This reminds of a joke I heard the late Peter Baxandall say at an AES meeting many many years ago. Q: What do you call very low-frequency noise? A: Weather.
John Atkinson
Editor, Stereophile
Hmmm... What do you call AC in this statement?
Edit: according to John's remark, weather change is already AC.
Last edited:
You need to separate Closed Loop AC gain (which indeed you ideally want to be unity a reg) and the AC Loop gain, which is not going to be unity in a typical design.
The reason the reg needs to be comp'd is because the AC loop gain has to meet the Bode criteria in order to avoid oscillation.
The reason the reg needs to be comp'd is because the AC loop gain has to meet the Bode criteria in order to avoid oscillation.
Hi,
I wrote "AC within the band of interest"...
To wit, for audio it is often considered 20Hz - 20KHz. A wider bandwidth will not do harm.
Ciao T
I wrote "AC within the band of interest"...
To wit, for audio it is often considered 20Hz - 20KHz. A wider bandwidth will not do harm.
Ciao T
Hi,
To me in the application only the closed loop AC gain is of concern. If some-one makes a regulator chip that is inherently unstable as shipped from factory then it is my choice to use it or not and to then take the measures to ensure stability.
Let's be clear. The capacitor after the regulator has long been tradition, even in circuits not strictly requiring it and when the chip regulators where designed the designers elected to rely on the presence of this capacitor for stability as this easier/cheaper than to produce an inherently stable regulator.
In more recent times this bad habit has incidentally become a problem as miniaturisation and demands for good ripple rejection into the 100's of KHz have changed the game considerably.
Ciao T
To me in the application only the closed loop AC gain is of concern. If some-one makes a regulator chip that is inherently unstable as shipped from factory then it is my choice to use it or not and to then take the measures to ensure stability.
Let's be clear. The capacitor after the regulator has long been tradition, even in circuits not strictly requiring it and when the chip regulators where designed the designers elected to rely on the presence of this capacitor for stability as this easier/cheaper than to produce an inherently stable regulator.
In more recent times this bad habit has incidentally become a problem as miniaturisation and demands for good ripple rejection into the 100's of KHz have changed the game considerably.
Ciao T
I wouldn't say it has anything to do with a part being shipped that is 'inherantly unstable' or 'tradition'. A typical series regulator like a 317 uses feedback and the closed loop gain sets the output voltage. For the noise resons you state, you may well want the AC gain to be unity. However, if you want output errors (due to loading, line disturbances etc) to be removed, you will need a reasonable loop gain and you then cannot ignore what is going on inside the loop - and this is why the output cap is needed - its for compensation. Power amps that use feedback are no different - when you comp a power amp, you focus on the loop gain performance - and you can get some insight as to how its doing by looking at its closed loop performance and using spice analysis.
Cap free IC regulators are available, and typically targeted at low power (up to 100 to 200mA), miniturisation and portable applications. However, when a device is spec'd to be able to gracefully deal with a fast rise/fall time 0.5A to 1A transients, IC designers are going to struggle to do that without some help from an external capacitor. I know of very few high power linear regs that do not use output caps - otherwise a whole lot of other comprimises and tradeoffs will be needed. The output cap is a good engineering solution for these types of devices and the designer made sensible choices IMHO.
Cap free IC regulators are available, and typically targeted at low power (up to 100 to 200mA), miniturisation and portable applications. However, when a device is spec'd to be able to gracefully deal with a fast rise/fall time 0.5A to 1A transients, IC designers are going to struggle to do that without some help from an external capacitor. I know of very few high power linear regs that do not use output caps - otherwise a whole lot of other comprimises and tradeoffs will be needed. The output cap is a good engineering solution for these types of devices and the designer made sensible choices IMHO.
Last edited:
But noise below 20 Hz may do harm, obviously...
Also, my comment was about your interesting academic discussion with SY, About higher than 1 amplification factor on DC, but exactly 1 on AC.
It was rhetoric question of academic character from my side, of course.
Well, you have to deal with a trade off here. Either the reg slews because it cannot provide transient energy, or you use localized capacitor decoupling to solve the problem. Either way you are going to get some voltage disturbance on the supply line - that is unavoidable.
I would not use a regulator on a class AB amp - no need to use a pickaxe to solve a problem that needs a bit of finesse and a few resistors and small caps to create very high PSRR on the amplifer itself - ripple eater is another good technique. Douglas Self and I think Bob Cordell cover the PSRR and how to maximize it in their books.
Last edited:
A large cap in the power supply "freezes" the voltage, and gives the regulator time to catch up. This isn't like at the output of an amp driving a capacitive load, where it has to charge and discharge the capacitance at audio frequencies.
A DC regulator as a whole is supposed to have a high DC gain, as it should so it reacts to the smallest changes in DC. The output cap makes this so, and makes it more stable, as others have mentioned.
I think the regulator should act like a DC servo in an audio circuit. Audio signals as well as DC come into the circuit, but only DC appears at the output to correct the DC offset.
If there is only DC at the output of a regulator is doing a very good job.
My main concern is that the voltage disturbance on the power line is correlated to signal. I wonder if the discussion could be broadened to include the optimum tradeoffs in power supply bandwidth (small signal and slewing) vs. output impedance.
When I mentioned a classAB/B amplifier I meant a classic monolythic opamp, whose power supply is shared among all stages. Monolythic matching and modern design have made PSRR huge at audio frequencies, but the requirements are huge too.
Thanks,
Chris
Hi,
Unlike you, it seems I still remember "BIC" (Before IC's) and how discrete regulators where designed...
Ciao T
Unlike you, it seems I still remember "BIC" (Before IC's) and how discrete regulators where designed...
Ciao T
Hi,
I'm not afraid of it at all. Often I use only the Capacitor and trade the regulator for a choke or two and resistors...
Ciao T
I'm not afraid of it at all. Often I use only the Capacitor and trade the regulator for a choke or two and resistors...
Ciao T
I really wish people would study voltage regulators more, before coming to conclusions about how and how well they operate. If you place a cap on the output, you increase the current requirements to charge the cap, when it is depleted even a little. The Q of the cap will effect the AC noise of the regulator. There is NO FREE LUNCH! Everything is a compromise. Just LOOK at the spec sheets carefully. Live with them for at least 30 years, as well, then you can come to conclusions.
You must be older than me - as a schoolboy I was building power supplies based on uA723.
Now the uA723 was a nice quiet regulator. I think you can still get them - they are up on the National/TI site.
http://www.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=lm723&fileType=pdf
I was using them IIRC in the late 70's, but it may have been a bit later than that.
Output noise 2.5uV with the reference (6.9V) decoupled, and the line and load reg was also not bad.
http://www.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=lm723&fileType=pdf
I was using them IIRC in the late 70's, but it may have been a bit later than that.
Output noise 2.5uV with the reference (6.9V) decoupled, and the line and load reg was also not bad.
Last edited:
- Status
- Not open for further replies.