Even cheaper if you buy in US$.
Got mine for $24.69 = £16.23 at current PayPal rates.
http://www.ebay.co.uk/itm/400452242651?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1423.l2649
No mention of a certificate on any of the Ebay listings though.
even cheaper without the box @ US $15.99 (sold last month, don't know if more will be offered)
http://www.ebay.com/itm/New-AD584LH-4-Channel-2-5v-7-5v-5v-10v-High-Precision-Voltage-Reference-Module-/221210908532?pt=LH_DefaultDomain_0&hash=item3381329774They are a bargain with the calibration certificate, giving 6 usable digits, but without the calibration certificate , you could only use 4 digits. If you're looking for a unit to verify calibration, without the calibration certificate, you're a little short in accuracy to even check a 20,000 count Fluke 87. With calibration certificate, or if you have access to a calibrated device to give it 6 digits of accuracy, it's good for a 300,000 to 500,000 count meter. MJLorton's test was impressive with calibration certificate.
I considered what was on offer from a PentaRef ($61) to decade references and decided to watch out for decade references on Ebay as a better value for more versatility. Came across a 7 digit Kelvin Varley Divider (DAS57AL, $99.99), 0 to 1 volt DC and 0 to 10 volt DC, shown at auction dialed to 7.777700VDC and tested by an HP 34401A to 7.7776VDC, so I went for it.
Testing after receipt revealed that set to 10.00000VDC, both my HP 3478A units showed 10.0000VDC with no flickering, steady the whole day (still at 10.0000VDC 2 days later, never turned off, did drop down to 9.99999VDC or 9.99998VDC when the temperature plummeted 1 or 2 degrees). My Fluke 87III showed 9.995VDC. Even my old Philips nixie showed 9.95VDC. Will be trying to verify the DAS57AL to it's full 7 digits, but presently, I think I have confidence in it's ability to verify the HP 3478A's 300,000 count.