A Low Noise Balanced Input Moving Coil Preamp Using the ZTX851

[mediatechnology]

Thanks! And it works too....

This is the servo output measuring the long term (3:20) DC offset drift.

The first oscillogram is without the nut with the two ZTX851 being held together by only heatshrink.


ZTX851 Drift Heatshrink Coupling Only. 10 seconds per division. Measured at Servo correction output.

This is the ZTX851 pair in a "TO-M7" package with the devices expoxy'd together then epoxy'd into the M7 nut.


ZTX851 Drift Thermally Coupled with an M7 Nut. 10 seconds per division. Measured at Servo correction output.


A pair of ZTX851 thermally-coupled by an M7 nut.

[JR.]

Thermal mass is a good thing....

JR

[Hans]

All effort has been successfully invested in getting a very low noise input noise, but what about DC current flowing into the cart.
With 5 to 10mA collector current per ZTX851, Ibias and for that reason Ios must be in the many multi micro amps.
Isn´t that much too high and therefore over the top for an MC element?

Hans

[mediatechnology]

Thank you for joining us Hans!

I haven't measured the bias current but Ios - cartridge current - comes in at 80 nA.

In the test circuit the Ios was 80 nA.
Not all ZTX851/951 may match this well.
I expected worse.

I see moving coil preamps with input coupling capacitors in the thousands of uF as if they somehow actually block DC currents.
The leakage current flowing in both mic preamp and MC input capacitors are easily 10s of uA after they've formed for minutes.

The Nichicon UKA series for example has a leakage current which is 3 uA or 0.01*CV whichever is greater after 2 minutes.
A 2200 uF 6.3V UKA-series, typical of what you might see on a MC input, can therefore have a leakage current of 132 uA.
If the base of the input transistor is not at 0V, cart currents can easily form due to capacitor leakage alone.

[Hans]

I did some additional measurements with 56 dB gain and with the hum field nulled.

Rsource is 3R3
Rgain 1R
2X ZTX851 and NJM2068.
The OPA1612 performs identically to the NJM2068DD.

Ein of -143 dBu.
Voltage noise density 3.67 nV/√Hz
Env = 8.3 Ohms


Flat Balanced Input Moving Coil Preamp 2X ZTX851 NJM2068 Gain 56dB Rsource 3R3. O dBFS is +8 dBu.

The noise power sum of Rs+rb+rb+Rg (values equal to 3R3+2+2+1) is -143.492 dBu. (Assumes an rb of about 2 Ohms per ZTX851.)
The measured noise performance is pretty close to the theoretical being only about 1/2 dB worse.
For a 3R3 moving coil cartridge the noise figure should be about 4.5 dB.

Paralleling two additional ZTX851 - four total - provided about a 0.3 dB improvement.

I have a few questions regarding this posting and postings from a later moment.
The mentioned figures of -143dBu and 8.3 Ohm are easy to reproduce.
But then I read 3,67 nV/rtHz, a figure that I can’t reproduce ??

Looking at the graph, I assume your (unspecified) Filter BW is ca 4Hz, because the level for frequencies above 1Khz lies approx. 6dB above the -137.6 dBFS for a filter bandwidth of 1Hz.

The 80nA input Ios, an almost magic figure of -60dB below Ibias, has been achieved by adjusting the 100 Ohm pot in the collectors.
This would result in a difference between a short circuit and an open input of 25mV = (80nA * 1K * 50dB ) between TP1 and TP2. Is this how the 80 nA was measured ?

But how does this figure develop on the long run after 24 Hours or even a week, this would give a more representative impression of the input Ios .
In a later posting one can see the effect of thermal coupling of the two transistors on the Servo output.
Although a big improvement with the metal nut, nevertheless for a period of only 200sec, one can already see a steady up going drift of 25mV which could be a full 80nA extra.
But no mentioning is made whether the input was open or short circuited while recording, so it can just as well be the input Vos drift only.

Maybe I’m just worrying a bit too much, but when still using a ballpark figure for a max DC Cart current of 1% of the current flowing while playing under max modulation, this circuitry design could relatively easily pass this figure.
A few examples:
1) An Ortofon Anna produces 0.3mV peak at 5cm/sec@1Khz. Ri = 6 Ohm and minimum Rload = 10 Ohm. I max is then 3e-4/(10+6)= 19 uA. 1% of this figure is 190nA.
2) A Benz LP with Ri = 38 Ohm, minimum Rload of 500 Ohm and 0.5 mV at 5cm/sec@1kHz calculates with 1% even into 13nA.

Although I like the design very much, my feeling is that with regards to the input Ios, there is still room for improvement.

Hans

[mediatechnology]

I think it's 0.36 nV/√Hz because the decimal was moved.

The 80 nA was pure luck as I was expecting to see 10 uA or more and the measurement may be subject to errors.
Because I saw 80 nA on one I didn't get convinced they would all be that good.
In fact it is still an area of concern...

I measured the Vos across the inputs to indirectly measure current.
The shunt resistance in the 200 uA range of my DVM is about 1K so I could not read current through Rsource directly.
(The least-significant digit - the lowest I can directly measure - is 10 nA.)

I can read down to 10 uV. (Also the LSD so not with any accuracy.)
Reading the Vos developed across the 3R3 (or a 10R) was below range.
So the Ios had to be <3 uA or it would have at least made a blip on the DVM.
I could easily see 10 uA Ios develop at 33 uV.

Again none of this is with any high degree of accuracy but at least it establishes what order of magnitude the currents were.
To get any measurement at all I recall measuring the Vos developed across the 499RX2 bias resistors to get an estimate.
An 80 nA Ios across 1K is 80 uV.

Although I like the design very much, my feeling is that with regards to the input Ios, there is still room for improvement.

Thank you very much. And I agree.
Never said I was finished though.

In fact I wrote in my first post...

No effort has been made to address the DC performance - output offset is allowed to develop.

Though I did add common mode servos to pin the output Vos the input stage offsets are still allowed to develop.
But at least the pair of devices in the nut give me a less wobbly thermal pair to work with.
After posting this thread I went off to do other things.

So far the exercise is to see if the ZTX851/951 are promising enough in the noise department.
I think they are.
They may not be perfect but the do have low rbb and seem to have low 1/f noise.

[Hans]

To measure the Offset current accurately, you only have to measure between TP1 and TP2 with 1) the input shorted and 2) with the input open, i.e. only having the 2x 499 between the inputs.
Subtract the two voltages, divide by 158000 (=50dB*499)and you have the Offset current in nano-amps.
Thus 15.8mV difference means 100nA offset, etc. Cart current when connected will be exactly half this current or 50nA.
The only thing that adds to a measuring error is the galvanic effect of having different materials and also having different temperatures..
When you have gold plated input sockets and a rhodium plated plug, this already could act as a thermocouple, so try to use the same materials where possible.
When touching the plug with your warm hands, temp may increase with only a fraction of a degree, more than enough to see that the voltage between TP1 and TP2 needs some time to settle after short circuiting, showing how sensitive measurements are at this nanovolt level.
Even beter than using a short circuit plug is to use a switch preferably with gold plated contacts, because you won't touch the contacts when switching which prevents them to change their temp.

The other thing you mention is 1/f noise.
Even for a Riaa amp, although the low frequencies are amplified quite a bit, this effect never plays a role since an un-weighted (Riaa) noise spectrum does in no way reflect what we hear.
The best weighting is ITU468 and A-weighting coming as second.
When applying one of the two weighting curves, you will notice that even the more pronounced 1/f noise of Fet's hardly influence weighted noise negatively because frequencies below 1 kHz are firmly suppressed.

Hans

[mediatechnology]

I made the Ios measurement pretty quickly.
At the time my expectation was that it would be high enough to measure directly.
Next time I have it powered on the bench I'll measure it through the output.

[Hans]

I made the Ios measurement pretty quickly.
At the time my expectation was that it would be high enough to measure directly.
Next time I have it powered on the bench I'll measure it through the output.

O.K.
And please see my contribution as a way to contribute in a positive way.

Hans

[mediatechnology]

And please see my contribution as a way to contribute in a positive way.

I do, and appreciate it!