would only add, in case you agree, the suggested 0.1uF capacitor to be X7R, Ceramic. Do you agree?
Op-amp decoupling... further thoughts
Hi Steven,
If each rail is stiff wrt ground, then they will be stiff wrt each other.
Supply bypass is not supposed to be fixing noise from the power supply. It is supposed to fix two things:
1) Op amp problems due to too much impedance on the supply rails. RF oscillation is a typical problem.
2) Noise from the op amp driving its load. Note that current to the load is only drawn from one rail at a time, so bypass should be to GND.
There is one scenario where bypass between the rails can benefit: if you have a differential or dual op-amp driving a load bridged between the outputs. The current from/to each rail is then identical, and a C value between the rails is twice as effective as that C from rail to ground. In that scenario, I might put 0.1uF from V+ to GND, 0.1uF from V- to GND, and 0.1uF from V+ to V-. Only in that bridged scenario would a cap from V+ to V- be of benefit.
Hi Steven,
If each rail is stiff wrt ground, then they will be stiff wrt each other.
Supply bypass is not supposed to be fixing noise from the power supply. It is supposed to fix two things:
1) Op amp problems due to too much impedance on the supply rails. RF oscillation is a typical problem.
2) Noise from the op amp driving its load. Note that current to the load is only drawn from one rail at a time, so bypass should be to GND.
There is one scenario where bypass between the rails can benefit: if you have a differential or dual op-amp driving a load bridged between the outputs. The current from/to each rail is then identical, and a C value between the rails is twice as effective as that C from rail to ground. In that scenario, I might put 0.1uF from V+ to GND, 0.1uF from V- to GND, and 0.1uF from V+ to V-. Only in that bridged scenario would a cap from V+ to V- be of benefit.
Yes, I meant differential components and loads.
True, they can use rail to ground too, but, in such a case, the current would go from the positive rail through one of the capacitor, through ground and then through the other capacitor. Two capacitors add their ESR and ESL. Thus, one capacitor is preferable.
An amplifier is, in most cases, a differential load to the power supply and, yes, so are the speakers in a differential amplifier.
In regards to noises, I think, the decoupling capacitors near the amplifier have two tasks : to reduce any noise which comes from the amplifier and to reduce any noise which comes from power supply through the PCB track and or is electromagnetically induced in the tracks.
Also, unrelated to your post, I forgot to mention when I mentioned the star connection, in case of analogue and digital circuits AND an ADC, the star connection best be at the ADC. Originally, I said, I make the star connection at the overall outputs. This is when there is no ADC. When there is, the most important is to eliminate the noise at the ADC to get a higher ADC acccuracy.
Please, post more.
True, they can use rail to ground too, but, in such a case, the current would go from the positive rail through one of the capacitor, through ground and then through the other capacitor. Two capacitors add their ESR and ESL. Thus, one capacitor is preferable.
An amplifier is, in most cases, a differential load to the power supply and, yes, so are the speakers in a differential amplifier.
In regards to noises, I think, the decoupling capacitors near the amplifier have two tasks : to reduce any noise which comes from the amplifier and to reduce any noise which comes from power supply through the PCB track and or is electromagnetically induced in the tracks.
Also, unrelated to your post, I forgot to mention when I mentioned the star connection, in case of analogue and digital circuits AND an ADC, the star connection best be at the ADC. Originally, I said, I make the star connection at the overall outputs. This is when there is no ADC. When there is, the most important is to eliminate the noise at the ADC to get a higher ADC acccuracy.
Please, post more.
Also, I have just found another piece on the Internet. This provides some information, which, I am not sure how true or false. Most importantly, there is a link to another article by Analog Devices :
https://www.analog.com/media/en/training-seminars/tutorials/MT-043.pdf
Here is the whole discussion :
How to use Decoupling capacitors for Op amp in dual rail | All About Circuits
Again, I do not know whether the information on the discussion is true or false.
Please, post more.
https://www.analog.com/media/en/training-seminars/tutorials/MT-043.pdf
Here is the whole discussion :
How to use Decoupling capacitors for Op amp in dual rail | All About Circuits
Again, I do not know whether the information on the discussion is true or false.
Please, post more.
Yes, I fully agree the decoupling capacitors of an amplifier increase stability and response as the influence of the impedance of the tracks.
Similar is the effect of the power supply capacitors when a momentary response by any load is required. The capacitors, with their low ESR, provide the voltage and the current for a short period, while the power supply has a higher output impedance.
Similar is the effect of the power supply capacitors when a momentary response by any load is required. The capacitors, with their low ESR, provide the voltage and the current for a short period, while the power supply has a higher output impedance.
Interesting thread like many on this forum 
I use some steps from large to individual :
- environment (home, automotive, professional) : agressive or not ?
- position : one or more chips (single or distributed power)
- design need : light to heavy consomption (high Z, buffer...), single-ended or differential, intrinsic chips specificities (speed, bandpass, suceptibility...)
Answer usualy call mixed caps (film, ceramic, tantalum) with some others parts (diode, choke coil...).
Attention was also focused on caps discharge path and to not exceed power supply perfomances (capacitive swing...)
I use some steps from large to individual :
- environment (home, automotive, professional) : agressive or not ?
- position : one or more chips (single or distributed power)
- design need : light to heavy consomption (high Z, buffer...), single-ended or differential, intrinsic chips specificities (speed, bandpass, suceptibility...)
Answer usualy call mixed caps (film, ceramic, tantalum) with some others parts (diode, choke coil...).
Attention was also focused on caps discharge path and to not exceed power supply perfomances (capacitive swing...)
I see very often modders placing consumer grade DIP sockets on their PCBs unaware of the parsitic C and L they are adding.
Now they want to operate this fancy high speed opamp which only comes in So8, so an adapter board is added and the device will oscillate at its transit frequency, no matter where and what capacitor is installed.
The scope is not capable to trigger at 100+ MHz, maybe they see an increased current consumption.
Finally they end up with a TL072 or NE5532, which are stable even without decouple Cs.
What I want to say: The lay-out is crucial, each current path has to be optimized, more so the gnd where the current flows back to the psu. No interruptions of the gnd-plane below the incoming current trace.
Measuring with a network analyzer with coax connectors soldered to the pcb fitting the tips.
Now they want to operate this fancy high speed opamp which only comes in So8, so an adapter board is added and the device will oscillate at its transit frequency, no matter where and what capacitor is installed.
The scope is not capable to trigger at 100+ MHz, maybe they see an increased current consumption.
Finally they end up with a TL072 or NE5532, which are stable even without decouple Cs.
What I want to say: The lay-out is crucial, each current path has to be optimized, more so the gnd where the current flows back to the psu. No interruptions of the gnd-plane below the incoming current trace.
Measuring with a network analyzer with coax connectors soldered to the pcb fitting the tips.
Illinois capacitors used to do very high quality and reliable industrial capacitors...i don't really know were your doubts come from...i think the ones i have from them and the Siemens(epcos) ones are the only ones i saw to speciffy the Ieff max value on them...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.