Prema 6048 teardown

[Kleinstein]

You are right, the PLL range adjustment changes the loop gain.
The way the PLL adjustment procedure is, they call for a certain VCO set voltage and this way also get an about fixed loop gain. The trimmer would mainly make up for variations in the VCO. Chances are all the trimmers would be set to rather similar values anyway. Changing to the new setup with a fixed relatively large R1 and adjust R2 (from the 4046 datasheet) would be a step to a different (e.g. a factor 30) loop gain that is than fixed. If it works the smaller VCO range can be good and this could be a change that is well worth it.
If the frequency noise can be seen with the scope, this is a bad sign or a very good scope. Especially some FM modulation (e.g. from hum effecting the PLL) could be an issue and effect the linearity.

In one photo I have seen U6 (the PLL loop filter) changes to an OP07. This could be change worth while testing. One is less LF noise and the other point is having some bias current, so that the phase comparator output is not in the dead zone. We don't know the details of this output of the prema chip, but chance are it is similar to the 2nd phase comperator in the 4046 and thus giving short pulses of either polarity.

Are the INL measurements shown with the origial PLL circuit or already with the modified PLL circuit ?

[dietert1]

The frequency/phase comparator output of the BK7 produces positive and negative pulses (2x open collector). During lock there is a small positive pulse of stable length, so that was well done. The Lecroy WR104 i am using was reporting a frequency RMS of 5.3 KHz, now it became 1.5 KHz. This is over short records of about 10 pulses = 5 usec. Another mod to try is moving C56 closer to the HC4046. The VCO mods did not resolve the nonlinearity issue.
That issue is in the interface between the amplifier and the ADC. We have seen that in one position of the polarity relay there was a hum issue, which means extra impedance where it isn't expected. My suspicion is with U11 and CR18/CR19 on the amplifier board. I have seen opamps with a dead zone near zero output current. Will do more tests.

Regards, Dieter

[Kleinstein]

I know that OP-amps may have some cross over distortion with near zero output current. However the ICL7650 should have very little of this due to it's high open loop gain, typical of AZ amplifiers.
U11 should have even less such effect as it is in a local feedback loop and inside the loop of U10.  So I think we can exclude the distortion from U11.
An error due to the amplifier should also be about the same for the 2 V and 20 V range. So this can not explain the observed INL near zero that is more similar with respect to the ADC counts.

Getting more hum in one polarity setting than with the other is not such a surprise for me. The polartiy relay changes which current path counts to the ADC input an which not. So it can effect the hum, especially if the scope is connected to the ADC.

For the VCO I see 3 paths to pick up interference: one is via the VCO input signal, the 2nd is via the supply and a 3rd is the reference switching part effecting the 50 Hz input signal to the PLL. The last would especially be an issue if the 50 Hz input signal to the PLL is switching right at the end of the reference pulse - which would be kind of a design blunder.

With the reduced VCO gain the sensitivty to the supply could well be larger than to the VCO input. A modulation in the frequency by 1.5 kHz looks way to high to give a good ADC. If it is white noise this would cause quite some noise. If this is hum, a pure 50 Hz modulation may not be that bad as the effect on both reference pulses per mains period would about compensate. Still the mains waveform is no longer a pure sine and especially 100 Hz hum could be bad and causs an INL error, though not so much the more localized to the extremes of the range type of error seen.
I think chances are that without the scope atached the VCO signal would show less frequency variations.

It is still a bit odd to me what is so special and bad if the reference pulses are getting shorter than some 15 µs or the input voltage getting below some 5 or 50 mV.
Maybe it is the R29*C90 time constant (6.8 µs), as the input amplifier is relatively slow with it's speed set by r30 and C22. So while the input amplifer in principle is a kind of current source, it could not compensate fully for a change in the integrator input voltage. For a test one could try something like doubling C90 and see if the problematic range gets larger

[dietert1]

Here i have some ADC signals of the Prema 6048 at 0 V. Channel 1 is the integrator input, Channel 2 is the switched reference source and F1 is a 100 trace average over the input signal of R29. That's the voltage over C90, which is meant to cover the missing bandwidth gap. It receives error voltage of about 7 mV from the ADC operation. 7 mV times 10 usec makes 7 uV in 10 msec, so that's about the right magnitude.

I increased R29 to 1 KOhm. This is possible, as the amplifier is an active current source. It will compensate for the extra voltage drop of up to 0.4 V. At the same time i added a 680 pF capacitor over the integrator input. In the HP 3456A ADC this capacitor is 470 pF. The "modded" scope dump shows the result. The error voltage on C90 and the high frequency spikes on the integrator input disappeared. Instead we can see a damped oscillation of the Integrator, caused by the 680 pF capacitive load. Just running another linearity test.

Regards, Dieter

[Kleinstein]

The signal seen at C90 looks a little that it could cause the INL problem. The amplifier is relatively slow and can not fully compensate for the changing voltage at C90. In addition there is a slight chance it may show some overshoot / ringing too. Per se ringing would not be that bad as it may not add area under the curve to add up to an error.

A RC combination in addition or instead of just a capacitors (e.g. the 470 pF) at the integrator input would likely dampen the ringing. The original R29+C90 may have done that job in addition to keeping the peak away from the input amplifier.

There is also the chance that the recovery / ringing at the integrator input could change with the load current to the OP77. With very low input it could be in a kind of dead zone and there react a little slower.

[dietert1]

The last mods did not resolve the nonlinearity issue. The nonlinearity is small. At zero input voltage in the 20 V range the ADC is measuring essentially the 118 mV offset. An error of about 20 uV is a 1.7E-4 effect, difficult to see on the scope. I think soon i will connect another DVM.
For the time being i looked at the switched reference. I added a p-channel JFET J175 and a 4.7 KOhm resistor to have a near constant current of 1.4 mA out of the reference buffer. The JFET gate is connected to the same signal as the four DG308 control inputs. It turns off when the DG308 switches turn on. I'm injecting the dummy current near the non-inverting input of the integrator.
One may also try another switch scheme with 2:1 muxes, replacing the two 1 MOhm resistors R53 R54 by switches. They were meant to reduce leakage. It works, but only after about 20 or 30 usec. The spare third MUX can replace the JFET.

Regards, DIeter

[dietert1]

Meanwhile i used the MAX5719 voltage source with two resistors of 208K to feed test current directly into the integrator current node. Thus i confirmed that the nonlinearity issue is in the integrator board and not in the amplifier/front end.
Next i will try to measure the current the time division DAC feeds into the integrator. I want to find out whether the nonlinearity stems from the DAC or the integrator (BK7 microcircuit).
There is an improved version DG308B of the DG308A with charge injection reduced from 10 pC to 1 pC. Also i ordered some DG309B to replace the two 1 MOhm resistors by switches. R63 in the DAC sense path causes a 30 Ohm / 1 MOhm = 30 ppm offset.
In another test i found that the scanner appears in the front panel display after closing the solder bridge labeled "SC" on the CPU board. Already ordered parts to make the scanner.

Regards, Dieter

[dietert1]

When making the scanner i tried to follow the original schematic as shown in the manual. If found that a CD4094 serves to replace the Prema C32 microcircuit. Extra features of the C32 are 4 more bits and relay driver outputs, both unused in the scanner.
In contrast to what they describe in the text of the 6047/6048 manual they did not use mechanically latching relays for the scanner. In the other boards relays are labeled S4L-6V and S2L-6V while in the scanner schematic they are labeled S4-12V. The drive circuit looks different, too. A CD4049 won't drive a latching relay due to its insufficient pull-high strength. Needs more work..

Regards, Dieter

[Kleinstein]

If the original circuit uses non latching relays the control would likely need also extra effort. This would likely mean using a µC  to  do the actual control. The SPI interface is kind of similar to the shift registers, but some may have trouble with 12 bit data.

[dietert1]

The schematic of the scanner board is the last page of the Prema 6047/6048 user manual.
It shows a resistor and a diode in addition to the capacitor as relay coil interface. This works because the pull-up of the CD4049 output is weak.
Meanwhile i think they used non-latching relays, as with latching relays there is a risk of creating shorts between different inputs of the scanner. During power-up their favorite simple capacitor method cannot switch off all relays to begin with. A scanner with latching relays driven by the capacitor method needs a power supervisor to disconnect the active channel during power-down. Then on the next power-on a short won't happen. Probably this is the way to go.

Regards, Dieter

[Rax]

But more seriously, it's actually in better shape than I thought, given what I paid for it. It seems to be measuring in the ballpark in most ranges - still running through the gamut - except current (which doesn't seem to work, though a big grain of salt on that as I've not had enough time to assess).

The 6048 has very few resistors in the gain path: the 200 mV/2 V and 20 V mainly have 1 pair of canned resistors and the LTZ1000 reference to produce drift. So I would expect it to hold the calibration very good, even without ACAL. The 2 V range may be the more stable range, as here it is the ratio of 2 equal value resistors.

For the current ranges there is the obvious fuse that can be bad and also relay contacts could be an issue. The terminals for current are separate, so a bit different than with most other meters. If there should be an issue this should be relatively easy to fix, though getting a relacement relay or shunt can be tricky (or ugly with a different form factor).

I'm only getting occasional opportunities to look at this. And I've never established if my current mode is working, so I'm trying to take advantage of this extended weekend in the US (national holiday) to register some progress.

The manual is a bit confusing (to me). It seems the blue function levels (where mAdc and mAac functions are) are activated by pressing the "2nd" key on the right keyboard (pressing the "TEMP/2nd" key on the left seems to always enable the temperature measurement function?...). Therefore, I assume if I see both the mAdc/Vdc and the "RATIO/2nd" LEDs on I am in the mA DC mode of operation, correct?

This is confusing because the "RATIO/2nd" seems to be the only way to enable the blue/top modes on both the left and the right keyboard clusters and I am not entirely sure I'm still in "MEAS" as opposed to "COMP" mode when the LED for "RATIO/2nd" is on. Does it matter the order in which I am pressing the "RATIO/2nd" to register an effect on a blue/top key on the right vs. the left keyboard clusters?

If I'm switching everything correctly, my current measurement function is behaving erratically. I'll have to figure the current signal path in the preamp, which exact relays are involved and what is the switching mode (open/close) for the current function.

[dietert1]

Yes, i also found it confusing. Until i noticed both groups have their own 2nd key. When a 2nd function is active, the corresponding panels 2nd key has its indicator led on. In order to measure mADC the VDC and the left 2nd indicator should be on. There is only one range = 2000 mA. Input on right connectors.
The ratio function should really be in the left group. You terminate the ratio function by selecting another function, e.g. VDC. Similarly the instrument leaves the compute mode when pressing the VDC key on the left. When you are in AC mode and press null, it switches to VDC to null the instrument and then goes back to VAC.
What i am missing is a way to get the instrument into continuous measurement mode after going local. Could be some reset button. It's a pitty that product was abandonded.

With my research on the nonlinearity near zero issue our 6048 proved to be rather inert to mods, which is something good. Their conversion method is rocksolid, except i have my doubts about integrating digital control of the integrator (counters, pll control) together with analog control (amplifier and comparator). Also i have some doubts about running the ADC without negative supply. The only negative supply comes from U9 to pin 11. The comparator won't fire on the positive glitch of the integrator when starting run-down. There seems to be a minimum run-down duration of about 7 usec, in case the comparator triggers early. In that occasion summing/averaging errors happen. I found a capacitor missing in the schematic and another one present in the temperature ADC but optional and not present in the main ADC.
The nonlinearity effect amounts to about 90 to 200 pA error current at the integrator summing point. With 10 pC charge injection and 10 msec integrator cycle time, charge injection offset current can be 1 nA. Cutting into that at short pulse lengths can produce that nonlinearity. I am still waiting for the DG308B switches with 1 pC charge injection spec.

Regards, Dieter

[Kleinstein]

The charge injection part should still be largely constant. The pulses are not that short, usually some15 µs and up. It would be rather unusual for the charge injection to change with pulse length on the that time scale. Checking a different switch chip is of cause an easy option (a good thing of having the chips in sockets).
A possibly nasty point could be the quite strong current pulse that the LT1007 sees to charge the switches. Having about half the capacitance with the DG308B can help too.
In the current circuit the LT1007 has to make a small step up in the voltage when the switch turns on. This is to copensate the drop on the switch resistance or some 140 mV. one could consider adding some large resistance from the inverting input to a positive voltage (e.g. some 33 M to the 12 V) to already have that extra voltage also in the off state.


I don't see a problem of not having a negative supply at the ADC chip. The interator is working with a positive output voltage only and the custom chip is mainly the comparator, logic and part of the PLL. What is a bit strange is that the integrator seems to be a simple 1 OP-amp type with the relatively slow OP77. I don't see a way to have much added gain in the ADC chip, as there is no input for the integrator input node and thus no way to react faster.
What I don't like very much is having the PLL together with the ADC part - if the timing is at a certain point the ADC may effect the PLL and errors in the clock can cause INL errors. Not seeing an effect from the change in the PLL indicates that the PLL may not be the trouble. How is the timing of the ADC reference pulses relative to the 50 Hz signal to the PLL ?

[dietert1]

I was not reporting all the mods i tried to let this continue as a Prema 6048 thread. One mod was a switched current source to keep the LT1007 at constant output current (see above). Also i found a simple method with two diodes and a 10K resistor to eliminate R63 and R64 using the PWM signal to provide switched pull-downs.

The OP77 U9 drives an amplifier of high voltage gain inside the BK7. U9 output is at -0.7 V and swings no more than 50 mVpp. A simple model would be a npn transistor inside the BK7 with base on Gnd, emitter (maybe with a small resistor) on output of OP77 and collector on pin 13 which is also the comparator input. I guess the other comparator input is pin 14. Pull-up cabability of the BK7 output pin 13 is somewhat limited, just enough to handle the -0.4 mA max input current from the frontend. The pull-down current to handle the 1.4 mA reference pulse comes from U9.

At zero input the integrator output remains close to Gnd. So that output transistor works near saturation. The comparator may be compatible with zero input levels, but again a negative supply could help. One could try to lift pin 14 to +1 V or so to get some margin on the output side of the integrator.

Regards, Dieter

[Rax]

Yes, i also found it confusing. Until i noticed both groups have their own 2nd key. When a 2nd function is active, the corresponding panels 2nd key has its indicator led on. In order to measure mADC the VDC and the left 2nd indicator should be on. There is only one range = 2000 mA.
[...]
Regards, Dieter

Thank you, Dieter! This is great. I fiddled with the controls and it seems when the LEDs for both Vdc/mAdc and the left side "TEMP/2nd" are lit is when the unit is in current measurement mode. I need to fiddle with it more to get it. I am not yet sure exactly if the "RATIO/2nd" need to be actuated in any way for this to kick in,* but it's not lit when the current is being measured.

The other thing I didn't realize is that there's just one range, so I was further confusing the unit (and myself) with trying to get it into AUTO.

I can report success, though, and my unit is able to measure current! Yeah! Nothing to see there.
________________________________________________________________________________________________________________

Later: * It seems to me the "left cluster 2nd" won't engage unless the "right cluster 2nd" is also actuated prior. I think I read something to this effect in the manual - though I can't seem to be able to find it now - but if so, this is really odd and unintuitive. Maybe my unit acts up in some way, though I lean to believe it's just weird controls logic.

[Rax]

if you search ebay for "Netzschalter main switch für Marantz 2216B-2218-2226-2238 - 2238B- SC 06" you will find a power switch that should fit. I think at the time everybody used these. I think the make was RAFI.

Regards, Dieter

I think I'll take a slightly different route with the switch. I enclose pics of mine, and I'm pretty sure this is one of those "ME5A" types used in some Commodore, Atari etc. monitors. I'll try one and post here if successful.

[dietert1]

The first scope dump shows the timing relationship between the BK7 PLL input pin 8 and the integrator run-down. The timing reference has a small asymmetry between positive and negative edges. The positive half is a little more than 10 msec. So i would say the PLL lock onto the negative edge and run-down starts at the same time.  After i removed the input short, run-down was slowly getting longer. At full range it reaches about 3 msec. With open inputs hum is so large that it is visible at this time scale: every second run-down is longer.

The second scope dump shows the LT1007 output during a short reference pulse. The reference voltage is 7,075 V. During the reference pulse the LT1007 output rises to 7.170 V under control of the sense connection (DG308 pin 11 to pin 6). After DG308 turn off, we can see the charge injection into a high impedance, some parasitic capacitance and the 1M resistor R32. The first rising pulse has an area of 10,1 V usec. Divided by 1 MOhm this makes a charge of 10,1 pC. The slower negative response with a duration of about 35 usec has a similar area and gives -12,5 pC. Something similar can be expected on the integrator summing point, where we have R34 750K as DC resistance. The net charge will be small and constant (e.g. -2.4 pC) if one waits long enough.

Regards, Dieter

[Kleinstein]

The PLL timing could be a slight issue. I see 2 possible weak points:
1) the reference switching may interfere with the PLL and possibly cause some frequency modulation.
2) the 50 Hz input signal could effect the comparator or intergrator. This would be the slope not used for sync.
However there is very little one can do about the timing. This should be more or less fixed in the chip.  Having U6 populated with an OP07( as in some photo) instead of the LF411 in the plan could make a small difference as this could shift the timing a little from the bias current that would require a small phase shift at the PLL. Still I would not expect a significant shift from this.

The settling of the LT1007 looks quite slow for the off part. However as this is the off phase there should be plenty (e.g. > 5 ms) of time for settling. The peak is quite large, but it should stay the same, no matter how long the reference pulse was.  The on phase settling looks fast.

[dietert1]

Changing the timing relationship becomes an all digital mod when using the HC4046 built-in phase comparators with the BK7 PWM output for comparison. Just a 9*8192 counter to divide 3.6864 MHz down to 50 Hz. A small CPLD can do that.

Regards, Dieter

[dietert1]

Meanwhile i found how to make a nonlinear integrator. As a low pass filter between PWM and integrator i added two parts: C31A and R31B in the schematic. This time the linearity test showed the opposite nonlinearity effect.
If this is the effect of a low pass filter, the P6048 nonlinearity is the effect of a high pass filter (charge injection). Using a DG308B did not help. In fact the parasitic capacitance of R31 gives more charge. Its metal can measures 5 pF to the resistor, so charge injection will be some fraction of 7 V * 5 pF = 35 pC. I will try and ground the can.
For the nonlinearity to disappear it needs to become a capacitive divider. Later one can add R31A to restore calibration.

Regards, Dieter

[Kleinstein]

Some charge caught at the metal can of R31 could indeed be an issue. So grounding the can is a really good idea to test.

I don't think adding R31A would make much sense - a new calibration is likely simpler.

[dietert1]

Meanwhile i solved the riddle of the nonlinearity and to make it short, the main problem was U9 OP77 that isn't fast enough to handle the PWM reference properly. 75 % of the nonlinearity vanished when i replaced it by an OPA140 (about 60x faster). I'm still using the filter with a 10 nF cap. As the time constant is about 1usec, it won't affect linearity in the range the instrument actually uses.

The other 25% of nonlinearity vanished after replacing U10 OP07 by another OPA140. This time the problem was in the input protection diodes of the OP07. When the four switches of the DG308 close at the beginning of the PWM pulse, it happens in arbitrary order and the OP07 can't follow anyway. A strong current pulse flows from the inverting input through the protection diodes to the non-inverting input of U10, hitting the reference output and causing U20 pin 7 to jump multiple times like a rubber ball on the ground (20 mV during 10 usec). The schematic shows a buffer cap C49 100nf not present in our P6048. For good reason as it causes instability of U20. I inserted a 10 uF cap in series with a 56R resistor. Also i added 100nF across diode CR14.
And another 100nF across R35.
The last sweep appears clean. INL axis is +/- 1 ppm of 2 V range.

Regards, Dieter

[Kleinstein]

Was U10 really an OP07 ? In the plan there is a much faster LT1007 for U10.

The OPA140 is a great JFET OP-amp, but in the LF range may still be a bit more noisy than the OP77 / LT1007. It would not show up much with a shorted input, but could be a slight issue when measuring a significant voltage.

[branadic]

That would be some serious improvement. What about the 20V range?

-branadic-

[dietert1]

With U10 the problem wasn't speed but the input protection diodes. LT1007 and OP07 both have those input protection diodes, so they are no good for U10.

As far as i remember OPA140 noise specs are all better than OP07. Anyway, since the SO8 OPA140 only sells in China right now, i will also try to use OPA189 chopper stabilized amplifiers (where we have some). In the OPA189 datasheet they discuss using the part with analog switches.

Yes, i will look at the 20 V range, too. That's why i did not want to compensate the nonlinearity but look for the origin.

Regards, Dieter