How to calibrate Siglent SPD1000X / SPD3303X / SPD3303X-E series power supplies?

[mawyatt]

Also noticed the long period to stabilize for the 34401A. Think this is because they don't have a fan, while the other high end modern DMMs all have fans which help achieve thermal equilibrium quicker.

Best,

[oz2cpu]

bad driver, you cant use the auto range.. need to make it 200uV always, else the curves will confuse us :-)

here is the left unit, maybe we should just come down back to earth :-)
noise of 0.1mV is NOT bad, not at all..

[C.J.S.]

maybe we should just come down back to earth :-)
noise of 0.1mV is NOT bad, not at all..

And well below the ripple+noise spec of ≤1mVrms（5Hz ～ 1MHz).
Although you would need an AC mV meter with adjustable bandwidth to measure this ripple+noise spec.

[Bad_Driver]

Ok, I used a boring TEAMs meeting to do a first measurement.

Siglent SDM3065, 2 V DC, 10 PLC, Auto-Zero, !0 Mohms
TestController software, 1 reading per second, 12 min

first photo: raw data
second photo: average over 20 readings

As stated by oz2cpu we see a kind of oscillation.

Edit: Pics in better quality

[mawyatt]

This does seem to have a low frequency and much higher frequency content in the waveform. An FFT on the data should revel the details and show these.

Best,

[oz2cpu]

exactly fft will tell us, and then it can be fixed (by modding or by siglent update the design?)
however if sample time is too low ? the resulting fft could fool us
this is why i played so much with recording time / sample interval

(all this amimated me to look into making a super nice stable and noise free reference for my test equipment experiments)

[Bad_Driver]

I started a 2h (3 measurements/sec) recording, than we have about 20.000 samples.
Has someone a suitable software for FFT by the hand?
I will provide the file as attachment asap.

Edit:
2h record, 3 Measurements/sec attached - feel free to do the FFT
(file renamed to txt, CSV-format)

[Bad_Driver]

ok, I've done a short self learning session regarding FFT with Excel.

Attached my first step. It needs deeper investigation and I'm considering a new series with 1 PLC / 50 readings per second.
As we learned it needs external triggering to get fast readings. I do not have time these days, maybe someone else can try it.

x=frequency, y=amplitude
As expected some peaks.
Any comments?

[oz2cpu]

in the quest to understand what is a good powersupply, i found a few more of my old things...

this one Oltronix stabpac mca 5-3 is from 1973 at least that is what I see written on the schematic
it is very hard to date this exact unit, from its serial number ? when was they sold ? start ? stop year ?
video detail : https://youtu.be/pspuqOsWG44
now see this power on curve from cold.. +/- 1mV scale..
also : its noise and regulation is in the crasy low, wtf :-)
cant even see 0.1mV drop when load change from 0-3A
it is possible... just saying

[Bad_Driver]

Hi oz2cpu

I saw your video regarding this old PSU, very impressive!

Since Santa Claus put a new toy (SSA3021X+) under the tree I didn't‘ spend more time with the SPD these days.

Yesterday I got an very old idea in mind for the simplest check at least (don‘t lough plz!) I connected the PSU outputs DC-BLOCKED
with the LF-input of my batterie powered speaker system (avoiding grounding problems) to listen to any noise - nothing!

I have to put more efforts in it, I was impressed in your video how good automatic triggering of your SDM3055 runs (the green flashing dot on the top left), with the 3065 it is a night mare…. as mentioned in the SDM thread I tried as suggested external triggering but did not get what I expected.

[oz2cpu]

>I saw your video regarding this old PSU, very impressive!

thanks alot, i had a lot of fun and work out of the making of that video, (if you like teardown videos, see my youtube)
if anyone ever seen this type of zener implementation before ?? please post link ?
I really like to know more about this genius opamp design.

>SSA3021X+
HAHA lucky you.. I got one too, super happy about it

>Yesterday I got an very old idea in mind for the simplest check at least (don‘t lough plz!) I connected the PSU outputs DC-BLOCKED
>with the LF-input of my batterie powered speaker system (avoiding grounding problems) to listen to any noise - nothing!

no one is laughing, this is a perfectly good idea,
I only see one problem : it reveal only noise in the audio range, that type of speaker system can handle,
small pc desk top speakers, are like 100Hz to 10kHz
if you cant hear anything, add more gain ? external opamp ?
actually such noise amp, headphones, could be used to debug alot of things.
but most problems with emc, switchmodes, regulation , stability, is not in the hearable range, either 0-100Hz or over 10kHz..

> I was impressed in your video how good automatic triggering of your SDM3055 runs (the green flashing dot on the top left)
I am not sure i understand what you mean ?
auto trig, is just freerun mode, the meter is not synced to anyting,
I just resently had an SDM3065 for loaner, did not see any problems with it, when my 3055 came back from repair, i had to let it go,
i actually missed it a bit, due to its one more digit resolution in low ohm range, and that is where i play the most at the moment

[mawyatt]

"thanks alot, i had a lot of fun and work out of the making of that video, (if you like teardown videos, see my youtube)
if anyone ever seen this type of zener implementation before ?? please post link ?
I really like to know more about this genius opamp design."

Yes this is a very old and clever technique, National even had a hybrid reference that used it back in ~70s. If you can find the old National Linear Databooks, it's shown in those. Almost all the zener based references use it.

Best,

[2N3055]

Bootstrapping zener like that by supplying it with self regulated voltage is used a lot.  Speaking of Walt Jung, he used it in his super low noise PSUs. OTOH, it might seem common now, but whoever though of it first to create what is basically a recursive regulator (in software parlance), that was quite clever...

[oz2cpu]

exactly 3055, it is clever designed,
and the performance speak for it self as explained in the video,
lets challenge any of your brand new, ultra expensive powersupplies, are they just as stable ?
from cold power on and a few hrs ahead ??

I have designed electronic products for 35 years
maybe i did not pay too deep attention to details, when i came to voltage zener refferences :-)
I have not seen this coupling before,
dont you just love it, when something as normal as a teardown, ends with wisdom,
It must be named something ? there must be a tech sheet puplished ?

[mawyatt]

Like 2N3055 mentioned, likely called a "Bootstrap Regulator or Reference". This has two stable voltage states, zero and the zener referenced, the 2 resistors and diode force the opamp into the desired zener state. The old 1N821 thru 1N829 series reference diode circuits used this technique as well.

Best,

[Bad_Driver]

The trigger problem is described in

https://www.eevblog.com/forum/testgear/siglent-new-bench-dmm-sdm3055/msg3466342/#msg3466342

With the SDM3065 you easily see it with the scope connected to Trigger Out. Thats why I fiddled with external trigger etc.
to get constant readings for FFT.

I know that the audio check has very limited BW. For PSU measurement I built a Low Noise preamplifier ( 10 Hz…. 100 kHz) some weeks ago
and with my 192 kHz, 24 bit semi professional external sound card and the software AudioTester you can do FFT up to 90 kHz. For higher frequencies -> SSA3021X/SVA1032X   

Unfortunately my design (2 stage OP amp for 60 dB) tends to oscillate and now I have a 3 stage design on the bench but not finished.

[2N3055]

Like 2N3055 mentioned, likely called a "Bootstrap Regulator or Reference". This has two stable voltage states, zero and the zener referenced, the 2 resistors and diode force the opamp into the desired zener state. The old 1N821 thru 1N829 series reference diode circuits used this technique as well.

Best,

Absolutely correct,  it probably originated with old ref zener diodes as a solution to run them at very specific zero tempco current, combined with hand picked good stability resistor..
But since even LM399 has some current dependency, it is used with LM399 too, to make it even more stable...
Walt Jung used it to make his quiet linear regulators have excellent PSRR by virtue of feeding voltage reference with already regulated voltage... It's a gift that keeps on giving :-)

[mawyatt]

I have not seen this coupling before,
dont you just love it, when something as normal as a teardown, ends with wisdom,

Very powerful learning experience, even for older folks

Recall tearing down the HP3325 Function Generator when it first arrived way back, and curious how they achieved the ultra linear triangle waveform. They used a couple ECL gates and a fixed 100MHz (think it was 100) crystal oscillator for one gate input, the other input came from the frequency synthesizer which used the same 100MHz as a reference. The ECL gate output pulse width @ 100MHz varied as the frequency difference and was low pass filtered by a passive LC filter into an almost prefect triangle wave, which was amplified and offset scaled accordingly. Brilliant design concept!!

Best,

[Bad_Driver]

I wanna come back to our SDP3303X.

I have to admit that it is not possible to get constant readings with the SDM3065, the time deviation between the readings during recording is very high - doesn't care if I use EasyDMM, USB-disk, internal recording which is missing a time stamp. Also external triggering doesn't help. Do the 3055 users have the same problem?

Once again, Siglent, shame on you!

So I used my 25y old 34401a, connected via Agilent GPIB-adaptor and the old but free software "Agilent DMM". I got very constant readings @ 16.5 Hz with 1 PLC and a far better FFT analysis of the SPD3303X output.
Voltage = 1 volts since this fits best to the 1V range of the 34401a. FFT was done with EXCEL. Both devices are running for 24 hours.

You see now much better the oscillating (yes, it is still in specs) of the output, I think it comes from a internal feedback loop of the regulator.
I'm interested what you think?

[oz2cpu]

is it recalculated to HZ the bins ? of the fft ?

and YES i also belive it is regulation loop that is not performing uptimal,
any one got a schematic ? come on siglent, we offer to fix this for you, for free,
so this super nice supply, can be 10 times better performance :-)

[Bad_Driver]

Yes, X is in Hz. Linear looks better than in Log.
A schematic would be helpful.

oz2cpu, can you be so kind to check the behavior of your 3055 during recording with respect to the time intervals?
Siglent should be able to get thinks done as a 25y old HP-DMM....

[oz2cpu]

i will be happy to help,
please tell me the most easy methode to test/reveal the time interval issue,
it is something i did not dig into yet,
so i need a little bit for dummies how to :-) 

[Bad_Driver]

Very easy, connect DMM to EasyDMM, settings for faster reading (1 PLC or so), No AutoZero and record some minutes.

Export to EXCEL, normalize Time Stamp (I think you must do a division by 1E6 or so to get seconds) and calculate the time intervals between readings.

With the SDM3065 the intervals have a deviation of more than 50% from one to the other, with the HP34401 the deviation is about 2%.
Than you can try to use external triggering with about 10 Hz to see if there is some improvement.

[oz2cpu]

i assume i need to input a known signal, with a known repeat freq ?
also ethernet and my routers and switches and my pc involved, is not a realtime envoiment
and cant be thrusted to the last ms.
 

[Bad_Driver]

No input signal can be DC (our SPD maybe). You get only a reading if the SDM is ready.
I assume network connection is better than usb, but every kind of connection is fast enough for some hundred bytes per second.
You can also try the same with recording to usb-disk as well.