Phono Stage w/split passive RIAA, Instrumentation input

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Here is my stab at a phono preamp, with differential MM and MC input stages, passive split RIAA, and twin secondary+output stages, inspired by the Jensen twin servo mic preamp.

I'm building in an industrial Viking RP120, a 60s-era tape preamp, with four XLR inputs and two out, which had four octal-socket input modules for balanced or unbalanced line/mic, that I am now using two of for the MC stage (could be either active, as designed, or transformer), and two for the RIAA modules. So far, these modules are all I have actually built, so scrutiny is only welcome everywhere else. :)

Input stages use an instrumentation amplifier topology: 3 op-amps (NE5532/4) for MM, and INA217 (or SSM2017) for MC. I'm using a mil-spec sealed relay with gold contacts for switching between the MC and MM stages' outputs, prior to DC filtering since I figure the more electrons flowing through the relay the better, at the expense of a potential thump when switching. RIAA EQ is done in two stages, and the values of the resistors in the DC filters are accounted for in each.

I'm not 100% sure of the input loading. I wanted a load across the input terminals, and I know I needed paths to ground (float->bal), so I did both equally. Is there a method to use here? Thanks in advance for taking a look!


  • PhonoPreamp_RP120.gif
    48.5 KB · Views: 1,488
I have built a similar one with LT1028 for low impedance MC (60dB)
This alowd for much smaller resistors on the Imput than SSM2017
It was 3/4 years ago and I will have to dig up the SCH
cartridge loading is just a reistor / capacitor in parallel to carttridge
No need to go to ground as you can keep cartridge floating and the third op amp is grounded
This configuration of the INstumentation Amplifier that let you change gain with just one resistor easy peasy to get total gains from 40 to 60dB and you can use paralel combinations to switch on and make signal path simpler
so MC MM is done at imput and passive RIAA s are not disturbed


  • 03046.png
    4.6 KB · Views: 1,207

signal- and noise-level-wise a symmetric input stage using an instrumentation amp is absolutely sufficient to accommodate for lowoutput MCs up to highoutput MMs. But if You intend to make switchtable inputs, I think it´s a good idea to have some gain first, since most relays are not even specced below 100µV. Also some relays like the Omron G6K react very sensitive to coil supply and layout and may show high crosstalk. I´d rather parallel the relay contacts and give each channel its own relay, than sharing the relay between the channels.
Since most MMs are not balanced sources, I´d choose an unbalanced MM input stage using a lownoise JFET OPamp like the AD745 that offers a unique combination of low voltage and current noise.
For the balanced MC-Input You may also look at the INA103 and INA163 of BB/TI. The ADI SSM2017 requires quite high input bias currents, app. double that of the BB/TI parts, is specced for >2kOhm output loads and requires larger Gain setting resistors. The BB/TI work very especially well at gains around 100x (+40dB). Due to the high gains, one must take into account considerable amounts of output offset voltage which may reduce the headroom. A active DC-servo, feeding into the ref-pin can take care of this and would render the additional LF-Filter between stages 2 and 3 obsolete.
If the -3dB freq is designed to be 20Hz instead of the typical 100mHz You meet the RIAA/IEC1963 requirements at the same.
I´d also have a lowpass after the input gain stage, but just the passive 75µ/2120Hz Filter. I´d then choose a gain stage with the 3180/318µ equalization in the feedback path. This would require 20dB less gain.
Imho any possible advantage of a pure passive equalization will be completely swamped by the disadvantages of +20dB higher required gain and more required parts As Opamp I´d choose a FET-type also, since it is more immune against the frequency dependant source impedance and impedance differences between the OPamps inputs. A OPA134 would come to mind. If You use a Dual Opamp like the OPA2134 You can use the second OPamp as the DC-servo for the input stage. The second (and 3rd.) stage don´t require varying gain. The gain setting of the input stage will be all You need. You wouldn´t need a Buffer stage after the OPamp but You could add a buffer, either on its own, or inclosed to the OPamps feedback loop. If balanced outputs are wished for a dedicated balanced line driver like the SSM2142 or DRV134 may be used.
All this put together resukts in a short, straightforward signal path with a low parts count, high flexibility and excellent results.
Examples for such phono stages can be found here at DIYaudio and on my website


ps. btw. You might rethink Your Pin1 connection.

I kindly disagree ;)
The FDW is kind of a 'questionable' overkill in the power supply. Stuffing a supply with tons of caps doesn't guarantee for best results. Imho it is rather wasted money and space.
The design of the Input stages is off of optimum. I don't know why one would want an OPamp for a balanced input stage if there are enough high quality INAs around? Even built from 3 OPamps or as with a discrete frontend it'll be very hard to beat the INAs.
A OPAmp input is also not balanced at all. The inverting and noninverting ins don't behave the same and gain setting will be problematic. INAs are much more conveniant not only in this regard.
Configured as they are in the FDW the high impedances add noise, and probabely are the input resistors the dominant noise source here. A low current noise OPamp would be needed, rather a FET-type than a bipolar.
The layout also leaves room for improvements. The supply current paths layout is highly important, the more so the 'faster' the OPamps are. Surprisingly each OPamp is just HF blocked, but no local Joule reservoir. Instead there are long inductive traces. Its sad to see so many precious devices beeing compromised in their possible performance by inferior implementation. :(
Besides, its claimed to be a low parts count design, but it certainly isn't.

You may want to read the thread "Schematic for Pro-ject phono box" from #62 for a INA based balanced Input stage.


I kindly disagree ;).... I don't know why one would want an OPamp for a balanced input stage if there are enough high quality INAs around? Even built from 3 OPamps or as with a discrete frontend it'll be very hard to beat the INAs.

Just have a look at the resistors values and as it seems that you know quite a fair bit about things do your own calculations for Jhonson noise.

I had 3 X LT1028 with resistors as low as 30 Homs and could have gone lower.

But this is just not what you want to ear so please do ignore it
Have a nice day again...;)

I kindly disagree ;)
A OPAmp input is also not balanced at all. The inverting and noninverting ins don't behave the same and gain setting will be problematic. Calvin

Sems to me that this go nought to do whit litle SCH I have posted
and the reason I posted that is that it does show the connection to ground on one resistor, so cartridge can be floating...and the loading part goes just in parralel as standard, which I think is just an answer to what was asked but if one realy realy need to argue about....

Why is a MM not balanced?
it is just a bit of vire wound up in to a coil same as MC
one got coil moving against a magnet other is a magnet moving against a coil

Same do have tag to Ground that can be cut by the way but the bit of vire wound up does not change so where did you read about it:D

yes, #5 commented only about the FDW, and on a strictly technical level, as I may add. :rolleyes:

If You want to use or build an instrumentation amp, the 3-OPamp structure is the most consequent and the LT1028 could be a good choice for a lownoise design. I'd probabely just choose a different OPamp for the 3.rd, the differential stage. Still though the Q remains, what could be the advantage of a 3-OPamp implementation against the integrated solution in form of the above named devices from ADI or TI? I haven't found any that is of major importance ;)

Why is a MM not balanced?
I didn't say that! The signal generating principle is of a balanced kind, same as for MCs. That 'most MMs are no balanced sources' (cited correctly) though is due to the ground tag you mentioned yourself.

LT1028 is a low voltage noise and on paper it looks good till one find out about the current noise it has

On a pure technical point using an complete INA chip limits the choice of values for the resistors in the circuit
INA163 use 3 K and 6K resistors for example, with LT1028 I was using 30 and 120 Hom's
this will also warrant the use of different values for the gain resistor
Thanks for all the tips, insights, etc.

Surface-mount fears aside, I like the idea of using something like the INA163 for the MM stage. In general it seems harder to find instrumentation amps that have both low current and voltage noise. I went with the 3-op-amp method largely because I have parts on hand, and I've used the NE5532 enough that I have a good feel for what it can do; it seemed pretty well suited to MM impedance in terms of noise specs. I'm not married to it, however, and it sounds like there are some other contenders.

On that note, I noticed that the LT1028 is gain of -1 stable, while the LT1128 is +1 stable. For the unity differential configuration (op amp #3 in the in-amp), which is preferable?

I'm now working on the pin-1 issue, and I appreciate the suggestion to rethink it. :) Given the inconsistencies in other gear I have, and cables/connectors, I'm thinking of just adding chassis/signal/float switches on the inputs' and outputs' #1 pins. Switches are cheap and allow for easy experimentation with a given configuration.

while it is clear to me what the 3k and 6k resistor positions are, Im not sure where You put the 30 and 120Ohms in Your design. The only position I´d expect such low values would be for the Rg. The LT1028 would be seriously loaded down if it had to drive a following load plus a lowimpedance network.
The Rg is also the dominant noise source resistor. For the INA103/163 the useful gain range is between 20 and 1.000, giving Rg values between 316 and 6 Ohms. So the noise contribution from Rg fortunately gets the lower the higher the chosen gain is. The 3k feedback resistors hardly effect noise at all in presence of such a low Rg.
Since the differential stage just buffers and sums its input signals its noisegain is negligible too. Higher resistor values only play a minor role here.

Easy enoug to see if one bothers to look at data sheets they are called R in litle dwg I posted

And if you listen to LT1028 it does like to run hot
How impedance network take any part in it as design is for split passive and has 2 more gain stages as in post 1 those wuld be the triangular shaped things in the drawing.

Here is data sheet and hey presto sch is on page 1
Last edited:

if it were that clear, I sure wouldn't have asked. :rolleyes:
First because such low impedance values are far off of the usual value range for the feedback resistors and are rather typical for Rg.
Such low feedback-R values would ask for Rg values of 1ohm or even less.
How do you deal with gain variations due to the increased sensitivity to tolerances, temperature and ageing and gain differences between the two stereo channels? Have you had the opportunity to verify the behaviour of the LTs regarding noise, THD and headroom under those heavy loading conditions?

Second because there are six resistor positions -or better two times three- for two resistor values, allowing for more than just one single combination.
The differential stage doesn't need to be a unity gain buffer as shown in the simplified schematics. It could as well be a differential gain stage. (remark: I refer to 'differential stage' as the third OPamp within the three-OPamp instrumentation amplifier schematic)

How impedance network take any part in it as design is for split passive and has 2 more gain stages as in post 1 those wuld be the triangular shaped things in the drawing
??? And the message is??? What?


ps: OMG, have been building balanced in phono stages for 20years now and never realized that triangles were gain stages or OPamps, nor that there are datasheets for info :rolleyes::rolleyes: I must be a truely dumb nut. thanks for clearing that one.
Well last post for me here
From the data sheet you could clearly see the resistors values used in the INA163
there are 4 X 6K and 2 X 3 K resistors and this IMO is limiting design choices

It did sound prety good with a prety 3 dimentional stage and good tollerance to scratches and such

Wow 20 years design experience I beter let you carry on as I think wagneric has the answer he required about loading

Yes, the DS states the resistor values as 3k and 6k.
The first value allows to calculate the gain for a given Rg or vice versa the required Rg for a given gain. The Q is, which resistor contributes how much to the output noise?
For noise calculations the 3k are in parallel to Rg at the negative input. For a gain of 100, Rg is 61Ohms, for 1000 its 6.006Ohms. So the 3k ´in parallel´ doesn´t affect the ´equivalent source resistor´ value (59.78Ohms and 5.994) in a noticable order. The Pickup impedance in parallel to the input loading resistor defines the equivalent source impedance for the positive input.
A AT33PTG has a coil dc resistance of 17Ohms, giving 14.52ohms with 100Ohm loading. Same result here....Pickup and Rg are the dominating factors of noise calculation, not the 100Ohm Input loading nor the 3k feedback resistor.
As explained in #12 the 6k resistors of the differential buffer also contribute negligible to noise, since thats gain is much lower than the input stage´s.
In fact the noise penalty of typical resistor values compared to the extreme low values in the LT1028 INA is < 2dB(!). The extreme low resistor values increase tolerance sensitivity and heat. THD is increased grossly (btw. there´s a graphic in LT1028s DS of THD vers. Frequency, vers. Load resistance), which may be the source of the sound You like about the hot running LT1028. ;)

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.