HP 33120A waveform is distorted. Video inside. What is going on?

[king.oslo]

Hello everyone,

I bought an HP 33120A on eBay, and when I got it, there was a problem with the output waveform. It seems the bottom of the waveform is being clamped somehow. 

I've made a short video introducing my problem. I hope you are able to watch it, and tell me what you think may be going on, because I have traced the problem to the output amplifier, and I cannot quite work out what is happening there which is causing the problem.

This is a link to the service manual: http://cp.literature.agilent.com/litweb/pdf/33120-90017.pdf
The schematic is on page 137. The block diagram is on page 87. (The diagram for the whole instrument is on page 129.)

This is the video:



Please, if you are able to make any suggestion, I will measure/check it out.

Thank you for your time.

Kind regards,
Marius

[w2aew]

Interesting that the problem goes away at the amplitude range change.  Is there some sort of output attenuator that switches at the range limit?  It's tough to see the schematic in the video, but it looks like a multi-stage differential amplifier.  Check to see what gets varied (bias voltages, DAC output, bias currents, etc.) when the amplitude is adjusted. 

I hope to look at the service manual over the holidays to see if I can give you specific things to look at.

[alm]

Based on the output clipping only as it approaches a certain fraction of full scale, my first guess would be that one of the transistor amplifiers responsible for the negative part of the signal is bumping its head, possibly due to an incorrect bias voltage (eg. failed bias transistor). It would be good to figure out whether the signal is inverted somewhere along the way before you waste time testing the wrong half of the circuit.

[c4757p]

one of the transistor amplifiers responsible for the negative part of the signal is bumping its head

I must add this to my growing list of EE anthropomorphisms.

[king.oslo]

Merry Christmas everyone,

I found a resistor (R771). It doesn't look healthy. Look at it. The blue label has flaked off. And the trace and pad underneeth has lifted from the board, yet seems to be connected to the circuit. But I desoldered it, measured it, and soldered it back. It was circa 28 ohms (the schematic specifies 26R1). The voltage drop across it is much greater than the resistor R772 which is the low side equivalent. The drop across R771 is aprox 800mV, while the drop across R772 is about 300mV.

Interesting that the problem goes away at the amplitude range change.  Is there some sort of output attenuator that switches at the range limit?  I hope to look at the service manual over the holidays to see if I can give you specific things to look at.

There is an output attenuator, but it only has a bunch of resistors, I do not think it could be clipping the waveform like that.

Second, yet most importantly, the error happens in the output amplifier. By the time the signal gets to the output attenuator, the waveform is already distorted.

Thanks.M

[free_electron]

Dc offset problem which makes the thing go in clipping. Ive seen this before.

The output amplifier works in current feedback mode. Measuring voltages gets you nowhere.... These things are hell to troubleshoot. Let me check my notes. Check the transistors in the first stage. Measure the voltage across collector emitter. If one clamps ( staying in full conduction ) you need to trace the base path to the output stage (the feedback). Most likely one of the sot223 or so8 transistors there is fried .

They inject the ac waveform and the dc offset as currents in this amplifier.

Another thing to check is : does it do this only on sinewave or also on square and triangle ? You may have a broken relay in the filterbank. Ive seen that happen too. One side goes wonky so the signal is no longer differential ... And then you get crazy stuff... These are simple telecoms relays with mechanical memory (set reset coil) and they do go bad over time.

[king.oslo]

Hello everyone,

I found something, but I don't want to attempt a desolder something in case I have made a blunder in my diagnostics (which I probably have)

Please watch the video and tell me what you think:

These things are hell to troubleshoot.

 

Another thing to check is : does it do this only on sinewave or also on square and triangle ?

Yes, it does.


Thank you for your time.

Kind regards,
Marius

[free_electron]

Most likely the base emitter of the bottom transistor in the bridge is shot. The bridge is the part where you have a transistor going to 3 resistors going to another transistor. About 2/3 to the right in the schematic .  You point at it at 1:31 in the video. But not the top transistor. My speculation is the bottom transistor ( move your pen down to the other side of the resistors.

This amplifier is special. The whole thing works in current domain. The goal is to keep opamp at the input keeps to get its output to be at 0. Any deviation there results in current injection.

That bridge you point at is the summing point. If you would carve those resistors in half the center tap should be at zero.  Think of it like an instrumentation amp. The resistors set the gain and the center holds the common mode ( which should be zero ).  This is exactly how this thing works but then in current domain.

Depending on the age of your machine these transistors will eithe be mpsa54.. Or bfy9. Series.  The mpsa54 is unobtainium... Discontinued by motoroil when they became on semi.  The good news is you can put in the bfy , but it is a different package..... However, it is possible on the pcb.  I have those transistors if you need them. Too bad i sold mine, i could have taken a picture on how to get it in there. I had one of these with the same problem and repaired it, sold it a couple months ago when i got the 33520.. I have the different versions of the schematic as well.

[king.oslo]

Dear Free_electron,

Are you referring to Q710 or Q715? I attached a schematic underneath to make it easy to decide 

I did the diode forward voltage test, and base collector of Q715, and its V_BE and V_BC were fine. However on Q710 V_BE was 0.223V! Is that it? That is one of those tiny SOT23-3 packages.

The transistor has this laser engraved on its outside: "3D 4"

Kind regards,
Marius

[Jay_Diddy_B]

Hi,

I think your drawing for the pin connections is wrong.

The E and C are switched.

The datasheet can be found here:
http://www.fairchildsemi.com/ds/MM/MMBTH81.pdf

This is the standard pin out for a SOT-23 transistor.

The MMBTH81 is a faster transistor. You can substitute a MMBT3906 (or similar) for now to see if the circuit works and then get the MBBTH81 later.

Regards,

Jay_Diddy_B

[free_electron]

The bridge is indeed q709 and q710.  If you got 0.2 there it means its game over for that transistor.

The way this thing works is that the whole amplifier pumps current through the three paralleled resistors. Of you were to cut this virtual 2.2kiloohm resistor in half so you get two 1.1k resistors in series , then the centerpoint should be ground if there is no DC offset requested from the output .

Depending on the current flowing through this virtual resistor the transistors q715 and the others are driven more or less. To create a dc offset at the output the regulation loop creates an imbalance by playing with q709 and q710.

All the transistors following this 'bridge' are purely for current gain so the machine can drive 50 ohms. There is no more voltage gain.

If you look at q711 and q715:  remove q709 and q710 (imagine they are not there. Gone ). Since q711 is a pnp there will be current coming out of its base , going through the three resistors in parallel and into the base of q715 which is an npn. So if q709 and q710 were not there , the transistors q711 and 715 would turn themselves on. This would also turn on q712 and q717 which will also deliver current through the three 6.8k resistors. A large current is now flowing through r726 and r727 which are a mirror of my two virtual 1.1k. If there is no dc offset then the voltage across the 1pf there is zero. If there is a dc offset then you will see it across the 1pf.

This whole stage is a current amplifier to drive the two push pull circuits.q712 q714 and q716 q717.
These two half bridges drive the real power transistors ( the thru hole beasties with a heatsink on them. )

So this whole thing forms a current mode very fast push pull amplifier. So all q709 and q710 do is inject current (the top one q709) and deviate current (q710).  By controlling how much qou control both ac amplitude and dc amplitude. If they conduct in pure antiphase you create an ac signal only. If you add a bit more to 709 than q710. You lift the voltage across the three 6k1 towards positive side thus making the top bridge conduct less and creating a dc offset one way. Suck more current than you inject by driving q710 harder and you drive the bottom bridge hards, creating a dc offset the other way.

The regulation loop is before q709 and q710. The rest is pure current gain to drive the power transistors.

From your schematic you have the newer version with the philips transistors . Mine had motorolas. Mrfa5415 which is obsolete. You have the pzt3906 and pzt3904. Those are standard and still in production.

Desolder q715 and its twin on top as well as the top and bottom transistor of the half bridges and see if you get a correct signal across the three 6.8 k resistors.  In my generator not only q710 was dead but also one of the ones in the half bridges.

[king.oslo]

I noticed another thing. Q709 V_BC is precisely .223V that is spot on the millivolt of Q710 V_BE. That is odd.M

[free_electron]

Replace the pzt3906 and 3904s as well as the two drivers 709 and 710. Itll cost you 3$ and the machine will work fine again. No point i. Spending 5 hours trying to find the exact dead transistor.  These things run hot in their lifetime and get a beating.... Hp should have put a heatsink on em...

[SeanB]

Just mount the TO92 parts there and they will be better at dissipating heat as well.

[king.oslo]

Thank you so much guys!  Absolutely, really grateful for your help. Thank you so much free electron I will replace all these parts, and let you know how it goes.

I also found some serious 2khz spikes on the +18V rail.

I suspect this is due to something switching on and off at 2khz. When I disconnect the front panel, it goes away, so I will work out what is causing it. But if you already know, please let us know if it is normal, or requires attention. Thanks!M 

EDIT: It seems the only thing the display board which is +18V is the anode and grid voltage. I must desolder the VFD, (which looks daunting) to be able to modify the circuit underneath. The VFD is soldered on top of all the display circuitry. If I did this I could have added more capacitance near the input to the display. I hope that you all agree that these spikes which BTW also supply the output amplifier is not problematic. Thanks.

[king.oslo]

Hello everyone,

Since last time, I have replaced a bunch of transistors. The rest were desoldered and measured before soldering back in.



The one marked in red was desoldered, no matter where I measured it base-emitter, emitter-base, emitter-collector, I measured in all the combinations possible, and always got my multimeter showing OL. I replaced it with another PNP transistor I had laying around.

The output is still clipping.

What do you advice that I check next?

Thank you for your time.M

[free_electron]

check on R704 ( at the opamp ) that you have an UNDISTORTED signal.
that will eliminate some other problems.

then check the collecotr resisotr sof the endstage transisotrs( like the red one ) as well as the zeners in the collecctor line... verify you got the right voltages there.

[Pat Pending]

I'm wondering if the DC offset has skewed the output signal too close to a negative supply rail. So I'd try this:

Set the waveform amplitude to zero, then sweep and measure the DC offset at the output. Should be linear.

Connect the opamp +, R701, R702, R703 node to AGND. Output waveform should be centered about 0V.

[king.oslo]

I'm wondering if the DC offset has skewed the output signal too close to a negative supply rail. So I'd try this:

Set the waveform amplitude to zero, then sweep and measure the DC offset at the output. Should be linear.

Connect the opamp +, R701, R702, R703 node to AGND. Output waveform should be centered about 0V.

Thanks for your reply. I'd really like to check what you are suggesting. My problem is first of all, I am unsure I understand what you are asking. Do you want me to set amplitude to 0? (this is not possible, it will only go as low as 50mV_PP


Second: The offset doesn't change at the output of the instrument. It only changes when the instrument changes range. Consequently, nothing changes when I short NI+ node of the opamp. Furthermore, when I probe the waveforms and ni+ and in- of the opamp, and I notice that the DC offset is only visible at ni+. The DC shows no offset. What are your thoughts?

check on R704 ( at the opamp ) that you have an UNDISTORTED signal.
that will eliminate some other problems.

then check the collecotr resisotr sof the endstage transisotrs( like the red one ) as well as the zeners in the collecctor line... verify you got the right voltages there.

All the resistors on in the output amplifier is the correct value.

I measured the input to the opamp. My Rigol scope says that Vmin is -424mV and Vmax is 432mV. Is this DC offset no problem, you think?

[Shuggsy]

check on R704 ( at the opamp ) that you have an UNDISTORTED signal.
that will eliminate some other problems.

then check the collecotr resisotr sof the endstage transisotrs( like the red one ) as well as the zeners in the collecctor line... verify you got the right voltages there.

All the resistors on in the output amplifier is the correct value.

I measured the input to the opamp. My Rigol scope says that Vmin is -424mV and Vmax is 432mV. Is this DC offset no problem, you think?

Is the signal at that node (or the other side of R704, net name of +AMP_IN on pg.137, the output amp schematic) UNDISTORTED, as free_electron said? That is, as you vary the output amplitude, you should essentially see whatever sine wave you're trying to output, I think... just at a lower gain than the final output stage. Best to measure that with a high impedance termination on the scope (in case you happen to be using a 50 ohm termination for these measurements. Not sure if you are! No worries ).

Since that is the (non-inverted) input to the output amp stage, if you see a clean signal there then the problem definitely lies within the output amp. You already noted the distortion/clipping exists at the input to the output attenuator, so if the signal is good when it enters the output stage and bad when it exists... the problem is within!

If the signal is clean at that node, then keep going with what free_electron said. If the signal IS distorted, then check back to the pre-attenuator stage... I'd just keep working your way back until you see an undistorted signal, if possible. Always good to check things from both the input and output sides... keeps narrowing down where the issue is.

My money is on some obscure behavior in the output amp or something odd in the pre-attenuator stage. Out of curiosity, does varying R710 have any affect on your issue?*

*EDIT: Just read a little more on the manual. There's a cool write-up on the theory of operation for the output amp on pages 87-88 of that manual. I'm doubtful varying R710 will do much now, but if it's not sealed in some set position then it likely won't hurt to check. Also, it may pay to check the bias current setting resistors of R734 and R727 on the low side of the output stage.

[king.oslo]

The waveform is not distorted before the output amp. It has is a few mV offset, but not clipped or distorted. I think the problem is in the output amp too. But now I have measured all the resistors, replaced more than half of the transistors and measured the rest of tem. Is my best bet to canalization the output amplifier with hot air, soak it in flux and rebuild the whole amplifier from scratch?   

Adjusting R710 offsets the output signal a bit, so it lifts itself a bit out of the clipped range, but only by about 3 percent of the amplitude.

Thank you for your time time

Kind regards,
Marius

[Pat Pending]

Since the output amp is symmetrical try taking voltage readings at node pairs relative to AGND, they should be similar in magnitude assuming the offset can be zeroed.

[king.oslo]

Hello again,

I made the decision to change both Q713 and Q718 (the big TO-39 output transistors with heatsinks), and I was very, very pleased to observe that the waveform was no longer being clipped.

But the offset functionality still doesn't work. When I offset the waveform, I expect that the current-injection-opamp should shift the output up, and both its inputs should be equal beause it is in negative feedback. This is not the case. The opamp doesnt have the same waveforms on both its inputs. I reckon I will try and replace it. One time in the past whilst probing the output of this opamp, I managed to short its V+ power rail to its output with the scope probe. I remember seeing on the scope that the offset of the output waveform made a small jump of about 10mV. A few days after I found that the offset functionality wasnt working.

Hopefully after I change this opamp, I will once again have a working instrument.

Thank you to everyone whom helped me; thank you!  %-B

Kind regards,
Marius

[fran]

Hi all,

I know this is a zombie thread, and I am new here but I am seeing a similar problem to the OP so I thought it best to post here rather than start a new thread.

I picked up this hp3320a locally, and it has been working fine until this week - the symptoms are clipping of the top of the sine wave (fairly constant once you go above a lowish amplitude). It is clearly happening on sine, square, sawtooth waves but its not visible on square wave. Amplitude is accurate enough, as is frequency.

So I pulled the manual and started looking around. The signal into +amp_in and -amp_in are clean. I have a DC offset of ~200mV - this jumps in steps as the relays switch to each output range - doubling to 400mV, then 800mV etc.
So I'm a bit confused as to where to go next - and I'm not very experienced at this, so apologies in advance if I have missed anything.

Measurements:
U702 - AD711 - +/-15V present on pins 4 and 7. Pin 5 (output) has -6.5V , pin 2 (-input) -2mV, pin 3 (+input) -2mV.
R702 - which I think is in the feedback circuit - one side has 2mV (which must be the U702 side) and the other side has 6.1VDC.
Q701 and Q702 - see photo, I have clean signal on the base of these, but some distortion on the emitter of each. I'm seeing +/-13V on the collector of each, and -6mV and -3.6mV on the base of each.

I've done the transistor checks (in circuit mind you) around the amplifier and all seem OK - ie no dead shorts.
 
The resistors around the output (R834/R815/R835 all 69r) at K804 have gotten hot at some point,  and I lifted these off the board and they all measured perfectly.

I pulled Q713 and Q718 which seemed to be the OP's problem, but they measure fine in a cheap transistor tester.


I'm a bit stuck as to where to go next to try save this nice piece of test equipment, any help is much appreciated.

[strawberry]

feedback loop measurements are tricky because feedback will compensate actual problem

output transistors are most likely