Re: FeelElec New Arrival FY-6900 Signal Generator

[masterx81]

Sorry, but who upgrade the opamps on this units, remove the original ths part or put the 3091/3491 in parallel to the pre existing part? Because my unit had the 2 100ohm output resistors per channel not put in parallel, but every resistor was directed to one output of an opamp, giving a 100ohm at the output  if only one of them is installed (like out of the box...) . Now i have put them in parallel (shorting pins 1 and 7 of the removed u9 and u5) for have a 50ohm impedance also when using the range above 5v. And i have also removed the 2 resistor per channel of the old opamps. I suspect an error in the pcb...

[Johnny B Good]

Thanks Johnny B Good.... i still miss what is a "50ohm pad" but i've roughly understood the problem, that if i've well understood was fixed in the 6900.

Another thing, someone know what calibrate every trimmer onboard? In the fy2300 was easy as the circuit was much simpler. I see 3 of them per channel. On the 3200 there was the offset, zero adj and gain, i suspect that on the 6900 are the same...

 Your " i still miss what is a "50ohm pad" " is a slightly different question to the original. The 'padding out' attenuators used to trim the signal levels going into or leaving a fixed gain amplifier block are typically  spoken of simply as being an XdB pad in the telecoms industry (where X is the number of dBs of attenuation). The 'system impedance' (in our case, 50 ohms) is normally only  mentioned if there's any chance of confusion.

 In this case (Feeltech's confusion) I could only guess at the intended 20dB or so of attenuation, so mentioned only the 50 ohm 'system impedance' value versus the actual impedance value of 85 ohms of this misbegotten attenuator network formed by the two triplets of resistors marked as RS1 through to RS6 on the FY6900 mainboard which so clearly showed the 4 x 100 ohm (marked 101) shunt resistors and the 2 x 510 ohm (marked 511) series resistors exactly matching the resistor values used in the 6600 and 6800 models.

 I've never dared to mess with any of these trimmers since the dc offset and gain calibration settings were pretty well spot on to begin with and I wanted them to stay that way. There did appear to be some confusion as to the effect of a supposed DC offset trimpot on the gain. Looking at Der Kammi's reverse engineered circuit diagram, I think I can understand why.

 The W16 and W26 trimpots have nothing to do with DC offset adjustment as might be implied by DerKammi's labeling and everything to do with trimming out common mode gain (or balancing the gain between the inverting and non-inverting inputs to maximise the CMRR. Thinking about it, having just now solved the puzzle of those trimpots' function, I guess a better way to have labelled them would have been to have named them as "OPA_CH1_CMR_ADJ" rather than "DC AMPL_CH1_OFFSET" (easy to be wise in hindsight, of course ). All opamps are, by definition, DC amplifiers so the need to mention "DC" as a part of the description wasn't really required in this case anyway.

 I don't believe DerKammi thought the W16 / W26 trimpots had anything to do with calibrating dc offsets, it was just bad luck that his choice of DC and OFFSET in the label could, upon cursory inspection, imply such a function. I can now see that it requires two trimpots in order to adjust the gain by equal amounts on each leg of a balanced input amplifier, hence the W26 and W5 trimmer combination to maintain a high CMRR performance on CH1 (ditto W16 and W3 for CH2).

 The only trimpots involved in calibrating the DC offsets on each channel are those marked W1 and W2. It is only when you understand this and the interactive nature of the W5/W26 and W3/W16 gain adjustment trimpots, that you have any hope of improving the gain and DC calibration before those trimpots fall to bits from excessive trial and error attempts to get back to where you'd started from.

 Whilst I'd normally disregard that age old adage, "If it isn't broken, don't fix it!" as advice for wimps, in this case I rather think discretion is most definitely the better part of valour. If the calibration seems to be seriously out of wack and you're feeling brave/lucky and heroic enough, then by all means have a go (but take heed of the following advice lest you find yourself chasing phantoms).

 And, please note that allowing an earth loop to exist will create mystery DC offsets so leave it disconnected from anything else connected to ground if you haven't already fixed the earth loop issue created by Feeltech's half assed attempt to eliminate the half mains voltage ESD hazard to any DUT being connected up that had existed with the FY6600 and its predecessors.

 The above comments weren't intended as a criticism of DerKammi's very fine work in producing this reverse engineered circuit diagram which had been intended to serve the needs of a group who'd spent weeks discussing exactly what had gone into and what had been left out of this diminutive box of tricks.

 As for the FY6900, you can be pretty certain that there'll be few (if any at all) departures from what DerKammi had revealed of the FY6600 circuitry. There are only a limited number of ways to build a signal generator of this quality at such a low price point. From what I've seen of the FY6900's innards, it seems Feeltech have stuck with a basic tried and tested circuit layout, even to the extent of retaining that cockamamie 85 ohm attenuator pad   

JBG

[Johnny B Good]

So, regarding this argument, i've discovered something.
If i set the waveform to dc, it uses the offeset circuit to output the signal.
So, setting the signal to square wave and dc 99.999%:

• If amplitude is > 5v / with 6v signal i get +3vdc without load. With 50ohm load i get only 1v instead of 1.5v expected
• If amplitude is > 0.5 and <= 0.5 / with 5v signal i get +2.5vdc without load. With 50ohm load i get  1.25v as expected
• If amplitude is < 0.5 / with 0.5v signal i get +0.275vdc without load. With 50ohm load i get  0.102v

I know that 50ohm, must be the impedance and i'm measuring pure resistance, but it's normal?

 Please keep in mind that the P-P thresholds of 500mV and 5v correspond to the DC voltage levels of plus or minus 250mV and 2.5V respectively.

 It looks like the 99.999% must be the duty cycle value so you can use a DMM to measure DC voltage. The readings you obtained are correct for the first two cases but the final case should have given an open cct reading of 250mV rather than the 275mV you saw. A quick calculation shows that the 102mV reading matches expectations when an 85ohm attenuator has been inserted into the signal path.

 It could be the DMM reacting badly to the transient dip when switched to its 2Vdc range. repeat the test on the DC waveform setting using the plus or minus 250mV and plus or minus 2.5V relay switching thresholds and report back.

JBG

[masterx81]

Dc waveform is generated by the offset circuit, and not pass in the attenuation circuit (no relay clicking varying the range). So it's not good for this measurement.
I was using rectangle with duty so high to be compared to a dc (but this time we are going in the attenuation network) for have a good measirement of the voltage drp across the load resistor. As i've posted after that comment the reading at the >5v range was off, and actually i had a 100ohm resistor to the output of the ths opamp instead of 50 (that's why i had only 1v with amplitude of 6vpp, with duty 100% is +3v). With 50ohm load and a 100ohm output impedance 1v is correct, but it's not correct the 100 ohm output impedance. With 50 ohm output impedance and 50ohm load i must have half volts (so 1.5v).

[Johnny B Good]

Sorry, but who upgrade the opamps on this units, remove the original ths part or put the 3091/3491 in parallel to the pre existing part? Because my unit had the 2 100ohm output resistors per channel not put in parallel, but every resistor was directed to one output of an opamp, giving a 100ohm at the output  if only one of them is installed (like out of the box...) . Now i have put them in parallel (shorting pins 1 and 7 of the removed u9 and u5) for have a 50ohm impedance also when using the range above 5v. And i have also removed the 2 resistor per channel of the old opamps. I suspect an error in the pcb...

 Correct me if I have misunderstood the above but, as best as I can interpret it, you've already tried upgrading the single THS3002i dual opamp (shown as U5A and U5B in DerKammi's reverse engineered circuit diagram) to a pair of THS3491 (or THS3091 or THS3095) single opamps (depicted in the diagram as U21 and U22 (THS3001)). Unfortunately, this attempt to include the circuitry around the empty U21 and U22 IC locations implies that the original THS3002i stays in place when upgrading to a pair of higher grade 3001, 3091, 3095, 3491 or 3495 opamps. This not the case. The original THS3002i must be removed when populating the empty U21, U22 IC locations (keep in mind that this was aimed at a group who were as much aware of this requirement as DerKammi had been when he'd originally traced out the circuit layout for his own needs).

 When I did my own opamp upgrade just over a year ago, it was simply just a matter of extracting the old THS3002i opamp from the board without damaging any circuit traces and soldering the replacements into the U21 and U22 locations. There had been no need to solder additional wire straps or re-arrange resistor placements. All the required circuit traces on the FY6600's main PCB were present to make this an either/or choice. You must remove U5 when you install single opamp ICs in the U21/22 locations.

 However, in the specific case of the THS3491 opamps I'd chosen to use, I ran a wire strap between pins 1 and 4 in order to take the ref pin (1) to the negative rail (pin 4) as per the the exhortation in the datasheet notes to connect this to a more negative potential than the intended switching threshold of the /PD pin. The purpose of this ref pin is to simplify interfacing between the control logic and the THS3491 when it is powered from dual rail +/-5 to +/-15v  supplies which share the control logic's ground return rail.

 The older THS3095 also has these extra pins but the datasheet was last revised December 2015 and neglects to mention the need to tie the ref pin to a negative potential (the 3491datasheet was revised more recently in July 2018) so those who'd used the 3095 had simply left these pins floating, seemingly without the ill effects implied by the THS3491 datasheet instruction not to allow the ref pin to float. Being aware of the implied risk of allowing any ref pin to just float, I played it safe and tied it to the -12v rail on pin 4. The /PD pin has an internal pullup which means it will float high, enabling the opamp by default in this case, so can safely be left open circuit.

 The circuit layout on the FY6600 main PCB to allow a simple swap between a single dual opamp and a pair of single opamps does introduce additional parasitic inductance and capacitance into the circuitry. It's just possible that Feeltech have made changes to reduce these effects which do now require re-positioning of the resistors when making such an opamp upgrade. It's impossible to tell from the photograph.

 The once relatively straightforward opamp upgrade modification with the FY6600 (and AFAIAA, the FY6800) boards may now be a little more complicated in the case of the 6900 boards. Since I don't own an FY6900, I've not felt any pressing need to track down any information in this regard so this is just speculation on my part. If you've been basing your upgrade on the FY6600 mods, you may have simply been caught out by some small but radical changes to the PCB layout on the FY6900 main board.

JBG

[masterx81]

As now i've done nothing to the ref pin. Simply removed the 3001 opamp, removed his set of feedback resistors, and shorted the two 100 ohm output resistors for have a 50ohm output impedance.
Need to check on fy6900 where the ref pin goes, but by now all seem works as it should.

I was also checking the attenuation pad for finding the correct values. Following this link:
https://www.electronics-tutorials.ws/attenuators/pi-pad-attenuator.html
It suggest that's a pi-pad, with 50 ohm input (the 49.9 ohm resistance at the output of the ad8009). Following the formulas, for a 20db attenuation (from 2.5v to 250mv) i need the 2 resistors to ground at 61.1ohm and the resistor in series at 247.5ohm, for have a Zi=Zo= 50ohm.
This calculation seem correct? This evening i'll try this combination. For sure the opamp ad8009 can source the required current (rated at 150ma).

[masterx81]

Those values works perfectly. Attenuation perfect. I've used for the 247.5ohm a series of 0805 220+27 ohm, and the 61.1 is a series of 56+4.7. Why they have not used this values?

[masterx81]

Attach a photo of the attenuation pad mod and a photo of the inside of the device, crumbled with heat sinks (everything got it heatsink, fpga, ldo, also glued something on the 2 ad8009, and for sure one big on the ths), 2 trasformers (one for +/-12 and one for 5v all linear). The unit got also a 0.9ppm 10mhz tcxo, ths3491, attn pad mod and the 50ohm mod for the ths output. Also added 2 multiturn pot for set the gain of the opamp that do the offset as was quite off. Original resistor is 3k, used 5k pot. Also used an ntc sensor + tc648 pwm fan controller + 40mm silent fan.
As hardware i hope that's all. I've noticed a lot of sw bugs. Sometime the sync not works,  in dc mode if the frequency is  set high i can't go over +/-2.5v (but it's dc....), somerimes does random weired thing on the settings. Next i try to do a fw upgrade (now i have 1.3.1).

[Johnny B Good]

As now i've done nothing to the ref pin. Simply removed the 3001 opamp, removed his set of feedback resistors, and shorted the two 100 ohm output resistors for have a 50ohm output impedance.
Need to check on fy6900 where the ref pin goes, but by now all seem works as it should.

I was also checking the attenuation pad for finding the correct values. Following this link:
https://www.electronics-tutorials.ws/attenuators/pi-pad-attenuator.html
It suggest that's a pi-pad, with 50 ohm input (the 49.9 ohm resistance at the output of the ad8009). Following the formulas, for a 20db attenuation (from 2.5v to 250mv) i need the 2 resistors to ground at 61.1ohm and the resistor in series at 247.5ohm, for have a Zi=Zo= 50ohm.
This calculation seem correct? This evening i'll try this combination. For sure the opamp ad8009 can source the required current (rated at 150ma).

 Prompted by your efforts at cracking this conundrum (converting from an 85 ohm pad to a 50 ohm one which matches the high impedance on the output condition that Feeltech corrected for in the firmware), I've girded my loins and faced this beast of a problem once more.

 This time, I think I've got the correct solution. Remarkably, it turns out that the attenuation required is very close to the 20dB figure you arrived at. It turns out to be 19.69dB (50 ohm pad). This was an "intermediate" figure which I'd confused myself into thinking required a 6.02dB correction factor (up or down - I was undecided initially but went for the extra attenuation which then required me to trim the series resistance down so low that I ended up with a 45 ohm pad).

 I think I'd overlooked the fact that the output voltage under open circuit (no loading on the output) in the range >0.5v <5 volt P-P when the OPA686 is directly driving the output via its pair of 100 ohm resistors which set the output impedance to 50 ohm, the voltage is that at the OPA686's output pin (theoretically, no volt drop in the 50 ohm resistor).

 Obviously, when the pad is switched into circuit at the transition 501mV to 500mV p-p whilst the output remains unloaded, the 85.9 ohms loading from the pad causes a volt drop in that 50 ohm resistor reducing the voltage going into the pad to 63.22% of the OPA686's output voltage which is reduced by another 16.39344% in the pad itself to give a total reduction factor of 0.10364 which corresponds to an attenuation of 19.69dB.

 When I'd reached this point in my earlier calculation attempts, I'd erroneously assumed I'd need to add (or subtract) another 6.02dB to account for the terminated output state on account these calculations had been based on voltage ratios in the unterminated state. Since I figured I'd not properly accounted for the actual parameters of this conundrum, I decided to run the volt drop calculations again with the 19.69dB resistor network values as a sanity check (it had just seemed too easy a solution to be true).

 The new resistor values I came up with are 62 ohms for the shunt elements with a 240 ohm series element which gives you a 19.68dB 50.34 ohm pad. If you can afford (or else select on test) the 61.9 ohm resistors, this gives you a 19.69dB 50.27 ohm pad with a 240 ohm series element. I used the following attenuator calculator:

https://chemandy.com/calculators/pi-attenuator-calculator.htm

along with this on-line dB ratio calculator: http://www.sengpielaudio.com/calculator-db.htm

 You might wonder why I'm chasing attenuation values to within 0.01dB and impedance values within half an ohm of the nominal 50 ohm system impedance. It's critical to exactly match the attenuation given by the 85 ohm in the unterminated state if you don't want to see the output jump up by 3 mV from the expected 500mV after dropping down from the 501mV mark using the 62 and 240 ohm values in place of the perfect resistor values of 61.562 and 238.636 ohms produced by the calculator (assuming 0.5% tolerance 49.9 (or 100//100) resistors on the opamp outputs).

 If Feeltech had used an actual 20dB 50 ohm pad, then the OPA686 output voltage would have been boosted from 500mV to 5v when switched from a 501mV setting to the 500mV level. However, to correct for the error introduced by this 85 ohm pad, they had to reduce the boost in voltage at this transition point to 4.8244v p-p which is what complicates the replacement of the 85 ohm pad with a 50 ohm pad that will minimise this voltage jump (which will now be correct for both unterminated and terminated cases - currently only correct for the unterminated case more commonly associated with audio frequency testing), hence the resulting 19.69dB pad.

 I'm hoping you can run through the calculations to check my own best effort. After getting it so wrong on my first and second attempts over 6 months ago, I don't fully trust my calculations and would prefer an independent check by others that aren't me!

 When I last asked if anyone else subscribed to the FY6600 topic thread could have a crack at this problem, I was met with resounding silence. The real issue with this "Skoolboy Howler" of Feeltech's is that for terminated conditions, this leaves a gaping hole in the range of output voltage levels just below the 500mV p-p settings range (250mV p-p terminated) that can only be filled by using an external 50 ohm 20dB attenuator.

 Thinking about this issue of gaping holes in the voltage level adjustment range, a modest 'backward jump' of a few mV at this 500mV transition point is rather more tolerable than the huge gap that currently exists since you can compensate by dialling the voltage level down a few more mV. If you substitute with a 20dB 50 ohm pad, you'll create a small hole in the voltage settings range whereas a 19.5dB pad will just give you a small overlap which whilst still a little irksome, at least doesn't prevent you adjusting to an exact voltage setting (within 1mV) required to execute a test.

 Anyway, that 20dB pad you're thinking of creating is only 0.31dB adrift from my own calculations so you might want to try the 62 and 240 ohm resistor values I came up with for a 19.68dB 50.34 ohm pad (a little overlap in voltage settings is better than a little gap imo ). This, of course, assumes you don't see a problem with any of my calculations.

JBG

[masterx81]

With rectangle wave set to 99.999% dc, passing from 0.5000vpp setting without load (0.250v at the output) to 0.5001 the measured voltage it's the same to the mv scale. A setting change of 0.0001v is under the DAC resolution. With 50 ohm load obviously the same.
Sorry, but why you speak that we need less than 20db of attenuation? With output voltage of 0.250vpp (0.500vpp setting with dc 99.999) i've measured exaclty 2.500v at the output of the opamp. A drop of ten times it's a perfect 20db attenuation. And actually with the 56+4.7 and 220+27 resistors all seem absolutely perfect.
The only imperfection that now have the device is when switch from the 5v range to the 5.0001, so when the ths kicks in. If i have a measured 1.2478v output voltage, when ths are inserted, increasing from 5.0000 to 5.0001 switch to 1.2635v. Around 16mv error, in any case not too bad. It's different from ch1 to ch2. About this, it's too difficult to fix. Need to made adjustable the gain of the ths, but a trimmer here means introducing some problems. Maybe it's sufficient to replace the feedback resistors with fancy precision units.
I not think that worth.

[masterx81]

Some measurements of my device now, with 50ohm load, 10khz, duty 99.999% (can be assumed dc).
Setting 0.0000v, measured 0.0002 (lost a lot of time with trimmers for have this!)
Setting 0.5000v, measured 0.1243v
Setting 0.5001v, measured 0.1247v
Setting 5.0000v, measured 1.2480v
Setting 5.0001v, measured 1.2648v
Setting 24.0000v, measured 6.036v

Same test, but without load:
Setting 0.0000v, measured 0.0004v
Setting 0.5000v, measured 0.2475v
Setting 0.5001v, measured 0.2497v (2mv error, quite normal without the matched output impedance, but enough precise for me)
Setting 5.0000v, measured 2.4964v
Setting 5.0001v, measured 2.5346v
Setting 24.0000v, measured 12.096v

Ch2 is a lot better as the ths resistor probably are a bit mor in tollerance.

By now, i'm quite satisfied. I hope to not open it for at least some time

[Johnny B Good]

With rectangle wave set to 99.999% dc, passing from 0.5000vpp setting without load (0.250v at the output) to 0.5001 the measured voltage it's the same to the mv scale. A setting change of 0.0001v is under the DAC resolution. With 50 ohm load obviously the same.
Sorry, but why you speak that we need less than 20db of attenuation? With output voltage of 0.250vpp (0.500vpp setting with dc 99.999) i've measured exaclty 2.500v at the output of the opamp. A drop of ten times it's a perfect 20db attenuation. And actually with the 56+4.7 and 220+27 resistors all seem absolutely perfect.
The only imperfection that now have the device is when switch from the 5v range to the 5.0001, so when the ths kicks in. If i have a measured 1.2478v output voltage, when ths are inserted, increasing from 5.0000 to 5.0001 switch to 1.2635v. Around 16mv error, in any case not too bad. It's different from ch1 to ch2. About this, it's too difficult to fix. Need to made adjustable the gain of the ths, but a trimmer here means introducing some problems. Maybe it's sufficient to replace the feedback resistors with fancy precision units.
I not think that worth.

 Many thanks to you for those test results. They more or less confirm my own recent findings (although I've yet to test the actual voltage change on the opamp output pins when the attenuator is switched in and out of circuit).

 The reason for the smaller (19.69dB) attenuation value was simply to match that of the 100/510/100 85 ohm pad's effect when the output is unloaded since Feeltech had seemingly corrected for this departure in the firmware from the 20dB they must have originally intended. My original tests had shown less than 1mV error in making adjustments around the 500mV relay switching threshold in the unloaded test case but a massive 64mV drop in the terminated test case when going from 501 to 500mV p-p settings.

 Both channels exhibited the same pattern to within a mV or two of each other so I assumed from this that Feeltech had applied a firmware patch to correct the error introduced by using an 85 ohm pad in place of a 50 ohm pad. This correction method can only apply to either the unterminated or terminated case not both and it seems Feeltech had elected to correct for the high impedance case typical of audio frequency kit.

 In hindsight, I suppose the 5% resistors they'd selected from the E24 range could have fortuitously created a 20dB pad on both channels in my case rather than the theoretical 19.69dB of a Pi attenuator network comprised of 100 ohm shunt arms with a 510 ohm series element.

 If I had been more thorough in my testing, I could have used the "DC Wave" option and test settings of 250 and 251 millivolts whilst probing the output pins of the OPA686 opamps with a DMM which would have provided definite proof as to whether Feeltech had actually noticed the problem and applied a firmware fix to compensate for the high impedance case or whether it had simply been the effect of wide tolerances in the components they'd used just happening to minimise the error by "dumb luck" alone with my particular example.

 Of course, it's never too late to run this check to properly verify what is actually going on. I may have been giving Feeltech far too much credit for their technical 'expertise' in kludging a firmware bodge to hide this error from the Audiophool community.

 If my (rather belated) testing shows the same voltage jump at the output of those OPA686 opamps of 10 to 1 and 1 to 10 as the attenuator is switched out and in respectively as you've discovered with the FY6900 (the FY6600 could have been different), all the complexity of calculating a resistor network for a 19.69dB pad vanishes at a stroke and I'm left with the much simpler task of calculating for a bog standard 20dB pad.

 Mind you, I'd still have to find four 61.2 and two 246 ohm resistors to make up 19.95dB 50.01 ohm pads out of the E192 0.5% range or else select on test combinations that will create a 20dB pad within 0.5 ohm of 'nominal impedance'. In this case, I think it's preferable to insert slightly too little than slightly too much attenuation since the resulting overlap can be worked around if you need to produce a specific level to within a mV or so.

 The similar issue when the 12dB boost amplifier is switched into circuit for the 5 to 20 volt p-p range is less of a problem since the 16mV error you mentioned with the output terminated corresponds to an equivalent of 1.6mV at the 500mV transition point and, unlike the 500mV case, represents the same dB change whether the output is terminated or not when adjusting the level around the 5v p-p point.

 Like you, I don't think this small error is worth chasing down. Very likely signal/function generators in the "Semi-professional Budget Class" costing ten times the price of the Feeltech will also exhibit a similar issue whenever their 12dB boost amplifier stages are switched in and out.

 In view of your findings regarding the tenfold increased voltage when the 85 ohm attenuator is switched into circuit in the case of your FY6900, the presence of an 85 ohm attenuator across three generations of function generator models defies all reasonable explanation. Its very existence appears simply to to be an oversight of utter technical incompetence of shamefully monumental proportions on Feeltech's part and nothing more (no clever firmware bodge to hide its consequences from the audiophool community after all!  )

 It seems highly probable that the 6600 and 6800 models also applied a 10 fold boost in output voltage to compensate for switching a 20dB attenuator into the circuit so as to sustain the signal to noise performance at output levels below the 50mV p-p mark.

 Of course, this might not be true for the FY6600 but I'll be testing this effect in my own function generator just to confirm the exact truth of the matter before attempting to re-work those "20dB" attenuator networks formed by resistors RS1 through to RS6. TBH, I'll be extremely surprised If my results, which I'll be reporting back to this topic thread (and the FY6600 thread even though they don't deserve to know ), fail to match yours.

JBG

[masterx81]

When attenuation kicks in, it switch from the 0.500vpp (0.5001 setting) to 5.000vpp (0.5000vpp setting). So the attenuator need to output 500mvpp.
I've neither calculated the attenuation factor of the original pad (but i've confirmed the 84.something Zout). I knew Zin, Zout, attenuation factor and done calculations based on this. I not know why feeltech chosen those values.

[BlackICE]

I have fy6900 also and it indeed has the 83 or ohm  resistor problem at switch over point of 0.5v. Also I find it annoying setting the voltage offset to zero you never really get a 0 offset. When trimming the offset to a true zero and changing the amplitude the offset drift off again.

I understand the recommendation to connect the ground lead through a 1k to 10K resistor to PE instead of direct connection. What are the resistor modifications to correct for the 83 ohm problem?

I created an esp32 circuit board to interface with a Silgent oscilloscope to do Bode plots. I initially used the 5 volts coming off the rear connector. But soon discovered that that didn't work too well it created huge amounts of amplitude and timing Jitter of the generated waveforms. I suspect the current available from the 5-volt output isn't nearly adequate for esp32 to that times can draw an excess of 200 milliamps when the Wi-Fi is active and up to 800 milliamps of you turn Wi-Fi and Bluetooth on at the same time. I also added some decoupling capacitors and that didn't seem to help at all. Does anybody have any idea about the true current capacity of the 5 volt output and Lead.

Bty this changing output impedance led me to falsely believe after doing Bode plots that my audio amplifier was changing gain based on the input amplitude.

[Johnny B Good]

When attenuation kicks in, it switch from the 0.500vpp (0.5001 setting) to 5.000vpp (0.5000vpp setting). So the attenuator need to output 500mvpp.
I've neither calculated the attenuation factor of the original pad (but i've confirmed the 84.something Zout). I knew Zin, Zout, attenuation factor and done calculations based on this. I not know why feeltech chosen those values.

 We'll never know for sure why they chose to approximate the required 20dB for the unterminated case using 100 ohm shunt resistors with a 510 ohm series resistor. We can speculate that it could have arisen from an act of bean counteritus (cost savings) and an overstock of 100 and 510 ohm resistors which just happened to fit the bill (give our audiophool customers the required unterminated attenuation and let the Devil take the hindmost (those needing to drive 50 ohm loads to get valid test results),

 An alternative speculation is that their computers and/or access to the on-line attenuator calculators were down that day and the "designer" had looked at the most likely and abundant resistors they had in stock (favouring the massively overstocked components) and trialled some test calculations that would at least serve the high impedance case and earn him a "Pat on the head" from the boss or chief bean counter for his ingenious cost cutting solution. 

 Yet another possibility is that an error in calculating the resistor values (the "Skoolboy Howler" hypothesis) had simply gone unnoticed for lack of proper qualification testing  before committing the design to production.

 And one for the more cynically minded is that it was a deliberate error, planted as an "Easter Egg" challenge to their largely cash strapped home hobbyist market demographic to solve as yet another warranty  voiding temptation to keep customer support costs as close to zero as they possibly could.

 That's four possible reasons for those chosen values (all pure speculation for my amusement off the top of my head) you can choose from. If you can think of any others to add to that list, then pray, do please tell.

JBG

[Johnny B Good]

I have fy6900 also and it indeed has the 83 or ohm  resistor problem at switch over point of 0.5v. Also I find it annoying setting the voltage offset to zero you never really get a 0 offset. When trimming the offset to a true zero and changing the amplitude the offset drift off again.

I understand the recommendation to connect the ground lead through a 1k to 10K resistor to PE instead of direct connection. What are the resistor modifications to correct for the 83 ohm problem?

I created an esp32 circuit board to interface with a Silgent oscilloscope to do Bode plots. I initially used the 5 volts coming off the rear connector. But soon discovered that that didn't work too well it created huge amounts of amplitude and timing Jitter of the generated waveforms. I suspect the current available from the 5-volt output isn't nearly adequate for esp32 to that times can draw an excess of 200 milliamps when the Wi-Fi is active and up to 800 milliamps of you turn Wi-Fi and Bluetooth on at the same time. I also added some decoupling capacitors and that didn't seem to help at all. Does anybody have any idea about the true current capacity of the 5 volt output and Lead.

Bty this changing output impedance led me to falsely believe after doing Bode plots that my audio amplifier was changing gain based on the input amplitude.

 Go to https://chemandy.com/calculators/pi-attenuator-calculator.htm and plug in the values 20dB and 50 ohm. This will give you the perfect values but you can also specify 'preferred' resistor values from the E12, E24, E48, E96 and E192 tables here: https://chemandy.com/technical-articles/single-pages/preferred-resistor-values.htm.

 You can also try different value combinations in the "Preferred" values entry boxes from the less exotic resistor tables or do what masterx81 did and make up the required values with pairs of resistors in series (or alternatively use parallel resistor combinations).

 The use of a 20dB attenuator for voltages of 0.5v p-p or less (and likewise, the use of a 12dB boost amplifier for the voltage range 5 to 24v p-p (5 to 20v p-p in the 6600 and 6800 models)) provides a 'gearbox' function for the 14 bit DAC 'engine' to maintain a better match to the available s/n performance (notably the issue of quantization error noise that becomes increasingly intrusive as lower voltage outputs are selected).

 This use of relay switched attenuators (and boost amplifiers for signal generators specified to deliver more than the 5v p-p limit of lesser specified devices) will often introduce a discontinuity of level as they're adjusted through the switching thresholds due to slight mismatches between the attenuator and tenfold boost of signal voltage intended to precisely match the additional 20dB attenuation when going from the 501mV setting to the 500mV setting (and vice versa for the opposite direction). This calls for a very tightly specified 20dB attenuator to hold such discontinuities to an acceptable level (and assumes a very precise tenfold increase or decrease of the output voltage from the opamp feeding this switched attenuator).

 masterx81 has already done the necessary testing to prove that the required attenuation is the classic 20dB with an associated order of magnitude step change in level  from the opamp's output applied at the switching points in the amplitude settings range so you can proceed with some confidence that replacing resistors RS1 to RS6 with values obtained using the attenuator calculator will resolve the problems you've been experiencing.

 As for the power issues you've been suffering with the esp32 module, all I can advise is that you stop abusing the poor old FY6900's PSU board and use a separate 5v psu to power your esp32. There was room for some doubt as to whether you were abusing the 'scope's or the signal generator's 5 volt PSU rail - the signal generator seemed the most likely candidate for this abuse. However, whichever it was, stop doing it! Use a dedicated 5v supply (and link the common zero volt rails as required).

JBG

[BlackICE]

Johnny thanks for the tips. I was only abusing the function generator which is cheap and disposable. The oscilloscope connects to the esp32 through Wi-Fi and have no physical connection whatsoever. It's too bad that the function generator cannot output enough 5 volts to run the ESP32. Otherwise it would be nice slick solution to just add a small board plugged into the back and have full support for Bode Plots. Are there any easy adjustments for the DC offset? It seems to be all over the place and amplitude dependent.

The thread below talks about how do Bode plot using a FY series function generator. The first solution was to have a PC act as a function generator simulating a Siglent function generator then translating those commands to an FY function generator through a serial port. The second version used a precursor to the ESP 32 the ESP8266 to act as a Siglent function generator on a LAN using Wi-Fi. I took the Wi-Fi code and modified it to use esp32 and added support for using DB units.

https://www.eevblog.com/forum/testgear/siglent-sds1104x-e-and-sds1204x-e-bode-plot-with-non-siglent-awg/

[masterx81]

The psu of the fy6900 is the same as the fy3200... not so nice. And for sure can't keep much more load on 5v line than the 500ma required by the logic of the  unit.

Question: it's safe to connect the ground through an 1k resistor? By now i have a linear psu with ground tied to earth. Would be nice to put a switch, but for now i not need it.
Higher end unit how handle the grounding problem?


For setting the offset and amplification i've used a long procedure.
For first, go in dc waveform, set volts to 0, and turn the w1 or w2 (depending on channel) for have 0v on the output. Then set rectangle waveform, 5.000v setting, and turn w16 or w26 until with duty set at 99.999 and duty set to 0.001 you not have a perfect specular value (for example 2.550 and -2.550). After this, set dc at 99.999,  move the gain (w3 or w5) for have a value half the error (as the previous example, 2.525). Then set the duty back to 0.001 and fix the offset as before until with dc at 0.001 and 99.999 the measure isn't the same. Then recheck with waveform in dc if the 0v is remained at 0 (very unlikely) and restart the "loop". I've lost a lot of time for a good result.
Moving the dc "wave" offset moves a bit also the rectangle wave offset.
Do it after at least 30min the device is powered on and at operating temperature. Adding heat sink to everything and a temperature controlled fan lowered the change with temperature, but something changhe

For attenuation pad i suggest a series of 56+4.7ohm instead of 100ohm, and a series of 220+27ohm  instead the 510ohm. Both 0.5ohm under the perfect match but for me are enough, are a bit a pain to be soldered. If you find a way with parallel resistor would be easier, and let us know

[BlackICE]

Thanks, masterx81

How close do you get your voltage offset to be for all that work?

I haven't opened the box yet so it would be helpful if you can tell me the correct size of the resistors to fit the pads. I don't think voiding the warranty is a big deal since I appears there isn't one! I asked my seller and email feelelec about my sin wave got corrupted. No answer from feelelec the seller said they will look into it and get back to me. Haven't heard anything since. Later when I manage to update the PC software (need to run in admin mode so that China can spy on my PC?) I found a menu item that as labeled "repair_sine" that fixed it. (for how long?)

Johnny was giving a number of reason of why the wrong resistor values may have been used. Given that they are using super old NEC relays, I would say what they can get cheap, or had plenty stock of, who cares no one will find out and if they do ha ha no warranty.

[masterx81]

Some measurements of my device now, with 50ohm load, 10khz, duty 99.999% (can be assumed dc).
Setting 0.0000v, measured 0.0002 (lost a lot of time with trimmers for have this!)
Setting 0.5000v, measured 0.1243v
Setting 0.5001v, measured 0.1247v
Setting 5.0000v, measured 1.2480v
Setting 5.0001v, measured 1.2648v
Setting 24.0000v, measured 6.036v

Same test, but without load:
Setting 0.0000v, measured 0.0004v
Setting 0.5000v, measured 0.2475v
Setting 0.5001v, measured 0.2497v (2mv error, quite normal without the matched output impedance, but enough precise for me)
Setting 5.0000v, measured 2.4964v
Setting 5.0001v, measured 2.5346v
Setting 24.0000v, measured 12.096v

Ch2 is a lot better as the ths resistor probably are a bit mor in tollerance.

By now, i'm quite satisfied. I hope to not open it for at least some time

The 0v is under the mv range.
My calibration procedure was also a bit harder than normal because i've added 2 trim pots for regulating the gain of the dc offset opamp (for have an almost perfect offset).
Resistors are 0805. If you see my previous post did you see how i've soldered them for have a series. All can be done also in parallel (that's better) but you have to do a lot of calculation do find the right values.

[BlackICE]

How about this combo.

250 plus 60.4 ohms

https://www.digikey.com/products/en/resistors/chip-resistor-surface-mount/52?k=805+resistor&k=&pkeyword=805+resistor&sv=0&pv2085=u250+MOhms&sf=0&FV=3%7C1131%2C3%7C1684%2C3%7C2106%2C3%7C2503%2C3%7C848%2C3%7C976%2C7%7C2%2C7%7C3%2C1291%7C39328%2C-8%7C52&quantity=&ColumnSort=0&page=1&stock=1&pageSize=25

https://www.digikey.com/product-detail/en/panasonic-electronic-components/ERJ-PB6D6042V/P21222CT-ND/6215477

[masterx81]

Would be ok but both aren't on common values (and i not have them in house). If you need to do an order in any case, as you haven't nothing in 0805 format, order them

Be aware! Those links are wrong! 250megaohm! 60.5kohm!

[BlackICE]

I clicked too soon without reading carefully. The 250 is pricey but only need 2 of those.

I don't have any SMD devices. Never soldered any SMD parts! To remove the old ones I planning to use 2 soldering irons to lift the old part off like using chopsticks. I haven't really done much electronics for decades and have avoided SMD better they seem to be a PITA compared to through hole. However I noticed that places like JLCPCB solders common SMD parts for you at very low cost that makes SMD more attractive.

https://www.digikey.com/product-detail/en/koa-speer-electronics-inc/RK73H2ATTD60R4F/2019-RK73H2ATTD60R4FCT-ND/10234995
https://www.digikey.com/product-detail/en/stackpole-electronics-inc/RNCF0805TKY250R/RNCF0805TKY250RCT-ND/4928401

[masterx81]

Here in italy you can buy a sample set of 25 resistors 1% 0805 composed of 170values (scale 1.0 1.2 1.5 1.8 2.2 2.7  3.3 3.9 4.7 5.6 6.8 8.2 and multiple) for around 35e shipped. In usa i suspect a lot cheaper.
If you plan to use them only for this work order only what you need.
If you plan to do some smd repairing a set i handy to have.
I use only one soldering iron, but need a bit of practice

[masterx81]

Thanks to the calculator in this software
http://lygte-info.dk/project/TestControllerIntro%20UK.html#Notes_&_Download
I've found some better values. For 61.1 ohm 39+22=61 (instead of 56+4.7=60.7) and for 247.5 a parallel of 390 and 680=247.85 (instead of 220+27=247), all in common E12 values. I not think that change that much, but it's a bit better