All,
I've been doing some measurement lately on pre-amp power supply regulators. I've read a lot on this forum about super regulators, etc and whether the regulators supply the amp or the op-amp decoupling caps, etc, etc. However, I'm just starting out, so need to build a firmer foundation for my own understanding before moving on.
I am new to measurements in audio and I don't have the equipment to measure the regulator PSRR, so I have constructed a mock-up pre amp. This consists of a TL071CN op-amp with a gain of 10, delivering 10v pk-pk, between 100Hz and 200KHz into a 1Kohm resistive load. I can power this pre-amp with a 78/79 series or a LM317/337 (implemented as data sheet; variable supply with improved ripple rejection) supply. I have decided to take my measurements at the positive supply pin of the op-amp; so that I can investigate what difference it "sees" with different supplies. Here are my results, all pk-pk mV:
Frequency_______ripple 78/79__________ ripple LM317/337
100 Hz _________1.8_____________________ 1
1 KHZ1__________10______________________2
10 KHZ___________8______________________4
20 KHZ___________6______________________5
200 KHZ__________6______________________6
The shape of the ripple varied between saw tooth and sine wave, with some weird shapes at certain frequencies; see below. All these waveforms had about 1mV hf superimposed on them, at about 10MHz, which is a bit dubious as the scope has only 15MHz bandwidth!
The 78/79 regulators gave a peak ripple of 22mV pk-pk at 3 KHz; the LM317/337 regulators gave a peak ripple of 6mV pk-pk at 200 KHz.
The 78/79 regulators gave the weirdest shape ripple at 1.2 KHz (pictured below), the LM317/337 regulators gave the weirdest shape ripple at 22.85 KHz (similar in shape to the picture below, although about 1/2 the pk-pk value).
The test was not that scientific as the regulators had different transformers and values of smoothing caps.
Listening tests using my pre-amp showed the LM317/337 as a clear winner across the board, much tighter and more powerful bass, the sound was less "box bound", better soundstage and better treble.
The questions i have are:
What is responsible for the weird shape waveforms I am encountering, is this normal?
Having determined by measurement and listening, that the LM317/337 route is best for me, what's my best way ahead?
Thanks for reading, Mark
I've been doing some measurement lately on pre-amp power supply regulators. I've read a lot on this forum about super regulators, etc and whether the regulators supply the amp or the op-amp decoupling caps, etc, etc. However, I'm just starting out, so need to build a firmer foundation for my own understanding before moving on.
I am new to measurements in audio and I don't have the equipment to measure the regulator PSRR, so I have constructed a mock-up pre amp. This consists of a TL071CN op-amp with a gain of 10, delivering 10v pk-pk, between 100Hz and 200KHz into a 1Kohm resistive load. I can power this pre-amp with a 78/79 series or a LM317/337 (implemented as data sheet; variable supply with improved ripple rejection) supply. I have decided to take my measurements at the positive supply pin of the op-amp; so that I can investigate what difference it "sees" with different supplies. Here are my results, all pk-pk mV:
Frequency_______ripple 78/79__________ ripple LM317/337
100 Hz _________1.8_____________________ 1
1 KHZ1__________10______________________2
10 KHZ___________8______________________4
20 KHZ___________6______________________5
200 KHZ__________6______________________6
The shape of the ripple varied between saw tooth and sine wave, with some weird shapes at certain frequencies; see below. All these waveforms had about 1mV hf superimposed on them, at about 10MHz, which is a bit dubious as the scope has only 15MHz bandwidth!
The 78/79 regulators gave a peak ripple of 22mV pk-pk at 3 KHz; the LM317/337 regulators gave a peak ripple of 6mV pk-pk at 200 KHz.
The 78/79 regulators gave the weirdest shape ripple at 1.2 KHz (pictured below), the LM317/337 regulators gave the weirdest shape ripple at 22.85 KHz (similar in shape to the picture below, although about 1/2 the pk-pk value).
An externally hosted image should be here but it was not working when we last tested it.
The test was not that scientific as the regulators had different transformers and values of smoothing caps.
Listening tests using my pre-amp showed the LM317/337 as a clear winner across the board, much tighter and more powerful bass, the sound was less "box bound", better soundstage and better treble.
The questions i have are:
What is responsible for the weird shape waveforms I am encountering, is this normal?
Having determined by measurement and listening, that the LM317/337 route is best for me, what's my best way ahead?
Thanks for reading, Mark
test set up?
"The shape of the ripple varied between saw tooth and sine wave, with some weird shapes at certain frequencies; see below. All these waveforms had about 1mV hf superimposed on them, at about 10MHz, which is a bit dubious as the scope has only 15MHz bandwidth!"
It is entirely possible.
Do you have decoupling caps on the op amp and three terminal regulators?
What is the ripple voltage on the input voltage to the regulator?
What supply voltages is the op amp running on?
What does the circuit for the gain of 10 op amp look like?
Do you have have a build out resistor between the output of the op amp and any cable capacitance it is driving?
How long is the ground lead from the board to the scope probe?
Make the measurement exactly like your test set up now but with the probe tip to ground any see if you see HF noise.
It really sounds like the circuit under test is not stable. You should see at least a factor of 1000 less output ripple than input ripple for either type regulator. You need to go ver your test set up if your not.
"The shape of the ripple varied between saw tooth and sine wave, with some weird shapes at certain frequencies; see below. All these waveforms had about 1mV hf superimposed on them, at about 10MHz, which is a bit dubious as the scope has only 15MHz bandwidth!"
It is entirely possible.
Do you have decoupling caps on the op amp and three terminal regulators?
What is the ripple voltage on the input voltage to the regulator?
What supply voltages is the op amp running on?
What does the circuit for the gain of 10 op amp look like?
Do you have have a build out resistor between the output of the op amp and any cable capacitance it is driving?
How long is the ground lead from the board to the scope probe?
Make the measurement exactly like your test set up now but with the probe tip to ground any see if you see HF noise.
It really sounds like the circuit under test is not stable. You should see at least a factor of 1000 less output ripple than input ripple for either type regulator. You need to go ver your test set up if your not.
Low Noise Regulators
You might want to try some Linear Technologies regulators, they are ultra low noise but are also surface mount and only put out like 200-500 mA where those National chips are about 3 Amps or so. If you are using opamps my evaluation of these was that most only require about 5-10 mA, so 200 or 300 mA in a regulator should be plenty. Try LT1763, LT1962, LT1964. for more info, see my Gainclone w/ LN Power Supply For Input Stage.
Looking at your graphs on the scope, I think these will provide at least as good ripple rejection if not better from what I recall seeing on my scope.
You might want to try some Linear Technologies regulators, they are ultra low noise but are also surface mount and only put out like 200-500 mA where those National chips are about 3 Amps or so. If you are using opamps my evaluation of these was that most only require about 5-10 mA, so 200 or 300 mA in a regulator should be plenty. Try LT1763, LT1962, LT1964. for more info, see my Gainclone w/ LN Power Supply For Input Stage.
Looking at your graphs on the scope, I think these will provide at least as good ripple rejection if not better from what I recall seeing on my scope.
Thanks for all your replies!
Dave, will investigate the -ve supply soon.
Stocker: I am happy to accept that my results are wrong, that's why i'm asking opinions, i could make exactly the same psu's and run the test again, it's a lot of effort though, do you think it's worth it?
Fred:
This is the setup i have:
and this is my ripple voltage on the input voltage to the regulator at .1 v/div(so less than 0.2v pk-pk):
Not sure how clear things are going to be from the photo: the regs have no decoupling caps as they are within 4" of the main filter caps. the op-amp supplies are decoupled with a 4.7uF electrolytic and a 100n bypass. The supply voltage is +/-15 to make it the same as the 78/915 supply.
What does the circuit for the gain of 10 op amp look like?
hopefully the photo will be clear enough for you to see!
Do you have have a build out resistor between the output of the op amp and any cable capacitance it is driving?
No, it's driviving a 1K ohm resistor only.
How long is the ground lead from the board to the scope probe?
a few inches.
Make the measurement exactly like your test set up now but with the probe tip to ground any see if you see HF noise.
It's the same with the op-amp supply and sig gen all off!
It really sounds like the circuit under test is not stable. You should see at least a factor of 1000 less output ripple than input ripple for either type regulator. You need to go ver your test set up if your not.
0.2v ripple on the smoothing caps, i can't measure anything on the reg o/p at 100 Hz, my scope only goes down to 2mV/div. The frequency of the ripple measured at the +ve supply pin of the op-amp, in my first post is the same as the signal it's passing, not 100Hz. sorry i should have been more clear!
Thanks again
Mark
Dave, will investigate the -ve supply soon.
Stocker: I am happy to accept that my results are wrong, that's why i'm asking opinions, i could make exactly the same psu's and run the test again, it's a lot of effort though, do you think it's worth it?
Fred:
This is the setup i have:
An externally hosted image should be here but it was not working when we last tested it.
and this is my ripple voltage on the input voltage to the regulator at .1 v/div(so less than 0.2v pk-pk):
An externally hosted image should be here but it was not working when we last tested it.
Not sure how clear things are going to be from the photo: the regs have no decoupling caps as they are within 4" of the main filter caps. the op-amp supplies are decoupled with a 4.7uF electrolytic and a 100n bypass. The supply voltage is +/-15 to make it the same as the 78/915 supply.
What does the circuit for the gain of 10 op amp look like?
hopefully the photo will be clear enough for you to see!
Do you have have a build out resistor between the output of the op amp and any cable capacitance it is driving?
No, it's driviving a 1K ohm resistor only.
How long is the ground lead from the board to the scope probe?
a few inches.
Make the measurement exactly like your test set up now but with the probe tip to ground any see if you see HF noise.
It's the same with the op-amp supply and sig gen all off!
It really sounds like the circuit under test is not stable. You should see at least a factor of 1000 less output ripple than input ripple for either type regulator. You need to go ver your test set up if your not.
0.2v ripple on the smoothing caps, i can't measure anything on the reg o/p at 100 Hz, my scope only goes down to 2mV/div. The frequency of the ripple measured at the +ve supply pin of the op-amp, in my first post is the same as the signal it's passing, not 100Hz. sorry i should have been more clear!
Thanks again
Mark
Did you not have a large capacitor on the outputs of both regulators before? If not, I highly reccommend it. I have found in my own experimentation that you can get by on a small input cap if you have a large cap on the output, since 3-terminal regulators have such high rejection rates. Specifically, with IIRC about 1000 or 2000uF bypassed with a couple of smaller electrolytics (2.2? 4.7? ) on the output of a 78L05, ripple went to ~0.1mV. Yes it's a bit big but it worked g-r-e-a-t . Try another several hundred uF on the outputs of both and see what you get.
Hi Jos/Stocker.
Jos. that could well be my prob with the 78/9 series, they're only driving an op-amp, although quite hard, not constantly, will introduce some base load. already have 1K ohm loads for my 3x7 setup.
Stocker. I've limited my experimentation to the 3x7 setup for now. I have had different results to you though! I have added a 470uF cap at the 3x7 regs o/p. Whilst this has reduced the positive supply pin ripple in my test setup 6x at 20KHz, the listening results have been quite negative. On the plus side; i have possibly the sweetest treble emerging from my scanspeak 2905/9500 tweeters. However, the overall balance has changed from punchy to thin, as well as a lack of depth and power in the bass, what little i have is muddled. I have 2200uF b4 the regs BTW.
Thanks Mark
Jos. that could well be my prob with the 78/9 series, they're only driving an op-amp, although quite hard, not constantly, will introduce some base load. already have 1K ohm loads for my 3x7 setup.
Stocker. I've limited my experimentation to the 3x7 setup for now. I have had different results to you though! I have added a 470uF cap at the 3x7 regs o/p. Whilst this has reduced the positive supply pin ripple in my test setup 6x at 20KHz, the listening results have been quite negative. On the plus side; i have possibly the sweetest treble emerging from my scanspeak 2905/9500 tweeters. However, the overall balance has changed from punchy to thin, as well as a lack of depth and power in the bass, what little i have is muddled. I have 2200uF b4 the regs BTW.
Thanks Mark
As always Stocker!
searching on this forum i found there does not seem to be any real consensus about which way to go regulator. I think results depend upon regulator use (digital/class A/op-amp). This is why my approach was to make measurements on a system (regulator with hard working pre-amp in tow) rather than just the regulators themselves. TBH, i think i was measuring the wrong thing-well obviously as the measured results were not sonically verifiable. How did your measured results influence sound?
Mark
searching on this forum i found there does not seem to be any real consensus about which way to go regulator. I think results depend upon regulator use (digital/class A/op-amp). This is why my approach was to make measurements on a system (regulator with hard working pre-amp in tow) rather than just the regulators themselves. TBH, i think i was measuring the wrong thing-well obviously as the measured results were not sonically verifiable. How did your measured results influence sound?
Mark
Mark25 said:
If not constantly then it hardly qualifies as a load. And why compare two regulators when their loads are fundamentally different anyway?
Most three-pin series regulators have an open loop output resistance and inductance heavily dependent on the current they source. You need some 10mA, preferably 20-30mA, DC, to approach a regulator behaviour as outlined in the datasheets. Below that ...
Output capacitors are a similar story: wrong value and wrong ESR and you turn your regulator into either a noise source or an oscillator.
As recently an ASIC Youngster and a Systems Youngster over here experienced, trying to characterise a new 12 bit aerospace ADC in the vicinity of 50mV supply noise at 30kHz. In came the Weird Old Audio Guy, pointing at the bank of LM317s they had, each one looking into the bottomless pit of a 2.2uF ultra low ESR ceramic cap, and more than one trying to regulate a purely switching 5kHz/10mA load.
Replacing the 2.2uFs with a 100uF medium-ESR elcap (*) fixed the situation entirely.
(* The only elcap of which the Company accidentially still has some, as the Systems Youngsters, in their Infinite Wisdom, decided that only aforementioned ultra low ESR ceramics are worthy enough to grace Their Board Designs.)
Werner said:
Yes, that was quite sloppy on my part
My only excuse is that i built the 3x7 circuit as per the data sheet, whereas the 78/9 was "borrowed", from a commercial intergrated amp.TBH, this doesn't seem to be the way to learn about regulation anyway, i can't make an equivalent circuit of the 3x7 in my head, so can't begin to understand how to use it optimally. i have started using 15v zeners now, measures bad, but supprisingly good listening results, bass is not as powerful as the 3x7's though.
Cheers
Mark
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.