Datron Calibrator 7v to 10v conversion with a transformer

[chuckb]

I seem to remember that Dayton had a calibrator (or DVM) that used an inductive voltage divider for the zener 7v to 10v conversion. It was probably a different voltage ratio. The inductive voltage divider provides a very stable but complex voltage ratio.

Does anyone remember the Datron Model number of this unit. Are any schematics available for this section of the unit?
Thanks 

[doktor pyta]

Wow, you read my mind!
Yesterday I did this for testing with LTC1043 as modulator and demodulator.

[doktor pyta]

Good idea. I have few H11F1.
I will have to provide some dead time between operation of two switches in the demodulator.
My first observation is that it is extremely easy to saturate the core of my divider and it won't work without say 2200uF capacitor connected in series with the input of the divider. The inductance of the input is about 3H. I used Nanoperm core.
Later I will share some results.

[Kleinstein]

Even with a high permeability core one should use a separate feedback on the actual voltage generated by the transformer. The Datron 1281 uses this to adjust the driving voltage. There are two reasons for this: one is the magnetizing current. The second thing is a not so perfect waveform due to parasitic capacitance. So there will not be a perfect square wave, but some ringing - this is why some dead-time is needed. Beside ringing, there will be a slight slope from the magnetizing current / winding resistance - so the exact timing and demodulator can have an influence.

With high permeability cores one should be careful about applying mechanical stress to the core - this can have an influence on the AL value and thus inductance. It really helps to have a core with very low magnetostriction. I would consider this more important that high permeability by itself. So the simple metgals is not a good choice.

The Datron circuit get's away without an DC isolating capacitor, by using a split coil for driving and having a strictly symmetric waveform (from a divide by 2 circuit).  Having a not so perfect 50% duty cycle and a coupling capacitor to compensate would cause non equal voltages for the two parts.

The H11F1 is good, but I don't think you have to care so much about charge injection, as the transformer usually is not that high in impedance. It would already takes 4 switches to build two demodulators.

[chuckb]

I got the Datron 1281 information, thanks!

Is there some parameter that indicates magnetostriction or is it something you have to test for different families of magnetic materials?

I was thinking about applying a symmetrical trapezoid waveform (around 400 hz) to control the bandwidth applied to the transformer. The 72% tap would pass through a buffer before it went to the a synchronous rectifier to demodulate it.

The Direct Current Comparators use many tricks to make the transformers "ideal". I will have to see what can be learned there. So much to learn, it's great!

[Kleinstein]

Some data sheets on the magnetic materials / cores  give data on magnetostriction. For the very high µ materials usually a low magnetostriction is needed. The metglas is one of the few exceptions where you get really high  µ and still quite some magnetostriction.

One could test the sensitivity to external stress to check if there is much trouble. So not a direct measurement (which is not that simple), but measuring the reverse effect that is possibly making trouble. With a sensitive ferrite I could change the AL value and thus inductivity just by pressing with the hand by something like a factor of 2, and winding too tight reduced it by a factor of 10.

I don't think you absolutely need a buffer before demodulation. The output impedance is about the same as the driving one.

[zlymex]

........The Direct Current Comparators use many tricks to make the transformers "ideal". .......

Which are all very pricy and very bulky.
The next best thing might be LEM IT 600-S with <1ppm linearity:
https://www.eevblog.com/forum/reviews/teardown-lem-ultrastab-1ppm-600a-current-sensor/msg721897/#msg721897

[doktor pyta]

First results are promissing.
Using VERY poor layout and WITHOUT any shielding I managed to  obtain ca. 10V from ca. 7.15V with accuracy of +/-2ppm during 12h -measured using K2001.

The problem I have is not exact ratio of the voltages- when I set input voltage so that output after 100 turns is exactly 10.00000V, the tap after 10 turns gives 0.999842V at 1kHz excitation.

Voltages measured at the output of second phase sensitive demodulator
while input of demodulator is connected to various points:

0 turn (GND) it reads 0.00000 V
After 10 turns I get 0.99984V
After 20 turns I get 1.99981V
After 30 turns I get 2.99981V
After 40 turns I get 3.99982V
After 50 turns I get 4.99984V
After 60 turns I get 5.99987V
After 70 turns I get 6.99990V
After 80 turns I get 7.99994V
After 90 turns I get 8.99998V
After 100 turns I get 10.00000V
After 110 turns I get 11.00005V
After 120 turns I get 12.00010V.

When the frequency increases, the ratio goes worse (as expected).

I think that more turns would help. This would improve coupling and increase primary inductance so the magnetizing current will be lower.

Wavforms captured @10kHz show:
yellow= voltage on secondary (tap after 70 turns)
two lower traces= D1 and D2 demodulator drive signals

*OP77 was finally removed from the output (problem with voltages close to 0V - I used single +15V supply)

[doktor pyta]

The CD4017 is a great part to get the needed pulse trains, but I would shit-can the "Mickey Mouse Logic" with the diodes-- as these will lead to some problems.  With all due respect to Don Lancaster, I would not use this kind of diode logic on serious stuff.  A better way might be to create some set/reset flip/flops using a pair of 2-input NOR gates.  Connect the set and reset inputs to the appropriate outputs of the 4017, and now you have a glitch-free true and complement of each needed signal.

If you want to keep what you have, consider using the square-wave output of the 4017 to drive the transformer driver instead of decoding this from the 4017 outputs.  IIRC, it is on pin-12.  This is high from counts 0..4 and low from counts 5..9.

Thanks.
I used 4017 since I was a child and I skipped that feature...
I learn something every day

Is there any reason you want the 12V output?

I used 25 wire litz. I used 10+10 to make ratio transformer so I have 5 spare windings.
This 12V is just for free.

P.S. I didn't use autotransformer, because I was afraid of the magnetizing current which would flow only through 0...70 windings. So 'upper' windings 70...100 would operate without this current passing. This could produce difference in ratio.
(My first thought was, why did guys from DATRON use separate windings...)
Or maybe I'm wrong ?

[Kleinstein]

The core material is really good and has low magnetostriction (e.g. < 1 ppm, which is a good value, compared to about 20-50 pm for metglass and ferrites). However it is conductive and there thus will be eddy currents if the frequency is high. A larger core might help to work with a lower frequency and avoid saturation. It is not only the switches that limit the speed. Of cause a low frequency makes capacitive coupling less practical.

To avoid an error from the magnetizing current it is right to use separate windings for driving and reading the voltages. It might not be much, but I think it is worth avoiding it. It might also be important how the winding is returned to the starting point. Going once around the core is like having a hidden turn in parallel to the core.

However the magnetizing current also has the effect that the square is not perfectly flat on top, but should have a slight down slope. The finite size of the coupling capacitor would cause a similar effect. With a low resistance switch like the one used, this might not be such a problem.  To compensate for this one would need to use the feedback at any time and not just after demodulation. However this could turn out to be difficult (e.g. avoid oscillations, ringing).  It might be possible to check for the importance of the magnetizing current by applying a transversal field in the direction of the core axis - this can reduce the AL value and thus increase the magnetizing current.

It might be an option to check conventional CMOS switches for demodulation, despite of some charge injection. But on the other side there is much less delay and thus uncertainty in time. However this would need a +- supply.

The diode logic could be a problem because of slightly different delays, though likely less a problem than the slow opto MOS switches. Ideally there would be D flip-flops for the outputs to bring the signals back in sync with the clock. The demodulating signals also shows some glitches (could be from the diodes), here just a smaller pull down / filter cap might help. One might also consider using a small µC to generate the signals.

[doktor pyta]

Little update:

results @ 198.7Hz excitation (200Hz will produce very low beat frequency with 50Hz mains)
+modification with CD4017 Carry Output, so the excitation has duty cycle 50%

0 turn (GND) it reads 0.000000 V
After 10 turns I get 0.999950V
After 20 turns I get 1.999924V
After 30 turns I get 2.99992V
After 40 turns I get 3.99991V
After 50 turns I get 4.99991V
After 60 turns I get 5.99992V
After 70 turns I get 6.99995V
After 80 turns I get 7.99996V
After 90 turns I get 8.99997V
After 100 turns I get 10.00000V
After 110 turns I get 11.00002V
After 120 turns I get 12.00003V

The core saturates when I work with ca. 160Hz excitation.
P.S. Adding another 2200uF in parallel doesn't change anything.

[doktor pyta]

OK. I won't spam the forum anymore.

Last results with  MAX430 auto zero loop amplifier.
130 turns were used on primary to improve coupling.

results @ 198.7Hz excitation

0 turn (GND) it reads 0.000000 V
After 10 turns I get 0.999960V
After 20 turns I get 1.999930V
After 30 turns I get 2.99992V
After 40 turns I get 3.99992V
After 50 turns I get 4.99993V
After 60 turns I get 5.99993V
After 70 turns I get 6.99995V
After 80 turns I get 7.99996V
After 90 turns I get 8.99998V
After 100 turns I get 10.00000V
After 110 turns I get 11.00002V
After 120 turns I get 12.00004V

Well, there is no big difference.

[chuckb]

I think a discussion is helped immensely with a practical circuit. It gets everybody focused instead of the subject being all theoretical. I spend to much time thinking about a solution and not enough time jumping in and learning by doing.
Thanks

[zlymex]

I had a North Atlantic AC ratio box(same as this: http://www.ebay.com/itm/291616067694) that they employ two-core winding technique.

[acbern]

Could you guys explain what you mean by autotransformer and two core winding technique and what else is critical for .
While I find this topic pretty interesting, I am not at all an expert in ratio transformers (just basics, use them; DT72) and how to achieve high accuracy there. thanks in advance.

[zlymex]

Could you guys explain what you mean by autotransformer and two core winding technique and what else is critical for .
While I find this topic pretty interesting, I am not at all an expert in ratio transformers (just basics, use them; DT72) and how to achieve high accuracy there. thanks in advance.

The primary and the secondary share the same winding in an autotransformer rather than separate and isolate.
North Atlantic uses two-core winding technique to solve the problem of resolution where one winding of say 1000 turns can only resolve 0.1% and it is not feasible to have a winding of a million turns in order to resolve 1ppm.

[Kleinstein]

As there are two identical demodulators used for the reference feedback and the output signal, I do not think the nonlinearity should be due to the transformer. I see really no way why the voltage from twice the number of winding should not give twice the voltage. If at all one would expect more loading to the transformer, when using more turns. But this would apply the same way to the reference winding too. The transformer however could be a source of the not so perfect square form.

It's more likely that the nonlinearity is due to the demodulator: the H11F1 is an on resistance in the 200 Ohms range, thus there is a time constant of about 0.2 ms to adjust the voltage in the capacitors. Due to turn resistance and output impedance of the transformer the time constant will be slightly longer for more turns. Due to the non ideal transformer and the coupling cap, the voltage is not perfectly square, but with a slight slope. So even in well settled state there will be a little current flowing in and out.

Then on resistance is not perfectly linear (2% limit in the datasheet), though only the rather low current range is actually used. This also could cause this minute nonlinearity.

It might be worth to check if an additional series resistance in the 10 to 100 Ohms range has an influence. At least the turn resistance could be compensated for (e.g. add resistors to the low turn taps) - for the transformed input impedance one might have to check this - it could get more complicated. Another possible variation would be drive strenght for the opto couplers.

For error estimates it might help to know the approximate size of the magnetizing current, with an inductance in the H range I would expect maybe 1 mA or a little less, but nonlinear.

[doktor pyta]

Just for curiosity I performed measurements in autotransformer configurations.
There are two basic possibilities of connection shown on the picture as case A an B.


Case A, results @ 198.7Hz excitation

0 turns   0.000000 V
10 turns 1.000111V
20 turns 2.000238V
30 turns 3.00038V
40 turns 4.00053V
50 turns 5.00071V
60 turns 6.00087V
70 turns 7.00104V
80 turns 8.00069V
90 turns 9.00034V
100 turns 10.00000V
110 turns 10.99967V
120 turns 11.99936V

Comment: if the magnetizing current flows only through 0...70 windings the ratio errors can be higher than 100ppm.


Case B, results @ 198.7Hz excitation

0 turns   0.000000 V
10 turns 0.999953V
20 turns 1.999925V
30 turns 2.99992V
40 turns 3.99991V
50 turns 4.99994V
60 turns 5.99992V
70 turns 6.99995V
80 turns 7.99997V
90 turns 8.99998V
100 turns 10.00000V
110 turns 11.00003V
120 turns 12.00005V

Comment: errors similar to the circuit with separated primary and secondary windings (only slightly worse).

*************************************************************************

And final test of this particular transformer:
case B, but 240 turns at the top of the transformer and a tap for feedback after 140 turns.
Frequency 102.4Hz (now the inductance seen from the excitation source is higher so the frequency can be decreased)

0 turns   0.000000 V
10 turns 0.500001V
20 turns 0.999975V
30 turns 1.499960V
40 turns 1.999948V
50 turns 2.49994V
60 turns 2.99993V
70 turns 3.49992V
80 turns 3.99994V
90 turns 4.49992V
100 turns 4.99993V
110 turns 5.49993V
120 turns 5.99993V
130 turns 6.49994V
140 turns 6.99994V
150 turns 7.49995V
160 turns 7.99996V
170 turns 8.49998V
180 turns 9.00000V
190 turns 9.50000V
200 turns 10.00000V
210 turns 10.50002V
220 turns 11.00003V
230 turns 11.50005V
240 turns 12.00006V

Conclusion: autotransformer can be a good option if properly used. Magnetizing current must just equally flow through all used windings.

Nevertheless in the future I must try this with a new, bigger core and more turns.

[doktor pyta]

For the H11F1, they need about 16mA on the LED current to be fully turned on, and I didn't see any kind of high-current LED driver in your schematic.  Perhaps you could address this and see what happens?  We will talk about slope compensation after that.

In my case LED current is close to 9mA. I will try to increase this current but it needs to wait for the weekend.

I'm wondering what benefits would I have if I used the two core method together with 'case B' connection comparing to one core method with 'case B' connection (assuming the same number of turns)? The inter-winding capacitance problem would be similar in both cases. Any ideas?

[zlymex]

It seems to be a pattern with the non-linearity.

[Kleinstein]

The two core method helps to reduce the effect of the magnetizing current. The 1.st core does most of the work, while the 2 nd. core is essentially only there to provide the very low voltage to compensate the error. So only very little current needed in that coil. This alone does not help with capacitive coupling.

To compensate for the capacitive coupling, is a second step, with coax cable. Here the 1. core also powers the shields and woks as an auxiliary divider for this. As far as I see this, the shielded cables only get practical with relatively large cores - as it needs space and a larger core can work with fewer turns. 

The outputs of the 4017 are not that strong - so I have some doubt they could provide two times 9 mA. With a 10 V driving voltage one could have the two LEDs driven together in series instead of parallel.  A weak drive could make the switch slower.

For me one possible cause for the nonlinear curve could be a combination of a not perfect square form (e.g.  due to magnetizing current, but also the coupling cap) and a timing that is slightly shifting with voltage. So it should help to make the signal more square and get a more stable timing (here more currents to the leds could help). Using the divider as an auto-transformer could help to make the signal more square, if the driving uses all the turns. As a side effect more turns are available, so less field and magnetizing current. However this still leaves the coupling cap at the drive side.

With a square drive and quite some dead time as used here, the capacitive coupling might not be that bad, as the charging currents will flow mainly directly after switching and not at the constant voltage level.

[doktor pyta]

Before I add LED driver stage I have some news.
I picked up 6 decade inductive voltage divider which uses two-core transformer in first stage.
So this factory made divider can be used as something I can be sure of.
I disconnected feedback amplifier and I connected VDD of TC4426 to output of my Data Precision 8200 DC calibrator.
I set voltage slightly above 10V.
CD4017 and LEDs of demodulator are powered from separate power supply- let's call it VLED.
Excitation frequency 102.4Hz

Now I will show measurement results for changing first decade of the divider from 0.0 to 0.9 for VLED=12V and from 0.0 to 0.9 for VLED=16V.
This way I can determine what is the influence of LED current on linearity of the demodulator.


for VLED=12V
set 0.0   measured 0.000000V
set 0.1   measured 0.999965V
set 0.2   measured 1.999938V
set 0.3   measured 2.99992V
set 0.4   measured 3.99991V
set 0.5   measured 4.99992V
set 0.6   measured 5.99993V
set 0.7   measured 6.99993V
set 0.8   measured 7.99996V
set 0.9   measured 9.00000V (voltage calibrator was used to provide exact value)


for VLED=16V
set 0.0   measured 0.000000V
set 0.1   measured 0.999968V
set 0.2   measured 1.999940V
set 0.3   measured 2.99992V
set 0.4   measured 3.99992V
set 0.5   measured 4.99992V
set 0.6   measured 5.99992V
set 0.7   measured 6.99994V
set 0.8   measured 7.99996V
set 0.9   measured 9.00000V (voltage calibrator was used to provide exact value)

Conclusion: for both supply voltages the linearity (related to full scale of 9.00000V) is very close.
However when I set 9.00000V for VLED=12V and then I set VLED=16V the K2001 reads 8.99962V so there is some dependency.

*Additional info:
In my case Excitation frequency is ca. 100Hz so the period is 10ms and one H11F1 is turned on for Ton=3ms.
Assuming H11F1 Ron resistance equal 400 ohms and 1uF capacitor it gives Tau=0.4ms. This gives Ton/Tau =7.5
Let's take 8 for clarity.
Now we know that for this ratio, the voltage on capacitor will settle to 335ppm.
Only increasing this ratio above 14 will cause settling to 1ppm.
The charging repeats each cycle so it might be not so important.

Does it make sense to lower the series capacitor of the demodulator to 68nF ?
What do You think ?

EDIT:
You don't have to answer.
The linearity is worse with 68nF capacitor in the demodulator.
Only 3 measurements :

0.00000V
4.99913V
9.00000V


EDIT 2
series capacitor of the demodulator was increased to 2uF

for VLED=16V
set 0.0   measured 0.000000V
set 0.1   measured 0.999981V
set 0.2   measured 1.999962V
set 0.3   measured 2.99996V
set 0.4   measured 3.99995V
set 0.5   measured 4.99995V
set 0.6   measured 5.99995V
set 0.7   measured 6.99998V
set 0.8   measured 7.99999V
set 0.9   measured 9.00000V (voltage calibrator was used to provide exact value)

[Kleinstein]

The slightly nonlinear curve seems to be very similar with the commercial transformer too. This somewhat indicated that it is not directly connected to the transformer, but more to the rest of the circuit, especially the demodulation circuit.

The change in output from just changing VLED is quite large. This could also be a problem, as the H11F1 may be slightly different and temperature dependent (e.g. less light from LED at higher temperature). With the lower frequency the size of the coupling capacitor could also get more important, as it is one reason for a not so perfect square signal.  I would expect the output to be less dependent on VLED when the signal is more square, like that directly from the TC4426 (before the coupling capacitor).

[doktor pyta]

OK, I added TC4427 dual non inverting drivers for driving LEDs.
I decided to put ca. 30mA through single LED.

Measurement using commercial IVD.
series capacitor of the demodulator was increased to 2uF

for VLED=16V
set 0.0   measured 0.000000V
set 0.1   measured 0.999988V
set 0.2   measured 1.99997V
set 0.3   measured 2.99996V
set 0.4   measured 3.99996V
set 0.5   measured 4.99997V
set 0.6   measured 5.99998V
set 0.7   measured 6.99999V
set 0.8   measured 7.99999V
set 0.9   measured 9.00000V (voltage calibrator was used to provide exact value)

It is slightly better.
The reading changes much less when I change VLED from 16V to 12V.
This must be a result of flat resistance vs. LED current curve in this region.

BTW--- The input bias current of the OP77 is around 1.2nA at room temperature... I didn't bother with the math, but I wonder if this is enough to cause this problem?

Yesterday I changed opamp to MAX430 with typical 30pA input current. There was no significant difference between OP77.

@DiligentMinds.com
Slope compensation question is very interesting, but It might be difficult to build this and prevent it from oscillating at least in this simple version. Do You have any example to show this slow and fast opamp solution ?

I wonder if it makes sense to continue this very simple, single +15V supply solution -or-  just to try to copy DATRON 1281 circuit.
As I understand, they used mid tapped transformer on primary and secondary. This makes possible to do this slope compensation fairly easy with one opamp. However they used loop filter with quite high time constant (dual RC circuit with Tau=1ms). I wonder what frequency they use for modulation. I wouldn't be surprised if they used something like 10...20Hz.

[zlymex]

It is better as in green line.

[Kleinstein]

I first was a little suspicious about the H11F1, as it is relatively slow and not that low in impedance, but it has two big advantages: no charge injection and it is independent of the DC level at the switch. Both could be a big problem with normal CMOS switches, and might require a bootstrapped power supply. During the time the switch is turned on or off the voltage at the switch is rather small (only residual from the slope) and thus not much effect on the switching time is expected. So I changed my mind and now like the H11F1 as a switch.

I don't think the inter winding capacitance should be that bad, as this would be an effect only for the time just after switching. The dead time before the signal is actually used should be large enough so that the capacitors should be well charged. This could be different if used for AC with a sine wave.

I would expect most of the slope to come from the coupling capacitors. So the split turn and separate drive like the Datron circuit might be good enough. If DC magnetization is a problem one might use a separate correction loop (or only adjustment pot), by checking for the 2 nd harmonic (or positive and negative peak values) in the driving current. For the secondary side the current solution looks quite good and might even be an advantage when using the split turn for driving, as it is not sensitive to uneven voltage for both halves.

Making things worse first to find weak spots is a good idea, especially a main cause of trouble is likely a combination of the slope from driving the transformer and non ideal demodulation that is sensitive to the slope. One source could be the nonlinear resistance of the H11F1, that sees about the voltage of the slope. So it might already help to have a resistor in the 1 K range in series from the transformer tap / capacitor, so that less voltage drop is at the switches itself.
If I understand the specs of the H11F1 right, nonlinearity can be up to 0.1% at 25 µA. This would be something like a 10 mV change from the slope, than could contribute something in the 10 µV range.

[doktor pyta]

Specially for Kleinstein I measured voltages with 1k resistor in series with both demodulator switches H11F1.

Measurement using commercial IVD.
series capacitor of the demodulator 2uF

for VLED=16V
set 0.0   measured 0.000000V
set 0.1   measured 0.999983V
set 0.2   measured 1.99997V
set 0.3   measured 2.99995V
set 0.4   measured 3.99996V
set 0.5   measured 4.99997V
set 0.6   measured 5.99998V
set 0.7   measured 6.99999V
set 0.8   measured 7.99999V
set 0.9   measured 9.00000V (voltage calibrator was used to provide exact value)

Comment: no visible change in results.


EDIT:
Before stopping this experiment I measured some parameters of the divider shown in Reply #2.
Single winding (10 turns): DC Resistance 0.17ohm, Inductance @120Hz: 20mH
Capacitance between adjacent windings: ca. 100pF

[doktor pyta]

In an old Polish book about metrology 'Etalony...'  I found information about inductive voltage dividers.
I took two pictures of it.

First chart shows 'relative internal current level vs. distance from the beginning of the winding for multifilar winding'. As I understand we are talking about current due to inter winding capacitances

Second picture shows how to deal with it: to connect calculated values of admitances (external capacitors).

Maybe the same error of ratio is present also in DATRON circuit, but as long as it is stable in time few ppm's of (constant) shift may not be so important.

[Kleinstein]

Thanks for doing the extra measurement with series resistors. So the slightly nonlinear resistance of the H11F1 seems to be not the culprit for the nonlinearity.

As the errors got larger with higher frequency, it might be a good idea to do future tests at a higher frequency. The H11F1 has a specified response time of 15 µs. Assuming at least 3 time constants, this would be about 50 µs or a 20 kHz limit for the clock to the 4017 and thus 2 kHz for the transformer.

For the effect of inter winding capacitance, it should be possible to test this, by adding capacitive loading and watch the change in voltage. Due to the nearly square waveform, I have some doubt that this is the problem. Anyway it might be interesting to know / check the output impedance of the transformer. I would expect something in the 1-10 Ohms range - so capacitors should charge quite fast.

Form the Datron circuit it looks like they use a frequency of slightly less than 1 kHz. But they also use faster switches and I don't think they want absolute accuracy, but just stable values. For just doing the 7 to 10 V step for a 10 V reference, there is no need to have an absolute value, just stable would be good enough. However understanding why the value is different from ideal can help to make it more stable and having something like accurate 1,2,....,10 V steps would be really nice.

Edit:
I just realized one source of error:

The capacitance in off state together with the now 2 µF capacitor make up a voltage divider that reduces the output voltage. The capacitance of the H11F1 is supposed to be in the 15 pF range (at 15 V), but it is expected to get larger at lower voltage. This error component should be independent of frequency and scale with the capacitance. The shape and size of the observed nonlinearity looks reasonably for this effect. One might be able to measure the H11F1 independently and than calculate the effect. The effect might be smaller with other (e.g CMOS) switches, but the principle problem should be the same.