Hello, I am a new member here and this is my first thread. It’s about RIAA amp topology...
If RIAA amp is designed using feedback principle, why not use shunt feedback?
The reason is the noise that is produced by 47K resistor being in the signal path, they say.
But in my experience that noise is not obtrusive at all and stays well below the
surface noise of the record. In my amp (using op amps with FET input, see attachment)
there are two inverting stages so that the first one is producing RIAA curve with
dc amplification of 40dB and the second one acting as a buffer giving extra
amplification of 20dB. It is well known that shunt feedback handles transients
better than series feedback and in my experience it can be heard too. In a way
shunt feedback concept is half passive by keeping the op amp out of the signal
path as much as possible.
...MK
If RIAA amp is designed using feedback principle, why not use shunt feedback?
The reason is the noise that is produced by 47K resistor being in the signal path, they say.
But in my experience that noise is not obtrusive at all and stays well below the
surface noise of the record. In my amp (using op amps with FET input, see attachment)
there are two inverting stages so that the first one is producing RIAA curve with
dc amplification of 40dB and the second one acting as a buffer giving extra
amplification of 20dB. It is well known that shunt feedback handles transients
better than series feedback and in my experience it can be heard too. In a way
shunt feedback concept is half passive by keeping the op amp out of the signal
path as much as possible.
...MK
Moved to analog source.this includes a useful discussion:
http://www.angelfire.com/az3/dimitri/images/riaa.pdf
this one looks like it implements one of the options discussed in the previous.
http://www.redcircuits.com/Page154.htm
http://www.angelfire.com/az3/dimitri/images/riaa.pdf
this one looks like it implements one of the options discussed in the previous.
http://www.redcircuits.com/Page154.htm
Last edited:
http://www.redcircuits.com/Page154.htm
It seems to me that two of the RIAA time constants in this design are incorrect.
They are 404µs and 3900µs as they should be 318µs and 3180µs. That is quite a lot
of deviation…
R6xC4 = 3900µs
(R3 + R4 + R5)/(R3 + R4 + R5 + R6)xR6xC4 = 404µs.
And this amp inverts the phono signal. I wonder, is that audible??
It seems to me that two of the RIAA time constants in this design are incorrect.
They are 404µs and 3900µs as they should be 318µs and 3180µs. That is quite a lot
of deviation…
R6xC4 = 3900µs
(R3 + R4 + R5)/(R3 + R4 + R5 + R6)xR6xC4 = 404µs.
And this amp inverts the phono signal. I wonder, is that audible??
47k produces about 4uV of thermal noise, although that is then reduced by the RIAA equalisation. Better without it, in my view.
Note that RIAA time constants and circuits are connected in a non-intuitive way, as different CR networks interact.
I am unsure what you mean by this. The op amp is fully in 'the signal path'. The feedback is not 'passive'.acmn said:
Note that RIAA time constants and circuits are connected in a non-intuitive way, as different CR networks interact.
hi all
the topolagy of shunt feedback RIAA, preceeded with a series amp was used in some top level designs. Meridian used it and John Linsley Hood was famous in championing this sort of circuit.
link to jllh designs
A Paul Kemble web page - John Linsley Hood preamp designs.
A Paul Kemble web page - John Linsley Hood preamp designs.
the all descreate version i have built quite a few of these, sounds very good.
this design even has a moving coil setting (although i have not tried it with MC cart myself)
the topolagy of shunt feedback RIAA, preceeded with a series amp was used in some top level designs. Meridian used it and John Linsley Hood was famous in championing this sort of circuit.
link to jllh designs
A Paul Kemble web page - John Linsley Hood preamp designs.
A Paul Kemble web page - John Linsley Hood preamp designs.
the all descreate version i have built quite a few of these, sounds very good.
this design even has a moving coil setting (although i have not tried it with MC cart myself)
Hi DF96.
What I meant is that the input current from the cartridge bypasses the first op amp.
So it can be described as being out of the picture, sort of. As I said there are views
that shunt feedback amp handles transients better than series feedback amp.
And if that thermal noise is covered by vinyl surface noise, it can be accepted, I think.
Anyway my amp sounds really good.
Are you saying that my calculations of RIAA constants are incorrect?
What I meant is that the input current from the cartridge bypasses the first op amp.
So it can be described as being out of the picture, sort of. As I said there are views
that shunt feedback amp handles transients better than series feedback amp.
And if that thermal noise is covered by vinyl surface noise, it can be accepted, I think.
Anyway my amp sounds really good.
Are you saying that my calculations of RIAA constants are incorrect?
With the combined RIAA network you are using, there will be a 75µS time constant and the other will be something other than what you might expect. I am not at home right now so I can't check the papers I have with the math, but I remember something like C1 will be 3.6X the value of C2 and R1 will be about 11.65 or so times the value of R2. This is because the two RC time constants will interact to produce the 318µS constant. Your 75µS values come out to 74.5µS, so your values appear to be very close to correct. I'll check for sure later on when I get home.
As far as shunt feedback goes, its biggest disadvantage is higher noise over series feedback. If we take the source impedance of a MM cartridge to be 10Kohms (of course this depends on frequency, cartridge model, etc.), then series feedback has the source impedance of 10K in parallel with 47K, or about 8.2K. In the shunt feedback example, the source impedance is the two resistors in series, or 57K. The source impedance is 6.95 times that of the series feedback circuit, so the noise voltage is also 6.95 times higher, or about 16.8 dB higher.
Shunt feedback does, however, have the advantages of the gain being able to go lower than unity, so there does not end up being a rising response at high frequencies, and it avoids common-mode distortion in the op amp.
My own opinion is to go for the lower noise of the series feedback circuit, although I will admit that surface noise from the record will dominate in either case.
-Erich
As far as shunt feedback goes, its biggest disadvantage is higher noise over series feedback. If we take the source impedance of a MM cartridge to be 10Kohms (of course this depends on frequency, cartridge model, etc.), then series feedback has the source impedance of 10K in parallel with 47K, or about 8.2K. In the shunt feedback example, the source impedance is the two resistors in series, or 57K. The source impedance is 6.95 times that of the series feedback circuit, so the noise voltage is also 6.95 times higher, or about 16.8 dB higher.
Shunt feedback does, however, have the advantages of the gain being able to go lower than unity, so there does not end up being a rising response at high frequencies, and it avoids common-mode distortion in the op amp.
My own opinion is to go for the lower noise of the series feedback circuit, although I will admit that surface noise from the record will dominate in either case.
-Erich
Played with shunt feedback nearly 30 years ago, when 5534 was pretty much as good as it got. Loved the sound, not because of any inherent advantage of shunt feedback, but because the 5534 sounded horrid with series feedback and much better with shunt.
Equalising impedances on both inputs made it sound even better, but at a further noise penalty.
Still, 47k of a noise resistance at input is just nonsense. A buffer is absolutely essential preceding the opamp stage. This brings the value of the series resistor to whatever the buffer is capable of driving, easily below 1k. As an added bonus the 75uS can now be made passive between the buffer and the opamp
Equalising impedances on both inputs made it sound even better, but at a further noise penalty.
Still, 47k of a noise resistance at input is just nonsense. A buffer is absolutely essential preceding the opamp stage. This brings the value of the series resistor to whatever the buffer is capable of driving, easily below 1k. As an added bonus the 75uS can now be made passive between the buffer and the opamp
A long time ago, there were 10 kOhm Grado MM cartridges.
I use one in a modified Linsley-Hood's circuit called 'Super -Liniac' using RIAA shunt feedback. The voltage gain consisted of only one BJT cascoded by an FET, it was followed by a buffer which could be made either of a push-pull of CFPs or of a Darlington pair.
No hiss, no annoying noise at all bar it was sensitive to a local radio FM emitter which was located at about 100 meters from my home.
Maybe using a 10-12 kOhm input resistor as in Bob Cordell 'Vinytrack' preamp and an adequate RIAA network could lead to a nice simple phono preamp having shunt feedback.
I use one in a modified Linsley-Hood's circuit called 'Super -Liniac' using RIAA shunt feedback. The voltage gain consisted of only one BJT cascoded by an FET, it was followed by a buffer which could be made either of a push-pull of CFPs or of a Darlington pair.
No hiss, no annoying noise at all bar it was sensitive to a local radio FM emitter which was located at about 100 meters from my home.
Maybe using a 10-12 kOhm input resistor as in Bob Cordell 'Vinytrack' preamp and an adequate RIAA network could lead to a nice simple phono preamp having shunt feedback.
Both shunt and series feedback circuits do this. In one the MM current flows to ground through a 47k shunt resistor. In the other the MM current flows through the 47k series resistor and then through the feedback resistor to the opamp output. So in both cases almost all the MM current does not flow into the opamp input. In one case the MM current does flow into the opamp output.acmn said:
The two configurations give different HF asymptotic behaviour so will handle fast transients differently. Series is more accurate, but shunt can be corrected by adding an extra HF rolloff and the result is then the same. Passive equalisation avoids the problem.
I have not checked your RIAA calculations. Find the famous Lipschitz paper and it will tell you all you need to know.
Vice versa, I think.
Series feedback is the one with the extra zero built-in.Of course, some folks think there should be an extra zero to compensate for the unwanted but unavoidable HF roll-off of the disc-cutting process. That's a whole other can of worms though.
OK, I checked my papers and found the RIAA component values correspond to 75 and 3180µS time constants, but the two R and C values need to be in a specific ratio to get the 318µS time constant right. If R1 and C1 are the 75µS network and R2 and C2 are the 3180µS network, then C2 needs to be 3.6 times the value of C1, and R2 needs to be 11.78 times that of R1. The OP's schematic has 202pF X 369K = 74.54µS and 736pF X 4.33M = 3186.9µS. This is pretty close to 75 and 3180µS. C2:C1 = 736pF/202pF = 3.64, and 4.33M/369K = 11.73, so the ratios are also close. The EQ should be fairly accurate.
I got the noise calculation wrong previously because I forgot the square root in the noise equation. The ratio of source impedances was 6.95, so sqrt 6.95 = 2.64, and the noise difference is 20log(2.64) = 8.4dB. Douglas Self in the "Small Signal Audio Design" book gives a difference of 14dB between series and shunt feedback, but does not explain why. I imagine it is mostly due to the higher source impedance in the shunt feedback configuration.
I got the noise calculation wrong previously because I forgot the square root in the noise equation. The ratio of source impedances was 6.95, so sqrt 6.95 = 2.64, and the noise difference is 20log(2.64) = 8.4dB. Douglas Self in the "Small Signal Audio Design" book gives a difference of 14dB between series and shunt feedback, but does not explain why. I imagine it is mostly due to the higher source impedance in the shunt feedback configuration.
I have not read all of Andrew T's references which look pretty comprehensive but I am very familiar with with the work of John Linsley Hood (in burbeck's schematics). His first published pre-amp around 1970 was shunt feedback and he continued to be a fan with various developments over the next couple of decades. Without going through all his comments I think the following is a fair summary of what he saw as the advantages.
(i) it gave accuracy to the RIAA curve in the higher frequencies compared to series f/b. This was both theoretically true and borne out by square wave responses on CRO pictures. (some series f/b schematics e.g Douglas Self then added some extra RC components to make their compensation more accurate) (Sorry hadn't read DF96's posts when I did this.)
(ii) Noise. He was frequently criticized (early '70's Wireless World) about the noise of the shunt network when compared to series f/b. His reply was that it was low enough when compared to record surface noise. In ETI in the '80's he made the observation that the "quality" of the noise was more acceptable in that the shunt f/b was more like a "rustle" than the "hiss" of series networks. His comments about the effect of the RIAA network on the frequency of the noise were interesting too but I can't recall them now. He also did listening tests and reported that most people could hear the difference and most preferred the shunt arrangement. (ETI around 1984 or the next version he did)
(iii) Distortion. He published a small article (WW) showing that the shunt arrangement had lower distortion in the higher frequencies. Since confirmed by other published work such as Geisberts (not sure of spelling) in "Elector".
(iv) He said that impulse noise was audibly more acceptable with shunt networks.
Now as the years went on he adopted systems where the gain was split and he would buffer the input with a "flat" series f/b opamp prior to the frequency changing network.
I don't want to get into a subjectivist/objectivist debate but I built one of his pre-amps in 1977 and I can recall that it just sounded far better than anything I had heard before......or since.
Cheers,
Jonathan
(i) it gave accuracy to the RIAA curve in the higher frequencies compared to series f/b. This was both theoretically true and borne out by square wave responses on CRO pictures. (some series f/b schematics e.g Douglas Self then added some extra RC components to make their compensation more accurate) (Sorry hadn't read DF96's posts when I did this.)
(ii) Noise. He was frequently criticized (early '70's Wireless World) about the noise of the shunt network when compared to series f/b. His reply was that it was low enough when compared to record surface noise. In ETI in the '80's he made the observation that the "quality" of the noise was more acceptable in that the shunt f/b was more like a "rustle" than the "hiss" of series networks. His comments about the effect of the RIAA network on the frequency of the noise were interesting too but I can't recall them now. He also did listening tests and reported that most people could hear the difference and most preferred the shunt arrangement. (ETI around 1984 or the next version he did)
(iii) Distortion. He published a small article (WW) showing that the shunt arrangement had lower distortion in the higher frequencies. Since confirmed by other published work such as Geisberts (not sure of spelling) in "Elector".
(iv) He said that impulse noise was audibly more acceptable with shunt networks.
Now as the years went on he adopted systems where the gain was split and he would buffer the input with a "flat" series f/b opamp prior to the frequency changing network.
I don't want to get into a subjectivist/objectivist debate but I built one of his pre-amps in 1977 and I can recall that it just sounded far better than anything I had heard before......or since.
Cheers,
Jonathan
Last edited:
Hi loepke72...again.
I see you have read my document thoroughly, thanks. According to my calculations
the exact values of time constants are 3186.88µs, 74.5380µs and 318.942.
R1=4.33M
R2=369K
C1=736pF
C2=202pF
R1xC1=3186.88µs (deviation to 3180µs about 0.2%)
R2xC2=74.5380µs (deviation to 75µs about 0.6%)
R1xR2/(R1+R2)x(C1+C2)=318.942µs (deviation to 318µs about 0.3%).
What these constants are in reality, depends on component tolerances (1%
in this case).
Regards...MK
I see you have read my document thoroughly, thanks. According to my calculations
the exact values of time constants are 3186.88µs, 74.5380µs and 318.942.
R1=4.33M
R2=369K
C1=736pF
C2=202pF
R1xC1=3186.88µs (deviation to 3180µs about 0.2%)
R2xC2=74.5380µs (deviation to 75µs about 0.6%)
R1xR2/(R1+R2)x(C1+C2)=318.942µs (deviation to 318µs about 0.3%).
What these constants are in reality, depends on component tolerances (1%
in this case).
Regards...MK
When the RIAA equalisation is achieved by a negative feedback network without an additionnal low pass RC filtering at the output, a series preamp circuit has a gain of 1 at high frequencies.
As a shunt circuit has a gain falling to null. This should imply that the high frequencies input level infinitely increases with increasing frequency, which can't be the case in reality.
If square waves are injected at the input of a phono preamp, they must be somewhat low-passed to reflect the facts that the record groove does not contain very high frequencies and that the phono cartridge is itself a frequency and slew limited device.
Hi DF96…again.
You are right, input current bypasses the first op amp in the series configuration too
and in fact it goes straight back to the cartridge trough that 47K resistor!
But the real beauty of the shunt feedback configuration is that the op amp’s input lines
are tightly grounded (minus line virtually grounded) all the time and in that sense
you can say that the op amp stays out of the picture. But of course it must output the
current that is needed to push the signal forward and in that sense it is very much present.
Series feedback configuration suffers from the fact those input lines are both
“reading” the voltage in the 47K resistor (i.e. voltage coming from the cartridge) which
produces common mode distortion with inferior transient handling.
…Regards MK.
You are right, input current bypasses the first op amp in the series configuration too
and in fact it goes straight back to the cartridge trough that 47K resistor!
But the real beauty of the shunt feedback configuration is that the op amp’s input lines
are tightly grounded (minus line virtually grounded) all the time and in that sense
you can say that the op amp stays out of the picture. But of course it must output the
current that is needed to push the signal forward and in that sense it is very much present.
Series feedback configuration suffers from the fact those input lines are both
“reading” the voltage in the 47K resistor (i.e. voltage coming from the cartridge) which
produces common mode distortion with inferior transient handling.
…Regards MK.